SPEECH AND FACSIMILE SCRAMBLING AND DECODING

• Document Type: CREST
• Collection: General CIA Records
• Document Number (FOIA) /ESDN (CREST): CIA-RDP10T02279R000100180001-2
• Release Decision: RIFPUB
• Original Classification: S
• Document Page Count: 141
• Document Creation Date: December 23, 2016
• Document Release Date: December 11, 2013
• Sequence Number: 1
• Publication Date: January 1, 1946
• Content Type: REPORT

Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100186601-2
130510
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
.
11144 t s i; t* *It
o
0
o
:0
!CD
C\1
C\1
0
vOrwelf
0
I 0_
a
<
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
SUMMARY TECHNICAL REPORT
OF THE
NATIONAL DEFENSE RESEARCH COMMITTEE
This document contains information affecting the national defense of
the United States within the meaning of the Espionage Act, 50 U. S. C.,
31 and 32, as amended. Its transmission or the revelation of its contents
in any manner to an unauthorized person is prohibited by law.
This volume is classified SECRET in accordance with security regula-
tions of the War and Navy Departments because certain chapters
contain material which was SECRET at the date of printing. Other
chapters may have had a lower classii.fication or none. The reader is
advised to consult the War and Navy agencies listed on the reverse of
this page for the current classification of any material.
eclassitiea ana pprove or e ease
SECRET
6c0-0(2&as)-4)
Ds+-r
. -
 Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Manuscript and illustrations for this volume were prepared
for publlication by the Summary Reports Group of the
Columbia University Division of War Research under con-
tract OEMsr-1131 with the Office of Scientific Research and
Development. This volume was printed and bound by the
Columbia University Press.
Distribution of the Summary Technical Report of NDRC has
been made by the War and Navy Departments. Inquiries
concerniing the availability and distribution of the Summary
Technical Report volumes and microfilmed and other refer-
ence material should be addressed to the War Department
Library,. _Room 1A-522, The Pentagon, Washington 25, D. C.,
or to the Office of Naval Research, Navy Department, Atten-
tion.: Reports and Documents Section, Washington 25, D. C.
Copy No.
149
This volume, hUie the seventy others of the Summary Tech-
nical Report of NDRC, has been written, edited, and printed
under great pressure. Inevitably there are errors which have
slipped past Division readers and proofreaders. There may
be errors of fact not known at time of printing. The author
has not been able to follow through his writing to the final
page proof.
Please report errors to:
JOINT RESEARCH AND DEVELOPMENT BOARD
PROGRAMS DIVISION (STR ERRATA)
WASHINGTON 25, D. C.
A master errata sheet will be compiled from these reports
and sent to recipients of the volume. Your help will make this
book more useful to other readers and will be of great value
in preparing any revisions.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
SUMMARY TECHNICAL REPORT OF DIVISION 13, NDRC
VOLUME 3
SPEECH AND FACSIMILE
SCRAMBLING AND DECODING
OFFICE OF SCIENTIFIC RESEARCH AND DEVELOPMENT
VANNEVAR BUSH, DIRECTOR
NATIONAL DEFENSE RESEARCH COMMITTEE
JAMES B. CONANT, CHAIRMAN
DIVISION 13
HARADEN PRATT, CHIEF
WASHINGTON, D. C., 1946
SECRET
N. C. A. RESEARCH LIBRARY
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
NATIONAL DEFENSE RESEARCH COMMITTEE
James B. Conant, Chairman
Richard C. Tolman., Vice Chairman
Roger Adams
Frank B. Jewett
Karl T. Compton
Army Representative'
Navy Representative2
Commissioner of Patents3
Irvin Stewart, Executive Secretary
nArmy representatives in order of service:
Maj. Gen. G. V.. Strong
Maj. Gen. R. C. Moore
Maj. Gen. C. C. 'Williams
Brig. Gen. W. A. Wood, Jr. Col. M. M. Irvine
Col. E. A. Routheau
Col. L. A. Denson
Col. P. R. Faymonville
Brig. Gen. E. A. Regnier
Wavy representatives in order of service:
!ear Adm. H. G. Bowen Rear Adm. J. A. Furer
Capt. Lybrand P. Smith Rear Adm. A. H. Van Keuren
Commodore H. A. Schade
3Comninissioners of Patents in order of service:
Conway P. Coe Casper W. Ooms
NOTES ON THE ORGANIZATION OF NDRC
The duties of the National Defense Research Committee
were (X) to recommend to the Director of OSRD snit-
able projects amd research programs on the instrumen-
talities ,of warfare, together with contract facilities for
carrying out these projects and programs, and (2) to
administer the technical and scientific work of the con-
tracts. More specifically, NDRC functioned by initiating
researcfri projects an requests from the Army or the
Navy, or on reqpiests from an allied government trans-
mitted through the Liaison Office of OSRD, or on its own
considered initiative as a result of the experience of its
members. Proposals prepared by the Division, Panel, or
Committee for research contracts for performance of
the won't involved in such projects were first reviewed
by NDECC, and if approved, recommended to the Director
of OSRID. Upon, approval of a proposal by the Director,
a contract pernnitting maximum flexibility of scientific
effort ,;was arranged. The business aspects of the con-
tract, including such matters as materials, clearances,
vouchere, patents, priorities, legal matters, and admin-
istration of patent matters were handled by the Execu-
tive Secretary off OSRD.
Originally NIDRC administered its work through five
divisions, each }headed by one of the NDRC members.
These were:
Division A?Armor and Ordnance
Division B?Bombs, Fuels, Gases, & Chemical Problems
DivishonC?Ccommunications and Transportation
Division D?Detection, Controls, and Instruments
Division E?Ptatests and Inventions
iv
In a reorganization in the fall of 1942, twenty-three
administrative divisions, panels, or committees were
created, each with a chief selected on the basis of his
outstanding work in the particular field. The NDRC
members then became a reviewing and advisory group
to the Director of OSRD. The final organization was as
follows:
Division 1?Ballistic Research
Division 2?Effects of Impact and Explosion
Division 3?Rocket Ordnance
Division 4?Ordnance Accessories
Division 5?New Missiles
Division 6?Sub-Surface Warfare
Division 7?Fire Control
Division 8?Explosives
Division 9?Chemistry
Division 10?Absorbents and Aerosols
Division 11?Chemical Engineering
Division 12?Transportation
Division 13?Electrical Communication
Division 14?Radar
Division 15?Radio Coordination
Division 16?Optics and Camouflage
Division 17?Physics
Division 18?War Metallurgy
Division 19?Miscellaneous
Applied Mathematics Panel
Applied Psychology Panel
Committee on Propagation
Tropical Deterioration Administrative Committee
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
1
0
NDRC FOREWORD
AS EVENTS of the years preceding 1940 re-
vealed more and more clearly the serious-
ness of the world situation, many scientists in
this country came to realize the need of organ-
izing scientific research for service in a na-
tional emergency. Recommendations which they
made to the White House were given careful
amd sympathetic attention, and as a result the
National Defense Research Committee [NDRC]
was formed by Executive Order of the Presi-
demt in the summer of 1940. The members of
NIDRC, appointed by the President, were in-
structed to supplement the work of the Army
and the Navy in the development of the instru-
mentalities of war. A year later, upon the
esttablishment of the Office of Scientific Re-
search and Development [OSRD], NDRC be-
came one of its units.
The Summary Technical Report of NDRC is
a (conscientious effort on the part of NDRC to
summarize and evaluate its work and to present
it in a useful and permanent form.. It comprises
sonne seventy volumes broken into groups cor-
responding to the NDRC Divisions, Panels, and
Committees.
The Summary Technical Report of each Di-
vision, Panel, or Committee is an integral sur-
vey of the work of that group. The first volume
of each group's report contains a summary of
thre report, stating the problems presented and
tine philosophy of attacking them, and summar-
iziing the results of the research, development,
anid training activities undertaken. Some vol-
umes may be "state of the art" treatises cover-
ing subjects to which various research groups
have contributed information. Others may con-
tain descriptions of devices developed in the
lalboratories. A master index of all these di-
viional, panel, and committee reports which
together constitute the Summary Technical Re-
port of NDRC is contained in a separate vol-
ume, which also includes the index of a
microfilm record of pertinent technical labora-
tory reports and reference material.
Some of the NDRC-sponsored researches
vvlhich had been declassified by the end of 1945
were of sufficient popular interest that it was
found desirable to report them in the form of
nronographs, such as the series on radar by Di-
vision 14 and the monograph on sampling
inspection by the Applied Mathematics Panel.
Since the material treated in them is not dupli-
cated in the Summary Technical Report of
NDRC, the monographs are an important part
of the story of these aspects of NDRC research.
In contrast to the information on radar,
which is of widespread interest and much of
which is released to the public, the research on
subsurface warfare is largely classified and is
of general interest to a more restricted group.
As a consequence, the report of Division 6 is
found almost entirely in its Summary Technical
Report, which runs to over twenty volumes. The
extent of the work of a Division cannot there-
fore be judged solely by the number of volumes
devoted to it in the Summary Technical Report
of NDRC; account must be taken of the mono-
graphs and available reports published else-
where.
Of all the NDRC Divisions, few were larger
or charged with more diverse responsibilities
than Division 13. Under the urgent pressure of
wartime requirements, the staff of the Division
developed navigation and communications de-
vices and systems which not only contributed to
the successful Allied war effort, but which will
continue to be of value in time of peace in the
fields of transportation and communications.
The work of the Division, under the direction
first of C. B. Jolliffe and later of Haraden Pratt,
furnishes a foundation for what promises to be
even more radical developments than those of
the war?for one example, direction finders
which will operate at all elevations and azimuth
angles, in other words, hemispherically.
The Summary Technical Report of Division
13 was prepared under the direction of the Di-
vision Chief and authorized by him for publica-
tion. The report presents the methods and
results of the widely varied research and de-
velopment program, and, in the case of work
with speech scrambling and decoding, it pre-
sents for the first time a comprehensive review
of the state of the art. The report is also a
notable record of the skill and integrity of the
scientists and engineers, who, with the coopera-
tion of the Army and Navy and Division con-
tractors, contributed brilliantly to the defense
of the nation. To all of these we express our
sincere appreciation.
VANNEVAR BUSH, Director
Offi-ce of Scientific Research and Development
SECRET
J. B. CONANT, Chairman
National Defense Research Committee
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
FOREWORD
FiARLY IN OCTOIRER 1940 a subcommittee on
speech secrecy was set up in the Communi-
cations Division of the National Defense Re-
search Committee [NDRC], later known as
Division 13 of NURC. This group was to con-
sider both the scrambling and unscrambling of
telephone signals. It was soon recognized that
the decoding problem was of primary impor-
tance both as a means of evaluating privacy
systems for possible use by the Services and for
decoding possible enemy signals. Thus the work
under Division 13 on speech privacy took on two
aspects, that of providing "secure" means of
voice, code, and picture communication, and
that of decoding transmissions for the dual
purpose of learning how useful were our own
secrecy systems and for learning what the
enemy might be saying in his scrambled mes-
sages.
Numerous projects were carried forward
under the guidaince of the Division, some of
them with the object of developing scrambling
or secrecy methods, others dealing entirely with
code cracking methods_ In this way the work
accomplished on methods of secrecy could be
tested at every phase by the decoding groups
which were also busy on other aspects of the
decoding problem. The summaries to follow will
disclose the objects and accomplishments of the
individual projects.
Work on speech secrecy problems went on in
the Division from 1940 quite to the end of the
war, the final report on Project 13-106 being
issued on August 18, 1945. Thus, this volume
summarizes the results of about five years' ex-
perience in developing secrecy systems and in
diagnosing, decoding, and evaluating these and
other systems submitted for study by the Army,
the Navy, and NDRC.
All these stuctiies are reported in detail in the
preliminary andl final reports on the projects
and are only stummarized here. The sum of
these reports, tlherefore, provides a record of
accumulated experience, much of which has
never been reported in any other way. In toto,
the reports make available information both
positive and negative which would have to be
accumulated by another group if they were to
embark on a similar project.
The immediate pressure behind these studies
was caused, of course, by the war. The work
summarized in this volume, therefore, should
serve as a guide to the individual reports and
should aid a newcomer to the field in becoming
properly oriented to the state of the art so far
as the NDRC project reports are concerned.
Other work, carried out in other government
groups, is not believed to duplicate this material.
In passing, it is worth noting that, in con-
trast to a rather extensive literature on code
and cipher systems, and on cryptanalysis and
cryptography, which apply to telegraphic types
of communication, very little has been written
on speech privacy systems or decoding methods
applying to them. Two moderately comprehen-
sive articles have been published. One appeared
in the Post Office Engineers Journal, October
1933; the other, in the Brown Boveri Review
for December 1941. The latter report is repro-
duced and discussed in Preliminary Report No.
5, Project C-43. It covers a number of basic
types of scrambling systems and, in addition,
discloses one that was new at the time. This is a
modification of the "time division scrambling"
or TDS system, and several of the Division 13
projects were concerned with this useful
privacy method.
Without any doubt, one of the major accom-
plishments of the Division's work and certainly
the high point in its efforts within the subject
matter of this volume was the development of
the sound spectrograph by Mr. R. K. Potter and
his co-workers at the Bell Telephone Labora-
tories. The efficacy of this instrument in ana-
lyzing scrambled speech will be revealed in the
chapters to follow.
SECRET
HARADEN PRATT
Chief, Division 13
vii
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
PREFACE
I N SUMMARIZING the several hundred reports
of contractors on the hundred-odd research
projects sponsored by Division 13 of the Na-
tional Defense Research Committee, the editor
has had to settle in his own mind how much or
how little of each project report should be in-
cluded; in ether words, how far the boiling-
down process should go.
The editor has an abhorrence for seeing good
scientific or technical material go unpublished.
Only by publication can the facts or methods
developed lay a few researchers become avail-
able for all researchers. On this basis, substan-
tially all of Division 13's program should be
included in the volumes, of which this is one,
summarizing the work of the Division. On the
other hand, time moves forward inexorably so
that it is qu4e likely that, by the day of publica-
tion, 'much 4of the data would already be out
of date. Furthermore, time and human energy
are always scarce. On these bases, all that might
be required -would be a paragraph or two sum-
mariiing flue aims of the project and its accom-
plishments.
A middle c ourse was steered, a course between
the easiest Solution of publishing practically all
cf each report and the more difficult job of
really digesting the project purpose and results.
The editor, however, deliberately chose to pub-
lish too much rather than too little. In most
cases it will be unnecessary for the reader to
search out the original source material unless
he wishes to dig deep into the subject. In those
cases where fundamental information was as-
sembled and printed in the project report, that
is, information on which future research might
be based, the summaries have been permitted to
take as much space as required.
The plan followed in this volume is briefly
as follows: After a comprehensive descrip-
tion of the several scrambling methods, brief
summaries are given of the work carried out in
the several projects dealing with scrambling.
Then follows a comprehensive description of
decoding or cracking methods employed, after
which are summaries of the projects dealing
mainly with decoding or code cracking. In other
words, most of the space is devoted to basic ma-
terial and the least space to details of the actual
work carried out. Thus this volume might be
considered as a basic text on speech scrambling
and descramhling.
The material is not arranged chronologically
with respect to the order of the work in the Di-
vision but, rather, in an arrangement which
appeared to the editor as giving the reader the
easiest and quickest approach to the whole sub-
ject of speech privacy and code cracking. Thus,
parts of the individual project reports will be
found in several places in this volume.
KEITH HEN NEY
Editor
SECRET
ix
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
CONTENTS
CHI A PTER PAGE
1 Speech Scrambling Methods
.2 Time Division Scrambling Systems 13
3 Speech Privacy System Development 25
4 Unscrambling and Decoding Methods 35
.5 Decoding Projects ? 100
6 Facsimile Privacy Systems 105
7 Miscellaneous Projects . 120
Bibliography 125 -0.SRD Appointees 127
Contract Numbers 128
Service Project Numbers 129
Index . 131
SECRET xi
Declassified and Approved For Release 2013/12/11: CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Chapter 1
SPEECH SCRAMBLING METHODS
AWIDE VARIETY of speech scrambling methods
will be examined in this chapter, taken
from Part I of the final report' of Project
C-43, in order to become familiar with the
devices which might be used alone or in combi-
nation to make up speech privacy systems.
Some. of these systems are in commercial or
military use, others exist only on paper, mostly
in the form of patents or patent applications.
It is not intended to include all the variations
of all the different methods but rather to cover
basic scrambling methods, with their most im-
portant variations, in which the original speech
is transmitted With its parts modified, dis-
placed, or interchanged.
The two main dimensions of speech which are
operated upon to make it unintelligible are the
frequency dimension and the time dimension.
Scrambling systems usually depend on re-
arranging the components of speech in either
or both of these dimensions. In general it may
be said that those that operate on the frequency
dimension alone are capable of the best quality
in the reproduced speech. A complete list of
the systems covered in the discussion is given
in Table 1 of Chapter 4, together with other
data concerning them.
1.1
SYSTEMS INVOLVING SINGLE
MODULATION
A basic device in privacy systems is the
modulator. One form of modulator, shown in
the coils as shown. In some cases the coils can
be omitted as shown in Figure 1B.
Figure 2 shows the method of producing
simple inversion. In this and in succeeding
illustrations the numerical values are not neces-
sarily the best values for practical operation,
but they serve to illustrate the manner in which
the device operates.
In the system shown in Figure 2 the speech
band ?is limited to 3 kc by a low-pass filter. It
is then modulated with a frequency of 3 kc.
This produces an upper and a lower sideband of
which only the latter is passed by the output
filter. The system is called inversion, because
the high frequencies in the original speech
appear as low frequencies in the output and
the low frequencies in the original speech
appear as high frequencies. At the receiving
end the inverted signal, in passing through an
identical system in the same direction, is re-
inverted back to normal speech.
A very commonly proposed variation of this
system involves using a variable frequency in-
stead of the steady 3-kc carrier. We might vary
the frequency continuously or in discrete steps.
It should be noted, however, that the cutoff of
the low-pass output filter is fixed, which limits
the variation permissible in the carrier fre-
quency. A wide variation would either permit
too much of the upper sideband to get through
or would cut off some of the lower sideband.
If the modulator in Figure 2 is of the type
shown in Figure 1A, speech can be scrambled
by introducing instead of the 3-kc carrier a
IN
CARRIER
OUT
8
FIGURE 1. Use of copper oxide varistors as modulators. In A, balanced coils are necessary; in B, coils are
not required.
Figure 1A, consists of four copper oxide
varistor units between two balanced coils. The
carrier frequency is fed into the midpoints of
square wave whose changes from positive to
negative value are irregular in time. Each one
of the reversals in the carrier wave causes a
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
4
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
2
SPEECH SCRAMBLING METHODS
reversal of phase in the speech wave. The pat-
tern of these? irregular reversals may be ar-
ranged so that the speech becomes unintel-
ligible. At the receiving end a coding wave must
be introduced: which is exactly in step with the
consists of the speech band right side up but
displaced from its normal position by 2 kc. If
the 16-kc value is used the output consists of
the 3-kc speech band inverted and displaced by
3 kc. We might use these two values alternately
at short intervals, or we might have the carrier
LP
30 00
MOD
LP
3000,m
7.10-1
D
LP
3 KC
3000 rti
FIGURE 2. Simple inversion produced by single
modulation.
one at the receiving end with proper alloWance
for any delay- there may be in the transmitting
channel.
A two-chamnel system using one modulator
for each chamnel is shown in Figure 3. In this
system the carrier fed into both modulators
is the same im frequency but differs 90 degrees
in phase. Twco separate speech channels can he
transmitted by this method without substantial
mutual interfeerence, but both sidebands as well
as the carrier- must be transmitted. At the re-
ceiving end tthe carrier must be split into two
components with the proper phases. Each com-
ponent will diemodulate its own portion of the
signal and dhereby separate the two speech
channels. On of the channels may consist of
noise or spurious speech from a recording or
the like, whical tends to mask the real message
if the signal ds demodulated with an ordinary
set. This scheme was originally proposed as a
multiplex sysltem, but an obvious variation is
to divide a siingle speech band into two halves
with filters amd then transmit the two halves
on carriers diiffering by 90 degrees in phase.
1.2 SYSTEMS INVOLVING DOUBLE
MODULATION
Figure 4 shiows a much more flexible systern.
Here the signial is modulated twice, with a
band-pass flitter between the two modulators.
With this arrrangement the carrier frequency
fed into the szecond modulator can be varied in
several ways._ In the illustration two carrier
frequencies aire shown for the second modu-
lator. If the 8-kc value is used the output
3LC
MOD
OSCILL 3 KC
PHASE SHIFTER
FIGURE 3. Split-phase multiplex used as two-
channel system.
vary continuously back and forth, say between
13 and 16 kc. Another variation is to use a
multiplicity of values, for example, 500 or 1,000
cycles apart, (not between 10 and 13 kc for
this illustration) and switch between these
values in a regular or irregular sequence. A
MOO
10% KC
MOD'
LP
6 KC
CI) 10 PLC (1318 TO 16 KC
FIGURE 4. System in which signal is modulated
twice, requiring wider frequency band than is
usually provided by radio sets or telephone lines.
disadvantage of these systems is that the trans-
mission channel needs to be wider than that
usually afforded by radio sets or telephone
lines. In all these systems, the speech is restored
by passing through identical equipment in the
opposite direction.
3-3 TRIPLE MODULATION?RE-ENTRANT
BAND SHIFT
Going back to Figure 2, suppose the carrier
frequency were made 4 instead of 3 kc, but
the 3-kc input and output filters were retained.
The output would then be an inverted sideband
ranging from 1 to 3 kc; that portion of the
sideband above 3 kc would be cut off by the
output filter. Since, however, there is a 1-kc
gap at the lower edge of the transmitted band,
the portion which would be cut off by the filter
might be modulated down and sent along with
the rest of the signal in this lower part of the
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
4,
BAND-SPLUTTING SYSTEMS
3
spectrum. In other words the portion of the
sideband which would otherwise disappear
above the upper edge of the transrmitted band
might be made to reappear at the bottom.
L P
3 KC
P.
output low-pass filter. A variation of this
arrangement is to allow the 7-kc carrier to vary
in discrete steps according to some regular or
irregular program or vary it continuously be-
MOD
NIP
31, KC
MOD
CID 7 KC
GD3 KC
Be
3-6 KC
MOD
FIGURE 5. System iinvolving re-entrant
Figure 5 shows ,a system of moduilators and
filters for accomplishing this kind of hand shift.
The first modulator is followed by at high-pass
filter which selects the upper sidefband from
3 to 6 kc. This is combined -with some of the
original signal which ranges from zero to 1 kc.
The second modulator is fed with a carrier
frequency of, for example, 7 kc which inverts
the whole band. This is followed by a band-pass
filter passing the range from 3 to 6 ic. A third
modulator with its carrier frequency placed
INPUT
inversion.
LP
3 KC
tween the limits of 6 to 9 kc. This provides a
variable band-shifting arrangement without
using more than the normal 3-kc transmission
channel.
1.4 BAND-SPLITTING SYSTEMS
A privacy system in wide commercial use,
known as the split-band system, involves split-
ting up the whole speech band into a number
SWITCH
6100".
OUTPUT
MOD
e 00 OR
650.
5550,4
MOD
5550 OR
1200,0
5000
MOD
5000 OR
1750.
4450",
3900"...
MOO
3900 OR
2650 r?,
* op FIALTERS ALL ALIKE ?3100 TO 3650 "4
FIGURE 6. One form of split-band system in commercial use.
at the lower edge of the 3- to 6-ke input band
moves the whole band, still inverted, down to
the usual range of zero to 3 kc. The upper side-
band of this modulation step is removed by the
of subbands and shifting these around out of
their normal positions in the frequency spec-
trum. Figure 6 shows one manner in which
this can be accomplished. The numerical values
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
4
SPEECH SCRAMBLING METHODS
are chosen so that the band from 250 to 3,000
cycles is divided into five subbands each 550
cycles wide..
The speech band is fed to five modulators
in parallel. The five band filters following the
modulators are all alike, passing the band from
3,100 to 3,650 cycles. It will be seen that the
uppermost modulator in Figure 6 with its
carrier of 6.1 kc will invert the speech band
and displace it by such an amount that the
frequency band which originally occupied the
space from 2,450 to 3,000 cycles will pass
through the filter. In other words, this modula-
tor in combination with its band filter selects
MOD
5800,u
MOD
5200"
BP*
An additional set of frequencies is indicated
in the drawing for the second set of modu-
lators. These frequencies will cause the output
subbands to be inverted instead of right side
up. One or all of these alternate frequencies
may be used as desired.
The "switch" may be changed as often as
desired. On an experimental basis the codes
have been changed as often as 25 times per
second without appreciable distortion in the
quality of the received speech, showing that it
is possible to shift bands as wide as 550 cycles
at a rapid rate without generating appreciable
distortion products.
MOD
4600,,
BP*
MOD
3600,,
MOD
4000'.
BP*
MOD
3500,u
BP*
MOD
MOD
3600'.
* BP FILTERS ALL ALIKE -3000 TO 3600,??
FIGURE 7. Time division multiplex in which N separate signals are sent over line, each signal being
transmitted only 1/Nth of time.
the uppermost of the five subbands from the
input signal. Similarly the lowest modulator
in combination with its band filter selects the
lowest subband from the input signal. The out-
puts of the band filters all occupy the same
frequency range, but they all came originally
from different frequency ranges. Similarly the
output modulators are so designed that each
one accepts the band from 3,100 to 3,650 cycles
and shifts it to a particular band location in
the output. The five leads going into the box
labeled "switch" may, therefore, be cross-
connected in any desired manner with the five
output leads. The resulting output will always
cover the complete range from 250 to 3,000
cycles and there will be no overlapping sub-
bands.
1.5
TIME DIVISION MULTIPLEX
Time division multiplex [TDM] is a system
in which N separate signals occupying the same
frequency range are sent over a single line,
each signal being transmitted only 1/Nth of
the time. This might be illustrated by showing
the N signals connected to the N segments of
a commutator. A rapidly rotating brush picks
up the N signals one after the other. For
acceptable quality, however, the brush must
make at least as many rotations per second as
the highest frequency in the transmitted signal.
This means that a mechanical brush is out of
the question and is used simply for illustration.
This kind of switching, however, can be
accomplished with electronic ring circuits.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
 Afformermemssmos- -
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
TIME DIVISION SCRAMBLING 5
Sirnce we are interested here in privacy sys-
tems rather than multiplex systems, we will
confirne ourselves to the use of TDM for trans-
mittirng a single speech channel. This can be
acconnplished by dividing the speech band into
a nurniber of subbands all occupying the same
frequency range, and connecting these to the
segments of our hypothetical commutator.. Re-
ferrimg to Figure 7, which is similar to Figure
6, thiis can be accomplished by feeding all the
outpult modulators with the same carrier, and
connecting each modulator to a commutator
segment. In this illustration, there are four
600-czyc1e subbands, oovering the range from
400 to 2,800 cycles. It has been shown mathe-
matically that the output of this system
consists of sidebands around a frequency oorre-
spondling to the rotation of the brush and also
sidelmands around frequencies corresponding to
odd Inarmonics of the rotation frequency. Each
sidebland, hovkver, contains components from
each of the subbands.. It has also been shown
that the total channel width required for good
transmissitm need be no greater than that of
the orriginal signal.
To, increase the privacy of this system one
of the subbands may be replaced by a band of
noise._ This can be filtered out at the receiving
end. Obviously this system requires a high
degree of synchronism between the two ends.
" SYSTEMS USING TAPE RECORDING
Leraving the frequency substitution systems
for tine time being, we will introduce a device
which permits operating on the time scale.. The
most versatile device for this purpose is the
magmetic tape recording and reproducing sys-
tem. This takes the form of a tape of magnetic
alloy a few mils thick either run as a loop over
pulleycs or attached firmly to the perimeter of
a dislk. The recording is done by means of small
electromagnetic pole-pieces. The signal is
picked up by similar pole-pieces which may be
placed at a distance from the recording pole-
piece depending on the amount of delay desired.
The outstanding advantage of the magnetic
tape system for this type of application is that
the Signal may be erased and the recording
medium be used over and over again. The
quality of this type of transmission can be made
very good with proper design.
Figure 8 shows a rather simple privacy sys-
tem using magnetic tape. The input signal is
passed through a 3-way pad, whereby it is
impressed on a band filter, and also recorded
FIGURE 8. Variable subband delay system using
magmetie tape.
on the rnagnetic tape. It is picked up by equally
spaced pole-pieces each associated with a dif-
ferent band filter. With the arrangement shown
in Figure 8 the band from 0 to 1 kc is trans-
mitted without delay. The band from 1 to 2 kc
is transmitted with 100 msec (0.1 sec) delay
and the band from 2 to 3 kc is delayed 200 msec.
At the receiving end the scrambled signal is
passed through an identical system in the same
way except that the two extreme band filters
are interchanged. In this way the band which
received no delay in transmission is given
maximum delay in the receiving machine, and
the band which received maximum delay in
transmission is given zero delay in the receiver.
In this way all the bands are delayed the same
amount and the speech is restored to normal.
This system alone does not provide any high
degree of privacy, but it can be combined with
other systems, as we shall see.
1.7
TIME DIVISION SCRAMBLING
An important class of scrambles involving
magnetic tape is known as time division
scrambling [TDS]. A simplified diagram of this
system is shown in Figure 9. There are a re-
cording pole-piece and a number of pickup pole-
pieces. There is also a commutator driven in
synchronism with the tape. The length (in
time) of each segment of the commutator is,
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
6
SPEECH SCRAMBLING METHODS
in general, equal to the delay between succes-
sive pickup pole-pieces. However, the number
of segments need not be the same as the number
of pole-pieces. A switch is provided whereby
any segment may be connected to any pole-
piece.
TAPE
Co)
4E RASE
FIGURE 9. One form? of time division scrambling
[TDS].
With this system the speech is cut up into
time elements corresponding in length to the
spacing of the pole-pieces. These time elements
are transmitted in a scrambled order. For
instance, six successive time elements which we
might label 1, 2, 3, 4, 5, 6 might be transmitted
in the order 2, 4, 1, 3, 6, 5. The possibilities
of TDS coding are far too complex to cover
here. The general requirements for all TDS
systems may be stated as follows: (1) Each
element of the original speech must be trans-
mitted once and only once. (2) The sum of
the delay in the transmitting machine plus
the delay in the receiving machine must be
equal for all elements. With these two require-
ments fulfilled it is obvious that the speech
comes out of the receiving machine in its
normal order. It is delayed, however, by an
amount equal to the sum of the transmitting
and receiving delay.
At the receiving end there are several ways
of handling the scrambled signal. (1) The pick-
up pole-pieces can be used as ?recording pole-
pieces and the signal picked up by an additional
pole-piece shown at the right in Figure 9.
With this arrangement the connections between
the commutator and the pole-pieces are the
same in the transmitting and receiving ma-
chines. (2) The signal can be recorded with
the same pole-piece used in the transmitting
machine and the connections between the pole-
pieces and the segments rearranged for re-
ceiving by a push-to-talk relay. (3) The codes
can be restricted to a particular class called
self-converse codes. These have the property of
being self-decoding, that is, the same code
which scrambles the speech in the transmitter
restores it in the receiver.
1.8
INTERLACE
An important variation of this system is
called "interlace." In this system the number
of segments on the commutator is doubled.
The odd segments are connected to the pole-
pieces according to one code and the even
segthents are connected according to a com-
pletely independent code. The reason for this
device is to increase the difficulty encountered
by the enemy in trying one code after the
other to find the right one, particularly if the
total number of codes available is small. With
the interlace system the total number of com-
binations possible is equal to the square of the
number of codes.
IN
TAPE
WHEEL
FLEXIBLE LEAD
OUT
FIGURE 10. Speed wobble in which speech time
scale is alternately compressed and expanded.
The rotating commutator shown in Figure
9 results in a repeated code, that is, each rota-
tion produces the same scramble. It is possible
to substitute for the commutator and switch
arrangement, shown in Figure 9, a more com-
plex arrangement whereby the speech is
scrambled in a never repeating manner. There
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
COMBINATIONS OF TIME AND FREQUENCY SCRAMBLING
7
are several ways of accomplishing this. Perhaps
the simplest way to represent it is as a punched
tape which permits the pole-pieces to be con-
nected to the output, one at a time, in any
desired order permissible under the restrictions
outlined above.
Another way of utilizing magnetic tape to
scramble speech is shown in Figure 10. Here
the pickup pole-piece is oscillated back and
forth along the tape mechanically. With this
arrangement, or other variations equivalent to
speech changes, the speech time scale is alter-
nately compressed and expanded. The fre-
quency scale is correspondingly expanded and
compressed, respectively.
With the arrangement shown in Figure 11,
speech is broken up into time segments each
s44E DIRECTION -AS TAPE
BUT TWICE AS FAST
FIGURE 11. Time inversion in which speech is
broken up into time segments, each of which is
transmitted backwards.
of which is transmitted backwards. The motion
of the pickup pole-pieces is twice as great as
the motion of the tape and is in the same
direction. Therefore, the relative motion of the
tape and the pole-pieces is the reverse of that
used in recording. This is the same as running
the tape backwards for reproduction.
1.9
COMBINATIONS OF TIME AND
FREQUENCY SCRAMBLING
Obviously the two kinds of systems described
in the previous sections can be used together.
For instance, some of the time elements of a
TDS system might be inverted according to
a regular or irregular program. The next more
complex step is to combine the band-splitting
system of Figure 6 with the TDS system. The
codes of the band-splitting system might be
fixed or might be switched in synchronism with
SPEECH(S)
MULTIPLIER
SVC S XC
CODE WAVE
(C)
MULTIPLIER
FIGURE 12. Multiplication system in which speech
wave is multiplied by coding wave.
the TDS elements, the time scale of the
scrambled speech not being further broken up.
If they are switched nonsynchronously, how-
ever, the time dimensions will be further broken
up, as will be seen subsequently. Combinations
of nonrepeated code TDS and rapidly switched
split-band coding can be made to afford a very
high degree of privacy. The two kinds of
coding, of course, must not be so interrelated
that one furnishes clues for the other. If, for
instance, a certain pole-piece were syste-
matically associated with a certain split-band
code the total privacy of the combination might
be impaired rather than enhanced. A coding
method for avoiding this difficulty is described
in Preliminary Report2 No. 21 of Project C-43.
A very special kind of scramble is produced
by a system which consists functionally of
Figure 6 (rapidly switched) in tandem with
Figure 8 (with five bands) followed by an
additional Figure 6: This is not the simplest
form of the system, but it serves to illustrate
the principle. Two frequency scrambles with
a time shift in between produce a particular
kind of two-dimensional scramble in which the
speech is broken up into both time and fre-
quency elements. Each of these elements may
be shifted both in time and in frequency so
as to be out of proximity with other elements
with which they were originally associated
either in time or in frequency. Another way
of accomplishing this kind of scramble would
be a combination of rapidly switched split band
with a separate TDS system in each subband.
A two-dimensional system was described in
the Brown Boveri article3 reproduced in Pre-
liminary Report No. 54 and analyzed in Pre-
liminary Report No. 93 of Project C-43.
For the sake of completeness two other sys-
SECRET
Declassified and Approved For Release 2013/12/11: CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
8
SPEECH SCRAMBLING METHODS
tems involving time and frequency shifting will
be mentioned, although as far as is known they
exist only on paper. Suppose a sample of speech
were recorded on tape and then reproduced
at twice the normal speed. It would occupy only
half the time it took to speak the words, but
I I
I I i
CODE WAVE eK)
I I 1
I I
I I
CODED SIGNAL p(S)
LILA 1641II
Pg 11
DECODING WAVE! ( )
KS
DECODED SIGNAL ( ?)
FIGURE 13. Example of multiplication system.
its frequency range would be twice the normal
range. Let the upper half of the expanded
frequency range be separated by a filter and
modulated down to the normal range and used
to fill up the unused time. The directly opposite
but analogous system would involve reproduc-
ing recorded speech at half its normal speed;
the frequency range would then be only half the
normal range. Alternate sections, therefore,
could be modulated up to fill the unused fre-
quency space, thereby keeping the total trans-
mitting time substantially unchanged. In both
of these systems, there would be a delay equal
to the length of one time element.
1.10 WAVE FORM MODIFICATION
1:21us far we have considered systems in
which frequency bands were shifted around
or time elements were rearranged. There are
a few privacy systems which make speech
unintelligible by a direct modification of the
wave form. One of these is shown diagram-
matically in Figure 12. It depends upon a
process whereby two waves are multiplied
together, that is, the instantaneous amplitude
of the resulting wave is the product of the
amplitudes of the two input waves (not the
sum or the difference as is the case in the simple
inversion methods described above). One of
the input waves to the multiplier is speech.
The other is a complex coding wave. If the
coding wave is sufficiently complex the result-
ing scramble is unintelligible. At the receiving
end a reciprocal of the coding wave is derived
and used as a multiplier, thereby restoring the
original speech. Naturally, the coding waves
at the two ends of the system must be in close
agreement, otherwise there will be considerable
VARIABLE GAIN AMP
CONTROL
CKT
FIGURE 14. Level modulation; form of wave form
alteration in which drastic changes in speech
levels occur.
background noise in the decoded speech. An
example of the multiplication system is shown
in Figure 13.
Another method for changing the wave form
is shown in Figure 14. The essential feature
of this system is an amplifier whose gain can
be varied rapidly with time. Drastic changes
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
MASKING SYSTEMS
9
in the level of speech, if they occur rapidly
enough, will make the speech unintelligible. The
level changes might be made according to some
program or they might be made to follow the
CONTROL CKTS
FIGURE 15. Level modulation system in which
speech is first divided into subbands, each of
which is subjected to level changes.
speech wave itself. For instance, extreme com-
pression or expansion could be used. Corre-
sponding gain changes, of course, must be made
at the receiving end.
A variation of this system is shown in Figure
15. Here the speech band is first divided into
subbands, and these are individually subjected
to level changes according to separate pro-
grams.
/N
MIXING
PAD
NOISE
GENERATOR
use very high levels of masking signals to hide
the intelligibility. This makes it difficult to
subtract out satisfactorily: the difficulties are
such that masking systems are more likely to
be found on wire lines than on radio. A few
speculative masking systems are outlined in
the following paragraphs.
One form of masking system is shown in
Figure 16. In this system, two telephone lines
are usNd. At the sending end, noise is added
to the speech in a mixing pad and the combina-
tion is sent over line 1. The noise alone is sent
over a second line and is used at the receiving
end to cancel the noise transmitted with the
speech by simple subtraction. This system has
the advantage that the noise can be completely
random. However, since the enemy might tap
both lines and thereby be able to make the
same subtraction, a variation of this system
consists in distorting the noise in some pre-
determined manner before sending it over the
second line. At the receiving end, this distor-
tion is first nullified so that the noise may be
subtracted. Naturally, the form of distortion
must be unknown to the enemy. It can, of
course, be varied from moment to moment.
Another masking system is shown in Figure
17, which uses only one line. In this system,
noise is added to the line at the receiving end
instead of at the sending end. Again, the noise
can be perfectly random. Since the noise is
generated at the receiving end, the process of
DISTORTION
FIGURE 16. Noise
MASKING SYSTEMS
LINE I
LINE 2
MIXING OUT
PAD
RECIPROCAL
DISTORTION
masking requiring use of two telephone lines.
cancellation can, theoretically, be made very
exact. This system, however, cannot be used for
radio at all because the level of the noise de-
creases with distance from the receiver, while
the level of the signal increases. The inter-
ceptor, therefore, will get good speech signals
if close to the transmitter. With telephone lines
this differential can be kept small.
One of the first schemes which is likely to
occur to a person considering how to make
speech private is to add noise or other disturb-
ing signal to the speech and remove it at the
other end, in other words, to mask the speech.
He will find, however, that it is necessary to
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
10 SPEECH SCRAMBLING METHODS
Another simple masking system is to have
a sequence of tones superposed on the signal
at the transmitting end. At the receiving end,
SPEECH ?D.
LINE
NOISE
MIXING
PAD
MIXING
PAD
NOISE
GENERATOR
FIGURE 17. Method of applying masking noise at
receiving end of communication circuit.
sharply tuned band elimination networks can
be synchronously switched so as to remove the The Vocoder system6, 7 may be made the
tones from the listener's ear. Similarly, short basis for privacy systems of various kinds. The
can be made to occur at irregular intervals
according to a never repeating program. Both
of these systems involve the loss of small por-
tions of the speech, either in the time scale or
the frequency scale.
A system described in Preliminary Report
No. 4 of Project C-43 might be classified as
a masking system, although it might be better
classified as a means of communicating without
the snemy's knowledge.
1.12
VOCODER SYSTEMS
1
BF .
m
1:>-
--
SF
C>.?????
BF
^
BF
,111.???
,-
BF
LINE
BF
SPEECH
TRANSMITS
INPUT
MODULATED
CARRIERS
BF"
????=???
BF
BF
..amen.?
BE
???????.
?????????
mim??>--
BE
BF
BF
BE
>
BF H
SF
BE
BE
BE
BF
I ?
BF
BE
????????
BF
BE
???.-?
BE
BF
.>'". SYNTI-..- . !C
BF
^
BE
BE
BE
di!
BUZZ.
BF
BF
SPEECH
OUTPUT
FIGURE 18. Vocoder system which may be made basis of privacy systems of various kinds.
spurts of noise covering the whole frequency system is shown schematically in Figure 18.
band can be applied at the transmitting end At the transmitting end the speech is passed
and shorted out at the receiving end. The spurts through a series of band filters, the outputs
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
CHANNEL-MIXING SYSTEMS 11
of which are individually rectified to form a
fluctuating d-c signal. These signals are indi-
vidually modulated in such a way that they can
all be sent over a single transmission path.
At the receiving end synthetic speech is
manufactured in accordance with the signals
transmitted over the line. A source of noise
which covers the whole frequency range is
passed through a set of band filters similar to
those at the transmitting end. The output of
each of these filters is controlled so that it is
the same level as the level of the speech in the
corresponding band at the transmitting end.
This is accomplished by separating the signals
in the various channels, detecting them and
using the resulting fluctuating direct current
to control the variable gain amplifiers in their
respective channels.
The noise is of two types, depending on
whether a voiced or unvoiced sound is to be
simulated. For an unvoiced sound, it is a hiss-
like thermal noise. For a voiced sound it is
a buzz which consists of a series of harmonics
covering the whole frequency range. A separate
carrier is used to transmit information for
operating this part of the system. At the trans-
mitting end the pitch used by the talker is
measured and this information is used to
control the pitch of the buzz sound. The absence
of a pitch signal switches the hiss sound into
the system.
This system by itself, of course, is not pri-
vate, since the enemy can build a similar system
and use the signals to regenerate speech.
Privacy must be achieved by operating on the
channel signals. One method is to permute the
channels at short intervals according to a
prearranged program. Another method is to
put a TDS system into the line, or into each
channel separately. A still more effective
method of this type is to apply a two-dimen-
sional scramble, such as was described earlier,
to the channels so that signal elements are
displaced in both time and frequency.
1.13
CHANNEL-MIXING SYSTEMS
Thus far, the methods we have examined
apply to a single transmission path. There is
another class of privacy system which depends
on using a multiplicity of paths. This is, of
course, inefficient if only a single message is
to be transmitted. However, the method can
be applied to cases where a number of channels
exist between two points and a number of mes-
sages would normally be transmitted over these
channels simultaneously.
CHANNEL 3
3, 4,I
FIGURE 19. Channel mixing in which multiplicity
of paths is involved.
Figure 19 shows one form of channel-mixing
system. Here three channels are shown
connected to the three segments of a com-
mutator. Three brushes on this commutator are
BF I
CHANNEL I
BF2
BF 3
BF I
CHANNEL 2
BF 2
BF 3
BF I
CHANNEL 3
BF 2
BF 3
SWITCH
FIGURE 20. Subband channel mixing.
connected to the outgoing channels which are
thereby caused to pick up one channel after the
other on a time division basis. Each channel
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
12
SPEECH SCRAMBLING METHODS
contains parts of messages from all three
channels. The commutator, of course, is too
simple to be very effective and would, in prac-
tice, be replaced by a permuting switch capable
of switching according to a more complex pro-
gram. One or more of the channels may be
filled up with the noise or spurious speech from
a recording or other similar source.
An analogous system which divides the mes-
sages on a frequency basis is shown in Figure
20. Here each channel is passed through three
band filters which divide the speech into sub-
bands. Each of the outgoing channels contains
subbands from each of the incoming channels.
To increase the privacy, a permuting switch
is shown which rearranges the subbands on a
time division basis. If only one message is to
be transmitted the other channels can be filled
in with noise or spurious speech.
1.14
SUMMARY
The above examples cover fairly completely
the range of schemes that might be used to
scramble speech at audio frequencies. In subse-
quent chapters we will examine each system
from the decoding standpoint. To facilitate
reference to the various systems, they are
summarized in Chapter 4, Table 1.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and A proved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Chapter 2
TIME DIVISION SCRAMBLING SYSTEMS
2.1 INTRODUCTION
THE TWO MAIN DIVISIONS or characteristics
of speech are the frequency dimension and
the time dimension. Either or both of these
components of speech may be rearranged or
altered before or during transmission to insure
privacy. At the receiver, the order of scram-
bling is reversed so that the original sense is
recovered.
A privacy system developed to a high order
during the war, known as time division scram-
bling [TDS], operates on the time character-
istics of speech. In this system successive
sections of speech, each m seconds long, are
divided into n short time elements, and these
n elements are sent in a scrambled time
sequence. The elements are much shorter than
a syllable, so that each word is cut up and
received as short bursts of energy in the wrong
order.
All the summaries which follow deal with
TDS systems.
4'2 PORTABLE TDS SYSTEMS?PROJECTS
C-1 AND C-1A
While several forms of speech scramblers
had been in use on radio circuits before 1940,
there was none which was small and light
enough to be suitable for mobile warfare. Work
done in the summer of 1940 at the Bell Tele-
phone Laboratories indicated that the need
might be met in the form of time division
scrambling by a small magnetic-tape recorder.
The purpose of Projects C-1 and C-1A, there-
fore, was to investigate the possibilities of pro-
ducing such a lightweight and effective privacy
unit based. on the TDS principle.
The fundamentals of this scheme were not
new. It was known that such privacy should
be very effective without appreciable expansion
of the original frequency band, but the idea
had not been developed for two reasons. First,
it could not be used for commercial telephony
because scrambling in time requires that the
speech be stored for a certain time which
introduced more delay than can be tolerated
by inexperienced users. Secondly, there were
several difficult technical problems - that had
not been solved.
Design ideas were collected by discussions
with various specialists and a particular ar-
rangement was visualized before the work was
started. This set an objective and helped in
segregating several problems that could be
handled more or less individually and simul-
taneously. The more important of these in-
cluded:
1. Development of a way to mount magnetic
tape on the edge of a disk without introducing
serious magnetic irregularities.
2. Development of a start-stop commutating
arrangement for rapidly switching magnetic
tape recorders and reproducers.
3. Development of a much more stable 24-
volt motor drive than was then available.
4. Development of a compact amplifier unit
to meet the special requirements of this system.
5. Development of a suitable switching ar-
rangement for setting up scrambled combina-
tions.
As the project progressed, the original de-
sign was changed where advisable and details
were added. The five months set for completion
of the project made it necessary to avoid any
suggested changes that would appreciably
delay the construction of models.
While development of experimental equip-
ment was in progress, a separate group was
investigating how various design factors would
affect possible requirements for future equip-
ment. These investigations were of two sorts.
The first was concerned with the degree of
privacy afforded by the equipment, and the
second with factors affecting the final quality
of the speech.
The degree of privacy was investigated as
a function of the number of scrambling inter-
vals within a cycle and the length of a cycle.
SECRET
Declassified and Approved For Release 2013/12/11: CIA-RDP10T02279R000100180001-2
13
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
*
14
TIME DIVISION SCRAMBLING SYSTEMS
In this connection a rather thorough study was
made of methods to "crack" the TDS privacy
under conditions ranging from limited facilities
and personnel to extensive laboratory equip-
ment handled by experts. The conclusions were
that cracking was rather unlikely unless
attempted by experts with special equipment,
and in this latter case the five-unit design
developed for experimental use might be
elements associated with the scrambling and
reconstruction process affected the final quality
of the speech. The most important conclusion
here was that the degree of stability afforded
by the special motor designed for the experi-
mental TDS unit met the requirements very
satisfactorily. The indications were that it
would not be difficult to provide a final design
which would introduce little degradation.
FIGURE 1. General view of TDS system developed under Project C-1.
cracked within minutes. In spite of this, it was
believed that the five-unit design would be very
effective when used on radio telephone circuits
for the direction of maneuvers that are com-
pleted within minutes. This would suggest use
for such links as plape-to-plane, .plane-to-
ground, and tank-to-tank.
The investigations concerned with speech
quality included such things as the way in
which accuracy of synchronism, start-stop
brush operation, and the position of various
With the experimental TDS unit produced
during the project about fifty useful combina-
tions could be set up on the five front-panel dial
switches. When used in association with radio-
telephone equipment, connections were required
to the transmitter, receiver, and power supply.
The system used 100 watts at 24 volts, but
would operate over a voltage range of approxi-
mately 22 to 30. The weight was about 40 lb
and the dimensions were 73/4 in. high, 9 in.
deep, and 20 in. long.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
PORTABLE TDS SYSTEMS?PROJECTS C-1 AND C-1A 15
2.2.1
Demonstrations
In March and April 1941, four experimental
TDS units were demonstrated in Washington
before members of the National Defense Re-
search Committee [NDRC] and representatives
of the Army and Navy. The units were set up
of a drum or wheel on the periphery of which
is mounted a wide magnetic tape alloy. Around
the wheel and in contact with the tape are
eleven pole-pieces, mounted so that all traverse
the same narrow band of the tape. The pole-
pieces comprise one eraser supplied with direct
current for magnetically saturating the tape to
FIGURE 2. Tape wheel of C-1 TDS unit showing placement of individual reproducers, etc.
first in separate rooms at the Carnegie Insti-
tution and later one unit was set up in the Navy
Building and another in the Munitions Building
with connections to telephone extensions in
both places. The equipment was demonstrated
to representatives of the Army and Navy as
well as to members of the British Military
Mission and the Canadian Air Ministry.
2.2.2 The TDS Units
The essential part of the TDS units consists
of the magnetic recorder-reproducer made up
obliterate previous recordings, one recorder
supplied with speech current, and a 1-ma d-c
depolarizing current and nine reproducer pole-
pieces spaced 36 degrees apart.
Associated with each reproducer pole-piece is
a segment of a commutator and a contact to a
switching system so that the order in which
the recorded speech culirents are taken off for
transmission by radio or wire can be changed
from the order in which the recording was
made.
The intervals of speech each 0.30 sec long
are divided into five sections of 0.06 sec dura-
SECRET
Declassified and Approved For Release 2013/12/11: CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
16
TIME DIVISION SCRAMBLING SYSTEMS
FIGURE 3. Commutator details.
tion. The instrument is described in consider-
able detail in the final reports of the project.
The final reports, 9 of Project C-1A, a con-
tinuation of Project C-1, describes methods of
cracking the TDS system, gives an estimate of
the time required to crack it, an evaluation of
the privacy secured by it, and recommendations
concerning future developments, some of which
resulted in Projects C-50 and C-65.
2.2.3
Conclusions
The project demonstrated that TDS could
provide a useful degree of privacy in a portable
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
CONTINUOUSLY CODED TDS?PROJECT C-50 17
device of reasonably small dimensions and
weight. Further developments to meet military
needs were pursued by the Services themselves.
The problem of speech privacy was attacked
in other NDRC projects along many other lines,
as evidenced by summaries of other Division 13
projects in this volume, but to the end of the
war TDS remained the only small and portable
PRESS TO TALK SWITCH
_
equipment by provisions for automatic code
changing every code cycle. In the winter of
1942 means for doing this were suggested and
Project C-501? was set up to develop these
means. The specific object was to provide Imodel
equipment so that the privacy obtained by
elaborating the TDS principle to its practical
limit could be studied.
?3
LATCH
BRUSHES
AMLLIFIER
2
RT
RT RT
..S11111.f.0.0315'
'
0 z
Cr 0 0?
Om
ct,
POWER
N I- SWITCH
m 0
Oz-
CODE
\ SWITCHES
- 7_
PULSING
CAM
RT
11T
TO RADIO TRANS
RELAY CIRCUIT
TO
FILAMENTS
00
MOTOR
2050
+ 250 V
0 0_
MOTOR
GENERATOR
FIGURE 4. Simplified circuit diagram of C-1 TDS system.
device for scrambling speech that was in fac-
tory production.
2.3 CONTINUOUSLY CODED TDS?
PROJECT C-50
The report on Project C-1A suggested an
improvement of security provided by TDS
VOLUME
CONTROL
HEAD SET
o
-z
o
CC
In the manually operated TDS system, the
nine pole-pieces which "take off" the recorded
segments of speech are connected to a commu-
tator and switch by which the operator can
change the order in which the successive ele-
ments are transmitted to the lines or to a radio
transmitter. The object in using different ones
of the various pole-pieces is to retard or delay
successive elements of the recorded message or
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
18
TIME DIVISION SCRAMBLING SYSTEMS
speech by different amounts. When two speech
elements which were originally in succession
are delayed by different amounts they are dis-
placed in time with respect to each other. They
may be merely interchanged with respect to
each other or they may be separated widely by
inserting other speech elements between them.
The scrambled speech elements are returned
to their proper order in the unscrambling proc-
42-t tIt]
Ft titi1111 LINE
0
0
og
of
08
0
Oc
00
IL
3 4 5 6 7 8 9 10
if /
COMMUTATOR
CODE
CONNECTOR
Code used is:
Commutator bars
Speech units
Coils
cording pole-piece, and an erasing pole-piece.
In the C-50 TDS system, each pole-piece is
capable of recording, of erasing, and of repro-
ducing. Such a system is known as a ten-ele-
ment system because ten speech elements are
scrambled in each code cycle. In the C-50 sys-
tem, two interlaced ten-element codes are em-
ployed in each code cycle.
If the interconnection of pole-pieces and
abc de 19h,
0-1. 1 I I 11111
1 2 3
4 2 1
beg
of
-x
X
X
oe
0 d
0
X
0 bX
a
X
X
4
6
5
8
6 7
3 10
h b
(I 234567891
8 9 10
5 7 9
h g f
CODE
CONNECTOR
COMMUTATOR
FIGURE 5. TDS coding connections. Transmit machine codes on "reproduce," while receive machine decodes
on "record."
ess by delaying each one in a selective manner
so that the total delay contributed by scram-
bling and unscrambling is the same for each
element. Elements scrambled with minimum
delay are unscrambled with maximum delay,
and vice versa. The speech, when reassembled,
has been delayed by the sum of the minimum
and maximum delays. In the C-50 TDS system
this total delay is 700 msec.
2.3.1
The C-50 TDS System
In the TDS units developed under Project
C-1, there were nine take-off pole-pieces, a re-
commutator segments remains fixed, the code
or key is fixed and the transposition scheme is
repeated with each revolution of the brush over
the twenty commutator segments. This takes
place in 750 msec. By means of punched code
cards, the C-50 TDS system can be operated in
this manner, if desired.
2.3.2 Continuous Coding
The continuous coding equipment makes two
changes in the interconnection of pole-pieces to
segments in each revolution. The connections
to odd-numbered segments remain fixed until
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
CONTINUOUSLY CODED TDS?PROJECT C-50 . 19
the commutator brush leaves the nineteenth
segment;? when the brush starts on segment
No. 1 an entirely different set of interconnec-
tions is used in the next revolution by the odd-
numbered segments. The pattern for the even-
numbered segments begins with the twelfth
segment and extends around the commutator
through the tenth segment; when the brush
required for the magnetic tape to move from
one reproducing pole-piece to the next. The
portions of speech available for reproduction
when the brush is on odd segments are, there-
fore, never available when the brush is on even
segments.
The patterns of interconnection between pole-
pieces and commutator segments must follow
FIGURE 6. Details of C-50 TDS Machine.
reaches the twelfth segment again a new pat-
tern is ready for the even segments. In this way
the coding for alternate segments constitutes
two independent systems of transpositions, each
changing every 750 msec but displaced from
each other by 375 msec.
The odd-numbered and the even-numbered
commutator segments can be treated separately
because the time taken for the brush to cover
one segment, 37.5 msec, is exactly half the time
certain rules in order that the TDS generate a
transposition which is valid, in the sense that
each speech element is transmitted once and
only once. Since speech on the magnetic tape
remains there as it goes past all nine reproduc-
ing pole-pieces, a random choice of reproducing
pole-pieces might pick up the same element of
speech twice or more, and some other element
would be omitted for each repetition.
The kinds of interconnections which give
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
20 TIME DIVISION SCRAMBLING SYSTEMS
useful codes are discussed in Appendix A of
the project final report. The rules worked out
in that appendix are basic in the design of the
automatic coding apparatus which must be so
arranged that the semirandom series of choices
called for in one part of the coding equipment
is scrutinized in another part and revised until
the interconnection finally set up is a valid
transposition.
2.3.3 Number of Codes Available
The complexity of the coding systems gen-
erated by the automatic coding equipment in
the C-50 TDS may be illustrated by the fact
that there are 1,625,702,400 different sequences
of codes for each of the two interlaced systems
or the square of this figure for the combination.
The choice of the particular sequence to be
used is governed by four punched cards. The
initial settings of ten selector switches then
determine the point in the sequence at which
the sequence will start. There are (3,282,972)2
such starting points for each sequence.
From the point of view of those operating
the equipment, the four punched cards and the
ten selector switch settings constitute the key,
since these choices must be agreed upon at both
ends of the circuit.
When the cards and initial settings have been
chosen, the equipment can be started, after
which it provides in each of the interlaced sys-
tems, an irregular order of valid codes from
a possible number of 60,316 codes. In a single
message, or even in a whole day's use with the
same initial settings, only an extremely small
fraction of any code cycle would ever be used.
Each particular sequence runs so long that it
would not begin to repeat before 6,400,000
years.
2.3.4 C-50 TDS Plus Frequency-Band
Switching
Although to the end of the project, decoding
methods developed under other Division 13
projects failed' to crack the C-50 system, it was
found that some trained individuals under fa-
vorable circumstances were able to understand
some of the scrambled speech. Less is under-
stood with fast talkers than with slow talkers
but the amount understood by such trained ob-
servers at normal speech rates leads one to the
conclusion that in a system intended to provide
long-term privacy the TDS principle should not
be used singly but in combination with other
principles.
A combined system should be so devised that
direct listening is of no value and should be
arranged so that an intercepter can not dis-
entangle the two types of scramble and obtain
intelligence by listening directly to one of them.
Such a system is provided by the C-50 TDS
combined with and controlling a rapidly
switched frequency-band shift system (A3) so
arranged that the C-50 TDS coding equipment
controls the sequence of A3 codes. In this ar-
rangement the A3 codes may be switched as
often as every 37.5 msec. No useful intelligence
had been extracted from the resultant scramble
in direct listening tests up to the time the final
report on Project C-50 was written. Further-
more the difficulty of decoding the scrambled
speech by other methods was expected to be
materially increased. Although the combined
C-50 TDS and A3 systems weighed about 2,200
lb and required a power of 1,500 watts, its
probable privacy and its availability were such
as to be recommended for use in corps-to-
division communications and in similar Navy
situations. This double scramble system is de-
scribed in Appendix F of the final report of
Project C-50. Actual work on such a system was
carried out under Project C-66.
2.3.5 Appendices to Report of Project C-50
The following titles of the several appendices
to the final report of Project C-50 give an indi-
cation of the material to be found in the report
itself but not summarized here.
Appendix A Basic TDS Principles
Appendix B Automatic Code Generating
Appendix C Number of Codes and Code Se-
quences
Appendix D Detailed Description of C-50
Equipment
Appendix E Operating Instructions
Appendix F C-50 and A3 Systems Combined
Appendix G Project C-50 Drawings
SECRET
Declassified and Approved For Release 2013/12/11: CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
FREQUENCY-TIME DIVISION SYSTEM?PROJECT C-66 21
2-4 FREQUENCY-TIME DIVISION SYSTEM?
PROJECT C-66
In 1942 it was known that the speech security
provided by the TDS units developed under
Project C-1 was short unless automatic coding
was used as developed in Project C-50. It was
known, too, that only short-time security was
provided by the A3 scramble based on permuta-
tions of frequency bands. There was reason to
believe that A3 in tandem with TDS might pro-
vide a much higher order of security and
Project C-66n was set up to study this.
2.4.1 Accomplishments of the Project
It was found that A3, switched even as often
as every 60 msec, in tandem with the portable
fixed-code TDS, gave an increase in privacy in-
commensurate with the circuit complexity.
When rapidly switched A3 was combined with
the continuously recoded TDS of Project C-50,
however, the degree of security attained ap-
peared to justify recommendation of the devel-
opment of the combination for truck-mounted
use in corps-to-division communication. So far
as is known, this recommendation was not acted
upon.
The actual security and the transmission
features of the combined systems may be sum-
marized as follows:
1. Although it is possible under some circum-
stances for trained observers to obtain some
intelligence from listening directly to the scram-
bled speech from either A3 or C-50 systems
alone, when these are combined practically no
intelligence can be obtained from the scram-
bled speech in this manner.
2. The unauthorized agent would require spe-
cial and complex analyzing equipment to restore
the scrambled speech from the combined sys-
tems. The cracking time had not been deter-
mined on the date of the final report (May 29,
1943). The C-50 TDS alone using nonrepeating
codes had been found very difficult to crack in a
form permitting reproduction of the message,
and it was the opinion of the personnel of
Project C-43 (which was largely concerned with
cracking methods and which became expert in
this technique) that the cracking of this TDS in
combination with and controlling the A3 would
be much more difficult, since the A3 codes ob-
scure the matching of the speech elements.
Even if this can be done successfully there ap-
pears to be a major development problem with
respect to reproducing the message from the
reassembled photographic traces.
3. The c-50 TDS plus AS provides transmis-
sion quality of a useful grade. Under limiting
conditions, however, the signal-to-noise ratio
must be about 7.5 db greater than that required
when the privacy system is not used. Under
average conditions articulation tests showed
that the C-50 TDS plus A3 system was equiva-
lent to a reference circuit without privacy hav-
ing a bandwidth from 250 to 3,000 cycles and a
signal-to-noise ratio of about 14 db.
2.4.2 Other Studies under Project C-66
The transmission quality of devices combin-
ing frequency scrambling with time delays pro-
vided by means of magnetic tape was found in
this investigation to depend upon control of the
following factors:
1. Modulation products giving rise to non-
linearity.
2. Flutter from speed variation in magnetic
tapes.
3. The signal-to-noise ratio of the overall sys-
tem.
4. The transmission frequency characteris-
tics.
5. Switching effects.
Detailed analyses of these controlling factors
are found in the final report of the project.
Among other systems and schemes for ob-
taining privacy by time and frequency scram-
bling considered in the course of Project C-66,
the following deserve mention:
1. A re-entrant frequency band shifter.
2. A two-dimensional scrambler employing
frequency band delay plus a re-entrant fre-
quency band shifter plus frequency band delays.
3. A two-dimensional scrambler employing a
re-entrant frequency band shifter plus fre-
SECRET
Declassified and Approved For Release 2013/12/11: CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
22 TIME DIVISION SCRAMBLING SYSTEMS
quency band delays plus a re-entrant band
shifter.
4. A two-dimensional scrambler employing
A3 privacy and five magnetic-tape delay circuits
of different delay values.
tion expires. If a mission of mobile units takes a
greater time than this protected period, changes
in the code should be made during the mission.
Only enough codes should be available to afford
the desired protection. More codes than neces-
FIGURE 7. Exterior view of Model B code-changing unit attached to D-150285 TDS unit.
2'5 CODE-CHANGING ATTACHMENT
FOR THE C-1 TDS UNIT?PROJECT C-65
The final report of Project C-1A" suggested
that the security of the TDS speech scramblers
would be improved by frequent changes of code
cards. In 1942 a variety of means had been sug-
gested for accomplishing this by an attachment
to the TDS unit (D-150285) then being pro-
duced for Army and Navy. Project C-65'2 had
as its purpose the development of such an at-
tachment. The background for this work is as
follows:
Since the security afforded by fixed-code pri-
vacy systems cannot be relied upon beyond the
period required by an unauthorized intercepter
to work out the code, it is desirable to change
the code before this minimum period of protec-
sary would provide the enemy with too much
information if a unit were captured.
Under the stress of the circumstances at-
tending an actual mission, the changing of codes
might be neglected unless the operation is fully
or partially automatic. If only partially auto-
matic, the operation should require only simple
and positive motions, such as pushing buttons
or pulling levers, to be carried out on command.
Two designs were selected from the many
considered and two units each of these designs
have been constructed. One design provides for
electrical operation under control of a timer,
though manual operation is possible; the other
design is arranged for manual operation but
could be converted to automatic operation by
further development. The correctness of the
operating principles has been checked by tests
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
s a
;-es
on.
3rd
es-
ch
at-
les
Ily
31e
ns ns
id.
fly
ns
or
Ler
nt
by
he
;ts
TELEGRAPHY APPLIED TO TDS?PROJECT C-55 23
on TDS units. Choice between the two will de-
pend on military requirements and manufactur-
ing considerations.
The two models are functionally equivalent
since each provides a choice of twenty double
codes, and the particular sets of twenty from
the several hundred available may be changed
between missions. In the automatic model, codes
are set up by means of ninety small relays, con-
trolled by contacts made through two perforated
FIGURE 8. Model C code-changing unit with case
removed; back and top view.
volume and, for comparable materials, about
half the weight of the TDS unit. Both models
could be operated by a single motion of a gloved
hand.
2.6 TELEGRAPHY APPLIED TO TDS--
PROJECT C-55
4 Project C-5513 was undertaken at the request
of the Signal Corps early in 1942 to determine
the advantages and disadvantages (if any) of
incorporating means for tone telegraphy in con-
nection with TDS equipment which had been
developed for the Services. At the same time
the question was raised as to whether the use of
telegraph might not jeopardize the privacy of
the device for speech. This was based on the
idea that the TDS code might be recovered more
quickly from scrambled telegraph signals than
from scrambled speech signals. It was impor-
tant to answer this question to determine
whether restrictions would be needed on the
use of telegraph.
The questions to be answered may be sum-
marized as follows.
1. Is TDS privacy less for telegraph than for
speech? That is, will the use of telegraph
sheets, similar to player-piano rolls, with twenty
codes per sheet. In the manually operated model,
a mechanism pushes two code cards, out of two
groups of twenty contained in loading boxes, to
close two sets of spring contacts according to
the arrangement of perforations in the code
cards.
The automatic model operates on 24 volts,
drawing 0.8 amp between changes and 1.5 amp
during changes.
The automatic model is about 10 per cent
less than the TDS unit in cubic contents, and,
for comparable materials, of about the same
weight. The manual model has about half the
FIGURE 9. Model B die for punching code masks.
through TDS tend to expose the code and thus
reduce the privacy for subsequent telephone
use?
2. Is TDS privacy for telegraph critically de-
pendent upon rates of hand sending and lengths
of TDS time elements?
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11: CIA-RDP10T02279R000100180001-2
24 TIME DIVISION SCRAMBLING SYSTEMS
3. What impairment to telegraph transmis-
sion is caused by TDS?
4. Is it feasible to apply machine telegraph
sending to TDS?
With regard to the relative privack of tele-
graph and speech, the data if taken literally
indicate a slightly greater privacy for telegraph
than for speech. The differences in the average
solving times are so small as to be negligible,
however. The solving times both for telegraph
and speech vary from sample to sample, so that
some telegraph samples were solved more
quickly than some speech samples, but the oppo-
site was likewise true, and the range of varia-
tion was about the same for telegraph and
speech.
However, there is a difference of another
nature which is inherent. The two main steps in
the restoration of scrambled speech are, first,
that the code be found and, second, that the
scrambled speech be reproduced through TDS
equipment arranged for the unscrambling code.
These steps also can be used for unscrambling
telegraph, and if they are used there is no dif-
ference in the degree of privacy for speech and
telegraph, since the times taken to ascertain the
codes were the same. But another method might
also be used for telegraph. If an oscillogram or a
paper trace were made of the entire scram-
bled message, then solution of the code would
supply the formula according to which trace
should be cut up and reassembled. If this were
done for speech, the result would still be un-
readable without fairly elaborate equipment;
but a reassembled telegraph trace could be read
by visual inspection of the dots and dashes.
As a practical matter this difference appears
to be of little importance, as the following cal-
culation indicates:
At 25 words per minute, a 100-word message
takes 4 min, or 240 sec. At 0.75 sec per TDS
cycle, this would cover 4/3 x 240 = 320 TDS
cycles, or supply 20 X 320 = 6,400 TDS ele-
ments for manipulation. If the trace were run
so as to allow I/4 in. to each element, or 48 ele-
ments to the foot, which would be extremely
compressed, the trace would need to be 6400/48
133 ft long. Even with practice and concen-
tration probably at least a minute would be
needed to reassemble each cycle, what with the
manual labor of cutting and pasting, or 320 min
in all; that is, it would take about 5 hr for one
person to recover a 4-min message. Additional
people or supplementary equipment would, of
course, reduce this time.
Thus, while the sense of the telegraph mes-
sage can be recovered with less elaborate equip-
ment than is needed for speech, the amount of
clerical work involved would "constitute a pro-
tection lasting longer than the privacy time
which ?should normally be associated with a
fixed-code system.
The answers to the questions posed above
were found to be about as follows.
1. Scrambled hand-sent telegraph signals re-
quire at least as much time to decode as scram-
bled speech; there is therefore no reason to be-
lieve that the application of manual tone tele-
graph to TDS will jeopardize its value for
speech.
2. The privacy of TDS for telegraph is not
critically dependent on rates of hand sending
and lengths of TDS time elements.
3. The limits of telegraph transmission are
reached with about 5 db less thermal (random)
noise with the TDS than without; this means
that the range is reduced unless the signal
strength is increased. The impairment is less
for short elements than for long TDS time ele-
ments.
4. Machine telegraph, both Boehme and tele-
typewriter, give generally legible, though not
letter-perfect, results with single-tone transmis-
sion, if care is used. Better results, though still
not perfect, may be obtained with two-tone
transmission.
The final report" of Project C-55 covers the
individual points raised in the above questions
in considerable detail and includes additional
material, not summarized here, on the question
of the quality of hand or machine sending as
affected by TDS.
SECRET
neclassified and Approved For Release 2013/12/11: CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Chapter 3
SPEECH PRIVACY SYSTEM DEVELOPMENT
3.1 RCA-BEDFORD SPEECH PRIVACY
SYSTEM
I N THE RCA-BEDFORD speech privacy system,a
a portable system having short-term secu-
rity, the speech wave is coded by multiplying
it by an audio-frequency coding wave. A con-
necting circuit between transmitter and re-
ceiver having essentially faithful reproduction
over a band of 100 to 4,000 cycles furnishes
the required intelligibility and reliability.
3.1-1 Basic Principles
At the sending end a sound wave So is first
"compressed" to a uniform average amplitude
SPEECH
CODING
WAVE
CODED
SPEECH
5K
DECODING
WAVE
1/K
DECODED
SPEECH
SI
FIGURE 1. Coding and decoding processes involved
in RCA-Bedford multiplication privacy system.
level by a special compandor circuit to form a
signal S. The signal S is then multiplied by
a suitable audio-frequency coding wave K to
form a coded wave SK which is unintelligible.
The expression "multiplied," as herein used,
a Project C-54, Contract No. OEMsr-592, Radio Cor-
poration of America.?
means that the product wave SK has instan-
taneous ordinates measured from its a-c axis
which are proportional to the corresponding
products of the original waves S and K. This
is illustrated by waves S, K, and SK, Figure 1.
At the seceiving end the coded wave SK is
multiplied by the "reciprocal" wave 1/K (which
is derived from the locally generated code wave
K) to produce the restored speech wave S',
which is ideally like S.
1
SK X ?K = S or S'.
As shown in Figure 1, the reciprocal wave 1/K
has instantaneous ordinates which are propor-
tional to the reciprocal of the corresponding
coding K.
The ordinates of the coded wave SK become
zero each time the coding wave becomes zero.
For these instants the reciprocal wave 1/K
ideally should become infinite. Since this is
impossible in practice, a narrow gap must occur
in the restored wave 5' as shown in the figure,
at each time the wave K passes through zero.
This amounts to an introduction of spurious
signal components, which are largely removed
by a low-pass filter.
The decoded wave 5', which is still com-
pressed, is "expanded" to its original level by
the receiving end of the compandor. The ex-
panded signal So', which is like So except for
losses, is heard in the phones.
The various treatments of the signal are in-
dicated in the block diagram, Figure 2.
3.1.2 Wave Multiplier
The wave multiplier used to multiply audio-
frequency waves S by K and SK by 1/K is really
a balanced modulator which is completely "bal-
anced" in the sense that only the instantaneous
product terms are produced. This balanced con-
dition can also be described by saying that the
output contains only the sideband frequencies
SECRET
25
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
- -
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
26 SPEECH PRIVACY SYSTEM DEVELOPMENT
of modulation; namely, those frequencies which
are the sums or the differences of the frequen-
cies of S and K. Ideally, no harmonics of either
S or K are present in the output. This is an
SEND
MICROPHONE
3 - KC
LOW PASS
FILTER
COMPRESSOR
lator followed by filters to suppress the unde-
sired frequencies. The present multiplication
process could not be carried out in this manner
because the frequencies it is desired to suppress
RECEIVE
RADIO
RECEIVER
So
SK
I- KC
REJECTOR
MIXER
VARIABLE
Isr
MU
TUBES
SK
MULTIPU ER
PILOT
TONE -\
-VARIABLE
WAS
I-K C
DETECTOR
DETECTOR
SOURCE
AND
OF
OSCILLATOR
FILTER
K-WAV
MULTIPLIER
IK
RECIPROCAL
CIRCUIT
SOURCE
OF
K- WAVE
E XPA NDOR
VARIABLE
MU
TUBES
PILOT TONE
I-KG
SELECTOR
DETECTOR
AND
FILTER
I-KC
REJECTOR
RADIO
TRANSMITTER
3-KC
LOW- PASS `?
FILTER
FIGURE 2. Block diagram showing compandor in RCA-Bedford privacy system.
important fact from the standpoint of security.
This process would correspond to modulation
as commonly used in radio transmission, except
that in the privacy system, both the carrier and
the speech waves are suppressed. In radio this
can be accomplished in an unbalanced modu-
s'o
PHONES
occupy the same frequency range as the desired
modulation sidebands.
The special wave multiplier developed for the
speech privacy system employs four small oxide
rectifiers (known commercially as varistors)
operating in their square law range.
SECRET
DeClassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
RCA-BEDFORD SPEECH PRIVACY SYSTEM 27
3.1.3
Compandor
The name "compandor" is derived from the
words "compress" and "expand." It is an old
device designed to improve the signal-to-noise
ratio in communication by raising the level of
transmission of the weaker sections of the
speech signal with respect to the strong sections
of the signal. This is equivalent to compressing
the higher sections. At the receiving end the
various sections of the signal are restored to
their original relative levels.
In the present case the compandor is used,
primarily, to destroy the cadence (or loudness
variation) of the coded signal, and thereby
improve the security. To further this end, a
special compandor circuit was developed, in
which a pilot tone is used to fill the space be-
tween words, and to provide control for the
expanding process at the receiver.
The curve So in Figure 3 represents the am-
plitude, or envelope, of a word of speech. This
is fed into the compressor (i.e., the sending part
of the compandor) as shown in Figure 2. A
filter removes any 1,000-cycle speech compo-
nents, and then the output of a 1,000-cycle
oscillator is added in the mixer. This gives the
combined envelope shown as "So ? tone" in
Figure 3. The amplitude of the tone is about
10 per cent of that of the loudest section of the
speech. The output of the mixer is detected,
filtered to remove the audio frequencies, and
then used to vary the bias of a pair of push-pull
variable-mu tubes which amplify the mixed
signal. This bias control is such as to make the
combined signal S have a substantially constant
loudness as shown at S in Figure 3. Note that
the lower speech levels are raised. This signal
contains a variable amplitude of 1,000-cycle
tone, as shown at T. The signal S is then coded
by multiplying by the code wave K, and trans-
mitted.
At the receiver, as shown in Figure 2, the
received signal SK is decoded to S'. In the ex-
pander, the 1,000-cycle tone is selected by a
filter, and then detected to control the gain of
a pair of variable-mu tubes. The variation in
gain is such as to restore the amplitude of the
tone component of the output to a constant
value, and simultaneously restore the speech
level to its original proportions. A filter then
removes the control tone, and a 3,000-cycle low-
pass filter removes the high-frequency distor-
tion components from the phone circuit.
So
TONE
50+ TONE
SPEECH
ist
\mialfth.
#.
,
'TONE
SPEECH
LOUDNESS OF CONTROL TONE
TRANSMITTED
\
FIGURE 3. Envelopes of speech and tone in
compandor.
When the signal transmitted in this manner
is received on a privacy unit, operating with
a code wave that differs substantially from the
code wave at the sending end, the pilot (1,000-
cycle) tone is spread irregularly over the audio
spectrum so that a filter cannot isolate it for
operating the expander to restore cadence.
Neither can a filter be used to remove it from
the phone circuit. The result is an unintelligible
mass of coded speech and noise. Tests show that
to get intelligibility from the decoded signal,
the sending code wave and the receiving code
wave must have a higher degree of similarity
than when the compandor is not used.
Inspection of Figure 2 shows that most of
the parts of the privacy unit, used in sending,
are the same as used in receiving. Therefore, in
a send-receive unit, the use of suitable switches
or a relay avoids duplicating these parts.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
28
SPEECH PRIVACY SYSTEM DEVELOPMENT
3.1.4 Generating Code Wave by Delay
Network
The coding wave K in the system is generated
by combining with different changeable polari-
ties the outputs from taps along a delay network
which is fed with narrow pulses from a fixed-
frequency multivibrator, illustrated in Figure 4.
The network contains 80 sections of series in-
ductance and shunt capacitance. A repeater and
an equalizer (not shown in the figure) are in-
serted at 16-section intervals to make up for
attenuation along the network.
f.S1(
SYNCH
A
Is
/
FIGURE 4. Generation of coding wave K.
The multivibrator [MV] supplies repeating
pulses to the input A of the delay network, as
shown by wave A in the figure. The frequency
of wave A is such that immediately after a
pulse has reached the far end of the network,
another pulse is fed into the near end. Then,
neglecting distortion for the present, the volt-
ages at tap points B, C, D, etc., are like A ex-
cept delayed various amounts in time, as shown.
Each tap is shown here with a SPDT switch
so that it can be connected through suitable
buffer resistors to point P, either directly or
through a polarity-reversing amplifier [RA].
(Actually, the switching arrangements are
more complicated than shown here, as will be
explained, but Figure 4 suffices for an ele-
mentary explanation.) The voltage at point P.
therefore, consists of the sums of the various
voltages along the network taken either posi-
tively or negatively, depending upon the switch
potions. Wave H illustrates various pulses
which would combine to form a wave such as I,
which has a variety of widths of lobes. From
this simple illustration, it is clear that a variety
of complex waves, having a repetition rate of
100 times per second, could be produced by
various switch settings.
Actually, the multivibrator tends to produce
a very narrow repeating pulse, as shown at J,
instead of a smooth, wide pulse, as shown at A.
However, the multivibrator shock excites the
first section of the network, resulting in an
oscillatory voltage wave, as shown at L, to
occur at tap A. As the pulse progresses along
the network, the phase distortion of the net-
work (which is very pronounced shortly below
cutoff frequency) causes the pulse to have an
extended oscillatory "tail" as shown by wave M.
The voltage at point P then is the sum of many
waves which are, themselves, quite complex in
shape, and may have lobes with a large variety
of amplitude, as well as widths as shown ,at K
in the figure.
In this privacy system, the instantaneous
voltage of the transmitted signal SK is propor-
tional to the instantaneous value of wave K.
Then, if the coding wave K were allowed to
have peak amplitudes far in excess of its aver-
age amplitude, the average utilization of the
available transmitter power would be very low.
Accordingly, to improve the power efficiency
of speech transmission, the high peaks of wave
K of Figure 4 are limited to amplitudes consid-
erably lower than the original amplitudes,
shown by the dotted peaks. The limiting action ?
is gradual in order to retain some amplitude
variation in the code wave. After the limiter,
a small shunt capacitor serves to smooth over
the discontinuities in the wave caused by lim-
--
iting.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
RCA-BEDFORD SPEECH PRIVACY
-
3.1.5 c,
Synchronization
of the Delay Network
Consideration of the decoding process shows
that it is necessary for the receiving code wave
K to have the same timing as the code wave at
the sending apparatus (after making suitable
allowances for the time of transmission of the
signal). To maintain this condition, a syn-
chronizing pulse is transmitted along with the
signal SK, to control the multivibrator pulsing
the network of Figure 4 when receiving.
Briefly the coded speech signal is inter-
rupted for about 0.001 sec each 0.01 sec and a
synchronizing pulse is inserted. At the receiver
this pulse is selected by virtue of the regularity
of its occurrence (as compared to any lobe of
the coded speech) and used to trigger the re-
ceiving multivibrator which pulses the delay
network which, in turn, generates the code
wave.
3.1.6
Code-Changing Method
The eight code disks which are part of a
rotary code-changing switch may be seen in the
FIGURE 5. Front view of Model RCAL-1 privacy
equipment.
photograph of Figure Each code disk has
twenty insulated silver segments, of three dif-
ferent types, which are located in irregular
5.
ocorddeerdiinskas; .tll
so
of Dthree switchb re ull
tive depending
under the brush
controls two tali's,
two buffer resisia
between the 80
f switches is If
:TI:he drum eiglitheyodee igtis
o
each disk can be 0,
twen*y possible posit.,1
the disks is rotatabl
solenoid in steps
d s of
energized at
controlled by a
fixed to to the right-band e
brated in 100 equal divis
3.1.7
each originalbot speech frecitleASocsoiwo701N
Auxiliary Cornpan?t)r 4 -i'
h upwar alld ? ir p a i
each frequericY. j, 000st of (11' AtiL
t
A Oila Efill 1
cOo theertt Al..4
iji MOd
icq.Ueerlr
csodifeteCe
wave. This Method , clifiti oite thtte?'
Therefore,
igibility obtained bY"
Ydforbey of the sPeee"
dwave SK is due to the
, it is believed tba,,. a to th?ti
Oderice ech to NI.
iistenill;f
the be q41
the til bled
the cadence of the sffe
ginning of
special
and therefore to irriPrevtpriefly seetN-
this suonnarY W3? - -cl,
p.ecial compandor described cievelopi,.:1 t
In the coding process used y c el
3.1.8 Ott oeVeati4
Weakness of Sil
thepliIend F
by a section of sPeeediled wave e; toilltilt,
codediguwreav6e, sac:. dTeb::: e:wc:rvoessilses7c:.?,,,u
Code Wave5 ._
each time the code W9' e 45. Si-
crossovers prodUced IV -1"-
d by the same sectiori .
when ,5,, crosses itS
corr ill thi-
-rn tiOS Wa 4v-kiti. '1..
Pend e) it"--
ti,_
41-1
different section of speech vi?V -xe
a e Sp whila},4, ..i;
axis
results in coded wave Sb-157: ""-- es v
tiplie of code t)
and. also
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
30 SPEECH PRIVACY SYSTEM DEVELOPMENT
with crossovers in wave SaK, but other cross-
overs do not agree.
At SK are superimposed several sections of
speech wave coded by identical sections of code
wave, such as would occur with the repeating
code wave of Model RCAL-1. This is substan-
tially the image seen in an oscilloscope picture
of the coded signal when properly synchronized
by the transmitted synchronizing pulses. In-
spection of this image readily reveals the prob-
FIGURE 6. Wave forms demonstrating weakness
of short repeating code wave.
able location of the crossovers of the code wave,
and also indicates the shape of the coding wave.
This is particularly true if a large number of
waves are superimposed, as seen on an oscillo-
scope with a long-retentivity screen. It has
been demonstrated that a rectangular wave, as
shown at can be used effectively to decode
a message transmitted by the code wave K.
Hence, it is only necessary to determine, and
make suitable use of, the location of the cross-
overs of the code wave to crack the message.
The code wave of Model RCAL-1 has (by
count), on the average, about 25 crossovers
which repeat at intervals of 0.01 sec. The cross-
overs are spaced at irregular intervals, depend-
ing upon the settings of the code switches. The
wave is composed of the various harmonics of
the 100-cycle fundamental from about 300 to
2,500 cycles per second. The harmonics have
various phases and amplitudes. This code wave
is changed, in part, each 11/2 sec through a
5-mhzr cycle by the clock-driven rotary switch
which contains the eight code disks mentioned
above. Ten of these small partial changes, oc-
curring in 15 sec, may cause a complete change
of all contacts. Therefore, only sections of about
5 sec of the message would be cracked by a
single code wave.
Although considerable apparatus of a very
special character would be needed to actually
crack a message coded with such a repeating
code wave by making use of the repeating pat-
tern of crossovers in the coded wave, as deter-
mined by statistical observation, nevertheless
it must be appreciated that a suitable accurately
calibrated oscilloscope could be used to observe
the coded wave and determine the timing of the
crossovers of the code wave, and that a suitable
special apparatus could then be adjusted to
produce a decoding wave such as K' of Figure 6
to crack the message. Furthermore, there is a
feasible plan for an apparatus to do the entire
job automatically. Hence, no coding wave which
repeats in a reasonable time can provide a high
degree of security against an able enemy.
For the sake of completeness of this discus-
sion, it should be pointed out that a message
coded by a code wave such as K, in Figure 6,
can be decoded by using a receiving code wave,
such as K", which differs considerably from K.
From observation, it is estimated that if most
of the crossovers of the receiving code wave
occur within 20 or 30 per cent of the time of
occurrence of the crossovers of the sending code
wave, the message can be cracked even when
the compandor is used. (The per cent value used
here is based upon the intervals between ad-
jacent crossovers in the sending code wave
being 100 per cent.) Also, a few extra cross-
overs in either code wave, having no corre-
sponding crossovers in the other wave, may be
tolerated. With such a wide tolerance in the
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
RCA-BEDFORD SPEECH PRIVACY SYSTEM
31
cracking code wave, it is not surprising that in
Model RCAL-1, which uses a repeating code
wave having an average of only 25 crossovers,
a short message can be cracked in a short time
by "playing" with the code disks, or switches,
at random.
Since all limitations of security of Model
RCAL-1 with the compandor arise directly, or
indirectly, from the repetitive character of the
coding wave, the use of a nonrepeating code
wave would appear desirable.
3-1.9 Dual Delay Networks for Generating
Long Code Wave
In a system providing longer security, the
period of the code wave could be made 0.4 sec
(as compared to 0.01 sec) by using two delay
networks operating at slightly different pulse
frequencies to generate the code wave. This is a
purely electrical device. Then the mechanical
apparatus for changing taps on the networks
can operate at the relatively low speed which
causes a substantially complete change in each
0.4-sec interval, and thereby prevent the code
wave from repeating itself until the .mechanical
system has run through its complete cycle. The
time for this is over an hour.
The dual delay network appears to be a very
economical electrical device for generating a
code wave with a period as long as 0.4 sec. To
generate a wave of this period and maximum
frequency directly by a single delay network,
would require about 3,000 LC sections.
3.1.10 Motor-Driven Tap Switch with
One-Hour Period
In,the dual delay system, it is required that a
critical number of taps on the dual delay net-
works be changed each 0.4 sec to prevent a repe-
tition of the code wave 2.5 times a second, which
would allow a statistical study of the coded
wave SK to reveal information as to the loca-
tion of crossovers in the code wave K. This is
accomplished by a three-speed motor-driven
rotary switch.
The switch, designed for the proposed long-
security system has nine code disks and brush
holders instead of the eight in Model RCAL-1.
The 45 SPDT switches provided are connected
in irregular order to control the 84 taps in the
dual delay networks. These nine code disks are
rotatably mounted with individual detents in
groups of three on three separate drums. The
three drums are geared to run at slightly differ-
ent speeds, which are, respectively:
iSi = 6 rpm,
82 20
= 6 X ?21 = 5.71 rpm,
20
Ss = 6 X ?19 = 6.31 rpm.
To the left of each group of code disks is a
timing disk which is fixed to the drum so that it
controls all three code disks in its group. Each
of three 40-tooth gears is rotatably mounted on
a counter shaft, but is restrained from relative
rotation by an eight-position detent. These
three detents allow the three timing disks to be
independently adjusted to their proper starting
positions by hand. The timing disks are cali-
brated, and numbered in 20, 21, and 19 equal
divisions, respectively, to agree with the detent
positions on the counter shaft acting through
the three different gear ratios. All code disks
have twenty positions on the drums, indicated
by letters.
The speeds of the three groups of code disks
are such that the system requires about 1 hr
6 min to run through a complete cycle, so that
all groups reach their starting position simul-
taneously. Therefore, by resetting the individ-
ual code disks to a new code each hour, a non-
repeating code wave is provided. Any two
groups of the disks go through a cycle in about
3 min. Then, if the effect of the other group of
three disks upon the code, wave were suffi-
ciently small, the code wave would effectively
repeat about twenty times each hour, and be-
come accessible to the enemy.
Tests, however, made with Model RCAL-1,
operating with the compandor and with a tem-
porary motor-drive for the disks, showed that
with three of the eight disks off code only 6 per
cent of words were understood. (Each word was
repeated three times.) With two of eight disks
"off," 25 per cent were correctly understood.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
32 SPEECH PRIVACY SYSTEM DEVELOPMENT
To keep the motor in the receiving apparatus
in synchronization with the motor at the send-
ing station, both synchronous motors are driven
by 60-cycle power whose frequency is controlled
by a 145-kc crystal.
Probable Security of the System
Based on tests made on the Model RCAL-1
with compandor and a temporary motor-drive,
it is concluded that, to obtain from a captured
machine adequate intelligibility to crack a mes-
sage, the enemy would have to have at least six
of the nine code disks in the correct position.
The probability of this condition being obtained
in a single trial, at random, is 1/889,000.
It is estimated that at least 3 sec would be
needed to set up and try each of the many pos-
sible code disk positions, even if elaborate spe-
cial means were built to allow orderly adjust-
ment of the relative positions of the disks while
running. (The time for trying each code setting
is made longer because of having to make a
short exploration for correct phase of the entire
disk assembly.) Hence, the average time for
reaching the condition for six code disks right
is:
3 X 889'000
Time ? ? 370 hr.
2 X 60 X 60
(The use of more cracking setups would shorten
this time.)
If this time is deemed inadequate to provide
the desired security, it could be greatly ex-
tended by adding a few more code disks. An
alternative method would use a selector switch
of the radio-band-selector type for manually
changing the connections between the motor-
driven switch and the taps on the delay net-
work. The setting of this switch would be part
of the code.
No reasonable method of cracking messages
sent by this system, other than the trial method
above, occurred to the RCA engineers. Bell
Laboratories' engineers (Project C-43) and
others were given complete verbal descriptions
of the proposed system, and invited to find some
property of the coded wave by which it might be
decoded. No likely property was revealed.27 Of
course, it cannot be known with certainty that
the signal could not be cracked eventually by a
shorter method.
By contrast, the basic weakness of the Model
RCAL-1, with its repeating code wave, was
known and reported very early, that is, well
before design was begun. The clock-driven code
switch was included with the expectation that
it would only make cracking a more laborious
prows. (After the model was built, it was
found that a message could be cracked by "play-
ing" with the code disks somewhat quicker than
originally expected. This was because it had not
been appreciated how different the receiving
code wave could be from the sending code wave
and still give appreciable intelligibility.)
In studying the probable security of the pro-
posed system, it should be remembered that the
signal S (which is coded to form SK) usually
contains a very substantial amount of the 1,000-
cycle compandor control tone. Offhand, this
fact would seem to offer the most likely ap-
proach to a reasonable method of cracking. If
this were found to be so, it would be necessary
to use some other apparatus to replace the corn-
pandor. Two different devices are available but
they are not attractive.
3.1.12 Present Status of the Development
Part I of the final report" of the project gives
an assessment of the speech quality transmitted
by the system and indicates that radio trans-
mitters and receivers somewhat better than
actually used during the war would be required
to deliver good intelligibility. Weights and di-
mensions are to be found in the final report.
At the close of the project most of the new
circuits required in the proposed high-security
system had been developed. Some of these de-
veloped and used in Model RCAL-1 are: the
wave multiplier, the reciprocal circuit, delay
network and tap switching circuits, synchroniz-
ing circuits, synchronizing blanking circuits,
and the compandor.
The following additional circuits were devel-
oped, or tested specifically to be used in the
proposed high-security units: the "clip, multi-
ply, and phase-distort" method for blending the
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
HAZELTINE BAND DISPLACEMENT SYSTEM 33
outputs of two delay networks was tested and verse direction for unscrambling and at the
shown to produce a coding wave with more
crossovers and with the initial crossovers ade-
quately "smeared." A synchronous motor suit-
able for driving the code disks had been driven
by an amplifier from a chain of frequency-
dividing multivibrators.
The rotary code switch had been designed,
but not detailed. It included code disks, brushes,
brush holders, and detents, which are very
lar to parts in Model RCAL-1.
3.1.13
Conclusion
It was concluded that the speech privacy unit
proposed would probably have a relatively high
degree of security, that it would have dimen-
sions and power requirements which allow it to
be portable, and that it could be developed with
reasonable expenditure of time and money.
3'2 HAZELTINE BAND DISPLACEMENT
SYSTEM
The Hazeltine speech secrecy systemb is of
the band displacement type in which the speech
is inverted and displaced in frequency to seven
successive positions in an extended a-f band, in
a sequence determined by a coding switch and
punched cards.
3.2.1 General Description of the System
In transmitting, speech is first passed
through a frequency selective network in which
the frequency spectrum is tipped up at the rate
of 6 db per octave. It is then modulated with a
13-kc carrier and the upper sideband extending
from 13 to 16 kc selected in a band-pass filter.
This is again modulated by a selected sequence
of seven displacement carriers, 1 kc apart from
17 to 23 kc. The lower sideband is then selected,
resulting in inverted speech in a sequence of
bands in a range from 1-4 to 7-10 kc. Amplifiers
are employed in the carrier channel to maintain
the output at the same level as the input.
In receiving, the same apparatus is used. The
speech is passed through the system in the re-
b Project C-15, Contract No. NDCrc-139, Hazeltine
Corporation.
normal speech end of the channel the frequency
spectrum is tipped down by 6 db per octave to
restore the original frequency characteristic.
An electronic switch for keying on the 17-
to 23-kc displacement carriers for 50-msec in-
tervals in a selected sequence comprises a chain
of seven switching tubes. Each of these is ar-
ranged to generate a 50-msec pulse. The termi-
nation of this pulse initiates operation of the
succeeding switching tube. In transmitting, the
pulse from the last of the switching tubes is
returned to the first to make the operation con-
tinuous. The pulses derived from the switching
tubes are delivered to the seven carrier fre-
quency oscillators as selected by the coding
switch and serve to key-on the oscillators.
To synchronize the receiver and the trans-
mitter, the pulse derived from the last of the
switching tubes in transmitting is suitably am-
plified, limited, and filtered to eliminate com-
ponents above 1 kc and then added to the scram-
bled speech at the end of the carrier channel.
In receiving, this synchronizing pulse is sepa-
rated from the scrambled speech and, after am-
plification and limiting, is applied to the first of
the switching tubes. The connection between
the last switching tube and the first is elimi-
nated in the receiver so that operation of the
electronic switch in receiving is initiated by re-
ceived synchronizing pulses only.
To reduce the possibility of actuation of the
electronic switch by spurious pulses, a barrier
circuit in the form of a pulse generator having
slightly less than a 350-msec period, is placed
ahead of the first switching tube. Once a syn-
chronizing pulse actuates the barrier circuit
(and simultaneously the first switching tube)
the barrier circuit becomes immune to further
pulses during its cycle. This prevents any pulses
from reaching the first switching tube until al-
most the entire cycle of the seven switching
tubes has been completed.
3.2.2 Performance of First Two Models
Two models embodying such a system were
completed, tested, and demonstrated to mem-
bers of Division 13. Although it was possible to
obtain speech of reasonably understandable
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
11
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
34
SPEECH PRIVACY SYSTEM DEVELOPMENT
quality after a complete cycle of scrambling
and unscrambling, it was felt that the perform-
ance could be appreciably improved in several
ways.
It was felt desirable to reduce the degrada-
tion of speech quality obtained with the equip-
ment. This degradation was probably caused by
the following factors. The oscillators did not
respond immediately to the keying pulses so
that intervals of several milliseconds were left
at the transition points for the displacement
carriers. This resulted in blanks in the speech
and in transient clicks. The frequency stability
of the oscillators was poor with the result that
synchronization between oscillators in the
transmitter and those in the receiver could not
be held. This resulted in slight changes in pitch
of the speech as the displacement carrier fre-
quency changed. There was also some nonline-
arity present in some of the amplifiers.
It was also found that when as little as 10 db
attenuation was inserted in the line between
the transmitter and the receiver, the receiver
had a tendency to drop out of synchronization
with the transmitter. It was felt that the equip-
ment should be designed to tolerate an attenu-
ation of at least 25 db below the normal level.
No experience had been obtained on the effect
of noise and code interference on the synchro-
nization of the two units. It was felt that rea-
sonable immunity to such interference was an
important requirement of the equipment.
3.2.3 Construction of Third Model
With the experience gained in the operation
of the first two models, the construction of a
third model was undertaken. The objectives
sought in this third model were as follows:
1. Improved oscillator frequency stability.
2. Fast keying on and off of oscillator.
3. Elimination of nonlinearity in amplifiers
in the carrier channel.
4. Substantially flat overall frequency char-
acteristic from 300 to 3,000 cycles per second.
5. Improved immunity to changes of level
and noise in the synchronization of the receiver.
6. Incorporation of a storage battery oper-
ated power supply.
7. Improved isolation and shielding of com-
ponents to avoid carrier leakages and cross
modulation.
8. 4ncorporation of a code switch suitable for
operation with a punched card.
9. Closer approximation to a production de-
sign.
The third model, completed in the latter part
of September 1941, met the above objectives
very well. Speech quality was materially im-
proved over that obtained with earlier models.
Switching transients between successive dis-
placement carriers could not be detected in the
output. The prominence of the sync pulse in the
output was greatly reduced. Measurements in-
dicated that the distortion of the system was
reasonably low. The input to the receiver could
be reduced more than 25 db below normal level
before the receiver dropped out of synchronism.
With high noise levels the receiver remained in
synchronism at least up to the point where the
noise was equal to the signal.
3.2.4 Apparatus Details
The final report15 on the project gives details
of the transmitting and receiving circuits,
means for providing nonscrambled speech when
desired, method of setting up the desired code
by means of punched code cards, an analysis
of the inherent secrecy of the system plus the
chances of cracking it, and suggested improve-
ments.
The equipment weighed 45 lb, required ap-
proximately 1.14 cu ft of space, operated from
6.3 volts and needed 10 amp at this voltage.
SECRET
Declassified and Approved For Release 2013/12/11: CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
4.1
Chapter 4
UNSCRAMBLING AND DECODING METHODS
HISTORY
LONG WITH PROJECTS and work primarily
concerned with the development of scram-
bling methods for speech, code telegraph, or
facsimile, projects concerned with unscram-
bling or code-cracking methods engaged a large
part of the time and energy of those associated
with this portion of Division 13 activity. In this
manner the effectiveness of the privacy methods
developed could be tested constantly.
Realizing early in the privacy research con-
ducted under the sponsorship of Division 13
that the ear had very limited capabilities for
analyzing scrambled speech, the sound spectro-
graph was developed by the Bell Telephone
Laboratories to provide speech patterns which
could be interpreted by the eye. In effect this
valuable instrument divided scrambled speech
into its three important dimensions of fre-
quency, amplitude, and time. By its means, any
alterations to the original speech in either fre-
quency or time could be detected and analyzed
with the object of adjusting unscrambling ap-
paratus so that the original speech could be
recovered.
Early in 1941 a rough laboratory model of
the sound spectrograph was demonstrated to
the National Defense Research Committee
[NDRC] and as a result Project C-32 was or-
ganized with the immediate object of producing
such a device in a form that would be useful
for diagnosing and decoding speech scrambling
systems. Such a model was produced and suc-
cessfully demonstrated to representatives of
NDRC, Army, and Navy.
Upon the termination of Project C-32 on
February 1, 1942, it was decided that the work
initiated under that project should be continued.
Accordingly Project C-43, "Continuation of
Decoding Speech Codes," was authorized. The
project anticipated some routine decoding, the
production of duplicate equipment to be used
by the Army and Navy intelligence services,
and further studies of decoding tools and meth-
ods. At that time the Army and Navy were
relying almost entirely upon this project to
furnish the above services until they could be
provided with suitable equipment and could
obtain trained personnel. Based on the needs
of the military, this project was thrice ex-
tended.
.4.
Under the guidance of NDRC Division 13,
the emphasis was placed at any given time on
what was deemed to be most urgent. This is
reflected in the subject matter of the prelim-
inary reports of Project C-43 which were issued
from time to time and which form the appendix
to the final report of that project. In addition
to the specific investigations covered by these
preliminary reports much work was carried on
as the basis for more general coverage of the
field of interception, diagnosis, decoding, and
evaluation of speech privacy systems.
In addition to the general studies mentioned
above, decoding equipment was developed and
models furnished to the Army and Navy. This
decoding equipment included (1) two models of
the sound spectograph, (2) a variable-area pat-
tern machine, and (3) equipment for decoding
two new enemy privacy systems intercepted by
the project personnel at Point Reyes, California.
In each case Army and Navy personnel were
instructed in the operation and maintenance of
these equipments.
Intercept activities of the Project C-43 per-
sonnel included (1) the study of recordings sub-
mitted early in the project by the Federal Com-
munications Commission, (2) exploratory work
at the Bell Telephone Laboratories experimental
radio receiving station at Holmdel, New Jersey,
and (3) exploratory work and routine intercep-
tion of radio telephone transmissions at the
American Telephone and Telegraph Co. radio
receiving station at Point Reyes, California. Re-
ports of the results of the above studies and
recordings of intercepted material were sub-
mitted directly to the interested military au-
thorities.
Many speech privacy schemes were submitted
through NDRC during the course of this proj-
ect. These were studied and evaluated. This
SECRET
35
Declassified and Approved For Release 2013/12/11: CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
36 UNSCRAMBLING AND DECODING METHODS
work led directly to the continued improvements
of the sound spectrograph and the development
of supplementary decoding tools and techniques.
As the Army and Navy became able to carry
on decoding activities themselves with the aid
of equipment and information furnished by
NDRC as the result of work outlined above, the
activity on Project C-43 gradually decreased.
The final reports on the several projects cover
all phases of the work on the general subject
and constitute a reference work for future
studies of speech privacy systems.
In this chapter will be found, first, some gen-
eral observations on the intercept problem and
on methods of cracking scrambled speech, a
general description of the sound spectrograph
and examples of its applications, and some ma-
terial on the practical evaluation of privacy
systems, all taken from the final report on
Project C-43.1 Then follows a summary of the
work accomplished in the several decoding proj-
ects working under Division 13 sponsorship.
4.2
INTERCEPTION
Speech privacy systems may be used in con-
nection with radio telephone systems or with
wire systems. The unauthorized interception of
wire communications in wartime, however, was
beyond the scope of the work done for Division
13. These notes therefore are confined to radio
interception problems and expands the material
in Preliminary Report No. 25.45 The decoding
techniques to be described subsequently, of
course, apply to wire as well as radio communi-
cations.
TYPES OF RADIO SYSTEMS
Radio telephone systems range in size and
complexity from high-power point-to-point sta-
tions operating over great distances to the low-
power, short-range sets carried by individual
soldiers. The high-power systems are usually
designed to operate between specific points,
using specific assigned frequencies. They are
equipped with elaborate fixed antennas, which
are usually of the directive type. Privacy equip-
ment associated with such terminals may be as
large and complex as desired to achieve virtual
secrecy. A major consideration in such systems,
of course, which adds to size and complexity,
is that the privacy must not degrade the quality
of the received speech to any appreciable extent.
On the other hand, anyone can intercept these
high-power signals at great distances, where he
can have a well-equipped centralized decoding
laboratory, with no limitation on the size and
complexity of the decoding equipment he might
bring to bear. This laboratory can be adequately
manned by a relatively few highly trained de-
coding specialists not necessarily members of
the Armed Services.
In contrast with this situation, the low-power,
short-range radio sets used in military opera-
tions are severely restricted as to size and
weight, and these restrictions also apply to
privacy equipment. The smallest privacy set
submitted to Project C-43 for study was roughly
a 10-in, cube, and was designed for mobile ap-
plications like tanks, planes, and command cars.
While it is difficult to achieve a high degree of
inherent privacy in mobile equipment, it should
be noted that the very mobility of such systems
adds to the security, because the signals can
not generally be picked up at great distances,
and whatever equipment an intercepter might
use to crack the privacy must also be mobile.
Furthermore, the decoding equipment must be
operated by military personnel, a large number
of whom may be required if the enemy is making
extensive use of mobile privacy.
Intermediate types of radio systems are used
for the higher echelons of command. For such
applications, the radio equipment is semimobile.
It can be transported in trucks and set up very
rapidly, and may have a considerable range.
For such applications, a high degree of privacy
is required, and a truckload of equipment might
be justified, because the enemy could afford to
devote considerable time, personnel, and equip-
ment to decoding the kind of messages which
would be transmitted over such systems.
INTERCEPTED SIGNAL QUALITY
Since most of this chapter deals with decod-
ing, the material from this point on will be
written from the point of view of the unau-
thorized rather than the authorized listener. It
is first of all desirable to get a good signal, as
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
INTERCEPTION
37
free as possible from interference. There are
several reasons for this. First, the process
which unscrambles the speech also scrambles
any noise such as static which has been super-
posed on the scrambled signal. This changes the
time or frequency distribution of the noise,
breaks up harmonic relationships, etc., thereby
increasing the interfering effect of the noise.
Second, the decoding is apt to be less perfectly
accomplished than at the authorized terminals,
which tends to make the speech harder to un-
derstand. Finally, there are usually language
differences which still further add to the diffi-
culty of understanding the message. Conversa-
tions can be carried on under extremely un-
favorable conditions by people speaking their
own language, but noise and poor quality rap-
idly degrade the intelligibility of a language
foreign to the listener.
In this connection it might be noted that it is
very desirable to be able to hear both sides of
the conversation without interruptions, in order
to follow the context. In the case of the point-
to-point systems, this will in general require
two receivers because the two directions are
transmitted over separate channels at different
frequencies. If the two outputs are mixed for
listening or recording, however, it should be
kept in mind that the noise on the weaker signal
will be superposed on the stronger signal and
may seriously degrade it. Putting the two sig-
nals on two headphones will improve this situa-
tion, because noise in one ear does not seriously
affect the intelligibility of a signal in the other
ear. This problem does not arise in the case of
the smaller radio systems, because these are
generally operated on the basis of switching
between transmitting and receiving conditions
on the same carrier frequency.
Methods of obtaining a good signal are the
same for the interceptor as for the intended re-
ceiver. A few of the important considerations
are listed here; further information on any or
all of them can be had from radio reference
works. (1) Point-to-point systems usually em-
ploy directive antennas; the intercept station
should therefore be located along or near the
line of the radio beam. (2) In locating stations
to intercept radio transmissions in the h-f range,
account should be taken of the skip distances
of the frequencies involved. Better signals will
sometimes be obtained by moving farther away
from the transmitter rather than closer. (3)
The use of directive antennas, directed towards
the transmitter being monitored, will improve
the signal-to-noise ratio by discriminating
against noise which is nondirectional. These
antennas of course should be designed for the
frequency and polarization of the signal, and
propeitly coupled to the receiving set. (4)
Stronger radio signals will be received if the
antennas are located in the open, with no trees
or other obstructions in the foreground. This is
particularly important in the v-h-f range. (5)
Radio signals increase in intensity as the height
of the antenna above the immediate foreground
is increased, particularly for v-h-f transmission.
Thus better results are obtained with the an-
tennas located on high masts or on hills over-
looking the foreground in the direction from
which the signal is arriving. If the signal is in
the v-h-f range and other measures are inade-
quate, it may even be desirable to consider
receiving the signal in an airplane and record-
ing it or retransmitting it for decoding. (6)
Noise improvement can generally be obtained
by keeping the receiving equipment away from
sources of man-made noise, such as ignition
systems and power lines.
RECEIVING SETS
With regard to the receiving sets, a distinc-
tion must be made _between the various ac-
tivities of an intercept station. One important
activity is searching for possible enemy trans-
mission channels. The object is to determine all
the channels in use, the location of their ter-
minals, the type of business transacted, and,
most of all, whether any special form of pri-
vacy is used on the channel. Some preliminary
searches of this type are described in Pre-
liminary Reports No. 246 and 23.47 If no privacy
is used, other than the usual commercial types,
it is unlikely that information of military im-
portance is transmitted over the channel, and
it may not be necessary to monitor it continu-
ously. If a new privacy system is located, how-
ever, it is very likely to be worth monitoring
and decoding continuously.
For the searching and scanning activities, the
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
38 UNSCRAMBLING AND DECODING METHODS
ordinary commercial sets of the "communica-
tions" type, equipped with a beat-frequency
oscillator, will serve very well for all types of
transmission. Even the suppressed carrier type
can be handled very well provided the signal
is fairly strong. It may require continual
manual adjustment of the local oscillator, but
sufficiently good reception can be obtained to
determine the nature of the channel. Cases of
extreme spread-band transmission can also be
handled in this manner.
If a particular channel employing suppressed
carrier is determined to be worth monitoring
continuously, then a single-sideband receiver
will give improved reception. These receivers
are equipped to amplify the partly suppressed
carrier, or supply a new one with great stability,
and they may provide as much as 15-db improve-
ment in signal-to-noise ratio in some cases.
They also permit selecting either the upper or
the lower sideband of double-sideband systems,
which may be of advantage in cases where in-
terference occurs on one or the other sideband
of such systems. However, these receivers are
not suitable for searching.
TYPES OF RADIO TRANSMISSION
A knowledge of the types of radio transmis-
sion which may be encountered is very im-
portant to the personnel of an intercept station.
Experience has shown that without such knowl-
edge, the nature of intercepted signals may be
completely misinterpreted. It is possible to mis-
take certain normal types of transmission for
new systems, or conversely to fail to recognize
new systems which should be monitored at once.
Double-Skleband. The commonest type of
transmission is the double-sideband type in
which the carrier is transmitted along with the
sidebands, which are usually about 3 kc in
width, and are located immediately adjacent to
the carrier. These are readily demodulated by
the ordinary receiver. This is true even if the
carrier is rapidly wobbled, provided the wobble
does not cover too great a frequency range.
Such wobbles are sometimes used in combina-
tion with simple inversion, to prevent reinvert-
ing with a locally supplied carrier at the edge
of one sideband.
Spread-Band. In this system, some or all of
the sidebands are displaced from the carrier.
Demodulated signals of 'this type will cover an
a-f range greater than 3 kc, usually as high as
6 kc. It is essential, therefore, that the receiver
be capable of handling such a band. To obtain
the intelligence, the signals must be further
demodulated (B1 in Table 1, page 47).
Suppressed Carrier. In the ordinary trans-
missions described above, the carrier level is
high compared to the speech sidebands. To
avoid loading up the transmitter with carrier,
and thereby permit radiating a higher sideband
level, many channels operate on the "sup-
pressed carrier" basis. In this system the carrier
is either eliminated completely, or transmitted
with greatly reduced level. To demodulate such
signals properly, the weak carrier must first be
greatly amplified, or a new one supplied locally.
If this is not done the signals will demodulate
themselves around whichever component in the
sideband happens to be predominant, produc-
ing thoroughly scrambled speech which can
thereafter not be restored. This condition can
be recognized by its characteristic sound to the
ear, together with wide syllabic fluctuations of
the meter which ordinarily indicates the carrier
level.
Twin Channels. With suppressed carrier sys-
tems, usually only one of the speech sidebands
is transmitted. However, a second sideband,
transmitting a second speech channel, is some-
times added, usually displaced from the carrier
by about 3 kc, to avoid crosstalk between the
channels. This is called "twin-channel" opera-
tion, and gives on demodulation an audio signal
covering about 6 kc. The two channels must be
separated and placed in their normal positions
by the methods cited under spread-band sys-
tems.
The above systems are the main types of
radio transmission used commercially with am-
plitude modulation. In addition, in the v-h-f
range and above, there are frequency-modula-
tion systems, and also pulse-modulation sys-
tems, both of which require receivers specially
designed to handle their particular types of
signals. This is too large a subject to cover
here, and reference must again be made to the
radio literature.
Finally it should be mentioned that in addi-
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
INTERCEPTION
39
tion to speech a great deal of telegraph trans-
mission will be found. There are several types
of telegraph signals, including hand-keyed,
such as Morse code, or machine-keyed such as
Boehme and teletype. Any of these types may
be transmitted by keying the carrier, or by key-
ing a tone modulated on the carrier. The marks
and spaces may be represented by changing the
amplitude (on-off) or by changing the fre-
quency (two-tone). Finally, since telegraph re-
quires a much smaller band than speech, it is
often operated on a multichannel basis, that is,
a voice channel will be divided into a number of
telegraph channels. In addition, there are fac-
simile transmission systems, which also may
be operated on an a-m or f-m basis. If a new
signal is encountered whose nature is in doubt,
these possibilities should be kept in mind for
further investigation when the need arises.
RECORDING
The same considerations which make it de-
sirable to obtain a good intercepted signal, apply
also to recording and reproducing scrambled
speech. In addition to the requirements as to
quality and noise, there is an even more serious
one concerning speed regulation. In general,
systems designed for a high degree of privacy
require a high degree of synchronization, and
in many cases ordinary recording methods are
not good enough, not only in long-time average
speed r,Fulation, but also in the steadiness of
the instantaneous speed. In the case of some of
the systems the requirements are so severe that
even the best commercial recorders will not
meet them.
The best solution of this problem is to decode
before recording. This will be possible in many
cases, although it may sometimes entail the
loss of parts of the message while adjustments
are being made or the code is being deter-
mined. It happens that some systems which im-
pose the severest requirements on speed regu-
lation (B3 in Table 1), can be handled in this
way. When this method is feasible, even poor
quality recorders, such as those designed to
record a great deal of material in a small area,
may be good enough.
In some of the systems it will not be possible
to decode before recording. It happens, however,
that in the case of the only known system for
which this is true (F3 in Table 1), the require-
ments as to quality and speed can easily be met
by good commercial type recordings.
The matter of convenience or ease of use of
the reproducing system is very important in
decoding work. In this respect also, the require-
ments are different for different privacy sys-
tems. The recording systems using the emboss-
ing pr4cess, for instance, are convenient
because they produce no thread, and they re-
quire little attention. However, they all suffer
from poor tracking during reproduction, which
can be exceedingly burdensome, especially
where the material must be reproduced many
times over. Recording magnetically on wire is
attractive from the standpoint of convenience
and also quality, but back-tracking is very time-
consuming and laborious.
The best solution, at the present writing, ap-
pears to be disk recording on acetate, with a
machine capable of recording at various speeds.
Low speeds can be used where quality need not
be too good, and a long record is desired. Higher
speeds can be used where better quality is
needed. Such recording systems are commer-
cially available.
DECODING TOOLS
In addition to the facilities discussed above,
an intercept station, if it is to be prepared to
diagnose and decode intercepted enemy signals,
must be equipped with a considerable variety of
special tools. These should include such well-
known devices as oscilloscopes, amplifiers, oscil-
lators, modulators, rectifiers, fixed and variable
filters, and a supply of components for con-
structing special circuits that may be required.
Some of the less well-known devices include
magnetic tape or wire recording and reproduc-
ing equipment in the form of loops with mul-
tiple pickups, commutators for sweep or timing
circuits, variable-speed drive mechanisms, chan-
nel shifters, the variable-area pattern machine,
and the sound spectrograph. There should also
be models of the more important types of exist-
ing speech privacy systems. Finally, and per-
haps most important of all, there should be sta-
tioned at the intercept location a group of
highly trained technicians, who should be thor-
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
40
UNSCRAMBLING AND DECODING METHODS
oughly familiar with radio transmission prob-
lems, radio facilities, cryptanalytic procedures,
and diagnosing and decoding methods. If these
technicians are not conversant with the lan-
guage encountered in intercepted communica-
tions, interpreters should be continuously avail-
able.
Even with all the special tools and personnel,
decoding in many instances will be a difficult
problem, and patience and painstaking effort
will be required to obtain useful information
from scrambled speech. Unless the needs have
been anticipated the enemy may have secret
communication for a considerable period of time
as a direct result of unpreparedness.
4.3
NONCRYPTOGRAPHIC TOOLS AND
METHODS
Beginners in the study of privacy systems
never fail to be amazed at the difficulty of
scrambling speech sufficiently to destroy the in-
telligence. The ear can tolerate or even ignore
surprising amounts of noise, nonlinearity, fre-
quency distortion, misplaced components, gaps,
superpositions, and other forms of interference.
Very often partial or even complete intelligence
can be obtained from a privacy system by par-
tial or imperfect decoding, and this in turn can
often be accomplished by operating on the
scramble in some way which the designer did
not contemplate.
The fact that the ear is such a good decoding
tool in combination with noncryptographic
methods makes the production of privacy sys-
tems very difficult. Scrambling systems which
look very effective on paper sometimes turn out
on trial to degrade the intelligibility very little,
although the scrambled speech usually sounds
unpleasant. Most methods pushed to the point
where they succeed in hiding the intelligibility
so distort the speech that it is impossible to
restore the speech with good quality. In fact,
there are very few speech privacy systems
which achieve a high degree of privacy with
acceptable quality.
Noncryptographic methods are very impor-
tant, because they may reduce the delay in ob-
taining the intelligence substantially to zero.
Furthermore, they may render completely futile
the most elaborately irregular code changing
systems which could be handled only with the
greatest difficulty by straight cryptographic
methods. A number of noncryptographic meth-
ods are given below. Some of them, of course,
result in poor quality, but the saving of time,
labor, and equipment may be very great.
CAPTURED SET OR FUNCTIONAL EQUIVALENT
With many privacy systems all that we need
to listen in is a captured set or its functional
equivalent built from knowledge of the scram-
bling method. An extreme example of this is
simple inversion. In this case the scrambled
speech is quite unintelligible to direct listening,
but if we know it is inversion, we can find the
inversion frequency very quickly by trial. An-
other example is the split-phase system (A5).
The phase-shifting network in the captured set
could readily be adjusted to demodulate either
of the two overlapping sidebands.
Slightly more complicated systems are those
with a simple program. Again with a captured
set or its equivalent it is usually easy to find the
program by trial. The only possible difficulty is
in keeping step with the sending end, particu-
larly if there is no synchronizing pulse. An
example of this is a wobble band displacement
(B3). If, for instance, the wobble is sinusoidal,
with the frequency and the sweep limits known,
the problem is to keep in synchronism. In this
ADJUSTING CHANGE
jHANDLE
POSITION
CHANGE
ROTATING
SWEEP
GENERATOR
DIFFERENTIAL
GEARING
FREQUENCY
VARIABLE
OSCILLATOR
YNCHRONOUS,
MOTOR
FIGURE 1. Method of aided tracking.
connection "aided tracking" might be men-
tioned, a device which is familiar in gunfire
control circles. With this system changes in
both frequency and phase are made simultane-
ously. This is illustrated in Figure 1. Suppose we
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
"
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
NONCRYPTOGRAPHIC TOOLS AND METHODS
41
find ourselves slightly out of step with the sig-
nal. By rotating the adjusting handle forwards
or backwards we can get back into step. Suppose
this adjustment was in the forward direction.
The fact that we had to catch up is an indica-
tion that the motor is slow. Therefore, some of
the motion of the handle required for catching
up is used by means of gearing to change the
frequency driving the motor. The gear ratios
are chosen to suit the particular problem. With
this method it is possible to get into step with
and stay in step with systems such as alternate
displacements and regular wobbles.
COMPROMISE DECODING METHODS
All the methods outlined in this section have
been tried, at least in the laboratory. Their suc-
cess, however, naturally depends to some extent
on the switching rates and similar variables. It
is possible, therefore, that a method might
MOD
1 . 13 TO 16 KC 10 TO I6KC
FIGURE 2. Band-shift filter; an important decod-
ing tool.
BP FILTER
UPPER CUTOFF 13 KC
LOWER CUTOFF 11.5 TO 12..0_ KC
MOD
prove unsuccessful against a scrambling system
which seems to be in the same general class as
the one that was tried in the laboratory.
Consider, for example, a system (A2) which
involves inversion about a number of frequen-
cies in succession. If these frequencies are not
too far apart we can choose a single frequency
somewhere in the middle range and demodulate
the whole signal with this one frequency. The
resulting band will be right side up, but dis-
placed by varying amounts not exceeding half
the total range. This has been found to be quite
intelligible, provided the switching rate is not
too high or the range of frequencies too wide.
With some systems it is expedient to listen
to only a portion of the frequency range rather
than the whole range. An outstanding example
of this is the system in which the subbands are
variously delayed (F1). Conceivably, these de-
lays could constantly be changed with time ac-
cording to a never repeating program. This,
however, would be futile because with a band
filter we need listen to only one of the bands;
disregarding the others. Unless this band is
very narrow the intelligibility may be practi-
cally complete. Similarly in band-splitting sys-
tems if the switching is not rapid (DI) we can
follow one of the bands around the frequency
range. The lowest or second lowest band is usu-
ally the best. Another example is the tone se-
q4ence (J3) ; instead of trying to filter out one
tone at a time as it occurs, we can leave all the
filters in all of the time and still have enough
speech coming through to yield the intelligence.
A special case in which the rejection of a
part of the frequency band of the scramble
makes decoding easier concerns those systems
such as A5 which depend on carrier phase to
mix and then separate components. There is no
phase requirement imposed on the demodulat-
ing carrier unless both sidebands are trans-
mitted. Therefore, either sideband of such a
system may be suppressed with a filter, and the
remaining sideband demodulated with a carrier
of any phase. The two signals in the sideband
will then be simply superposed.
For purposes such as those outlined a valu-
able tool is the band-shift filter illustrated in
Figure 2. With this device a band of adjustable
width can be taken from any portion of the
signal frequency range (0 to 3 kc) and relocated
in any other portion of the same frequency
range either straight or inverted. One form of
band-shift filter is described in Preliminary Re-
port No. II of Project C-43.36 It consists essen-
AIR CONDENSERS
FIGURE 3. Band-pass filter in which pass band is
variable.
tially of a double modulator, but with a band
filter of variable width. If the frequency loca-
tion of the band is not to be changed, the switch
in Figure 2 should be in the left-hand position.
One form of variable band filter is shown in
Figure 3. This tool has also proved useful in
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
42 UNSCRAMBLING AND DECODING METHODS
certain other systems such as the multiplica-
tion system (HI) and the time division multi-
plex [TDM] system (El).
Sometimes it is expedient to listen to a scram-
ble only part of the time. Some of the simpler
coding programs can sometimes be broken down
in this manner by trial. For instance, if a cod-
ing cycle has N elements we can listen to every
Nth element and make whatever adjustments
are needed to make this sound natural. We can
then listen to the next adjacent element and
adjust the system so that these elements blend
A
PEAK CHOPPERS
COMPRESSOR
FIGURE 4. Elementary circuits of peak choppers
and compressor.
properly. This attack applies to a system in
which several different displacements are used
(B2). A captured set, of course, is the easiest
way of selecting every Nth element because it
is usually easy to make the other time elements
inoperative.
Another useful device is the limiter, or peak
chopper. In this same class is the compressor.
These are illustrated in Figures 4A, B, C. They
all tend to equalize the successive lobes of a
complex wave. The peak chopper simply chops
off any peak which exceeds a certain instan-
taneous voltage. The compressor operates more
gradually and leaves the waves well rounded.
If straight speech is put through any of these
devices, distortion products are generated be-
cause the wave form is radically modified. It is
found, however, that this kind of distortion
damages the intelligibility very little. These de-
vices should be useful against any privacy sys-
tem in which sudden changes of level occur. A
good example is the subband level modulation
system (H3). A separate limiter or compressor
in each of the subbands will tend to smooth
out the level variations and make the speech
intelligible.
Another nonlinear device is the rectifier. Two
forms are shown in Figures 5A and B. The
rectifiea? as used here should not be confused
with the detector. The latter device also recti-
fies, but it then has a time constant incorporated
in the output circuit which tends to smooth the
output and give the envelope wave. The rectify-
ing action which is wanted here simply takes
all the negatives lobes of the signal and turns
them over. As in the case of the limiter, straight
speech put through a rectifier of this type is
about 95 per cent intelligible.
In the privacy system designated A4 the
phase of the speech signal is reversed at short
irregular intervals. If this signal is now recti-
fied, all the negative lobes will be made positive
and the resulting wave will be indistinguishable
from rectified straight speech except for slight
discontinuities at the points where the reversals
PAD
MOD
A
FIGURE 5. Two forms of rectifier circuits.
occurred in the privacy system. This is illus-
trated in Figure 6. Therefore, a simple phase-
reversal system, no matter how irregular,
should yield to rectification except that distor-
tion in the transmission process tends to change
the wave form and thereby degrade the quality
of the resulting speech. It should be noted that
the multiplication process (111) also results in a
phase reversal every time the coding wave
passes through zero. It has been found that rec-
tification tends to make this kind of scramble
more intelligible also.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
NONCRYPTOGRAPHIC TOOLS AND METHODS
43
A very useful noncryptographic device is
suiSerposition. For instance, with a three-
channel mixing system such as Li or L2, if
all three channels are listened to simultaneously,
three conversations will be heard at once, or
possibly one conversation with two noises super-
posed. Experience has shown that under such
conditions it is usually easy to concentrate on
the desired channel and ignore the others.
Another form of superposition is illustrated
by the following: Consider a split-band system
ORIGINAL SIGNAL
PHASE REVERSED
RECTIFIED
ORIGINAL RECTIFIED
FIGURE 6. Action of rectifier, useful in breaking
down multiplication scrambles.
(D2) in which six different codes are used in a
never repeating sequence. This would be rather
difficult to handle by cryptographic means. Sup-
pose, however, we had six separate decoding
units, each set to decode one of the six codes. If
the scrambled signal were fed into all six of
these decoders simultaneously, one of them
would always have straight speech in its output.
The other five would be scrambled. If these six
outputs are all superposed, we will hear straight
speech with five scrambled superposed. This
straight speech can be understood quite easily.
It will be noted that the unwanted components
in this kind of suDerposition are derived from
the wanted omponents, and always vary in
level simultaneously with the wanted compo-
nents; it appears that under these conditions
they do not do much damage.
The split-band equipment illustrated in Fig-
ure 6 of Chapter 1 is adapted for this kind of
superposition. A multiplicity of cross connec-
tions is made from each of the band-pass filters
to the output modulators whereby each of the
bands in the signal is placed in the desired bands
in the output. Steps should be taken to see that
these cross connections do not interfere with
each other. An amplifier after each band filter,
for instance, will perform this function. Figure 7
illustrates a simpler case of superposition ap-
plied to a system using two band shifts (B2).
4 It may be noted here that superposing time
displaced elements does not appear to be suc-
cessful. For instance, if all the segments of the
commutator jn a time ,division scrambling
[TDS] machine are connected to all the pole-
pieces, the output will be straight speech with
several scrambles superposed. This has been
found to be completely unintelligible.
In certain cases which have been met in
Project C-43 the privacy sets are equipped with
dials or similar means which were intended to
provide an easy method for obtaining a large
number of different codes. In some cases the
different codes may not be sufficiently different
to be mutually private. That is, while there may
be literally millions of different combinations,
it sometimes happens that there are thousands
FIGURE 7. One form of superposition decoding.
of combinations which will decode material
scrambled with one of the combinations. Vari-
ous degrees of quality will result from these
partial or incorrect decoding operations. How-
ever, as long as intelligibility can be extracted
the codes cannot be considered mutually private.
In such cases it is possible with a captured
machine simply to manipulate the dials sys-
tematically or unsystematically and listen to
the result. When the speech begins to sound
somewhat natural, systematic trials of each dial
in turn will sometimes steadily improve the
quality. Something of this sort could be done
SECRET
Declassified and Approved For Release 2013/12/11: CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
44 UNSCRAMBLING AND DECODING METHODS
with simple TDS systems also, except that the
use of interlaced codes makes this somewhat
more difficult.
In certain cases where there are a large
number of codes possible but only a few of these
codes are good codes from the standpoint of
direct listening, it would seem reasonable that
any code applied to the scrambled signal should
turn the good code into a poor code. In the five-
band split-band system for instance, there
are some 3,840 possible codes but only twelve
or so are considered really good. Any code in
SPEECH
NOISE
4-- SPEECH
ONLY
FIGURE 8. Use of directional discrimination
against noise.
the decoding machine, therefore, should de-
crease the privacy for direct listening. This has
been tried in the laboratory but has not been
pushed to the point of determining whether it
could compete with the superposition method.
The idea may possibly apply to other systems
which may be encountered.
A very specialized device, which applies to
wire line communication only, should be men-
tioned here because it is not very well known.
It distinguishes between the two directions of
FIGURE 9. Use of relay to disconnect receiver
from line during heavy bursts of noise.
transmission over wires. In the masking pri-
vacy system J2, for instance, the clear signal
in one direction is masked by noise sent in the
other direction. The device illustrated in Fig-
ure 8, however, discriminates against the noise,
allowing the speech to be heard. It requires a
small series resistance, which is built up by a
step-up transformer to the line impedance. The
secondary is connected to the other side of the
line. The direction of discrimination depends on
the phasing of the transformer windings.
AUTOMATIC DECODING
Whether speech is intelligible or unintelli-
gible is purely a subjective matter. However,
the method of making speech unintelligible in-
volves making physical changes in the speech
wave. Certain kinds of physical changes can
be detected quite readily by objective means
and utilized to undo the scramble automatically.
The most elaborately irregular code program
is completely futile if this kind of decoding can
be applied.
A very simple example of this is shown in
Figure 9. Suppose the system consists of short
spurts of noise applied in an irregular manner.
The noise must be high in level compared to the
speech in order to mask the speech. Therefore,
if the signal is applied to an amplifier-detector,
connected to a relay (or electronic switch) the
relay can be so biased that it operates only on the
noise spurts. The receiver is momentarily dis-
connected from the line whenever a noise spurt
occurs. The same method can be used for level
modulation systems (H2 and H3). Instead of
disconnecting the receiver, the high-level por-
tions of the signal cause the receiver to be con-
nected to a parallel path containing the required
amount of loss to equalize the levels. In the case
of subband level modulation (H3), a separate
device of this type must be used in each sub-
band.
The system just described operates on a total
energy basis. Sometimes it is possible to obtain
a switching signal on the basis of energy fre-
quency distribution. Consider, for instance, the
system using two different displacements (B2).
The alternate positions of the speech band are
illustrated in Figure 10. In one position, the
band is right side up and occupies the range
from 2 to 5 kc. The alternate position is in-
verted occupying the range from 3 to 6 kc.
Since most of the energy in the speech band
is concentrated in the low-frequency part of the
original spectrum most of the time, the system
illustrated in Figure 11 can be used to decode
this material automatically. The signal is ap-
plied to two band filters, one covering the range
2 to 3 kc, the other passing 5 to 6 kc. The out-
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
NONCRYPTOGRAPHIC TOOLS AND METHODS
45
puts of these band filters are rectified individu-
ally and fed to the two windings of a polar
relay. The relative energy in the two band
filters will be different for the two displace-
ments and the relay in Figure 11 will be oper-
ated alternately in the two directions, thereby
automatically connecting the proper carrier to
the input modulator in Figure 7 to put the
speech band in its normal position. This will
not be infallible, but with displacements as
different as the ones used in the illustration, it
should operate sufficiently well to yield most of
the intelligence of the message. Naturally, the
smaller the physical difference between the two
positions being distinguished, the more false
operations there will be. However, this method
is instantaneous even with an irregularly
switched system, whereas cryl?A9a-a-phic-metti--------d.e
un
ods would be very time-consuming.
Another variation of this general technique
might be mentioned for the sake of complete-
ness although it is somewhat more speculative.
Consider a privacy system which depends on
speed changes (F4). Changes in speed cause
changes in the pitch of the voice. Suppose we
apply this signal to a circuit which measures
the voice pitch. This technique has been worked
out in connection with the Vocoder. The output
of this circuit, which is a varying frequency,
is used to change the speed of a motor. The
motor is part of the drive of a magnetic tape
recording and reproducing system through
which the signal is passed. As the motor speed
is made to change, the tape speed changes in
such a direction as to tend to keep the derived
frequency constant. This takes out not only the
speed variations, but also the voice inflections.
However, a monotone is quite intelligible.
The following method, which has not yet
been tried out, is intended to apply to irregular
band displacements or wobbles (B4), which
would be exceedingly difficult to handle any
other way. Consider a system in which the band
is kept right side up, but is wobbled over a
range sufficient so that demodulation with some
intermediate carrier frequency will not give an
intelligible signal. Suppose the wobble follows
an irregular, nonrepeating program. The fol-
lowing decoding method is proposed.
The signal is impressed on a network having
a very steeply rising loss characteristic. If the
speech band were not wobbled, this network
would simply tend to make the lowest harmonie
of all voiced sounds the strongest component.
With the wobble, the same thing will be true
except that the level of this component will
TIME
4
I
5 KC
0
FIGURE 10.
10 KC
FREQUENCY
Sidebands in two-position displace-
ment system.
suiting
o severe fluctuations. Therefore, the re-
nal is subjected to some form of
automatic vIume control and also a limiting
action, tending to derive a single frequency.
Forgetting voice inflections for the moment,
this derived frequency would fluctuate up and
down (in frequency) in step with the band
wobble. In fact, it could be used as a subcarrier
in a double modulation decoder to demodulate
the signal to approximately the correct position
in the frequency range. It will be in error,
TO INPUT
-0 MOD IN
(I)FIG 7
16 KC
POLAR RELAY
FIGURE 11. Automatic decoding system depend-
ing upon energy-frequency distribution.
however, by an amount equal to the instantane-
ous voice pitch. Possibly this amount of error
will not prevent the signal from being intelli-
gible (we know that this amount of displace-
ment does not destroy the intelligence of other-
wise normal speech).
If it is desired to correct for this error, two
methods suggest themselves. One possible
method is to subtract from the derived fre-
quency, by a modulation process, an amount
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
46 UNSCRAMBLING AND DECODING METHODS
equal to the average pitch of the voice being
monitored. This will leave a small fluctuating
error. Another possibility is to derive the actual
instantaneous voice pitch, by difference tone
methods, and subtract this amount from the
derived subcarrier frequency.
If the displaced band is inverted instead of
right side up, a similar procedure can be used,
with a network of opposite loss characteristics.
In either case this method will correctly demod-
ulate only the voiced sounds, but experience
suggests that this is sufficient. If not, some kind
of carry-over effect might be incorporated in
the system to prevent sudden changes in the
subcarrier frequency, thereby tending to hold
over correct demodulation for short unvoiced
sounds also. This method has not been tried,
but is felt to be worth recording because of the
great difficulty of handling irregularly wobbled
systems by any other method.
Another rather speculative automatic method
might be mentioned because some form of it
might prove useful against certain multiplica-
tion systems such as H1. The code wave in the
particular case encountered was repeated many
times per second, and there was a synchroniz-
ing pulse ahead of each cycle. If the signal is
applied to a synchronized cathode-ray oscillo-
scope with a highly persistent screen, a definite
pattern appears because the coding wave al-
ways passes through zero at the same time.
Also, the speech energy tends to average out
after a few cycles so that the pattern reflects
the amplitude of the coding wave. It is quite
conceivable that this pattern on the screen
could be scanned optically and used to generate
a decoding wave for automatically unscram-
bling the signal. Obviously, if the coding wave
is changed periodically, a new decoding wave is
automatically produced. The only requirement
is that the coding wave persist long enough to
form an average pattern on the screen.
Another variation of automatic decoding
methods might be termed "parallel-automatic"
because two or more complete decoding units
are used in parallel but only the correct one is
applied to the listening receiver. To emphasize
the difference between this method and the one
previously discussed, we will use the same ex-
ample, namely, the system with two band dis-
placements. Referring to Figure 7, suppose
instead of the parallel modulators, there were
two complete units in parallel including the
band filter, the second modulator, and the out-
put filter. One of the waits is fed with the8-kc
carrier, the other with the 16-kc carrier. Each
unit will have straight speech in its output half
the time, and the other half the time will have
inverted speech displaced by 1 kc. A 1-kc low-
pass filter can then be used in a device similar
to Figure 11 to switch the listener to whichever
one of the decoding units has the straight
speech. For the particular system used in the
illustration, there does not appear to be any
particular advantage of one method over the
other. However, the latter system can be ap-
plied in cases where the other method might
not be feasible.
The parallel-automatic method can be made
to give a different type of switching signal. For
instance, use might be made of the harmonic
relationship between the components of speech
when the speech band is in its normal position.
If the voice pitch happens to be 100 cycles, then
all the harmonics will be multiples of 100 cycles.
If this speech is put through a suitable non-
linear system such as a rectifier or limiter,
difference tones will be generated which will
also be multiples of 100 cycles. If, however, the
speech band is displaced from its normal posi-
tion in any way, the difference tones' will not
coincide with the speech components. If, for
instance, the whole band has been displaced by
50 cycles, then the speech components will be
150, 250, 350 cycles, etc. The difference tones
generated by a nonlinear system will be 100,
200, 300 cycles, etc. If we now take a second
difference between the output of the nonlinear
system and the original components, there will
be generated multiples of 50 cycles. The lowest
component of this series will be lower than the
pitch of the voice. This will be true regardless
of how far the original band has been shifted,
except for the special case where the shift
happens to be an exact multiple of the voice
pitch. Since, however, the pitch is constantly
varying, this coincidence is of very brief dura-
tion. Theoretically, at least, a low-pass filter
with a cutoff lower than the normal range of
voice pitch can be used as a clue to determine
whether a speech band is in its proper location.
The method then would consist in having sev-
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
NONCRYPTOGRAPHIC TOOLS AND METHODS 47
eral decoders in parallel, listening only to the
one which did not generate a component in the
low-pass filter.
The above illustrations will serve to show
the possibilities of noncryptographic types of
attack on privacy systems. When a new system
is encountered, this type of attack should be
given serious consideration because of the
saving in time and equipment. Naturally, as
pointed out above, straightforward crypto-
graphic attack can be made to yield a bet-
ter quality signal. However, experience has
shown that the ear can become familiar with
certain kinds of distortion and learn to extract
the intelligence more and more readily with
practice.
TABLE 1. Summary of scrambling systems.
Noncrypto-
graphic
attacks
(Table 2)
Capable of
crypto-
graphic
attacks
Noncrypto-
graphic
attacks
(Table 2)
Capable of
crypto-
graphic
attacks
A. Single modulation
1. Inversion
2. Variable frequency
version
3. Alternate inversion
4. Phase reversal
5. Split phase
B. Double modulation
I. Fixed displacement la
2. Stepped displacement lb, 2a, 2f, 3b, 3f
3. Wobbled displacement,
regular lb, 2a
4. Wobbled displacement,
irregular 3d
C. Triple modulation
1. Re-entrant inversion,
steps
2. Re-entrant inversion,
continuous
D. Band splitting
1. Slowly switched
2. Rapidly switched
E. Time division multiplex
I. 4-band system lb, 2b
2. With noise channel lb, 2b
F. Magnetic tape
1. Delayed subbands 2b
2. TDS, repeated code 2c, 2g, 2h
3. TDS, nonrepeated code
4. Speed variations lb, 2a, 3c
5. Backwards trans-
mission lb
6. Alternate backwards
and forwards lb, 2c
la
in-
lb, 2a, 2f
lb, 2f
Ib, 2e, 2g, 3e
la -
lb, 2c, 21
lb, 2a
Ib, 2b, 2f, 2g, 2h
2c, 2f, 2h
?
G. Tape plus modulation
1. TDS and inversion
2. TDS and split band,
synchronous
3. TDS and split band,
nonsynchronous
4. Two-dimensional
5. Double-speed split lb
6. Half-speed split lb, 2c
21, 2g, 2h
H. Wave form distortion
1. Multiplication 2b, 2e, 2g, 3e
2. Level modulation 2d, 3a
3. Subband level modula-
tion 2d, 3a
J. Masking methods
1. Signal plus inoise, same
dixection '
2. Signal plus noise,
opposite direction 2i
3. Tone sequence lb, 2b, 3a, 3b
4. Noise spurts 2c, 2d, 3a -
5. Nonlinear distortion la
K. Vocoder methods
1. Permute channels
2. Invert channels lb, 2g
3. TDS, channels lb, 2g
4. Two-dimensional
scramble lb, 2g
5. Time division multiplex lb
L. Channel mixing
1. Time division mixing
2c, 21
2. Subband mixing
2b
3. Combination
2c, 21
TABLE 2. Noncryptographic decoding methods.
1. Captured set or functional equivalent
a. Fixed condition?find by trial
b. Simple program?get into step
2. Compromise decoding methods
a. Intermediate condition
b. Listen to portion of frequency band
c. Listen part time
d. Limiter, peak chopper, compressor
e. Rectifier
f. Superposition
g. Approximate Code by trial
2. (Continued)
h. "Spoil" good code by recoding
i. Directional discriminator
3. Automatic decoding
a. Total energy
b. Energy frequency distribution
c. Pitch-change corrector
d. Wobble corrector
e. Code wave generator
f. Parallel automatic
g. Inharmonic detector
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
48 UNSCRAMBLING AND DECODING METHODS
In general, noncryptographic methods re-
quire that the signal, as received, be of fairly
good quality. In some cases, the saving in time,
labor, and equipment would be so great that if
the signal, as received, is too poor to permit
noncryptographic attack, the most reasonable
thing to do is to move the intercept station to
get a better signal.
In Table 1, there is listed for each privacy
system, the type of noncryptographic attack
which might apply. It should be emphasized
once more, however, that the method which
succeeds at one switching speed may fail at
another. The list, therefore, should be taken
only as a recommendation of systems which
should be considered. The noncryptographic de-
coding methods are summarized in Table 2.
4.4
CRYPTOGRAPHIC TOOLS AND
METHODS
A cryptographic decoding method' involves
(1) actually determining a code which will
undo the scramble, and (2) restoring the
speech by means of this code. In the case of
repeated codes, this can sometimes be done
rather quickly. An example is the repeated-code
TDS system. The actual codes used can be
found in about 15 min. Having found the code,
we can set it into our receiving machine and
thereafter listen to the speech directly. In the
case of nonrepeated codes, every bit of the
message must be handled individually. It may
take a thousand or even a hundred thousand
times as long to decode as it did to speak the
words. It may take hours or even days to obtain
the intelligence from a short message; mean-
while other messages will have been sent and we
get farther and farther behind. The only way
this could be avoided would be to have approxi-
mately as many teams working in parallel as the
ratio of decoding time to message time, which,
of course, is impractical if the ratio is large.
As stated previously, the sound spectrograph
?described in detail later?is of tremendous
assistance in recognizing the nature of an un-
known scrambling system. The ear can usually
recognize the presence of time discontinuities.
It can also recognize the peculiar quality which
results from band-shifting systems. The exact
nature of the scramble, however, is usually
impossible to establish with the ear. Even
scrutiny of the wave form may yield no clue.
The strikingly graphic analysis provided by
the spectrograph, however, usually takes the
mystery out of the scrambling method immedi-
ately.
For example, speech privacy systems having
frequency subbands will show horizontal dis-
continuities or boundaries in their spectro-
grams.4Similarly, systems employing time
division will show vertical boundaries. A con-
siderable variety, of systems display both hori-
zontal and vertical boundaries. Methods of
telling these systems apart are described in the
final report on Project C-32.44
PROGRAM DETERMINATION
The simplest cases to handle are those in-
volving a program which can be determined
directly from spectrograms by inspection or
measurements. The re-entrant inversion system
(Cl). might be used as an example. Suppose a
multiplicity of displacements were used in some
irregular sequence. Discontinuities marking the
inversion frequencies appear in. the spectro-
grams and mice they have been determined by
measurements on a large number of spectro-
grams, the program can thereafter be deter-
mined quite readily by using a template. This
template can be marked directly with the set-
tings of the decoding machine which will restore
the speech to its normal position.
Another example involving a program would
be one like B2, in which two different displace-
ments are used alternately, with the intervals
irregular in duration. Here the time boundaries
will be quite apparent and they can be measured
with a suitable time scale.
In all likelihood changes of the above types
will occur in discrete steps for practical reasons.
The use of a program involving continuous
changes with time presents formidable tech-
nical difficulties at the authorized as well as the
unauthorized terminals.
MATCHING SPECTROGRAMS
In cases where the scrambling system in-
volves rearrangement of the speech elements
in time or in both time and frequency, the basic
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
CRYPTOGRAPHIC TOOLS AND METHODS 49
method for determining the codes involves cut-
ting up spectrograms along the element bound-
aries and rearranging the elements so as to
restore the straight speech. An example is
shown in Figure 12. The criterion for rear-
ranging the elements is that there should be
continuity at the boundaries. This continuity
includes the position and direction of the indi-
means for making a mechanically inverted pat-
tern as well as a normal pattern. The spectro-
gram at the top of Figure 13 shows a normal
pattern. Directly below it is an inverted pat-
tern of the same material. A mechanically in-
verted pattern is indistinguishable from a
pattern produced by electrical inversion of the
speech. Similarly, if the whole inverted spectro-
FIGURE 12. Method of matching spectrograph patterns of nonrepeated code TDS.
vidual harmonics, the position and direction of
the resonance areas, and, in general, the ampli-
tude as represented by the darkness or lightness
of the patterns. The pieces are numbered before
the matching process begins and when the
matching has been completed, the numbers on
the pieces determining the code.
If the scrambling process involves inversion
of the time or frequency scales, straight speech
can be restored for matching purposes by mak-
ing two spectrograms as shown in Figure 13.
Present models of the spectrograph include
gram is turned through 180 degrees, so that the
base line is at the bottom and towards the ob-
server, the result is indistinguishable from the
case in which the speech is transmitted back-
wards. Therefore, if an element in the scram-
ble is inverted, it may be recovered as straight
speech for matching purposes by cutting the
element from the mechanically inverted pat-
tern. If an element has been transmitted back-
wards, it can be restored to normal by cutting
it from the inverted pattern and rotating it 180
degrees as described above. If it is both back-
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
50 UNSCRAMBLING AND DECODING METHODS
wards and inverted, it may be restored by cut-
ting it from the regular pattern and turning it
around.
It has been found from experience that
matching is facilitated by enlarging the spectro-
grams by a factor of about two to one. Not only
is the increased size easier to handle, but the
heavy photographic paper is much better to
handle than the facsimile paper employed in the
poses regularly, then it may pay to adopt the
technique described in Preliminary Report No.
1317 (Project C-43) for producing large spectro-
grams photographically.
To facilitate matching, appropriate means
should be used for handling the elements. It
has been found that a slightly adhesive surface
is advantageous. In the illustration of Figure 12
this surface was provided by coating the boards
FIGURE 13. Inversion of time and frequency scales in spectrograms. If scramble contains inverted ele-
ments, these will appear right side up in mechanically inverted spectrogram. Time scale may be inverted
by rotating elements 180 degrees. Note position of base lines in examples.
spectrograph. The latter is delicate in texture
and its surface is easily stained. In this con-
nection it should be noted that the process of
enlarging the spectrograms does not appreci-
ably affect the decoding time in the case of
nonrepeated code systems. There will, of course,
be an initial delay, but in general, the matching
time will be controlling. Spectrograms can be
made, enlarged, and cut up faster than they
can be matched. If it is found necessary, how-
ever, to use spectrograms for matching pur-
and also the backs of the elements with ordi-
nary rubber cement. This is also the case in
Figure 14. This latter example shows a two-
dimensional scramble. Horizontal strips of rub-
berized Bristol board were provided for match-
ing along the time axis.
Once a system has been thoroughly diagnosed
certain numerical properties of the coding proc-
ess will be known. Advantage should be taken
of this knowledge to supplement and check the
matching process. Examples are given in Pre-
SECRET
Declassified and Approved For Release 2013/12/11: CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
CRYPTOGRAPHIC TOOLS AND METHODS
51
liminary Reports No. 10," 14," 22,20 and 2621 of
Project C-43.
EFFECT OF TRANSIENTS
The two examples thus far cited of spectro-
gram matching were artificially produced by
cutting up spectrograms of straight speech, and
the boundaries are therefore clear and sharp. In
practice the time and frequency boundaries will
be obscured by- transients. Frequency bound-
aries are filter cutoffs, and they are marred by
overlap or underlap and by phase distortion.
embodiment of this improvement in a spectro-
graph has not been accomplished because the
need was not sufficiently pressing in Project
C-43.
The basic idea for avoiding the obscuring
effects of spillover is to permit the spillover to
take place in such a way as to be subsequently
removable. For instance, suppose a sample of
TDS were recorded on the magnetic tape and
suppose0 the spectrograph were equipped with
a suitable switching arrangement such that
only every alternate element was reproduced.
FIGURE 14. Matching spectrograph patterns of two-dimensional scramble.
This, however, is not as serious as the transients
occurring at the time boundaries. There is a
basic difficulty here, arising from the desire to
obtain a high degree of frequency resolution,
which entails the use of a narrow scanning
filter. The response and decay time of such a
filter is appreciable in comparison with the ele-
ment length in many scrambling systems. The
decay time produces the more serious of the two
effects. It causes energy from a strong element
to spill over into the adjacent following element
in the spectrogram. This difficulty is unlikely to
cause trouble in any application of the spectro-
graph except decoding. Therefore, it is felt that
means for alleviating this difficulty should be
recorded here. A small amount of exploratory
work has been done along these lines, but the
The spillover from each element would then
occur in a blank area, and it could subsequently
be trimmed off, leaving a sharp, clear boundary.
A second spectrogram could then be made of the
alternate elements, again trimming off the spill-
over.
USE OF TWO SCANNING FILTERS
A logical extension of this idea, which would
save some time, would be to have two scanning
filters and use them alternately by suitable
switching means. Both the inputs and the out-
puts of the filters would have to be switched,
and the two switches should be separated by the
appropriate time delay to take account of the
transmission time through the filter.
A third variation of this idea which requires
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
52 UNSCRAMBLING AND DECODING METHODS
less equipment, is to make one spectrogram in
the usual manner and then make a second spec-
trogram with the machine running backwards.
The spillover always occurs into the leading
edges of the elements in spectrograms. Cutting
the first spectrogram in the proper places will
result in clear, sharp right-hand edges on each
element, but each left-hand edge will be ob-
scured by spillover. Cutting the second spectro-
gram in the proper places will give clear
left-hand edges on each element. Matches could
then be made between elements from the nor-
course, might be aggravated intentionally as
part of the privacy feature of the system. On
the whole, however, it looks as though ampli-
tude representation should be an improvement
in decoding work.
MATCHING VARIABLE-AREA PATTERNS
For some purposes it has been found that
wave form patterns offer certain advantages
ove4 spectrograms. They can be made more
rapidly and they can be played back directly to
reproduce the original speech. Intrinsically,
........VyftrIRMOIMI movaidonatavreiv
Rwitvvra eviwnworwifireftweramamilm
444gat4Alciwaviagag .4,0=ialooraism gamma
0~4VVP f4144V,W
ME= nAkel.,,un
021.0,1.1.
Ramona sereaseemargammagresseamerearmeassmare gaessamessa
101,411=11, 1.6001Vram reoroaraor
tZVX14
alblAtdivia
FIGURE 15. Matching variable-area patterns of nonrepeated code TDS.
mal and backwards spectrograms, in such a way
as to utilize the good edges of the elements.
In other respects it is to be expected that the
patterns produced by the spectrograph can be
improved. For instance, studies have been made
which show that amplitudes can be represented
in such a way that they can be interpreted
quantitatively. This is an improvement over the
rather indefinite shades of gray in the usual
spectrograms. It would provide another criter-
ion for matching. In some cases, however, this
might be a handicap. For instance, in TDS sys-
tems the pole pieces are not all of equal effi-
ciency. The amplitudes of adjacent speech
elements are affected by this change in effi-
ciency and they might not appear to match
when they really should. This condition, of
wave form patterns are not as good as spectro-
grams for diagnosing frequency shifts and the
like. However, they present the time scales more
graphically and they are not subject to tran-
sients at time discontinuities such as the spill-
over effects previously discussed.
The particular type of wave form pattern
found most useful was a variable-area pattern
similar to the sound track used in moving pic-
tures. Variable-area patterns are more distinc-
tive to the eye than oscillographic traces. They
form geometric designs that catch the eye and
facilitate matching. The manner of producing
and playing back these patterns is described in
Preliminary Reports No. 1,22 7,23 and 1224 of
Project C-43. An example of variable-area pat-
terns in process of matching is shown in Figure
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
I
.q.
CRYPTOGRAPHIC TOOLS AND METHODS 53 111
;
15 taken from Preliminary Report No. 2621 of
Project C-43.
? Variable-area patterns of this type have been
found particularly good for decoding TDS sys-
tems, especially repeated-code systems. Ampli-
tudes are clearly represented in these patterns.
By matching a multiplicity of cycles of a re-
peating-code system simultaneously, it is pos-
sible to take advantage of this amplitude
representation even though the wave form itself
band. Changes in the split-band code will then
have no effect on the wave form of patterns pro-
duced in this manner.
It was also proposed at one time that the use
of a whisper or monotone instead of normally
inflected speech would increase the privacy of
TDS systems. Again this is true in terms of
spectrograms, but it was found that variable-
area patterns could be matched almost as easily
for whispered speech as for normal speech, and
4
FIGURE 16. Oscillographic traces of Vocoder channel signals.
might be obscured by other features of the
privacy system. For instance, the use of split-
band coding was once proposed to increase the
privacy of TDS systems. This combination
would be much more private than plain TDS if
judged on the basis of matching spectrograms,
particularly if the split-band codes were rapidly
switched at intervals not simply related to the
TDS elements. No difficulty, however, was found
in matching the variable-area patterns to find
the TDS code. This is described in Preliminary
Report No. 1925 which also describes a scheme
for nullifying the effect of split-band coding on
the wave form. This consists of modulating all
the frequency bands down into one frequency
with the monotone it was actually easier. This
is described in Preliminary Report No. 1626 of
Project 0-43.
Another feature of the variable-area patterns
which might be useful is that the patterns have
characteristic shapes. Usually they look like a
series of damped oscillations with the highest
amplitude at the beginning of each fundamental
period. This should enable the recognition of
cases in which speech is transmitted backwards.
The characteristic periodicity of the patterns
might also be used to recognize whether a fre-
quency band is in its proper location.
Toward the end of Project 0-43 it came to be
recognized that there would be considerable ad-
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
54
UNSCRAMBLING AND DECODING METHODS
vantage in using a compressor in the produc-
tion of variable-area patterns. This tends to
bring out low-level sounds. The distortion of the
wave forms resulting from instantaneous com-
pression is immaterial if they are to be used
only for matching. This kind of compression,
however, should be sharply distinguished from
automatic volume control action. The latter is
relatively slow acting and it is obvious that in
TDS systems, for instance, it would alter the
amplitudes of certain elements in such a way as
to make matches impossible.
MATCHING OSCILLOGRAMS
Oscillographic traces can be used instead of
variable-area patterns, although in general
permuted at short intervals provides a rather
difficult privacy system to decode.
It has been found that compression enhances
the value of oscillographic traces of this type.
Without compression the lower amplitudes are
obscured by the width of ?the traces. Instan-
taneous compression makes changes in the
magnitude or direction of the traces apparent
in the lower level sounds. The patterns shown
in Figure 16 were produced in this manner.
4
INDICATOR METHODS
In the following methods a visual indication
is obtained denoting which of several possible
choices puts the speech elements in their proper
order. These methods are applicable only to
FIGURE 17. Time division scrambling [TDS].
there will be a disadvantage. For Vocoder
privacy systems, however, oscillographic traces
are required. The signals in Vocoder channels
are essentially fluctuating d-c signals after they
are modulated down to their normal frequency
location. They can best be examined in the
form of oscillographic traces. Figure 16 shows
a set of undistorted Vocoder channel signals.
It will be noted that there is a tendency for
the amplitudes to vary simultaneously in the
several tracks. It has been found that if the
signals in the various channels are permuted,
even with the sharp edges resulting from arti-
ficially produced scrambles, the number of
mismatches tends to be about 40 per cent. This
means that a Vocoder system with its channels
cases where the possible number of choices is
not overwhelmingly great. A natural example
of a visual indication occurs in the illustration
of TDS in Figure 17. Whenever two originally
adjacent speech elements remain adjacent in
the scramble the two elements are not separated
by a time boundary in the spectrogram. Ele-
ments which do not belong in adjacent positions
have a boundary resulting from discontinuities
in the harmonics and from spillover effects.
The absence of a time boundary can be taken
as an indication that the two adjacent elements
belong together. To make use of this effect the
following procedure is suggested. Record a
sample of the scramble on a loop of tape. Re-
produce this sample through a TDS machine
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
CRYPTOGRAPHIC TOOLS AND METHODS
55
and make a spectrogram, noting any adjacen-
cies which occur. Change the code in the TDS
machine and make another spectrogram again
noting adjacencies. A systematic set of codes
should be worked out in advance which explore
all the possible combinations of elements. At
the end of such a cycle of operations it should
occur at the boundaries of elements which do
not belong together. These will generate fre-
quencies higher than the cutoff of the high-pass
filter and will appear as pulses on the scope.
The absence of a transient will indicate either
that the elements belong together or that no
energy was present. Again a systematic cycle
FIGURE 18. Effect of rectification on normal and band-shifted speech. A, straight speech rectified; B, six-
code split-band scramble; C, effect of rectifying six-code split-band scramble.
be possible to place a large percentage of the
elements correctly. This can be applied to non-
repeated or repeated code TDS.
A variation of this method, which was sug-
gested but not tried and which should be much
faster, is as follows: Reproduce the recorded
sample through a low-pass filter, say 2,500
cycles. Pass it through a TDS machine and then
through a high-pass filter with the same cutoff.
View the output of the high-pass filter on a
cathode-ray oscilloscope whose sweep is syn-
chronized with the TDS cycle. Transients will
of codes should place most of the elements cor-
rectly.
Another example of the indicator method
is the following: Suppose in a split-band D2
system six known codes are used in an irregular
sequence, and it is desired to determine the
sequence. The following procedure is sug-
gested: Record a sample and reproduce it
through a decoding machine equipped with one
of the proper decodes, and make a spectro-
gram. Certain elements in the spectrogram will
be seen to be normal speech. These elements,
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
56 UNSCRAMBLING AND DECODING METHODS
of course, are the ones to which the particular
code applies. It is much easier to determine
whether a particular element consists of
straight or scrambled speech, than to deter-
mine which particular code was used. Repeat
this procedure with each of the other five
codes. Each element can thereby be identified
with a particular code.
USE OF RECTIFICATION
A variation of this procedure, which should
give more positive results, is as follows: The
NON-LINEAR DEVICE 1
OR DEVICES
LP FILTER
100
BASIC METHOD
SCRAMBLED
INPUT
MODULATOR
Fl
11.11
OSCILLATOR
5 FREQUENCIES
BAND FILTER
550 WIDE
TO INDICATING
DEVICE OR
RELAY
A
MODULATOR
TO
FIG. A
INPUT
11111
F2
1111
OSCILLATOR
10 FREQUENCIES
APPLICATION TO SPLIT BAND DECODING
FIGURE 19. Band-shift detector.
output of the decoding machine used as above
is rectified before making the spectrogram.
Rectifying normal speech does not add inhar-
monic components, whereas rectifying speech
which contains band shifts results in inhar-
monic components. This is illustrated in Figure
18. The upper spectrogram shows rectified
straight speech. This looks perfectly normal
,except that the frequency range is somewhat
more completely covered with harmonics than
is the case in normal speech. The second specto-
gram shows a sequence of split-band scrambles.
The third spectrogram shows a similar sample
rectified, with none of the elements decoded.
Rectifying the undecoded elements results in
a complete smear in the spectrogram compared
to the rectified straight speech. Properly de-
coded elements will stand out more clearly
? against the background of rectified scrambled
speech.
Another variation of the indicator method
consists in subjecting the scrambled speech to
a nonlinear device or devices in such a way as
to obtain difference tones between the compo-
nents. In normal speech, in which all com-
ponents are harmonically related, there will be
no difference tone lower than the pitch of the
voice. In scrambled speech the components are
not harmonically related and there will be dif-
ference tones lower than the pitch of the
44.
voice. The output of a 100-cycle low-pass filter
therefore, can be used to indicate whether a
band of speech is in its proper frequency loca-
tion or not. This is illustrated in Figure 19A. The
importance of this method lies in the fact that
each frequency band can be examined sepa-
rately. It might therefore be used to determine
for each element in a two-dimensional scramble
which frequency band it came from.
Figure 19B shows how each band can be
lifted out of the scramble and placed in each of
the five possible positions either straight or in-
verted. The spectrograph might be used to
speed up the analysis process as illustrated in
Figure 20. The output of the low-pass filter is
fed to the marking amplifier. Whenever the
output of the low-pass filter is zero there will
be no mark produced. Whenever there is an
output a mark will be produced. The procedure
would then be as follows: Set oscillatbr Fl at
one value and then set oscillator F2 succes-
sively at each of its ten values (or five if
inversion is not required). Repeat with oscilla-
tor Fl at each of its other five values. For each
of these 50 settings allow the spectrograph
drum to rotate two or three times with a few
blank rotations between each setting. The
traces on the drum will then look something
like the drawing. The time axis is as usual
disposed lengthwise. If all the traces in a
given time interval are blank it is presumed
that this represents a silent interval. Single
blank intervals in otherwise continuous marks
indicate that these settings were the correct
ones. If none of the marks for a particular
element are blank the indications are that at
that particular moment a consonant occurred
which of course is composed of inharmonic
components. This system was not actually tried
in this complete form but enough work was
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
CRYPTOGRAPHIC TOOLS AND METHODS 57
done to show that it is possible to make use
of the presence of inharmonic components in
some such manner. It appears therefore that
a substantial fraction of the elements in a two-
dimensional scramble might be identified as
to frequency location.
One other possibility of this type might be
mentioned. Variable-area patterns of vowel
sounds have characteristic configurations.
These configurations depend on their harmonic
structure, and a disturbance of this structure
SAMPLE OF SCRAMBLED
SPEECH ON TAPE
FROM TO
2
2
1 4
4;
RESULTING PATTERN
CODE SWITCHING POINTS
I I I I I I I I I I I I 1111
\-INDICATES
SILENT
INTERVAL
INDICATES
TRUE
POSITION
FIGURE 20. Adaptation of spectrograph for de-
coding switched split-band scramble.
should change these patterns in a recognizable
manner. For instance, if the components are
inharmonic there will be no periodicity at the
fundamental pitch rate. It might therefore be
possible to use variable-area patterns, which
can be produced much more rapidly than spec-
trograms, as indicators along the lines of the
above discussion.
APPLICATION TO TABLE 1
In this section we will examine the applica-
tion of cryptographic methods to the specific
scrambling systems listed in Table 1. In this
table the systems which might require crypto-
graphic attack are indicated. The following
paragraph numbers refer to privacy systems in
Table 1.
A4. Among the systems listed under single
modulation the only one that might require
cryptographic treatment is the phase-reversal
system. This system is a special case of the
multiplication system which will be treated
later.
B4, C2,4 Among the double and triple mod-
ulation systems, the irregular continuous
displacements were not handled by noncrypto-
graphic methods. It might be necessary to make
a continuous series of spectrograms to deter-
mine the displacements as a function of time.
This might someday be done continuously and
instantaneously, in which case compensating
frequency changes might be made continuously
by hand to decode the material.
DL D2. Among the band-splitting systems,
the fixed or slowly switched codes can be solved
by inspection. If the code is rapidly switched,
however, single elements seldom contain suffi-
cient information to determine the codes. If
the switching sequence is a repeated sequence,
it may be worthwhile for the sake of quality to
determine the sequence and get in step with it.
In this case the methods described under the
heading "Indicator Methods" should be of
assistance. If the switching sequence is never
FIGURE 21. Repeated code multiplication system.
repeated the indicated noncryptographic meth-
ods appear most reasonable.
F2, F3. TDS systems yield very poorly to
noncryptographic attack. For repeated-code
systems, however, the code can readily be de-
termined by matching either spectrograms or
SECRET
neclassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
1
1111
1111
!
II
1,11
58 UNSCRAMBLING AND DECODING METHODS
variable-area patterns, taking advantage of the
numerical properties of the codes. These meth-
ods are covered in Preliminary Report No. 1415
of Project C-43. Nonrepeated-code systems,
however, have thus far been found exceedingly
difficult to handle, although the methods of
spectrogram matching, matching variable-area
patterns, and indicator methods apply. Efforts
in this direction are described in Preliminary
Report No. 2621 of Project C-43.
F4. Speed variations, according to some pre-
liminary laboratory tests, are rather ineffective
in masking the intelligence of speech unless
the variations are exceedingly wide and rapid.
Technical difficulties then become so great that
this appears to be an unlikely privacy system
by itself. Small variations in speed, however,
might be used to make spectrograms of TDS
systems more difficult to match. In this case,
however, it will be unnecessary to determine
the speed variation program if the TDS
scramble can be removed.
Gl, G2, G3. Combinations of TDS and fre-
quency scrambles are interesting from the
cryptographic standpoint. Since repeated-code
TDS systems were found easy to break, it was
proposed to add various forms of split-band
scrambles. It was argued that the continuously
changing frequency scrambles would alter the
shapes of variable-area patterns so that
they could not be matched. Furthermore the
changing frequency scrambles would make
spectrograms unsuitable for matching, espe-
cially if the split-band codes were switched
nonsynchronously compared with the TDS
boundaries. Each time the frequency code was
switched a new vertical boundary would appear
in the spectrogram, and in combination with
the TDS boundaries the spectrograms would
be very severely broken up in the time scale.
It was found, however, as discussed in Pre-
liminary Report No. 1925 of Project C-43, that
if the TDS code is a repeated code the fre-
quency scrambles can be practically ignored in
matching variable-area patterns. Having found
and removed the TDS code the remaining fre-
quency scramble can be solved by noncrypto-
graphic methods.
In the case of nonrepeated TDS, however,
the addition of split-band coding would increase
the difficulty considerably, provided that the
two coding systems do not provide clues to each
other. The most promising method for handling
this system appears to be to determine the split-
band codes first by the indicator methods
previously discussed. If the split-band codes are
then removed the remaining scramble can be
handled as straight TDS. Another possible
method is to make variable-area patterns with
all,Athe decodes superposed. The resulting pat-
terns, however, will not be as satisfactory for
matching as patterns of straight speech.
G4. The two-dimensional scramble can be
handled by matching spectrograms if a re-
peated code is used. Experiments along these
lines are described in Preliminary Report No.
2220 of Project C-43. If the code is nonrepeated,
however, it would be exceedingly difficult and
time consuming to handle by unaided match-
ing. It would help considerably if the original
frequency location of each element in the
scramble could be determined. This might be
accomplished by the methods described under
the heading "Indicator Methods."
Hl. Determining the code for multiplication
or phase-reversal systems can be accomplished
quite readily if the code is repeated at suffi-
ciently short intervals. In the one system which
was met in Project C-43 (Preliminary Report
No. 1827) the code wave was repeated 100 times
per second. In this case the scrambled signal
could be applied to the vertical plates of an
oscilloscope with the horizontal sweep syn-
chronized with the code cycle. It is obvious that
every time the coding wave passes through
zero the scrambled signal also passes through
zero regardless of the value of speech signal
at the moment. If several cycles of scrambled
speech material are superposed, therefore, they
have the appearance shown in the photograph,
Figure 21. The superposed traces show a defi-
nite pattern, with regions of high and low
amplitude, and also sharp indentations. These
latter are the crossover points of the code wave.
There is also a marked tendency for the peaks
to occur alternately above and below the center
line, but the amplitudes of the peaks are not all
alike. Since the speech amplitudes tend to aver-
age out over a number of cycles, the amplitudes
of the superposed peaks reflect pretty accu-
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
CRYPTOGRAPHIC TOOLS AND METHODS 59
rately the amplitudes of the coding wave at
those points. The probable shape of the coding
wave based on this evidence, has been partly
traced ins
It has been found experimentally that if
only the crossovers of the coding wave are
reproduced the speech will be intelligibly de-
coded: The decoding wave need not be the
reciprocal of the coding wave. It can be like
the one drawn in at the right in the photograph.
It is only necessary, therefore, to generate a
wave having its crossovers at the indicated
points, and reverse the phase of the scrambled
signal of these points.
112, 113. Level modulations by themselves are
not private, but they might very well be used in
combination with other systems in an attempt
to foil the matching of speech patterns. The
level modulations themselves, however, need
not be solved cryptographically.
J1. There appears to be no method either
cryptographic or noncryptographic for break-
ing the noise-masking method if the noise is
predistorted, random, and sufficiently high in
level to really mask the speech. These require-
ments, however, make the technical difficulties
for system operation very great and it is un-
likely that this method can be used over radio
channels. Cracking this system therefore be-
comes a matter of solving the noise-distorting
system. Project C-43 had no experience along
these lines.
K1, K2, K3, K4. Scrambled Vocoder channels
can theoretically be solved by matching oscillo-
grams. Actually as mentioned under the head-
ing "Matching Oscillograms" this procedure is
very difficult because the channels look so much
alike.
L1, L2, L3. Channel-mixing systems would
be exceedingly difficult to handle crypto-
graphically if a sufficient number of channels
were involved so that noncryptographic meth-
ods were inapplicable. The only possible method
of attack appears to be matching spectrograms.
Since, however, about 25 per cent of normal
speech consists of pauses, many of the switch
points will occur in these pauses and it will
therefore be difficult to establish continuity
by matching.
DETERMINATION OF THE MESSAGE
The objective of decoding work is usually not
to determine the codes used, but to learn the
intelligence which was transmitted under these
codes. In the case of repeated-code systems,
the procedure for obtaining intelligence is
obvious once the code has been determined
by the methods outlined above. It is only neces-
sary to set this code into a machine similar
to that ttsed at the receiving end of the system
being monitored, and listen directly to the
transmitted speech. If the material has been
recorded while the code was being determined,
the recorded material can in general be de-
coded in the same way.
In the case of nonrepeated-code systems, the
determination of the code sequence leaves us in
general a long way from the determination of
the message. All the material must first be
recorded in scrambled form. It is necessary
during this process to establish time reference
points in the scramble, perhaps by superposing
clicks or spurts of tone during the recording
process, and referring the code sequences to
these points. A decoding machine must be
available, such as the one described in Pre-
liminary Report No. 1528 of Project C-43, which
is adaptable to a variety of coding systems.
The code sequence must be set into this ma-
chine, perhaps in the form of a punched tape.
The scrambled material must then be repro-
duced and fed into the machine, maintaining
proper synchronism between the reproducing
and decoding systems. This is a formidable job.
There are some alternative possibilities
which may apply in special cases. In the case
of nonrepeated-code TDS, for instance, the
process of matching variable-area patterns has
actually restored the speech in reproducible
form. Variable-area patterns can be played
back just like the sound tracks used with
motion pictures.
USE OF PLAYBACK
A playback machine of this type is described
in Preliminary Report No. 1224 of Project C-43.
The rearranged elements are mounted on a
strip of adhesive, and scanned with a light slit
and photocell. Considerable noise is caused by
SECRET
Declassified and Approved For Release 2013/12/11: CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
10;
!1'
1
UNSCRAMBLING AND DECODING METHODS
RECORDER UNIT
AMPLIFIER ANALYZER UNIT
' 6
RECTIFIER UNIT
FIGURE 22. Sound spectrograph (D-165529) on a "push-around."
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
THE SOUND SPECTROGRAPH 61
the joints between the separate elements, but
this could be largely eliminated by a specially
designed squelch circuit, perhaps controlled by
a separate light beam and photocell to cut off
the output wherever a joint is passing under
the scanning beam. The first attempt to use
this decoding method was unsuccessful, as dis-
cussed in Preliminary Report No. 2621 of Proj-
ect 0-43. However, there is nothing basically
wrong with the method; it simply needs better
execution than it received in the first attempt.
If the solution of the coding system requires
spectrograms rather than variable-area pat-
terns, it is still theoretically possible to play
back the rearranged pieces. A playback ma-
chine for spectrograms is described in Pre-
liminary Report No. 1729 of Project 0-43. This
first model requires a negative transparency
of the spectrograms, to be scanned by a light
slit and photocell, with a multifrequency light
chopper interposed ahead of the photocell.
Again the method is basically sound. The ex-
perimental machine described in the report
needs considerable improvement before it will
yield adequate quality for the purpose described
above, in order to overcome the degradation
of quality caused by the joints, by slight mis-
placements of the elements, by spillover at
the boundaries, etc. Furthermore, to get good
patterns for matching, the signal must be
subjected to a very high degree of compression,
which distorts both the time and the frequency
distribution of energy. It may be necessary
to make one kind of pattern for matching,
and another kind for playback, as was done
with the variable-area patterns described in
Preliminary Report No. 2621 of Project C-43.
As a final alternative, it is possible to learn
to read speech spectrograms by visual inspec-
tion. Theoretically, therefore, the rearranged
spectrograms might yield the message directly.
Here again, however, the boundary distortion
will increase the difficulty of reading the pat-
terns. It has also been found that the best
patterns for matching are not the best for
reading, and it' may be necessary to make two
sets of patterns. However, since spectrograms
have been continually improving, the possibility
of visually determining the intelligence from
rearranged spectrograms must be listed as a
distinct possibility, and one which, if it is
feasible, is the most general of all methods
since the basic procedure is the same for all
the scrambling methods which can be handled
in this manner.
4.5 THE SOUND SPECTROGRAPH
In the material to follow describing the sound
spectrograph developed in Project 0-32 and
used continuously in subsequent decoding and
evaluating projects, a brief and general
description will be followed by a more detailed
analysis of this important visual aid to the
study of privacy systems. The sound spectro-
graph analyzes speech (or other sounds) in
terms of its three basic dimensions of time,
frequency, and amplitude. Such analyses,
shown visually on a graph or chart, are helpful
in understanding the complexities of sound and
what various scrambling methods do to speech
to make it unintelligible.
In March 1941 an early laboratory model of
the sound spectrograph was demonstrated as
an instrument that with further development
might be useful in studies of telephone privacy.
It was appreciated at that time that the need
might arise for intercepting communications
in scrambled speech and decoding them. It was
also appreciated that new scrambling systems
might be encountered and that means would
be needed for diagnosing such systems. For
such a purpose the unaided ear has very limited
capabilities. Such things as oscillograms, which
show the wave form, provide few clues as to
the mechanism by which the wave form was
changed. Project C-32, the forerunner of Proj-
ect 0-43, was organized in the fall of 1941 to
product a sound spectrograph useful for
diagnosing and decoding speech scrambling
systems.
About a month before the attack on Pearl
Harbor, patterns that could be used for de-
coding work were being produced with a bread-
board model, and the first finished model of the
spectrograph was available by the end of that
year. Additional models of the spectrograph
were built for the Armed Services, incorporat-
ing improvements in operation and in rugged-
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
62 UNSCRAMBLING AND DECODING METHODS
ness. The model, described in the final report
of the Project C-43 and shown in Figure 22,
has been used in studies of various privacy
systems submitted by the Army, Navy, and
NDRC for the purpose of evaluating the degree
of security which they afforded. Improvements
in the form of a calibrating circuit built into
the spectrograph and control circuits added in
the form of an appliqu?nit, were made as the
work progressed.
4'5.1 How the Sound Spectrograph Works
A schematic diagram of the sound spectro-
graph is shown in Figure 23. The signal to be
analyzed is recorded on a loop of magnetic tape
at a speed of 25 rpm permitting a sample 2.4
REPRODUCING
AND RECORDING
AMPLIFIER
MECHANICALLY
COUPLED
VARIABLE
CARRIER
OSCILLATO
12-23 KG
MODULATOR
FREQUENCY
RECORDING
DRUM
BAND
MANNING) -
FILTER AT 12 KG
MARKING
AMPLIFIER
FIGURE 23. Schematic diagram of sound spectro-
graph.
sec long to be recorded. The recorded material
is then reproduced at 78 rpm. Because of this
speedup, the original signals which may have
filled the frequency region between zero and
3.5 kc now extend to about 11 kc. The signal
is modulated with a carrier which gradually
changes in frequency from 23 to 12 kc as the
recorded material is reproduced repeatedly.
The lower sideband of the resulting signal is
passed through a band-pass filter with a center
frequency of about 12 kc.
The output of the filter is amplified and fed
to a stylus bearing on facsimile paper, making
a trace varying in density with the instan-
taneous energy passed by the filter. The paper
is mounted on a drum which is geared to the
turntable rotating the magnetic tape. As the
frequency of the modulating carrier changes,
the stylus moves along the drum laterally. The
resulting spectrogram is built up line by line.
In this manner a pattern is produced which
shows by its light and dark areas how the
intensity in the signal varies as a function of
time and frequency.
The change in frequency produced by modu-
lating the voice signal with a varying carrier
signal of suitable high frequency would not be
necessary if it were possible to make a band-
pass filter whose center frequency could be
shifted easily. In this case the actual voice fre-
que2cies could be scanned by the filter to de-
termine the characteristics in frequency with
time. It is easier, however, to accomplish the
same object by the method actually used in the
spectrograph.
4.5.2
Operation
It is the fact that both time and frequency
variations are simultaneously displayed which
makes spectrograms so valuable for decoding
work.
Scanning filters of various widths can be used
for different purposes. If the filter is wide, it
will give an analysis which is limited in the
amount of detail it can portray in the frequency
dimensions, but it will respond quickly to
changes in amplitude with time, and will there-
fore give sharp time resolution. The narrower
the filter the more frequency detail is shown in
the spectrograms, at the sacrifice, however, of
some of the time resolution. With all the filters
thus far used, the shift in frequency range
from line to line is only a fraction of the width
of the filter. Successive lines in the spectro-
gram, therefore, do not represent successive
frequency bands. They represent frequency
ranges which overlap by a large fraction of
their total width. The density of the patterns,
therefore, changes very gradually along the
frequency dimension.
The kind of patterns produced by this method
of analysis is illustrated in Figure 24. The
upper spectrogram in the figure was made with
a scanning filter about 300 cycles in width. The
separate words can be plainly distinguished.
The vowels are distinguished by dark bands
with vertical striations. The consonants are in
general less intense and show a different type
of structure. It will be noted that the dark
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
THE SOUND SPECTROGRAPH
bands are different in the different vowels, and
they change not only from one word to the next
but also within each word. Analyses of this
type, therefore, graphically portray changes in
the energy frequency distribution of a complex
signal with both time and frequency. It should
be emphasized, however, that the relative in-
tensities of the various components of this
particular sample of speech, notably the con-
sonants, differ to a far greater extent than
would be judged by the relative blackness of
their patterns. In other words, a very large
discrete harmonics causes the vertical stria-
tions in the patterns made with the wider filter.
Whenever the filter is wide enough to pass
several harmonics at once, these harmonics
beat with each other and form maxima and
minima in the output of the filter. The fre-
quency of the beats corresponds exactly to the
frequency of the voice pitch.
It will be noted in the 45-cycle spectrogram
that he harmonics rise and fall in frequency
from moment to moment. This reflects the
changing pitch of the voice known as inflection.
FIGURE 24. Spectrograms of normal speech, words being "one, two, three, four, five, six." In this case,
spectrograms are somewhat smaller than normal size due to photographic reduction and some trimming
at ends.
amount of level compression is incorporated
in these patterns.
The lower spectrogram in the figure shows
the same words analyzed with a filter only 45
cycles wide. This filter is narrow enough to
resolve the individual harmonics of which
vowel sounds are composed. The harmonics
consist of the fundamental voice pitch together
with both odd and even multiples of this fre-
quency. Some of the harmonics are stronger
than the others, because they are reinforced
by resonance in the oral cavities as the words
are formed. It will be noted that the dark
areas in these patterns correspond in frequency
and in trend with those in the upper spectro-
gram. The fact that vowel sounds consist of
Inflection is normally used in connected speech;
and this fact is of assistance in decoding work,
because the spacing and trend of the individual
harmonics in spectrograms provide important
clues in diagnosing privacy systems.
4.5.3 Level Compression
In normal speech there is a tremendous
change in level from moment to moment par-
ticularly in the level of consonants as compared
to vowels. There is also a considerable differ-
ence in the average level at low frequencies
as compared to high frequencies. This latter
difference can be corrected by predistortion,
SECRET
Declassified and Approved For Release 2013/12/11: CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
UNSCRAMBLING AND DECODING METHODS
and present models of the spectrograph contain
shaping networks for this purpose. There are,
however, changes from moment to moment in
the relative levels of high and low frequencies
in different speech sounds which cannot be
corrected by shaping networks. The facsimile
paper on which spectrograms are made has a
range of between 10 and 15 db. The range of
levels in speech greatly exceeds this value. This
means that if the average level is adjusted
so that the highest components appear at maxi-
mum blackness, the lowest level components
will be invisible. Conversely if the level is so
adjusted that the low-level components appear
in the pattern, the high-level components will
severely overload the recording paper. To show
both the high- and low-level components occur-
ring in speech, therefore, it is necessary to corn-
pressl the instantaneous signal into a much
narrower range.
In the earliest models of the spectrograph the
marking amplifier shown in Figure 23 was
given a compressing action by means of a
thyrite varistor across the grid of the output
stage. Whenever the output of the scanning
filter was low the gain of the amplifier was
effectively raised from an average condition
and whenever the output was high the gain
was effectively lowered. This tended to equalize
changes in level with both frequency and time.
The compressor was replaced by devices which
can exercise certain types of discrimination in
controlling the instantaneous gain of the mark-
ing amplifier. These devices are known as con-
trol circuits. They provide patterns with better
resolution in both time and frequency than
can be obtained with the compressor. The pat-
terns shown in Figure 24 were made with these
control circuits in operation. The circuits are
described in Preliminary Report No. 27" of
Project C-43.
4.5.4
Possible Improvements
The spectrograph patterns underwent con-
tinual improvement in the course of this work,
but probably they can be still further improved.
The control circuits thus far produced are by
no means the final word. Circuits of this type
can be adapted to affect the patterns in various
ways, and it is conceivable that different con-
trol circuits could be developed for decoding
different types of scrambles.
One definite line of improvement concerns
the time resolution. Many scrambling methods
produce sharp discontinuities of the scrambled
speech in the time dimension. The process of
analyzing the scrambled signal in such a way
as4 to obtain high frequency resolution tends
to obscure the signal at these sharp boundaries.
This is a basic situation which affects not only
the spectrograph, but also all types of ana-
lyzers. To obtain a high degree of frequency
resolution, a narrow filter must be used. The
narrower the filter, however, the longer its
response and decay time, that is, the output of
the filter cannot be made to change as rapidly
in level as the instantaneous level of the signal.
This causes strong components to spill over
across the time boundaries. In general this
spillover does not interfere greatly with the
recognition of various privacy systems, but it
does interfere severely where spectrograms are
to be used for decoding work. Several possible
remedies for this situation have been devised
and are recorded in Chapter VI of Part I of the
final report of Project C-43.
4.5.5 Amplitude Representation
In the patterns thus far discussed the in-
stantaneous intensity of the signal is repre-
sented by the lightness or darkness of the trace
in the spectrograms. This representation is
inherently nonlinear and practically impossible
to make quantitative. For some types of work
it would be highly desirable if the amplitudes
could be represented in such a way that they
could be interpreted quantitatively.
Figure 25 shows a spectrogram which upon
close inspection will be seen to be made up of
discrete dots. The dots are close together in
the dark portions of the spectrogram and
farther apart in the light portions. The dots
themselves are all of equal blackness. The
spacing of the dots is in fact quantitatively
related to the instantaneous level of the signal.
The level at any point in the spectrogram can,
SECRET
Declassified and Approved For Release 2013/12/11: CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
;
THE SOUND SPECTROGRAPH
65
therefore, be measured by measuring the dot
spacing with suitable equipment and comparing
it with a scale showing dot spacing vs level.
Another type of representation is shown in
Figure 26. Here the levels are represented by
the type of technique used in representing
topographical variations in contour maps. The
contour lines each represent regions in which
the signal reaches a particular fixed level. The
lines may be spaced so as to represent steps
of any desired number of decibels, or any
number of volts. In the lower spectrogram the
features designed for the specific purpose in
mind.
4.5.6 Spectrograph Details
As described in general terms above, the
output of the scanning filter of the sound
spectrograph is recorded continuously on fac-
siwile paper wrapped around a drum rotating
with the magnetic tape so that one revolution
of the tape corresponds to one rotation of the
FIGURE 25. Method of representing amplitudes in such a way that they can be interpreted quantitatively
by use of discrete dots all equally black. Dots are closely spaced in dark regions and widely spaced in
light regions. There is definite quantitative relation between dot spacings at any point and level of
signal at that point. Level could, therefore, be determined by measuring dot spacing.
spaces between the contour lines have been
filled in with various densities of dot spacing.
This permits instant recognition of equality
of level in different portions of the signal.
Quantitative amplitude representation may
or may not prove useful in decoding work. For
certain kinds of signal it should prove useful,
because it provides another dimension besides
time and frequency which can be used for
determining continuity or discontinuity in the
signal. In other cases, however, it may prove
useless, because changes in level have arbi-
trarily been introduced into the scramble.
The developments mentioned above empha-
size the fact that the sound spectrograph is a
highly flexible device and its capabilities along
any line can be greatly increased by adding
drum. The drum is moved laterally by a lead
screw as the modulating frequency is slowly
changed.
The analyzer circuit comprises a variable-
frequency oscillator, a balanced modulator in
which the output of the oscillator is mixed with
the voice frequencies, and the scanning filter.
The balanced modulator automatically prevents
any signals from getting into its output circuit
except the resultant sidebands produced by the
modulating or mixing process. Thus neither
the original voice frequencies nor the oscillator
frequencies appear in the output. Since the
band-pass scanning filter has a mid-band fre-
quency of about 12 kc it automatically selects
or passes the lower sideband produced by the
modulating process. The sidebands appearing
SECRET
Declassified and Approved For Release 2013/12/11: CIA-RDP10T02279R000100180001-2
11
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
66
UNSCRAMBLING AND DECODING METHODS
in the output of the modulator have energy-
frequency distributions identical to the energy
of the modulating signal and occupy a position
in the frequency scale corresponding to the
carrier frequency. A change in the carrier or
modulated signal of, for example, ?200 cycles,
will cause the two modulator output sidebands
to shift to a position in the frequency scale
200 cycles lower.
The scanning filter has a mid-band frequency
subsequent extensions of this project, the spec-
trograms are slightly over 12 in. long normally
representing a recording of 2.4 sec making the
time scale approximately 200 msec to the inch.
The records are normally 2 in. wide covering
3.5 kc making the frequency scale about 1/16
in. per kilocycle. The frequency dimension can
be expanded if desired but this requires that
a longer time be available for making the
recyd.
FIGURE 26. Amplitude representation by contours, every point on any one line representing equal signal
level, and successive lines starting from blank background representing successively higher levels. In
upper spectrogram, it is not immediately apparent which regions are peaks and which are valleys. In
lower spectrogram, areas between successive lines have been filled in with patterns made up of discrete
dots. Closer dots indicate higher level, with regions of equal level in different parts having same dot
spacing.
such that when the carrier has one extreme
value, only the lowest frequency components of
the lower sideband will fall within the pass-
band of the filter. When the carrier has its
other extreme value, only the highest frequency
components of the same sideband will pass
through the filter. As the carrier frequency
is slowly changed, all frequencies of one side-
band (in this case the lower sideband) will be
scanned by the filter and the relative strengths
of the signals from moment to moment will be
impressed upon the recording facsimile paper.
The record is made on what is known as
Teledeltos paper whose light-colored surface is
blackened by the passage of an electric current
from the metal stylus through the paper to
the metal drum on which it is wrapped.
In the model emplqyed in Project C-43 and
4.5.7 Applications of the Spectrograph
Although the spectrograph was developed for
use in analyzing speech privacy systems and in
decoding scrambled speech records, it has more
general application. The spectrograms shown
here are characteristic of what the instrument
can do in analysis and illustrate the results
obtained with different filters, different scan-
ning rates, and different types of material. The
illustrations (taken from the October 1, 1943
report41 on Project C-43) are for the most part
familiar sounds selected to permit a mental
comparison of the sound and its time-frequency
pattern. These illustrations are followed by ex-
amples of the effects of various scrambles upon
speech.
Perhaps the most familiar example of a wave
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
THE SOUND SPECTROGRAPH
which is complex in both the frequency and
time dimensions is . speech, which therefore
provides excellent material for illustrating
what the spectrograph can do. Figure 27 shows
a spectrogram of the sentence, "We shall win
or we shall die," spoken in a normal manner by
a male voice and scanned with the narrow
filter. The time and frequency axes are indi-
67
lines, they are the separate harmonics of the
voice pitch, flowing up and down as the voice
is inflected. The unstriated sounds are unvoiced,
such as the "sh" in "shall."
Figure 28 shows spectrograms of the same
sentence made with three different filters. In
the one made with the widest filter, the separate
harmonics are no longer visible, but the areas
FIGURE 27. Spectrogram of "We shall win or we shall die," with "gray scale" below, showing intensity
variation relation.
cated by appropriate scales. The intensity
variations are indicated by variations in shade
as shown in the scale below the spectrogram.
It should be emphasized that there are about
400 horizontal scanning lines in this 4-in.
spectrogram; too close together to be seen
individually. The horizontal striations which do
appear in the spectrogram are not scanning
of resonance are brought out more clearly.
Figure 29 shows a comparison between normal
speech, a monotone, and a whisper, all by the
same voice. In the monotone the harmonics
are straight horizontal lines, and in the whisper
the harmonics disappear almost entirely, but
the dense regions by which the different sounds
are recognized still persist.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
II
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
68
UNSCRAMBLING AND DECODING METHODS
Figures 30 and 31 show some musical effects.
The soprano solo shows how the vibrato affects
both pitch and iptensity. The harmonics, it will
be noted, are much farther apart than in the
male voice. In the piano music, the notes show
tapering traces, as would be expected from
their nature. The other illustrations require
no special comment, except to note that the
kc. The lowest spectrogram shows an accelera-
tion measurement. It was desired to find how
long it took a phonograph record to come to
full speed after being released, the turntable
running at full speed all the time. This was
accomplished by simply recording an 8-kc tone
on a sample record, and capturing the output
of the reproducer during acceleration. The time
FIGURE 28. Same sentence as shown in Figure 27 recorded through filters of different bandwidths. A, with
45-cycle filter; B, with 90-cycle filter; C, with 300-cycle filter.
telephone bell was analyzed with the wide filter
to show the time pattern more clearly. It may
be seen that the clapper hits two bells alter-
nately, but somewhat irregularly.
Figure 32 shows some applications in which
the spectrograph provides a convenient method
of obtaining a graph of frequency versus time.
The upper spectrogram shows the output of a
particular slowly warbling oscillator, and the
second shows a more rapid sawtooth sweep
frequency, the latter spectrogram covering 11
consumed in acceleration, 130 msec in this case,
is directly indicated.
Figure 33 shows some more 11-kc spectro-
grams, illustrating the fact that some common
sounds cover a very wide frequency range.
Figures 34 and 35 return to the 3.5-kc range
to show longer samples. They require no special
comment.
pontrasting with these illustrations, Figure
36 shows an example of a steady wave, namely
thermal noise. Two different levels are shown,
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
IT
3
THE SOUND SPECTROGRAPH 69
one of them analyzed with both the "narrow"
and the "wide" filter. It will be noted that some
of the sounds in the previously discussed spec-
trograms have components which are of the
same nature as thermal noise, while some have
components of definite frequencies. The latter
can be recognized by their solid texture in the
spectrograms as compared to the characteristic
well as complex frequency structure because
the ability to show both at once is the unique
feature of the spectrograph.
SPEECH PRIVACY PROBLEMS
Even in a steady flow of speech, the distri-
bution of energy over the frequency range is
constantly changing. Voiced sounds have a
FIGURE 29. Spectrograms showing comparison of: A, normal speech; B, monotone; C, whisper.
texture of a random wave form exhibited by
thermal noise. The "sh" in "shall," and the
roar of the flame, are examples of random
sounds, while components of definite frequency
may be seen in the jingling of keys (Figure 33)
and the filing of metal.
These illustrations, which were chosen for
variety and for interesting features, show what
the spectrograph will do. Most of them, as
emphasized before, show time variations as
definite structure, consisting of a series of har-
monics of the fundamental voice pitch, the
harmonics being stronger in some frequency
regions than in others. Unvoiced sounds have
no such definite structure, but show a "smear"
of energy which may or may not be concen-
trated in definite frequency regions. Words and
sounds are recognized by their energy pattern
in both time and frequency. Different speakers
uttering the same sentence will produce pat-
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
70 UNSCRAMBLING AND DECODING METHODS
terns which show, a distinct general resem-
blance, but also marked differences. The speech
pattern also may be considerably distorted by
artificial means before the ear fails to recognize
the speech, provided the distortion is not too
discontinuous in frequency or time. Since
privacy systems depend for their effectiveness
on distorting the speech pattern beyond the
possibility of recognition by the ear, it seems
? (I
?:1 401.4.1,0 '
?
speech than the low=frequency regions. In these
patterns, as before, the horizontal scale is time
(about 1.8 sec is represented in each example),
the vertical scale is frequency (the upper limit
is 3 kc), and the density or blackness represents
the intensity of the energy in a given regiion.
It should be noted that the resolution of the
process is sufficient to separate each harmonic
in the voiced sounds. This is important, as will
4
C..4,1 ?
jr**.mono.y..
?
?
izSitr3.04.
FIGURE 30. Musical effects. A, soprano voice unaccompanied; B, piano music ;r C, orchestra music.
reasonable that this distortion would also be
visible to the eye if the scrambled speech pat-
tern could be reduced to suitable graphic form.
Examples of such patterns are shown in the
attached photographs taken from the final
report" on Project C-32. Figure 37 shows some
normal speech, undistorted except that a slop-
ing network was introduced in the electrical
circuit to bring out the high-frequency struc-
ture, since this is always weaker in normal
be seen later, because normally the voice funda-
mental is constantly changing, that is, the voice
is inflected, and since the harmonics are mul-
tiples of the fundamental, the higher harmonics
show progressively more change than the
fundamental. For instance, if the fundamental
goes from 100 to 200 cycles the tenth harmonic
goes from 1,000 to 2,000 cycles, a difference of
1,000 cycles as compared to 100 cycles for the
fundamental. The traces of harmonics in the
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
THE SOUND SPECTROGRAPH 0 71
visual speech patterns will, therefore, have
greater slopes at the high end of the pattern
than at the low end. This may be seen quite
clearly in the examples of Figure 37.
Figure 38 shows patterns of some vowel
sounds. In making these patterns, an attempt
was made to enunciate clearly, and also to keep
the pitch constant (monotone), so -,as to show
1
7777?--
?
showing transitions from one vowel to another.
These pairs were chosen because their time
characteristics are direct opposites.
SCRAMBLED SPEECH PATTERNS
Figure 40 shows the output of a privacy sys-
tem -which depends on simple inversion. In the
inverted speech, the slopes of the harmonic
?I
qoe.?,:_?0-., ,,..-tty':,31.'._ .?:4.*I-4!;:rA, ...,,,,r--,,--;
? . ' -45 . 4 ' ??? = 4 .
A.
FIGURE 31. Sound effects. A, whistling with warble or rising note; B, police whistle; C, telephone bell.
the difference in energy distribution for these
sounds. These are of interest because of the
possibility that the visual patterns themselves
might give clues as to the words they contain.
Figure 38 also shows the effect on the frequency
resolution of widening the band-pass scanning
filter. The wide filter gives much better reso-
lution in time, however, as will appear subse-
quently.
Figure 39 shows some diphthong patterns,
traces become greater towards the bottom of
the pattern, which is the direct opposite of
normal speech, and is therefore a definite sign
of inversion. The pattern also thins out at the
bottom, but this could be altered by a suitable
distorting network. No network, however, can
change the slopes of the harmonic traces. Inci-
dentally, the carrier' "leak" shows up in the
pattern, giving a direct indication of the fre-
quency about which the inversion was per-
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
72 UNSCRAMBLING AND DECODING METHODS
FIGURE 32. Spectrograph applications in which frequency versus time curve is presented. A, output of
warbling oscillator; B, sawtooth generator output; C, acceleration of phonograph record on moving
turntable.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
THE SOUND SPECTROGRAPH
73
formed. If the carrier is completely suppressed,
however, its location may be determined by
trial.
A more complicated privacy system is the
split-band system, used in transatlantic radio-
telephony, in which the frequency range is
divided by filters into several bands, which are
then arranged in a different order, and some
are inverted. Figure 41 shows patterns of the
Looking at the inflected portions, it is quite
easy to find one which is either definitely in-
verted or definitely erect. The other bands can
then be immediately labeled inverted or erect
depending on whether they have the same
direction of curvature as the band previoUsly
identified. Now, the relative slopes of the har-
monic traces in the different bands indicate
their original position in the frequency scale;
4
A
. 44%,
. .
itft
0 , ' .?
=.F`'0 6 f
Si..
FIGURE 33. Spectrographs showing that sounds exist in high-frequency region. These are 11-kc, 1,000-
cycle filter records. A, sound of crumpling paper; B, sound of tearing cloth; C, jingling keys.
output of such a system with two different
codes. Both of these samples contain portions
in which the voice was quite markedly inflected.
The fact that the frequency range has been
divided into five bands is quite apparent from
discontinuities in the energy distribution, and
also from discontinuities in the harmonic
traces. It is also quite apparent that some of
the bands have been inverted because the voice
cannot be inflected both up and down at once.
the band showing the least slope (or curvature)
must originally have been the lowest band, and
the band with the greatest slope (or curvature)
must have been the top band.
The scrambling methods thus far illustrated
alter the frequency characteristics of the speech
patterns. The TDS privacy system, discussed
in detail in Chapter 2, operates on the time
characteristics of speech, dividing successive
sections of speech, each m seconds long, into n
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
74 UNSCRAMBLING AND DECODING METHODS
short time elements which are sent in a
scrambled time sequence. Each syllable is cut
up and received as short bursts of energy in
the wrong order. The number of scrambled
orders available increases very rapidly with n.
Systems have been developed in which m is as
short as 0.6 sec, and n is 20, making each
element 30 msec. A pulse of tone is sent every
known. Means must then be found for deter-
mining the code, and this decoding process
must be repeated every time the code is
changed. If the code is changed often enough,
the decoding will lag far behind the message.
It is essential, therefore, that every artifice be
employed to increase the speed of decoding.
Figure 42 shows some speech patterns scram-.
FIGURE 34. Common sounds recorded in 3.5-kc range, 300-cycle filter. A, striking match and flame; B, filing
on metal; C, machinery noise; inset is same with 45-cycle filter.
n-t seconds to keep the transmitter and receiver
in synchronism. In one such system over 60,000
codes are available, and they may be changed
quite readily.
DECODING THE TDS SPEECH
Obviously, if it is desired to decode such a
privacy system, it is necessary first to evaluate
m, and n. Presumably a machine could be built
to unscramble the speech if the code were
bled by a TDS system. It is apparent that the
speech has been chopped up on the time scale
rather than on the frequency scale. It is easy
to determine n by the length of the elements,
and since the synchronizing pulse shows in the
pattern, the most natural assumption is that
m is given by the distance between these pulses
or some multiple of it. In the illustration of
Figure 42, duplicate patterns were made, each
element was numbered, and one of the
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
THE SOUND SPECTROGRAPH
75
scrambled patterns was then cut up and re-
assembled, giving the code. It should be noted
that in the scrambled patterns a few elements
within each code cycle immediately stand out
as probably belonging together, particularly
when voice inflection occurs. Usually the other
elements in a section cannot be positively
matched. It is of tremendous help, therefore, if
identical patterns are mounted on movable
slides, and viewed through a system of mirrors
which superposes the two, but all the upper
pattern to the right of a definite line is blocked
out, and all the lower pattern to the left of this
line is blocked out, so that effectively any two
elements may be juxtaposed to see whether
they look as though they were originally con-
FIGURE 35. Additional 3.5-kc sounds. A, splashing stream of water; B, air bubbles blown through water;
C, riffling stack of file cards.
the scrambling order is repeated over and over.
A doubtful match can be checked in another
section, and matches which are impossible to
spot in one section can be readily spotted in
another.
Rather than cut the pattern up as in the
illustration, an optical system has been built
for viewing two duplicate patterns simultane-
ously. This is shown in Figure 43. The two
secutive. If a match is discovered in one section,
the viewer may be shifted without moving the
slides, for an immediate check in other sections.
Instead of dividing the scrambled pattern with
lines and numbering them as in Figure 42,
suitable scales and numbers for a particular
TDS system can be incorporated in the slides.
Speech patterns have been shown to be useful
in decoding all systems of speech privacy known
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
76 UNSCRAMBLING AND DECODING METHODS
to be in use. This method, however, which is
quite general, may not necessarily be the speedi-
est in all cases.
SPECIAL METHODS FOR TDS
The TDS system appears to be the most diffi-
cult to decode of all the speech privacy systems
known to have been reduced to practice, par-
time characteristics have been scrambled, the
wave form itself provides evidence which can
be visuall interpreted. Various methods have
been tried, the first being the ordinary oscillo-
graph which discloses that there are discon-
tinuities in time more sudden and frequent than
occur in normal speech. It is possible to cut up
such traces and piece them together. Examples
FIGURE 36. Thermal noise as example of steady sound. Two upper records made with wide filter, two
levels, 12-db apart. Lower spectrograph made with narrow filter. These are 11-kc records.
ticularly if a large number of codes are avail-
able and if they are changed often. A great deal
of emphasis naturally was placed on TDS in
Division 13 work. The following sections dis-
cuss methods particularly applicable to TDS,
investigated in parallel with the speech pattern
development.
Wave Form Traces. Speech patterns of the
type thus far discussed are particularly de-
signed to display the frequency composition of
speech, so that distortion of the frequency scale
could be recognized visually. Where only the
of such records are given in the final report" of
Project C-32.
It was thought that a variable-area sound
track would provide more distinctive patterns
than oscillographic traces, and would have the
additional virtue that they could be played
back, and could, therefore, serve perhaps as the
primary record of the intercepted message.
This scheme has proved to be useful. Examples
of records made in this manner will be found
in the final report on Project C-32 and in sev-
eral of the preliminary reports of Project C-43.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
THE SOUND SPECTROGRAPH
77
? Partial Matching. The above methods would
' not serve if the TDS code were changed very
often, in the extreme case if it were changed
every cycle. One branch of the investigation
has, therefore, attacked TDS from a statistical
national Business Machine punched cards,
thereby enormously reducing the number of
codes remaining possible. The most complex
TDS system under consideration in Project
C-32 has twenty elements per code cycle with
ft
FIGURE 37. Normal undistorted speech with sloping network to bring out high-frequency structure.
angle. In a system with a sufficient number of
elements the total number of available codes
is very large. If, however, in a given code cycle
a few elements can be visually matched, the
others being inconclusive, it appears possible
to tabulate in advance all the codes which will
satisfy the observed matches, perhaps on Inter-
over 60,000 good codes available. This may be
reduced to only eight possible codes by match-
ing two groups of three elements. These eight
codes might conceivably be tested successively
by automatic means, the correct one being rec-
ognized by ear. Presumably, the message could
thus be decoded cycle by cycle. A cycle contain-
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
78
UNSCRAMBLING AND DECODING METHODS
ing insufficient material for visual matching
may contain no indispensable portion of the
Message either.
FURTHER POSSIBILITIES
Assuming that the methods outlined apply to
all privacy systems which it is desired to crack,
further work would be directed toward speed-
ing up the processes. Improvements in speed
can, it appears, be made in all of the processes
Desirable sections could be stopped at any time.
This appears, however, to require considerable
equipment as well as considerable development,
and would be undertaken only if it appeared
quite certain that speech patterns afforded the
best means of keeping up with a rapidly chang-
ing code.
Large Variable-Area Patterns. The variable-
area pattern method discussed above did not
revire photographic film, but the patterns had
FIGURE 38. Patterns of vowel sounds showing effect of wide and narrow filters.
outlined previously. These improvements may
be summarized as follows:
? Instantaneous Speech Patterns. With the
sound spectrograph the selected sample of
speech is scanned at more than twice its normal
speed, the filters, etc., being designed for this
purpose. It appears quite feasible, by pushing
this process up into television frequencies, to
obtain speech patterns of the same type on .the
face of a cathode-ray tube, for instance, repre-
senting either a "still" or a "moving" picture
of, for example, 1 sec of speech. In the latter
case the picture would be running off one edge
of the screen and onto the other continuously.
to be photographically enlarged for easy in-
spection and handling, which is a slow process
at best. It does not appear impractical, particu-
larly if it. is not necessary to play the record
back, and if the high frequencies are going to
be modulated down, to develop a simple record-
ing system to produce patterns of the variable-
area type big enough to see and handle without
enlargement, thus providing instantaneous pat-
terns. For cases where the code is repeated, a
cylinder, revolving synchronously once per
code, with a corresponding lateral movement
would give a spiral record on which the sinu
larly located elements of each code cycle woul
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
THE SOUND SPECTROGRAPH 81
ime.
able
ent,
tred
the?
trig-
ble-
not
had
44T 5L-te, 44k. ;
?
in-
ess
cu-
ard
? to
rd-
out
.at-
, a
per
?nt,
ni-
ule
NUM5ER5 INDICATE ORI6iNAL ORDER.
INDICATES IA/VERTEX) 13AND.
FIGURE 41. Patterns from split-band privacy system.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
;
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
82
UNSCRAMBLING AND DECODING METHODS
I,3-SCRAMBLED
2,4-REARRANGED IN ORIGINAL ORDER
I MARKS SYNCHRONIZING PULSE
?
FIGURE 42. Examples of TDS speech with elements cut apart and reassembled.
SECRET
2
4
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
THE SOUND, SPECTROGRAPH
83:
be vertically arranged so that several matches
could be seen simultaneously.'. 22-24
Decoding by Automatic Trial. It appears
quite feasible to combine a modified irDs re-
ceiving machine with a crossbar switch system,
actuated by punched cards, perhaps, so as to
try successive codes until one unscrambles the
speech. This is particularly applicable to cases
where the code is changed often, but where the
number of possibilities can be greatly reduced
by visual means. It is also applicable to cases
where the total number of available codes is
small.
eral ways. First a visual inspection will tell
whether portions of the speech are inverted or
not and the discontinuities in frequency or time
Or pattern will reveal the coding system used
at the remote transmitter. The individual seg.,
ments of the spectrograph may be cut apart
with shears and reassembled so that the un-
scrambled pattern may become evident. Having
determined the nature of the scramble, equip-
ment can be assembled or existing equipment
caribe adjuited so that future samples of the
scrambled speech may be translated audibly as
they come in.
FIGURE 43. Optical system for comparing duplicate patterns simultaneously, avoiding necessity of cutting
up spectrograph.
Decoding Equipment. Equipment Might be
assembled for actually decoding scrambled
speech. For instance, a system of adjustable
filters and carriers might be built to take care
of all split-band systems. This includes shifting
and inverting frequency bands, introducing dif-
ferent delay into various bands, removing
bands of noise, wobbling the carrier, and what-
ever other frequency distortion may be in-
cluded. A TDS decoding system with adjustable
elements and codes also deserves consideration.
4.5.8
Additional Material in Project C-32
In actual use, the spectrograms made of rec-
ords of scrambled speech are analyzed in sev-
In the final report" on Project C-32^ will
be found a discussion of methods for making
beech patterns other than those described
here, and the results of using certain foreign
language records (Linguaphone) in an effort
to discover if the spectrograph was universally
usable without the necessity of having foreign
language experts on the decoding staff. It was
found that the fundamental characteristics of
speech are universal and that the means by
which a recorded sample of scrambled speech
was distorted could be worked out even if those
in charge of the equipment did not understand
the language employed.
a Project C-32, Contract No. OEMsr-230, Western
Electric Company, Inc.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
J
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
84 UNSCRAMBLING AND DECODING METHODS
4'6 PRACTICAL EVALUATION OF PRIVACY
SYSTEMS
? Experience has shown that there is a strong
tendency to underestimate the security or mili-
tary value of a given privacy system as soon
as laboratory studies have indicated that the
system can be . cracked. An attempt will be
made here to point out the great difference
between what might be termed theoretical or
laboratory evaluation and practical or field
evaluation, written from the standpoint, not of
the man interested in decoding a system, but
of the man interested in getting a practical
)
rivacy system into use in the field.
RACKING TIME
The objective of a laboratory study of a pri-
vacy system is to obtain some kind of quanti-
tative measure of the time or effort required
to decode the system. The questions are: "How
long does it take to determine the code, and
how much equipment and how many people are
required?" The coding and decoding processes
are studied in detail, possibly with the aid of
mathematical analysis, to determine whether
there are any weaknesses or any characteristics
of the coding process of which advantage might
be taken to assist in the cracking process. Pos-
sibly a noncryptographic method will be found
to apply. In this case the cracking time reduces
substantially to zero. If noncryptographic meth-
ods are not applicable, available cryptographic
tools and methods are brought to bear. Usually
a new scrambling system will require modifica-
tions or changes in the existing tools or tech-
niques. Possibly the basic methods can be im-
proved for use against this particular system,
or possibly new methods can be devised. Pre-
sumably after all this development work the
project personnel will have become skilled in
the art of decoding this particular system. The
cracking time can then be determined quanti-
tatively, perhaps with estimates as to how far
this may be reduced by further skill.
In the case of repeated-code systems, the
cracking time determined in the above way
substantially represents the total decoding time,
because, as mentioned previously, this code can
be set into a receiving machine and the message
obtained directly. Some additional time might
be added, however, for determining what was
said during the time that the code was being
determined.
The procedure outlined above is very well
illustrated in the series of preliminary reports
on Project C-43 covering the development of
cracking methods for the repeated code TDS
system. They include mathematical analysis,3?, 31
the development of a new decoding too1,22, 23 and
the reduction of the decoding technique to a
routine." In the case of the multiplication sys-
tem, the chronological steps are in one report.27
Too often the cracking time, as determined
above, is quoted without qualification to de-
scribe the security of a system. It is, of course,
usually understood that the use of this figure
involves the following assumptions: (1) that
the enemy knows all about the coding system,
(2) that he is equipped with an adequate supply
of the machines (our own models may still be
far from the production stage), (3) that he
has developed the same decoding tools and tech-
niques that we have (some of our tools may
be entirely new and secret), (4) that he is
equipped with an adequate supply of the de-
coding tools, (5) that he has men trained in
their use, and (6) that he is in a position to
receive a good signal free of interference. Such
assumptions certainly represent an extreme
possibility. Experience has shown that there is
a strong tendency to forget just how extreme
a condition such assumptions represent. Even
if the assumptions are valid there are still other
factors which affect the military value of a
privacy system.
NONREPEATED CODE SYSTEMS
If the code is changed periodically it may be
necessary to have several decoding teams work-
ing in parallel to keep up with the transmitted
material. The number of teams which will be
required depends on the relation between the
intervals of the code changes and the cracking
time. No particular difficulty presents itself in
expressing the decoding effort under these con-
ditions in terms of man-hours. The evaluation
is complicated, however, by the necessity for
additional equipment, not only for decoding but
for recording.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11: CIA-RDP10T02279R000100180001-2
PRACTICAL EVALUATION OF PRIVACY SYSTEMS
85
? In the case of nonrepeated code systems, the
cracking time for any given portion of a mes-
sage will, in general, be long compared to the
duration of that portion of the message. Every
portion of the message must be cracked indi-
vidually, and the decoding effort can be ex-
pressed as a ratio of decoding time to message
time. This ratio may be 1,000 or 100,000 to 1,
that is, each second of message will take 1,000
or 100,000 teams to keep up with the messages
as they are spoken.
This kind of evaluation is somewhat unsatis-
factory, because the length of time it will take
the enemy to determine the intelligence in a
particular sentence which might carry military
information will depend on whether or not he
I v
is at the mo nt working on this sentence or
whether he is asting his time decoding previ-
ous material which might contain no informa-
tion of value to him. In fact, it has been pro-
posed that the security of such high-privacy
systems could be materially enhanced by keep-
ing the circuit 100 per cent busy with all kinds
of material, possibly even from recordings,
-making certain that the enemy has no way of
determining when the circuit is being used for
passing important information. As in the case
of nonrepeated code systems, it seems a bit
unrealistic in evaluating such a system to as-
sume that the enemy will seize upon the few
seconds of message time which are important,
and to compute the length of time it will take
him to decode that portion of the message.
CODE ANALYSIS
Many schemes have been proposed for gen-
erating ever-changing codes by a combination
of short cycles geared together in such a way
that the number of elements in the cycle is the
product of the number of elements in the indi-
vidual cycles. One scheme is to use odd ratios,
such as 99 to 100, so that the code cycle will
not repeat until the smaller wheel has made
100 revolutions. In other words there are 9,900
steps in the code cycle before it repeats. Another
scheme is the cyclometer type in which one
wheel rotates one step for each revolution of
another wheel. Again the total cycle is the
product of the number of steps on the indi-
vidual wheels.
Such schemes should be distinguished from
truly nonrepeating codes, because wherever
cyclic processes are used, they are subject to
analysis. This is a matter pertinent to the field
of cryptanalysis and will not be discussed here.
In general, it may be said that the difficulty of
solving such long cycles is not determined by
the total length but rather by the length of the
individual subcycles.
Systems designed to produce a long code
setuence usually contain provision for read-
justing or realigning certain elements period-
ically or from day to day. Assuming that we
know all about the system except the momen-
tary settings, estimates can usually be made of
the length of time and the number of people
it would take to determine the unknown set-
tings by analyzing a given sample of the code
sequence. The analyst requires a knowledge of
the code for a long sequence of scrambled
speech before he can begin the work aimed at
determining the unknown settings. He must
obtain and solve a sufficiently long sample of
the scramble and then analyze this sequence
to obtain the settings. Too frequently the eval-
uation of a coding system is based on the
analyzing time alone whereas the time required
for solving or unscrambling a long sequence of
scrambled speech may be overwhelmingly
greater than the analyzing time. In fact if there
is no way of solving the code sequence from
the scramble alone, then the analyst can con-
tribute nothing, and the system is still secret
regardless of any inherent weakness of the
cyclic coding system.
FIELD EVALUATION
The continuously changing military situa-
tions of modern warfare require rapid means
of communication in order that the required
military actions can be taken. A perfectly secure
speech privacy system is of no military value if
it requires so much time for encoding and
decoding that it slows up the communication
system to the point where appropriate steps
cannot be taken when needed. Similarly, a
cracking system is of no value if it is too slow
to permit countermeasures to be taken accord-
ing to the intercepted intelligence. For certain
purposes 15 min or even 5 min cracking time
SECRET
neclassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
86 UNSCRAMBLING AND DECODING METHODS
is much too slow. Where this is true, a privacy
system giving 15 min or 5 min privacy is just
as good as one with an hour's security. This is
important because systems affording a few
minutes of privacy were developed in portable
form, whereas those providing longer privacy
were not.
Consider also the equipment and trained
personnel required for decoding intercepted
communications. As a specific example, the
small TDS unit required about 15 min for de-
coding but it required a van-load of highly
specialized equipment.32 Suppose the small port-
able TDS unit were used in many planes and
tanks and other mobile equipment that required
some privacy. Suppose also that different codes
were used within different groups of units and
that the codes were changed at some reasonable
interval. Would it be worth the enemy's while
to provide enough decoding equipment and
enough trained personnel to follow these units
around and decode their messages? If it is not
worth his while, then units rated as low in
privacy may provide high-grade privacy under
such conditions.
Obviously the foregoing does not apply if the
units are used to convey messages between the
higher echelons of command. In such cases
the messages have a longer term significance
to the enemy, and he can afford to devote con-
siderable time and equipment to intercepting
and decoding them.
Advantage might also be taken of the element
of surprise. Suppose we suddenly introduce in
the field a low-grade privacy system in large
quantities. How long would it take before the
enemy diagnosed the system, developed a de-
coding method, manufactured receiving sets of
..?\:ihe proper type and also decoding equipment,
istributed these where needed and organized
and trained personnel to use them? Until he
has done these things the units provide com-
plete secrecy. A different kind of system might
then be introduced which would again provide
secrecy for a time..
It is intended simply to point out that there
are other considerations in the evaluation of
privacy systems than the time it takes a highly
specialized group, such as the personnel of
Project C-43, to decode the system under the
ideal conditions of a laboratory. The decoding
time alone is often quoted, because it is the
only element which can be described quanti-
tatively. While there is always theoretical
agreement about the existence of the other
considerations, they cannot be pointed out too
often or too strongly.
4.7 DIAGNOSIS OF UNKNOWN SYSTEMS
Before discussing the diagnosis of speech
privacy systems it should be pointed out that
facts concerning the origin of unknown signals
are often very necessary to their correct in-
terpretation. Such things as the frequency,
strength, and direction of the signals, the loca-
tion and type of receiver, and the manner in
which the signals were recorded, can be very
important data. That is why interceptors should
be equipped with complete knowledge of the
various kinds of radio systems and transmis-
sions used by both sides, including jamming
and radar signals as well as telegraph and fac-
simile signals. Some of these signals, particu-
larly if transmitted with suppressed carrier,
can give extremely puzzling results if demodu-
lated with an ordinary radio set. These possi-
bilities should be taken into account if signals
are found which do not seem to fit into the
classes discussed below.
As stated before, the spectrograph is of tre-
mendous assistance in recognizing the nature
of an unknown scrambling system. The ear can
usually recognize the presence of time discon-
tinuities. It can also usually recognize the pe-
culiar quality which results from band-shifting
systems. The exact nature of the scramble, (--
however, is usually impossible to establish with
the ear. Even scrutiny of the wave form may
yield no clue. The strikingly graphic analysis
provided by the spectrograph, however, usually
takes the mystery out of the scrambling method
immediately.
Speech privacy systems having frequency
subbands will show horizontal discontinuities
or boundaries in their spectrograms. Similarly
systems employing time division will show ver-
tical boundaries. A considerable variety of sys-
tems display both horizontal and vertical bound-
SECRET
Declassified and Approved For Release 2013/12/11: CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11: CIA-RDP10T02279R000100180001-2
DIAGNOSIS OF UNKNOWN SYSTEMS 87
aries. How to tell these different scrambling
systems apart is the subject of the discussion
and illustrations to follow.
MEASUREMENTS ON SPECTROGRAMS
Since an important part of the diagnosis
procedure consists in determining the length
of time elements and the location of frequency
boundaries, let us first examine the procedures
whereby the time and frequency scales of the
The application of this method to 11-kc spec-
trograms is not explicitly stated in the figure.
A value of K for this condition can be found
by the same formulas. This establishes the time
scale for the 11-kc spectrograms. For the fre-
quency scale the same pattern is used as for
the 3.5-kc spectrograms. However, each hori-
zontal striation is labeled with a frequency
obtained by multiplying the normal frequencies
by the ratio of the two K's.
FIGURE 44. Calibration of spectrograph scales.
The upper spectrogram (narrow filter) shows all the odd har-
monics of the 60-cycle input to a special harmonic generator. In the
wide filter spectrogram (below) the striations represent a beat note
of 120 cycles.
At the left is a portion cut off and inverted. The fact that the
harmonics can be lined up in this as well as other shifted positions
illustrates the linearity of the frequency scale. At the right is a
portion cut off and shifted downward by one component. Since the
harmonics are odd, the base line will fall exactly between two har-
monics if it represents exactly zero frequency.
spectrograph can be established. The spectro-
graph is equipped with a calibrating device
which consists of means for producing a com-
plex wave rich in harmonics from the 60-cycle
power supply. Spectrograms of this wave made
with both the 45-cycle filter and the 300-cycle
filter are shown in Figure 44. If the power
frequency is known, the horizontal and vertical
striations in these patterns provide the time
and frequency scales. If the power frequency is
not known the scales may be established by the
formulas given in the figure. This involves ad-
ditional measurements with a stop watch.
If the power frequency is exactly known, both the time scale and
the frequency scale are determined by the two patterns above. If
the power frequency is not known, the time scale factor can be
determined by the equation K..= LR/T inches per second, and the
frequency by F= KS/N.
L -= Total length of the spectrogram (circumference of the re-
cording drum).
R = Number of rotations of the drum in T seconds.
S ---- Number of striations in N inches.
Figure 45 shows how these scales can be
used to measure the time and frequency bound-
aries in a scramble. It will be noted that for
measuring the time elements spectrograms
made with the 300-cycle filter are best because
they have sharper time boundaries. For meas-
uring frequency boundaries the same filter must
be used as was used in obtaining the scale. It
may be noted here that in present models of
the spectrograph, the wide filter has a different
absolute location than the narrow filter and
therefore should not be used to estimate the
frequency of components or boundaries.
SECRET
neclassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
88
UNSCRAMBLING AND DECODING METHODS
ILLUSTRATIONS OF SCRAMBLED SPEECH
SPectrograms illustrating a large number of
privacy system scrambles are shown in Figures
46 through 65. In so far as possible, these
spectrograms were obtained with actual work-
ing models or systems. In some cases they were
stroy the typical harmonic structure of speech
leaving structureless patterns which cannot be
interpreted. This indicates a distortion of the
wave form. One of these systems, which had a
repeating code and a synchronizing pulse, could
be resolved by the method shown in Figure 55.
FIGURE 45. Time and frequency measurements.
Upper spectrogram, 95-cycle filter; lower spectrogram, 300-cycle
filter. The frequency boundaries are determined by comparing them
with the harmonics of the calibrating wave. These are all 120 cycles
apart, but the lowest is only 60 cycles from the base line.
The element length is best determined by using the 300-cycle filter
made with a laboratory setup simulating the
systems under scrutiny. In a few cases also the
illustrations were made by cutting up spectro-
grams and rearranging the parts. It should be
noted in these latter cases that the boundaries
are unnaturally clear and sharp because in
practice any discontinuity causes a transient
which tends to obscure the true speech along
the boundaries.
It will be noted that some of the spectrograms
in the illustrations were made with the 45-cycle
filter and some with the 300-cycle filter depend-
ing on what features were to be brought out
most clearly.
In some cases the spectrograms alone are
not sufficient to determine the exact nature of
the scramble. Certain systems completely de-
which gives sharp time boundaries, comparing them with the stria-
tions obtained by making a spectrogram of the calibrating wave
with the 300-cycle filter. Each one represents 420 second. Ten of
the above elements cover 70 striations. The length of each element
is Mo x 7912o second.
No general rules, however, can be given for
diagnosing this type of system.
SYSTEMS NOT ILLUSTRATED
Examination of Table 1 shows that there are
a few scrambling systems which are not repre-
sented in the illustrations. These will be dis-
cussed in the following paragraphs. In most
cases, the appearance of the spectrogram pat-
tern which would result can be visualized by
analogy with other systems.
The phase-reversal system (A4) will produce
a scramble indistinguishable from the multipli-
cation system (Hi) provided the phase re-
versals occur at irregular intervals and about
as rapidly as the crossovers in the coding wave
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
DIAGNOSIS OF UNKNOWN SYSTEMS
89
involved in Hi. It is probable that they would
have to occur about that often to make speech
unintelligible.
The split-phase system (A5) involving car-
riers 90 degrees apart was tried out in the lab-
oratory. The output appears just as if two
speech channels, or a speech channel and an
interfering noise, were simply superposed and
then modulated with a single carrier.
The stepped displacement system (B2) is
rather easy to visualize. There will be time
boundaries, with two or more discrete condi-
tions of displacement. Obviously, there are a
great number of possible sequences, including
the possibility of some of the conditions con-
sisting of inverted displacement.
The irregular wobbled displacement (B4)
will be similar to B3 except that the wobble
pattern will not be as simple.
The continuously varied re-entrant displace-
ment (C2) is practically impossible to simulate
artificially, as was done with Cl. If Cl is thor-
oughly understood, however, the appearance of
a wobbled instead of stepped re-entrant condi-
tion is not difficult to visualize.
Nonrepeated-code TDS (F3) will have the
same general appearance as repeated-code TDS.
It may or may not have the synchronizing pulse.
There will of course be no regularity in the
patterns such as was pointed out in F2.
The spectrogram for TDS plus inversion
(G1) is not difficult to visualize. Some or all of
the elements might be inverted, as in A3.
The systems listed in both G5 and G6 will
show equally spaced time boundaries corre-
sponding to the length of the elements. In G5,
the harmonics would be spaced much farther
apart than in normal speech, and show greater
slopes and curvatures. Alternate elements
would show rather consistent differences in
frequency distribution and in the degree of
slope or curvature. In G6, the harmonics would
be spaced abnormally closely, and show very
little slope or curvature. Words and spaces
would be very long. There would be a horizontal
boundary in the middle of the band, and the
patterns in each half would appear like corn-
plete spectrograms, with vowel and consonant
structures apparent. In both of these systems,
if the elements were cut apart, they could be
rearranged to form continuous speech with the
time and frequency scales compressed or ex-
panded from the normal condition.
Level modulations (H2 and H3) would hardly
show up in spectrograms because of the level
compression incorporated in the spectrograph.
This has been verified experimentally.
In J1 alit J2, if the noise were sufficient to
mask the speech effectively, the speech could
not be seen in spectrograms. Patterns for J3
and-J4 are easy to visualize. If the noise spurts
are sufficiently close together, however, they
may produce a pattern like Hi. As far as is
known, J5 exists only on paper.
In Vocoder types of scrambling systems the
spectrograph would show only the channel sig-
nals, which might be either amplitude or fre-
quency modulated. For this type of scramble,
oscillograms of the wave form of each separate
channel signal provide the best means for diag-
nosis and for decoding. A sample of such oscil-
lograms, which was obtained from an actual
Vocoder system, is shown in Figure 50. The
various methods of scrambling such signals
(KI, K2, K3, K4) will produce discontinuities
in these traces which are easy to visualize. A
sample of K5 has not been available.
Channel mixing (L3) can be done in various
ways and at various speeds. It will not be easy
to recognize if done rapidly. No actual systems
are in use, as far as is known.
It is felt that the above illustrations and dis-
cussions cover the known scrambling methods
fairly thoroughly. It is hoped that with their
help any system which might be encountered in
the future can be recognized. Certain additional
spectrographic material appearing in Part I
of the final report on Project C-43b is useful
to anyone wanting all possible data on the sub-
ject. Part II 33 includes all the preliminary re-
ports dealing with specific phases of the work
carried out under the project.
b project C-43, Contract OEMsr-435, Western Elec-
tric Company, Inc.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
90
UNSCRAMBLING AND DECODING METHODS
FIGURE 46. Illustrating modulation sidebands.
The upper spectrogram shows speech modulated with a carrier of
2,000 cycles. Note the symmetry of the pattern around this frequency.
Each harmonic, and each resonance area, is duplicated on both sides
of the carrier. This clearly shows that the two sidebands are exactly
alike, except that one is inverted. The recording from which these
spectrograms were taken has somewhat attenuated the frequencies
near the base line.
The carrier itself is largely suppressed by the double balanced
modulator; high level sounds, however, occasionally unbalance it
sufficiently for the carrier to show through.
In the lower spectrogram, the carrier frequency has been wobbled
at a rather slow rate. Note that as the sidebands move up and down
with the carrier, the harmonics remain parallel, as at point a,
except when marked voice inflections occur, as at point b.
In the clear space at the left, the wobbling carrier can be seen,
together with its second harmonic. In the upper part of the spectro-
gram at point a, the second harmonic of the carrier can also be
seen, with its own set of sidebands.
When the carrier frequency is low, as at points c and d, the lower
sideband can be seen folding back; the folded back portion is right
side up and overlaps the regular sidebands.
4,040W. e
FIGURE 47. Simple inversion.
The upper spectrogram was made in the normal manner, showing
inverted speech: the lower one was made with reversed oscillator
sweep, producing a mechanically inverted spectrogram in which the
speech appears right side Ina.
This sample contains harmonics with marked curvatures. These
are voice inflections, and their occurrence can easily be recognized
by ear. In general, samples with such voice inflections should be
captured because they are most useful for diagnosing scrambling
systems.
In the upper spectrogram, at points a and d, note how the curva-
ture of the harmonics is least at the highest frequencies and pro-
gressively greater toward the lowest frequencies. Similarly, at points
-
b and c, the slopes of the harmonics are least at the top and greatest
at the bottom of the spectrogram. This is directly the reverse of
normal speech and definitely indicates inversion. The lower spectro-
gram illustrates the normal slopes and curvatures.
There is obviously a low-pass filter in the system, at about 3,000
cycles, as indicated by the rather abrupt change in intensity. Such a
filter is normally used to cut off the upper sideband. It is usually
also designed to cut off the carrier. In this case, its cutoff frequency
is lower than the carrier frequency. This shows up at points a and b
in the harmonics, fading out before they completely flatten out.
However, the inversion frequency is not far from 3,000 cycles, be-
cause the slopes are substantially zero as they approach this fre-
quency.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
DIAGNOSIS OF UNKNOWN SYSTEMS
FIGURE 48. Alternate inversion.
Here the speech is divided up into sections by sharp vertical
boundaries. When the individual elements are examined, a, b, and c
show unmistakable signs of inversion. Elements d, e, and I are
definitely not inverted.
The regularity of the dark areas in the lower spectrogram suggests
that only two conditions are involved. Note also that in general,
where slopes can be clearly discerned, the harmonics slope in oppo-
91
site directions in adjacent elements. These indications point to
alternate straight and inverted transmission.
This diagnosis could be confirmed by making a mechanically
inverted spectrogram and matching together alternate pieces from
the two spectrograms.
The switching intervals are irregular, with no repetition apparent
within the time covered by this spectrogram. Additional spectro-
grams, covering a longer period, might show a repeated cycle.
FIGURE 49. Fixed displacement.
In the upper spectrogram, the speech band has been displaced
from its normal location by 1,000 cycles; in the lower one, by 2,000
cycles. Recognition of this condition is aided by familiarity with the
appearance of normal speech in spectrograms. All vowels have
characteristic resonant areas close to what would be the base line
(zero frequency) in normal speech, and the glides such as occur at
a and b tend to start from this region.
At c, the harmonics look as though they could meet at a point if
extended to the right. This point would be the true base line.
A displacement of this type could be produced by modulating with
a carrier of 1,000 or 2,000 cycles, and suppressing the lower (in-
verted) sideband with a high-pass filter. In practice, however, a
double modulation process is used, because the displacement may be
changed at will without changing the filter cutoff.
In both spectrograms, a small amount of lower sideband can be
seen. This incompletely suppressed sideband would look the same
whether produced by single or double modulation.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
? -1
1;
?!?
92
UNSCRAMBLING AND DECODING METHODS
FIGURE 50. Wobbled displacement.
The upper spectrogram shows an example of wobbled inversion.
Note the occurrence of harmonics symmetrically disposed about a
suppressed carrier frequency which is varying between 2,500 and
3,500 cycles. Note how the harmonics remain essentially parallel.
At points a and S the sidebands appear to consist only of low-fre-
quency noise. The carrier wobble is irregular in shape but regularly
repeated in time.
If there is a low-pass filter in the system, its cutoff frequency is
higher than 3,500 cycles. Note that if it were lower, it would occa-
sionally cut off some of the wanted (lower) sideband.
The lower spectrogram shows a wobble covering a much wider
frequency range. The lower sideband dips below the 3,500-cycle
range of the spectrogram only part of the time. This would cer-
tainly be diagnosed as a band displacement system involving double
modulation. If it were encountered in practice, wide-band spectro-
grams would be used to determine the exact displacement.
???: ? A..
Itet*
FIGURE 51. Re-entrant inversion.
The upper spectrogram was artificially produced by cutting up
and rearranging a spectrogram of simple inversion.
There are vertical boundaries about 275 msec apart. There are
also horizontal boundaries, but these are not continuous.
Elements 3 and 6 show no horizontal boundaries, but they show
the signs of simple inversion. All the other elements show horizontal
boundaries, with higher slopes above the boundary than below. If
the two portions of each element were interchanged, the slopes would
be in the correct order (for inverted speech).
It can also be seen that if the elements were thus rearranged, the
harmonics of each element would match those of the preceding and
following elements. This, of course, should be confirmed in practice
by trial.
These configurations would be produced by re-entrant inversion.
In terms of these spectrograms, this process results in inverting
successive elements about frequencies of 1,000, 2,000, and 3,000 cycles,
respectively, removing the upper sideband, and replacing it with
that portion of the lower sideband extending below about 200 cycles.
The lower spectrogram is a duplicate of the one above. The
boundaries have been marked off to show that the elements rear-
ranged, as suggested above, would form continuous inverted speech.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Ap?roved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
DIAGNOSIS OF UNKNOWN SYSTEMS
93
FIGURE 52. Fixed split-band scramble.
These are two samples with the same code. Note the horizontal
discontinuities in the frequency distribution of energy, which can
best be seen by looking endwise along the patterns. These are the
filter boundaries. This system shows five bands, covering the range
from 250 to 3,000 cycles. In split-band systems, the subbands are
generally equally wide, for practical reasons.
There are no vertical discontinuities, other than the normal
sequence of words and spaces.
Note that the harmonics slope or curve in opposite directions
within a word or syllable. This indicates that some of the subbands
have been inverted. The first and fourth bands are clearly normal,
the others are inverted. This shows up clearly at point a.
Note that the fourth band shows the least slope or curvature;
this must have been the lowest band originally. The middle band
shows the most slope or curvature; this must have been the highest
originally. The others can be similarly located, combining the indi-
cations from all the indicated points. Any one point is sometimes
misleading due to the proximity of a harmonic to the filter bound-
ary, as in the top band at a.
At points d and e there is a double inflection. This can be mis-
leading, unless the slopes are estimated for simultaneous instants.
The vertical line was drawn as a guide.
The code is 4, 2', I, 5', 3', the primes denoting' inversion.
FIGURE 53. Rapidly switched split-band scramble, example 1.
These spectrograms show both horizontal and vertical boundaries.
However, the level of the elements as a whole shows a rather smooth
flow, as of words and spaces. Note the clear space ahead of the first
word. These indications suggest that the elements have not been
shuffled in time.
The presence of band shifts, however, is quite obvious. Harmonics
can be seen sloping in both directions within the elements, particu-
larly in the first word group.
The checkerboard effect in the lower spectrogram suggests that
only two codes were used alternately. This is corroborated by the
fact that the middle band shows no vertical discontinuities, indicating
that it was never switched.
SECRET
Declassified and Approved For Release 2013/12/11 CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
94
UNSCRAMBLING AND DECODING METHODS
FIGURE 54. Rapidly switched split-band scramble, example 2.
Here there are both horizontal and vertical boundaries. There is
no evidence, however, that the elements have been time shifted.
There are unbroken clear spaces as at a, the long word groups such
as b show no abrupt changes in level or in pitch, and elements
having different characteristic appearance, such as those at e and d,
are not intermingled.
There is abundant evidence of band shifts: harmonics sloping in
opposite directions, indications of inversion, and abrupt changes in
the resonance areas. Close scrutiny of long word groups such as b
shows that several codes are being used, although it would take
several samples to establish just how many.
This sample illustrates that if the codes are rapidly shifted, any
one element seldom contains enough clear information to determine
which code applies to it. However, if accumulated information about
the system can be brought to bear, two clear bands may sometimes
be sufficient to identify the code.
FIGURE 55. Time
At first glance these look like split-band scrambles. The horizontal
boundaries are quite evident. The bands are 600 cycles wide, begin-
ning at 400 cycles. It will be noted, however, that all four bands
look alike, except that alternate ones are inverted. Otherwise the
slopes and curvatures are alike in all bands. An outstanding example
is at point a. There is no gradation in slope.
The components are not uniformly spaced within a band, and
they frequently go in both directions within a band, as at b and c.
division multiplex.
These are the characteristics of TDM scrambling. In the particular
system illustrated, the frequency range was divided into four bands,
and all were modulated down to the lowest frequency. The switching
rate was 600 per second, but the entire band was then modulated
up 900 cycles, to avoid having the lowest sideband extend to fre-
quencies too low for transmission channels to handle.
It is characteristic of TDM to produce upper and lower sidebands
around the switching frequency and around its odd multiples. The
sidebands differ only in the phases of their components.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
DIAGNOSIS OF UNKNOWN SYSTEMS
95
FIGURE 56. Time division multiplex with noise channel.
This is the same TDM system as in Figure 55, but a band of noise
has been added to increase the privacy. One half of the highest of
the four frequency bands into which the speech channel was divided
has been filled with thermal noise. In the upper spectrogram of
the figure, this thermal noise was steady: in the lower spectrogram,
V?..` ? . -5.7.1.19" "t-j-r;"
the noise was turned on and off about four times per second.
Note that although the noise was introduced into only one sub-
band, it appears in each of the four sidebands in the above patteims.
This shows that in TDM, each sideband contains components from
each subband.
.z - 4 ;
4-2 - ? r,..41 .4.4',.;?:.; ? -41,1;.:V..-1;:?.4:4".
? },
? .
'
a.
180'MSEC
?
? 4.
: eV' '434 .; " 4 ? ' i1,80
MSEC-
FIGURE 57. Subbands variously delayed.
Here there are horizontal boundaries, but the filters apparently
do not cut off very sharply because the bands appear to overlap
occasionally.
Note the staircase effect in the upper spectrogram, each syllable
in the uppermost band appearing somewhat later in the middle
band, and still later in the lowest band. This condition has been
rectified in the lower spectrogram by cutting the frequency bands
apart and shifting them relative to each other, thereby restoring the
normal appearance of words and spaces.
There has been no shifting or inversion of the subbands. Note that
the filter crossovers have been made very deep, as evidenced by the
gaps between bands, and the lowest band has been severely cur-
tailed in width, probably in an effort to reduce the amount of
intelligence which may be gained by listening to any one band.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
r.;
96 UNSCRAMBLING AND DECODING METHODS
FIGURE 58. Combination of TDS and
In this sample the evidences of band shifting are clear enough.
That several different codes are used is suggested by the irregular
distribution of the resonant areas over the frequency range. It is
more conclusive, however, to examine the slopes of the harmonics
in the upper spectrogram. At a, for instance, the lowest band shows
more slope than either the second or fourth; at c, the opposite is
true. At a, the harmonics in the second and third band slope in the
same direction; at b, in opposite directions. To tell how many codes
are used would require additional samples.
rapidly switched split-band scramble.
Evidences of time shifting are also clear. Elements a and b, for
instance, are both strong, but are surrounded by gaps. At d, an
element with energy distributed over the whole frequency range is
surrounded by elements with entirely different distribution. There is
also a marked difference in pitch between el and the surrounding
elements. An even more marked change occurs at e.
This scramble, therefore, is the result of both band shifting and
time shifting. It differs from a complete two-dimensional scramble
in only one resPect, which is described in a separate illustration.
FIGURE 59. Nonsynchronous combinations of TDS and split-band scramble.
These are two combinations of TDS and rapidly switched split-
band scrambles. They differ from the previous illustration in that
the two switching systems are independent. The split-band code is
changed at intervals of about 90 msec, whereas the length of the
TDS elements is about 34 msec.
Each of these switching systems produces its own set of vertical
boundaries. The distance between successive boundaries in the
scramble, therefore, varies irregularly from a value corresponding
to 34 msec down to zero.
In the upper spectrogram, the speech was first subjected to TDS,
and then to the split-band scramble; in the lower spectrogram, the
two scrambles were applied in the reverse order.
Only two split-band codes were used, alternately. Since the second
scramble tends to hide the first, the upper spectrogram shows the
characteristic checkerboard effect noted in a previous illustration.
In the lower spectrogram, the checkerboard effect is broken up by
the TDS.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
DIAGNOSIS OF UNKNOWN SYSTEMS
97
FIGURE 60. Test for two-dimensional scramble.
Spectrograms which serve to illustrate how a two-dimensional
scramble might be recognized. The scramble contains both time
shifts (TDS) and band shifts (rapidly switched split band). Note,
however, that in either of these scrambling systems, and in the
combination of both, elements which are simultaneous in the scram-
ble (that is, subbands within any vertical section) were also
simultaneous in the original speech.
In this example, therefore, a decided tendency may be seen for
the character of a vertical section to remain constant over the
frequency range. That is, low level elements are low all over the
frequency range; high level elements tend to be high all over.
Furthermore, subbands from voiced sounds do.not occur in the
same vertical sections with subbands from unvoiced sounds. Voiced
sound may be recognized by the presence of harmonics in the
45-cycle spectrograms, and regular vertical striations in the 300-
cycle spectrograms.
The most conclusive test for a two-dimensional scramble, however,
is based on the fact that there will be differences in pitch within
a vertical section. This can easily be tested as illustrated above.
The spectrogram is cut down the middle of a vertical section, and
the pieces shifted by one harmonic in either direction. If there is
no change in pitch, the harmonics will still match all over, as above.
lf there is a change in pitch, the shift which is correct for one
subband will be wrong for another. TWo-dimensional scrambles,
therefore, will not pass the above test.
FIGURE 61. Speed wobble
The curvatures of the harmonics in the upper spectrogram look
like voice inflections, except that they are abnormally frequent and
rapid. The resonance areas, as shown best in the lower, also show
a marked degree of curvature. Also, marked correlation may be
noted between the resonance areas and the pitch, that is, they reach
their high and low points simultaneously. In normal speech, the
frequency and trend of the resonance areas are independent of the
pitch trend. Wobbling the speed of a phonograph record or magnetic
tape, however, multiplies the frequency of resonances as well as mul-
tiplying the pitch. These spectrograms were produced in this manner.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
98
UNSCRAMBLING AND DECODING METHODS
2
...-.._ _..._ ... _ -____?7_,. ., 774r.,.. ::_-_________m....,...,,_..: .?-re..-...-....,-L..
,__..........,07,,......,.......,? ? -----. .,,---,,.---.
. ,. k.?,4_711..t-s,iLT--?..,..iir.E... - J.,z4,,,, . ,.e6 ts...alk, ;.rell,..; 1,,C....P.*...4. ...:........?.., .....'..,7.,....774h,r7..,...2,..,:;rL . ,yt:. ..?.t.,.... ..,,:.,$.::::.;......, .;:.",;---.:,.
.:.,.:4.?-,s-1-_-- :.....-:_.--,4 .4::..-1.:0-01,-,..s.--: :4--72:411,;:rt'-1 v.--e . el ::::7-::::::::4,' ' 41.
"r? -i'-'4^:t7,. --."":74. '.':--71,At?- ..=...,,,' .-:.4.M4--'-'"e:,, '--- ""i, ,ky. ,t0?__.--,...."????-?
' ,;-..41., t".4,.;; ,'''- re.,.....,' ? --,:?:".,4?V-c"...4.-n :'.::: .......".^."--
,...,..4........=,,,,,_ ....:-. ,...:,. ,......4--,.. ............____,......4.--''''' ' ' --"" -
1,
Y , ',-4?It? 11.4....-6:44, -,,,I.-Tz4 c-7-2-'.: .......?: ,_, ,---.1-1,-....- ?,,4, ,
?,....,...,......?.....,. ...?..,.....1,4 . ,-....?.....- .,
E._
- ..=--e "--
. .-
?...? ......,....,....-.0....s....44,,,,, ,-;,......?........,,,,,,,?,,-,,- ........,?........,,,..-0
.....,..1.:..- .,..,.,..............?......,,,,.....-A..._,_ ..........----.................1-0:4---- -'464.-.r,--....""'",. ''.-""4"`-'",-?-??"......."-?,'7?
......C...-,_..., _.ce.,-__,a,... ..-...W.,_.,.........
...-...-.................--
Z. ,.__...._... -r, -....?,,,........- - --......--..,== ,.."..: -,..' ----..-.. -4......-, -...4.4....rov!.."4..., .............-- ..."-.....' ' ..- ..-.. - .
. 1..."?-,- ....N. G.....1....,, , ,.. ? . ..
l'KS ....., .. . ,..? L : . ,A7:?;fr.. ' t....'' . r 47,1k.-' 4...-C3V.; ,
,,,, ?....._ _ ..,?.,,?..%,....:...i.."3,_ . . 3 ''',71,?,,,,,)--",f _ ?.....6.,,,,?.4.?..(.17.:,,,?,-k.. ?.? . -??,7?.., --.... _2:2
..
....,.'f.,??_;::- -
w ?
e. .,,i,., .0,-..,..`,..1,....1.........41?, ........?,?41....4,, , ........e..0?4,-,........7,,,c.?.....1. 1:4....:,,, ? ,...,....:,..-?
'cjif'.
;..., .,,nr,...v....,...?,? ?WV ? .p.,,...': ?47,:fil:, di...-;.d.,,,,,.......'???-..-- ,....
''`. -44
...., ....
,.?.
-,-....,..
_ .._...,
...
,,
,,,......... '?......,......... - S'.'1.4S--..? ;17 ..t 1 Z"...1..........7.......,*?".. ....---i 44-- ....4,-..................... . ........:?:.1, . . ..- 170, r ..r...
ak.....- % . . . -.,......111,1110.A.4? E.--..,,,..:i;..t -.- ,.j...,, 4............,......-......e.. 5-,ver --..- ? -------..-? -.", e "'"`-' - ?? "'" .....? tr..* "ii'-'1,
..............,......zN...b.....--
? ......?:????,?........,....
,...t.4.....04.....................41....t... ........ .._=....-.2.7.7 ... .1.1........., .r -
- --- , . .. - ?-? ?" ..-. A01110_ ?ki.??????..............?
FIGURE 62.
The above spectrograms were produced artificially by cutting up
and rearranging spectrograms of normal speech.
The upper spectrogram shows the speech transmitted in sections
about 160 msec long, each section transmitted backwards. If the
sections are as long as this, the condition can be recognized by
familiarity with the normal speech formations, that is, by the way
words normally start and end, and by the trend of the 'resonant
areas. The slopes and curvatures of the harmonics, however, look
perfectly normal.
If the elements are cut apart and matched, it will be found that
Backwards.
the right-hand edge of each element matches the left-hand edge of
the preceding element. This can be seen by inspection of the above
example. The order of the pieces will be completely reversed after
matching. If all the pieces are inverted, however, they will be found
to match in their present order.
The lower spectrogram shows the same material, but in this case
alternate elements are transmitted forwards and backwards. It will
be found that none of the elements can be matched together at all.
To match, alternate elements must be taken from a mechanically
inverted spectrogram, as described in another illustration.
.7.---- --..T.?,---r.------?---?-.
' '....;;., i --,--:?;?.- .
' N, .....-... ,.
, -
, ''?-: ,.4 .4
4, : . ..
' . ,
? J W . ? C.,....'r:7.',,.
FIGURE 63. Multiplication.
The upper spectrogram shows a continuous noise, with several
words or syllables showing through. Counting the harmonics shows
that the fundamental of the noise is about 100 cycles.
Examination of the signal with an oscilloscope shows that the
noise consists of short ?pulses about 10 msec apart. These can be
removed by a blanking circuit.
The lower spectrogram shows a sample (not the same as the one
above) without the pulses. The outstanding characteristic, as in the
sample above, is an almost complete lack of the harmonic structure
of normal speech. Also, the energy is distributed more or less evenly
over the frequency range for each word or syllable. There are no
characteristic resonance areas.
There are no regular boundaries, either vertical or horizontal.
1=1.:
- , ? -
The sequence of words and spaces looks normal in the spectrogram,
and has the normal cadence of speech to the ear.
These characteristics are to be expected when the .scrambling
system operates on the wave form directly. In this particular system,
the speech wave was multiplied by a coding wave. The latter was
'repeated 100 times per second, with a pulse between each cycle. It
is obvious that a high degree of synchronism is required to remove
the coding wave at the receiving end, which accounts for the high
frequency of the synchronizing pulses.
It may be noted that phase reversal (at a sufficiently high and
irregular rate to achieve privacy) is also essentially a multiplication
process, except that the coding wave has no values other than plus
and minus unity. Spectrograms of such a system would be expected
to look like the above.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
:".01
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
DIAGNOSIS OF UNKNOWN SYSTEMS
FIGURE 64. Time division channel mixing.
Here two talking circuits have been switched between two trans-
mission channels on a time division basis, at 300-msec intervals. The
vertical discontinuities can usually be seen, but point a is an out-
standing example of, apparent continuity in pitch, inflection, and
resonance.
The ear can usually recognize the fact that two voices are present,
99
at least at this switching rate. If the voices are nearly alike, or if
recorded samples of the same voice are used, the nature of the
scramble can be determined by cutting the pieces apart and attempt-
ing to rearrange them into continuous speech. This, of course will
be found impossible in channel mixing. Another transmission channel
should be found with the complementary elements.
FIGURE 65. Subband channel mixing.
Here the harmonics occasionally curve in different directions, as
at point a. A horizontal discontinuity is quite apparent at the
indicated frequency, above which there are changes in pitch. These
are not always readily apparent to the eye, but can be established
by measurements.
In general the syllables seem to begin and end at different times
in the two bands. Formations such as c and d do not occur in
normal speech.
No vertical discontinuities are apparent in either band, which
indicates that if any time delays are involved, they apply to the
whole bands. Yet, by trial, the speech in the upper and lower bands
cannot be matched by shifting the bands relative to each other.
It is apparent that two talking circuits are being switched be-
tween two transmission channels. Another channel should be found
which will contain the complementary subbands. This was produced
by a laboratory setup. In practice, to obtain sufficient privacy, it
would probably be necessary to combine this subband channel mixing
with time division channel mixing illustrated elsewhere.
Point b, as a matter of interest, marks an outstanding example
of apparent continuity in both pitch and slope.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
5.1
Chapter 5
DECODING PROJECTS
SPEECH PRIVACY PROBLEMS
STUDIES AND ANALYSES of several privacy
systems of special interest to the Bureau
of Ships were made under Project 13-106,a,"
a continuation of the work carried forward
under Project C-43. In each problem the prin-
cipal objectives were the determination of the
security afforded by and the transmission per-
formance of the privacy system in question.
NATURE AND SCOPE OF THE WORK
The material submitted for study and analy-
sis under this project comprised working models
of two privacy systems, recordings of speech
scrambled by three privacy systems, and paper
proposals for two systems.
Security evaluations were made under favor-
able laboratory conditions. It was assumed that
the enemy (1) was thoroughly familiar with
the speech privacy system, (2) had the neces-
sary intercept, recording, and decoding equip-
ment, (3) had trained personnel, (4) was in a
position to receive adequate signals, and (5)
was completely organized so that no time would
be lost in obtaining and making use of intelli-
gence from the decoded message. The security
ratings assigned to the several systems evalu-
ated did not take into consideration any prac-
tical difficulties which might be encountered in
the field or under combat conditions where the
work of intercepting, diagnosing, decoding, and
obtaining intelligence from scrambled messages
must be carried on under stress.
The following assignments of Work on this
project were authorized by Division 13:
1. British Modulator Type 2C (manually
switched) working models.
2. British Two-Dimensional Privacy System,
recording.
3. British Modulator Type 2C (rapidly
switched), recording.
a Project 13-106, Contract OEMsr-1440, Western
Electric Company, Inc.
100
4. New Zealand Switched-Band PriVacy Sys-
tem, recording.
5. New Zealand Switched-Band Privacy Sys-
tem, working models.
6. Proposals of L. E. Gabrilovitch for pri-
vacy systems.
4
SUMMARY OF RESULTS
The results of the work done on the several
assigned problems are given in detail in Reports
1 to 5, inclusive, which form an appendix to
the final report of the project.34 These reports
cover the six assignments listed above with the
fourth and fifth assignments, relating to the
New Zealand switched-band system, being com-
bined in one report. All of the systems consid-
ered under this project were of the short-term
privacy variety. Of these, the British two-
dimensional system appeared to be the most
promising.
BRIEF REsumt OF SYSTEMS
British Modulator Type 2C (Manually
Switched), Working Models. This system pro-
vides four fixed speech scrambling conditions,
each involving either one or two simple modu-
lating processes; the choice of any one of the
four scrambling conditions or clear speech is
under the control of a manual switch.
When a receiving unit, or its equivalent, is
at hand, there is no difficulty whatever in dis-
covering the proper decoding condition in a
matter of seconds. The security afforded by the
system is, therefore, almost nil.
The fixed-code scrambles can be demodulated
satisfactorily by a single modulation process
without filters, it being necessary only to use
the appropriate frequency of the demodulating
carrier. For this reason it is possible to obtain
intelligence from radio transmissions of the
scrambled speech by means of an ordinary type
of radio receiver equipped with a beat-fre-
quency oscillator. The efficacy of this method
will, of course, depend upon having adequate
relative stability of the radio carrier and the
beat-frequency oscillator.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
SPEECH PRIVACY PROBLEMS
101
These units were well constructed and oper-
ated satisfactorily from the standpoint of over-
all speech quality.
British Two-Dimensional Privacy System,
Recording. This system utilizes both frequency-
and time-division scrambling. It employs three
frequency bands and ten time elements of 0.065-
sec duration in a repeated code. The time delay
in one direction of transmission (exclusive of
delay of the transmission path) is 0.65 sec.
The evaluation of this system was based on
the study of one recording bearing a single
sample of scrambled speech, together with sam-
ples of clear speech and speech which had been
coded and decoded for comparison. An evalu-
ation based on such limited data is necessarily
tentative and should be supplemented by tests
on working models.
The speech scrambled by this system appears
to be invulnerable to direct listening and to
other noncryptographic attacks. It is, however,
vulnerable to cryptographic attacks" and a
working solution of the code can probably be
obtained in a matter of 3 or 4 hours. With
a model of the receiving equipment at hand, it
is conceivable, although it was impossible actu-
ally to try it, that a substantial amount of
intelligence could be obtained in the order of
half an hour. This latter method would involve
the use of spectrograms from which sugges-
tions are obtained for setting up partial decodes
on the receiving unit.
The most noteworthy weakness in the British
two-dimensional system appears to be the use
of a fixed repeated code. The addition of code
changing means would increase the crypto-
graphic security very greatly.
The quality of the restored (or decoded)
speech presented on the recording compared
favorably with the clear (or uncoded) speech
on the same recording.
British Modulator Type 2C (Rapidly
Switched) Recording. This system is the same
as the Type 2C discussed above except that
means are provided for rapidly switching from
one scrambling condition to another and that
clear speech is used as a fifth scrambling condi-
tion. The order in which the scrambling con-
ditions are selected is predetermined according
to a code which repeats after a sequence of 20
such selections, each enduring for approxi-
mately 0.065 sec.
A code-switching mechanism for use with
this system was promised, but was never re-
ceived. This would have made it possible to
make a more positive evaluation of the system
than can be made from the recording of scram-
bled speech. In fact, the results obtained from
noncryptographic attacks on the recorded
ssramble are considerably at variance not only
with what could logically be expected, but also
with the results obtained with working models
of a very similar system.34
Repeated listenings directly and also through
a two-path superposition circuit to the recorded
scramble yielded several words and phrases but
very little intelligence. Similar tests on the New
Zealand system (working models) yielded, on
the average, 40 per cent intelligence to direct
listening and 80 per cent intelligence with the
two-path circuit.
Repeated listenings through an automatic
analyzer-decoder circuit yielded approximately
60 per cent of the intelligence from the recorded
scramble of the British Modulator Type 2C
rapidly switched system. This same procedure
yielded practically 100 per cent intelligence on
the New Zealand system.
A cryptographic solution of the repeated-
code sequence used in making the recorded
scramble can be determined by inspection of
two spectrograms in about 15 min.
The quality of restored speech on the record-
ing compared favorably with the clear speech
on the same recording.
New Zealand Switched-Band Privacy System,
Recording and Working Models. Fundamen-
tally, this system is very similar to the British
Modulator Type 2C, rapidly switched, and dif-
fers in what appears to be only minor details:
The inversion frequencies in the scrambling
circuits are somewhat different; the duration
of each of the rapidly switched scrambling
conditions is 0.043 sec (rather than 0.065 sec) ;
a sequence of 18 selections of scrambling con-
ditions (rather than 20) comprises the coding
cycle. The New Zealand system is equipped
with an appliqu?nit for automatically chang-
ing the code each cycle for a total of 625 cycles,
or for a period of about 8 min, before repeating.
SECRET
im,,,inecifiarl and Annroved For Release 2013/12/11: CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
102
DECODING PROJECTS
A recording of two samples of speech scram-
bled by this system (using repeated code) was
received for analysis and most of the intelli-
gence was obtained by noncryptographic meth-
ods. Somewhat later, the scrambling equipment
for two terminal units was received and was
set up for tests and demonstration as a two-
way privacy system.
The security afforded by this system is very
low for military purposes and is inconsistent
with its size and weight.
Repeated listenings directly to the scramble
yielded on the average about 40 per cent of the
intelligence to experienced observers; repeated
listenings to the scramble through a two-path
superposition circuit yielded about 80 per cent
of the intelligence. A repeated code can be
determined by an aural method, using the ter-
minal equipment, or its equivalent, in about
7 min. An automatic analyzer-decoder circuit
yielded at least 50 per cent of the intelligence
from either a repeated or a nonrepeated
(8-min) code sequence on the first listening
and practically all the intelligence with few
additional listenings.
By cryptographic methods, a repeated code
can be determined in about 20 min and a non-
repeated code (including the starting point of
the automatic code-changing unit) can be de-
termined in about 1 hr.
Mechanically and electrically the units oper-
ated satisfactorily; the intelligibility of the
restored speech was good but the quality,
though fairly good, was somewhat inferior to
what might be achieved with improvements in
design.
Proposals of L. E. Gabrilovitch for Privacy
Systems. Of two proposals by Gabrilovitch, the
first, described as a "Screen Secrecy Set with
Narrow Audio Band," appeared to require
considerable equipment to obtain only a very
limited degree of security with probably poor
transmission performance and a sacrifice of
operating range.
The second proposal, described as a "Phase
Varied Inverter-Distorter" (simplified secrecy
set), although similar in basic principle to the
RCA-Bedford system developed under Project
C-5414 offered, theoretically, some possibilities
of obtaining a fairly compact and lightweight
set having somewhat better restored speech
quality than the Bedford system. There were,
however, a number of questions regarding the
degree of security, coding possibilities, and
practicability of some of the electronic proc-
esses.
When, after study, it appeared fairly evident
that the development of the second proposal
would tend more and more to duplicate that of
the Bgiflford system and would offer few, if any,
advantages over the latter when completed, it
was recommended that further study of this
proposal be discontinued.
GENERAL CONCLUSIONS AND REMARKS
In the course of the work done under this
project, a number of conclusions were reached
regarding the systems under consideration and
their evaluation based on the use of working
models as contrasted with phonograph record-
ings.
Switched-Band Systems. Of the systems con-
sidered, vulnerability to repeated listenings
directly to the scramble is attributed to the
inherent lack of privacy in some of the five
speech-scrambling conditions. The average in-
telligence obtained in listening to the five fixed
scrambles is 40 per cent, which is approxi-
mately the same as obtained (on the average)
when the scrambling conditions were rapidly
switched. The need for scrambling conditions,
each having an adequate degree of privacy, is
obviously indicated.
The high yield of intelligence obtained from
superposition listening is attributed mainly to
the fact that some of the five scrambling con-
ditions are not mutually private and effectively
decode one another (Codes A and B in the New
Zealand system and Codes 1 and 3 in the British
system). This effectively reduces the available
number of scrambling conditions. Hence, the
five scrambling conditions should be not only
inherently private but also mutually private.
The vulnerability of the systems to either
direct or superposition listening is independent
of whether a repeated or nonrepeated code is
employed.
The use of a repeated code makes the system
very vulnerable to methods of cracking wherein
the code is to be determined. It is necessary to
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Ap roved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
SPEECH PRIVACY PROBLEMS 103
decode only one cycle when successive cycles
can be used to obtain confirmation.
When the system is not vulnerable to non-
cryptographic attack, the use of nonrepeated
coding increases privacy. If the coding is truly
random, it is necessary to decode each indi-
vidual cycle with no opportunity for confirma-
tion from successive cycles.
Two-Dimensional System. This type of sys-
tem involving both frequency and time division
scrambling affords more security than can be
obtained by using either method of scrambling
alone. In the case of the British system con-
sidered, the privacy would have been materially
increased by using a nonrepeating code, more
frequency bands, and shorter and more time
elements.
Masking Systems. Systems of this type em-
/ ploy a screen of noise overlaying the signal to
be masked. This is accomplished in one system
by modulating the masking and masked signals
on a split-phase subcarrier. The discrimination
between these two signals at the receiver re-
quires absolute synchronization and proper
phasing of the demodulating carrier. Since this
is difficult to achieve in practice, because of
the distortions appearing in the transmission
channel, the restored speech will be of poor
quality, being distorted and noisy.
The relatively large amount of power re-
quired for the masking signal reduces the effi-
ciency of the radio transmitter in that smaller
transmitting ranges are obtained for a given
amount of output power.
Bedford Type Systems. Systems of this type,
of which the phase-varied inverter-distorter
system proposed by Gabrilovitch is one, depend
upon the modulation of speech by a complex
coding wave to obtain privacy. Clear speech is
obtained at the receiver by demodulation of the
scramble with an accurately synchronized de-
coding wave which effectively is the reciprocal
of the coding wave.
To avoid the possibility of partially cracking
the scramble by demodulating it with a single
frequency, it appears to be necessary that the
complex coding wave have no predominant
frequency components but, instead, should have
a fairly uniform spectrum of at least several
hundred cycles width within the limits of the
speech band. The resulting band width of
the scrambled speech exceeds the width of the
speech band by an amount equal to the highest
frequency in the coding wave. It follows, then,
that either the band width of the channel con-
veying the scrambled speech must be wider
than for normal speech bands, or the original
speech band must be made narrower than nor-
mal if distortion is to be avoided.
S4ince an accurately synchronized decoding
wave of proper phase is required for deriving
clear speech at the receiver, this system is sen-
sitive to distortions in the transmission channel.
The restored speech should not, however, be as
noisy as that of the masking systems. In the
Bedford type systems, imperfect demodulation
yields unwanted products which are propor-
tional to the speech energy rather than to the
relatively large masking energy.
Synchronization by means of a continuous
modulated wave is believed to be superior to
synchronization by pulses as proposed in the
RCA-Bedford system. In the latter instance,
the wave form of the transmitted pulse is both
important to the proper operation of the system
and sensitive to distortions over a large part of
the band of the transmitting channel.
Evaluation of Security of Systems from Re-
cordings. Phonograph recordings of speech
scrambled by a privacy system provide a less
desirable means for evaluating the security of
a privacy system than do working models of
the system. The results of analyses based on
phonograph recordings can be used for deter-
mining the nature of the privacy system and
the code but even though the quality of the
recording is good, difficulty may be experienced
in direct or superposition listening tests.
Very often it is found that recordings, which
are considered iyoderately good for clear speech
are surprisingly inadequate for storing scram-
bled speech for subsequent analysis and res-
toration. This appears to be due to (1) har-
monic distortion, which, when not too great,
passes unnoticed in clear speech, and also to
(2) irregular speed variations (in either the
recording or reproducing systems) which pre-
vent precise synchronization necessary in some
privacy systems. However, the fact that a high
quality recording is required in cracking a
SECRET
anri Annroved For Release 2013/12/11: CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
104
DECODING PROJECTS
given privacy system, is in itself, of consider-
able practical importance in evaluating the
system.
The most effective cracking techniques often
involve the use of a receiving unit, Or its
equivalent. When only recordings are available
for analysis, it becomes necessary either to
build an equivalent receiver or merely to specu-
late on what might be done with a working
model. Neither of these alternatives is very
satisfactory. It is, therefore, highly desirable
whenever possible, that evaluations be made by
tests on working models.
3'2 FIELD DECODING EQUIPMENT
The experience gained with the sound spec-
trograph in Projects C-32 and C-43 resulted
in the recommendation that the device be re-
designed for field use and included in field
decoding kits at radio intercept stations. Under
Project 13.3-86b eight of the units were built,
three for the Signal Corps, three for the Navy,
and two for the British under Lend-Lease
requisition. Together with operating and main-
tenance instiuctions, these eight spectrographs
(D-165-529) were delivered between January
and Ma. 1, 1944.
The units had the following weights and
dimensions :35
Unit
Recorder
Amplifier-analyzer
Rectifier
Weight
68 lb
64 lb
56 lb
Dimensions
(Inches)
171hx161/2x113/4
2014 x1114x1414
201/1x111/4x1034
b Project 13.3-86, Contract OEMsr-1110, Western
Electric Company, Inc.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Chapter 6
FACSIMILE PRIVACY SYSTEMS
6.1 INTRODUCTION
THE SCRAMBLING SYSTEMS heretofore de-
scribed have related to the transmission of
speech. Another important form of communi-
cation, however, is graphic copy such as maps,
drawings, and photographs, where a facsimile
of the original subject copy is to be transmitted
to a remote point by wire or by radio. In war-
time, of course, it is as important that privacy
be attained in this form of communicated in-
telligence as in the transmission of the human
voice.
Two projects under Division 13 were con-
cerned with the general problem of scfambling
graphic copy, Project C-73a being a survey of
all existing and proposed systems and Project
13.3-97b describing a system for scrambling
the copy before scanning it for transmission by
radio or wire.
6.2
6.2.1
FACSIMILE PRIVACY
Statement of the Problem
At the time of this project, all concerned
agreed that facsimile had attractive possibil-
ities as communication means, but up to that
time it had been handicapped by lack of privacy
for military use. Considerable information on
the several phases of the subject existed in
various places but no attempt had been made
to consolidate it. Accordingly Division 13 au-
thorized Project C-73 which had as its object
the survey of the general field of facsimile pri-
vacy and the preparation of a report on the
subject.
The project was completed between December
1942 and July 1943. A final report36 was pre-
pared on June 7, 1943, and a much abridged
report of this final report on October 15, 1943.
The summary to follow is taken from the
a project C-73, Contract OEMsr-837, Radio Corpora-
tion of America.
b Project 13.3-97, Contract OEMsr-1202, Faximile,
Inc.
abridged report. Since it is the basis of any
further work in this subject, this abridged
report is summarized in some detail. The com-
plete final report gives essential details of non-
private systems and furnishes complete back-
grdtt nd
The directive covering Project C-73 called
for:
1.A brief summary of basic facsimile mech-
anisms and modes of transmission.
2. Investigation of the degree of privacy ob-
tainable.
3. Adaptability of telephonic privacy systems
then existing and those under development.
4. Evaluation of means for cryptanalysis of
various graphic privacy systems the enemy
may use.
5. Suggested design of the most useful type
of equipment for both privacy transmission
and cryptanalysis.
The contractor made contact with all known
individuals and organizations in the United
States who might be able to contribute sugges-
tions which would be useful in providing fac-
simile privacy. A complete report on these
contacts is given in the appendices of the final
report. At the time of the survey, telephonic
scrambling methods were under active study
by Division 13 contractors and by the Military
Services, but up to that time no attempt had
been made to use telephonic systems for fac-
simile privacy.
Laboratory tests were made to determine the
effectiveness of telephone systems as applied to
facsimile. Throughout the investigation of all
known and suggested methods of privacy for
graphic copy, the attempt was made to con-
solidate the acquired data on means of apply-
ing the systems studies, and on the effectiveness
and availability for production and other points
of interest to the Services.
6.2.2 Privacy Defined
A coded facsimile subject may be considered
to have been rendered secret when it is so coded
SECRET
105
narinqcifiRd and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
106
FACSIMILE PRIVACY SYSTEMS
that no means may be improvised to decode it,
other than use of the applied code and equip-
ment. This is the ideal condition, but one which
was unattainable in practice.
A coded facsimile subject may be considered
to have been rendered private, from a military
point of view, when at least 72 hr are required
to reproduce the essential intelligence of the
original subject copy. If the subject copy is a
map, the essential intelligence is obtained when
locations are disclosed. If the subject copy is
type, the essential intelligence is disclosed when
the characters become legible, regardless of de-
coded subject quality.
The relative degree of privacy of a coded
facsimile subject may consequently be estimated
by the number of hours required to decode its
essential intelligence. This time may sometimes
be reduced by subdivision of decoding opera-
tions among several members of a decoding
staff.
6.2.3 Facsimile Coding Methods
A facsimile signal is less vulnerable to unau-
thorized reception than ordinary voide or code
signals because it is unintelligible unless one
has a recorder. With a recorder available, how-
ever, which can be readily adjusted over a wide
range of drum speeds, it would be a matter of
less than a minute to establish the correct op-
erating conditions to receive a picture. It must
be assumed that the enemy has such apparatus
and that, therefore, additional privacy means
are essential for military transmission.
Unlike voice transmission, the sending of
intelligence by facsimile depends on two para-
meters. The first is the continuous envelope of
signals representing the shading of successive
picture elements. The second is the information
as to where each successive picture element
should be printed on the recording sheet in
order that a picture may be formed. This is
normally supplied by moving the scanning spot
through a fixed and simple scanning pattern
at a known rate of speed. Obviously, privacy
can be secured
1. By scrambling the signal transmission,
2. By confusing the scanning pattern,
3. By both scrambling the signal transmis-
sion and confusing the scanning pattern.
The facsimile picture signal is ordinarily a
modulation on a subcarrier and can be trans-
mitted through the same apparatus and circuits
as are used for voice signals. For low-speed
facsimile the channel width required is also
of the same order. Thus, already developed
methods of voice scrambling can be considered
for f2csimile privacy. There are three of these,
the first three in the following tabulation.
SIX BASIC METHODS
The assigned designations below are used
throughout the rest of the report.
1. [A] Transposition of frequency bands.
The spectrum of signal fre-
quencies is divided into five bands
and these are manipulated to
secure privacy.
2. [B] Frequency multiplication.
All signal frequencies are mul-
tiplied by a complex and change-
able coding wave to produce a
new pattern of frequencies within
the transmitted spectrum.14 (See
Chapter 3.)
3. [TDS] Time delay system.
The signal envelope is divided
into intervals of time, which in-
tervals are variously delayed and
then transmitted in a new order.
When these three speech scrambling systems
are applied to facsimile, the first two scramble
the signal; the last one, however, is equivalent
to a confusion of the scanning pattern and it
can be most clearly thought of in that way. It
is just as if the scanning spot traversed one
part of a line of the subject, then jumped to
another part, and so covered the whole area
in an irregular sequence of partial scanning
lines.
To these already developed and self-con-
tained coding means there may be added the
following:
4. [PR] Polarity reversal.
Parts of the scanning line are
reversed in polarity, from black
to white and vice versa, according
to a coding sequence.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
FACSIMILE PRIVACY
107
5. [VS] Variable speed.
The scanning pattern is modi-
fied by changing the drum speed
by varying amounts according to
a coding sequence.
6. [PT] Pretransmission.
The subject copy is itself
scrambled bY optical or other
means before being placed on the
scanner.36 (See Section 6.3.)
Methods 4 and 5 involve special modifications
of the normal facsimile equipment. Method 4
can be thought of either as a distortion of the
signal envelope or a confusion of the scanning
pattern; Method 5 is obviously the latter.
Method 6 has grown out of the activity of this
project but is, of course, not limited to fac-
simile. It would be equally applicable to mes-
sages delivered by courier.
It should be emphasized that the very orderli-
ness of the normal facsimile process makes
difficult the successful application of privacy
methods to picture transmission. Any cyclic
switching operation will reveal its true time
sequence in the facsimile reproduction. Line
sections must be exactly fitted to avoid gaps
or overlap. Even the white picture background
may reveal coding changes due to slight ex-
posure variations. Long, straight lines normal
to the direction of scanning are particularly
revealing since they serve as time references,
and disclose periodicities. In fact, the repro-
duced facsimile copy is a permanent record of
all the optical and electrical operations which
have been performed on the subject copy. Un-
like the scrambled radio-telephone sound wave?
it is permanently captured to lure to the fullest
extent the ingenuity of the decipherer.
NOMENCLATURE FOR VARIOUS METHODS
To facilitate the designation of the various
secrecy methods, a nomenclature was adopted
in the final and abridged project reports in
which the basic systems are indicated by letter
combinations. In addition to the letters, a
numeral is appended to differentiate between
specific types of a basic system. If two basic
systems are operated in combination, the letter
group of each system will be used. These
designations are as follows.
Transposition of Frequency Bands [A]. The
method of subdividing the picture signal into
discrete frequency bands and interchanging
these bands in the frequency spectrum in ac-
cordance with a coding signal is designated by
the letter A. A modification of this method
includes the process of frequency inversion.
A-1 and A-2. These designations are left open
for experimental development.
A-3. This is the standard commercial system
of the Western Electric Company.
Frequency Multiplication [B]. The method
of transmission whereby a coding signal is
multiplied into an intelligence signal is desig-
nated by the letter B. Modifications of this basic
system are essentially modifications of either
the intelligence signal or the coding signal.
B-1. Straight multiplication without altera-
tion of the intelligence signal or the coding
signal.
B-2. An audio tone of fixed amplitude and
of a frequency outside of the picture signal
band has been added to the intelligence signal.
B-3. An audio tone of fixed amplitude and
of variable frequency outside of the picture
signal band has been added to the intelligence
signal.
B-4. The normal subcarrier frequency modu-
lation [SCFM] modulated with a 500-cycle
tone.
B-5. The normal SCFM picture signal is
limited; i.e., the sine wave of varying frequency
is converted to a square wave of varying fre-
quency before multiplication.
B-6. A very low frequency (less than 1 cycle
per second) has been added to the picture
before conversion to SCFM and subsequent
multiplication. This results in a slow shifting
of the picture signal band in the frequency
spectrum.
B-7. A method wherein the coding signal is
changed periodically. The equipment for this
system has been named "Myopia Mark I."
Time Delay System [T DS] . Essentially, this
method produces a continuous transmission of
coded signals by breaking down the normal
signal sequence into discrete time-signal inter-
vals and rearranging these intervals in a dif-
ferent chronological order. Modifications of this
system are made on the basis of the total
SECRET
npriacsified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
108
FACSIMILE PRIVACY SYSTEMS
number of signal intervals that can be switched.
TDS-1. This is the Model B magnetic-tape
system as developed by the Bell Telephone
Laboratories [BTL].
TDS-2. This is the D-Specification magnetic-
tape system as developed by BTL.
TDS-3. This is the C-50 magnetic-tape sys-
tem as developed by BTL under Project C-50."
TDS-4. This is a simple two-head magnetic-
tape system as used in the preliminary tests on
this project. The time delay between the two
heads was approximately 1/20 sec.
Polarity Reversals [PR]. This method is
designated by the letters PR. Basically, it is
a system which subdivides a continuously vary-
ing picture signal into time intervals and
determines the polarity of the transmitted
signal throughout these intervals. The process
can be visualized as a reversing switch con-
trolled by an auxiliary or coding signal which
PR-2. Polarity reversals in which the time
of transmission of one polarity is less than the
time of transmission of a scanning line. With
this arrangement, every scanning line of the
coded picture will resemble a chain whose alter-
nate links have reversed polarity.
PR-3. This method is similar to PR-2 except
that the duration of the switching polarity is
decreased until it approximates that of a pic-
ture elOment.
Variable Speed [VS] and [CVS]. In a
variable-speed transmission system, the linear
speed of scanning is varied in accordance with
a coding signal. In practice, this variation may
be in fixed steps [VS], or continuously variable
[CVS]. It represents a modulation of the scan-
ning rate, and may vary from a small per-
centage of the scanning rate to 15 per cent or
more.
Speed deviation classifications have been
FIGURE 1. Map uncoded (right), coded by frequency transposition (middle), arid decoded (left).
interchanges the black and white portions of
the picture. The degree of confusion will then
be directly proportional to the rate of switching.
Therefore, the applications of the basic method
are subdivided into the following forms.
PR-1. Polarity reversals in which the time
of transmission of one polarity is greater than
the time of transmission of a scanning line.
The coded picture will then be composed of
groups of scanning lines of alternate polarity
and the number of scanning lines in any group
will vary with the duration of the coding signal.
designated as indicated below:
VS-1. Speed change code of -?-1/2 per cent.
VS-2. Speed change code of ?-1 per cent.
VS-3. Speed change code greater than +1
per cent.
VS-4. Two dimensional VS scanning for
PT method.
CVS-1. Continuously variable speed change
code.
Pretransmission [PT]. Methods wherein the
subject copy is itself scrambled by optical or
other means before being placed on the scanner.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Si
Fi
de
sp
ti
fl
C(
ti
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
FACSIMILE PRIVACY
109
TYPICAL SAMPLES OF CODING AND
DECODING BY VARIOUS
SINGLE PRIVACY METHODS
Transposition of Frequency Bands [A].
Figure 1 is a typical example of coding and
decoding with the existing Western Electric
speech privacy equipment A-3. The right por-
Frequency Multiplication [B]. Figure 2 is a
typical example of coding and decoding with
the Myopia Mark I equipment. The right sec-
tion shows the coded picture using 3-sec normal
code shift; the center section, the decoded copy;
and the left section, the normal uncoded subject
copy.
FIGURE 2. Copy coded by frequency multiplication (right) and decoded (middle).
tion of the recording is the uncoded subject
copy; the middle section, coding by five test
frequencies; and the left section, the decoded
copy.
Mathematical analysis indicates that method
B should be subject to decoding by means of a
filter placed at the frequency corresponding to
either white or black, by virtue of the differen-
FIGURE 3. Effect of passing coded signal of Figure 2 through narrow-band filter. Left half employed only
one coding wave; right half used 3-sec code shifts.
The coded section is seen to introduce a
background, without scrambling of the essen-
tial intelligence. The system has no privacy
when applied to facsimile transmission.
tiating action of the photocell at the boundary.
Figure 3 shows the effect of passing the coded
signal through a 100-cycle narrow-band filter
having a 1,445-cycle mid-frequency, which cor-
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
110
FACSIMILE PRIVACY SYSTEMS
responds to white in the recording. The left
half of the recording utilized only one coding
wave, the right half employed 3-sec code shifts.
The essential intelligence of the subject copy
is seen to be revealed by filter decoding.
Figure 5 is the result of coding by polarity
reversals at the rate of approximately 50 per
scanning line. The left portion of the recording
is the result of applying a telegraphic signal
repeating every 80 bands, as an additional
FIGURE 4. Map scrambled by TDS system. Left portion is decoded version of map.
Time Delay System [TDS]. Figure 4 is a
typical example of coding and decoding by the
D-Specification TDS system. The right portion
of the picture shows a complete scramble with
little evidence of switching periodicities. The
left portion of the picture is the reconstructed
subject copy resulting from the decoding opera-
coding source. Boundary conditions are seen
to convey a considerable degree of intelligence
in both sections of the coded picture.
Figure 6 is the decoded version of the above
coded picture. It shows a reproduction of the
subject copy, with good quality.
The susceptibility of the PR method to de-
FIGURE 5. Appearance of copy scrambled by polarity reversal.
tion. D-Specification TDS is an immediately
available system, having a high degree of view-
ing privacy, and a considerable degree of de-
coding privacy.
Polarity Reversals [PR]. Figures 5 and 6
are typical samples utilizing the PR method.
coding by switching transients passed through
a narrow-band filter is shown in Figure 7.
In this recording the switch reverses polarity
at the rate of approximately 50 times per
scanning line. The left portion of the picture is
normal uncoded subject copy, the middle sec-
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
tic
pa
a.
fri
to
as 11
re
Fi
of
oi
P(
rc
cc
st
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
FACSIMILE PRIVACY
111
tion coded, and the right section decoded by
passing the 1.5- to 2-kc f-m signal through
a 100-cycle narrow-band filter having a mid-
frequency of 1,785 cycles. The recording is seen
to be completely decoded by the filter, as far
as essential intelligence is concerned.
drum at normal drum speed will be noted in
the coded section of the recording.
The CVS method, at the same maximum fre-
quency deviation of -?5 per cent, is illustrated
by Figure 9 made by Times Telephoto Equip-
ment, Inc., using continuously variable drum-
FIGURE 6. Decoded
Variable Speed [VS] and [CVS]. A typical
recording by the VS method is illustrated by
Figure 8. A speed deviation having a maximum
of .?5 per cent was applied for a fixed number
version of Figure 5.
speed shift, but with a random drum-speed
swing. The lower section of the picture is
normal uncoded, the middle section coded, and
the upper section synchronously decoded. The
FIGURE 7. Example showing susceptibility of PR system to decoding by passing coded signals through
narrow-band filter. Left portion, uncoded copy; middle portion, coded by PR; right portion, copy decoded
by use of filter.
of drum revolutions varying from 10 to 20
per frequency step. The lower section of the
recording is normal uncoded, the middle section
coded and the upper section decoded. A decoded
strip corresponding to ten revolutions of the
CVS method shows the highest degree of pri-
vacy of the single methods. No direct electrical
method has as yet been devised to break the
CVS code.
Pretransmission [PT]. The pretransmission
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
FACSIMILE PRIVACY SYSTEMS
method PT has not been developed experi-
mentally to the stage in which coded and ,
decoded subject copy is available by facsimile
transmission. The possibility of the method for
coding purposes is shown in Eastman Kodak
Company samples, Figure 10, in which a double
shredding process was used. At the lower left
NO
bination is illustrated by Figure 12. Mixing
of the two methods produces a high degree
of privacy for the combination. Difficulty of
synchronization and phasing at the scanner
and recorder, unfortunately, renders the com-
bination less attractive for practical circuit
applications.
77r""*.errxr:"
-
?
t
'icEituAri
FIGURE 8. Recording by variable speed [VS] sys-
tem. Lower portion, normal uncoded; upper por-
tion, decoded material. .
of Figure 11 is shown a coding by the single
shredding process, and at the lower right a
second shredding at right angles to the first.
The method is limited as to its basic privacy
by the problem of registration and skew which
will determine the practical element size. The
method gives promise of being particularly
valuable as a combination method with TDS
or CVS.
CODING AND DECODING BY TANDEM
COMBINATIONS OF PRIVACY METHODS
Polarity Reversal [PR] and Time Delay
[TDS] in Combination. The PR and TDS com-
i 7
? t.._.? , A
?, ..,
-- -0--..-
'
--"--: -,". -- ---^-4-A
?--
- --
.?-,..., --4 -.:7"-----72:::,-- -.F.-'' .....--..., ?, _ ,j. ..,-.:?,-;;?,---_,,
-41
?...-v.r...
^ - .- ..., . ? ..:-. re.... ----, - ^4.---,..-7:?. ?'-'-'7'.." re*,,,--? *`
t=r,..r.....?:.V1."' ? .,..'';.t.,:;"Z 4-
i. i,- ,;*.;;6,r.
...le ..?. ', . '. ._,.1 . 7'...r;:-_'' tr?-.2:4:e...-V1442.
-
?;,,,,.. -'
6..7.. ? ..__.._,,_-? ,
- ,..,.._ . * ;?*. **._,*?.41,-21."' k::`,,?'"'- "' -.L.:,,i..7.1
??.,.,.----1,7.--..--. 7-i -...--'"- - - '7:74' ' - - -. - ? '. -. -.1-'1,- 4t. ''''-, 7 P - ' . ,7.--* - ' - ? - . 7 ' ?
-; '
-3....:.**....te . .. "-^'4 .4*' - , J.^ . '4 et
M Jr
'........... ......-*?---- ? '',.:A.,-...*7*-- '5''' ,.".L...... r
........'.=-Z, -*--',"-. -,... '--,* '
t'?4"
?^-
Art-Efti:KNEA cam
FIGURE 9. Scrambling by
drum speed.
Transposition of Frequency Bands [A] and
Time Delay [TDS] in Combination. The A a.nd
TDS combination is illustrated by Figure 13,
made by using the standard A-3 Western Elec-
tric speech privacy system with the BTL Model
B TDS system. The right section shows normal
recording without A or TDS; the middle sec-
tion is the coded picture utilizing the coders of
A and TDS in combination; and the left section
is the decoded copy resulting from passing the
coded signal through the tandem decoders.
Reference to the coded middle section shows
that a single system having no privacy by itself
may be raised, by combination with another
continuously varying
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
FACSIMILE PRIVACY
113
single system, far above that of either compo-
nent system used singly.
Variable Speed [VS] and Time Delay [TDS]
in Combination. The TDS and VS combination
is illustrated by Figure 14. The single methods
used in this combination are D-Specification
TDS and a step-by-step variable speed VS at
the rate of ?-1/2 per cent. A remarkable in-
crease in visual scrambling is attained com-
pared to using the single method alone.
Frequency Multiplication [B] in Combina-
tion with Variable Speed [VS]. Figure 15 is
the result obtained with the B and VS combi-
and 'varying the drum speeds in synchronism;
and the left section, the uncoded subject copy.
The B and VS combination is less attractive
from an application viewpoint due to the rela-
tive ease with which B may be decoded by
a narrow-band filter intercepter.
PRACTICAL APPLICATIONS OF
PRIVACY METHODS
Transposition of frequency bands [A] is
represented by the Western Electric Type 3-A
speech privacy equipment. It is fixed station
equipment of large bulk and weight.
FIGURE 10. Example of pretransmission scrambling.
nation. The Myopia Mark I coder and decoder
was used in applying the B coding. SCFM
recording with a subcarrier frequency of 1,380
to 1,750 cycles was applied. The coder was
operated on a:?3-sec continuous-code shift with
constant amplitude into the unit. VS was
applied at the rate of ?? per cent speed
deviation.
The right portion of this illustration is the
coded subject; the middle section, decoded by
passing the coded signal through the decoder
The method depends upon fixed interchange
of frequency bands. There is. no automatic
synchronization or phasing problem. The
method has been shown to give no privacy
when applied to facsimile transmission.
Frequency multiplication [B] is represented
by Model RCAL-1" weighing 32 lb and having
a volume of approximately 3/4 cu ft. The code
is set up on numbered dials, with provision
for continuous change by means of a clock
mechanism.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
114
FACSIMILE PRIVACY SYSTEMS
FIGURE 11. Pretransmission coding by single shredding (left) and by double shredding process (right).
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11: CIA-RDP10T02279R000100180001-2
FACSIMILE PRIVACY
115
Phasing is obtained by starting the coding
dials at a given setting and time. Synchroniza-
tion between coding waves at sending and re-
ceiving points is obtained from reference pulses
transmitted over the circuit. These will require
precise frequency standardization.
_
code remains fixed until the cards are changed.
The receiving decoder is synchronized on a
start-stop basis with the commutator of the
sending coder by a regularly repeated pulse
signal. The code is repeated every 3/4 sec so
that no long time phasing is needed. The coder
FIGURE 12. Illustration of use of TDS
It is probable that multipath propagation will
destroy both the coded wave and the synchro-
nizing signal.
The unit as it stands can be used inde-
pendently for voice or facsimile transmission.
and polarity reversal in combination.
and decoder do not have to be started simul-
taneously. It is computed that the precision of
synchronization required is the same as now
obtains on facsimile apparatus. Any skew in
the received copy may be corrected by read-
FIGURE 13. Effect of using frequency band transposition [Al
security offered by either method.
Time delay system [TDS] is represented by
the BTL D-Specification model weighing 25 lb,
and having a volume of approximately 3/4 cu ft.
Codes are set up by inserting two perforated
cards in their appropriate boxes. A particular
and TDS in tandem, thus increasing
justment of the frequency standards between
pictures.
Distortion of the signal by multipath should
be the same on TDS as on straight uncoded
transmission. At the time of the project, D-Spec
SECRET
neclassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
!;
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
116 FACSIMILE PRIVACY SYSTEMS
TDS was available privacy equipment, an
independent unit, equally applicable to voice
or facsimile.
? Variable speed [VS] and continuously vari-
able speed [CVS] consists of deviating the
drum speed with respect to normal speed at a
be exactly the same at scanner and recorder.
This requires that the code sequence at the
two points should be started with an estimated
error of not more than 0.01 sec.
Radio circuit distortions will have no more
effect on VS and CVS than on normal facsimile.
FIGURE 19. Coded and decoded copy produced by combining VS and TDS.
coded rate, either in fixed steps or continuously
to a maximum of ?-15 per cent. The apparatus
added to a normal facsimile station will depend
upon the type of drum drive circuit being used.
It may, or may not, exceed in volume the 3/4
cu ft already mentioned for other methods.
They may, however, make it difficult to send
the timing signal at the start of each message
with the precision necessary to start the codes
in phase.
A special design of facsimile machine ap-
pears to be the most satisfactory solution for
FIGURE 15. Use of frequency multiplication (B) and variable speed (VS) in tandem. Right portion, coded
copy; middle portion, decoded copy.
A basic synchronization accuracy of 1 part
in 100,000 will be required. A differential be-
tween this standard and the scanner and
recorder must be taken to provide the required
percentages, slow Or fast. The critical condition
is that the sequence of speed variations must
the VS and CVS method. It is consequently not
an immediately available privacy system.
Pretransmission method [PT] is indicated
to be most useful as one of the basic methods
of a tandem combination. The application
greatly simplifies tandem transmission since
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
PRETRANSMISSIONTACSIMILE?PRIVATY?
111
the simultaneous use of more than one set of
coders and decoders is not required. The result-
ing privacy of TDS or VS transmission may
thereby be greatly enhanced.
The A-3 graphic method of Faximile, Inc.,
(Project 13.3-97)36 and the double shredding
method suggested by the Eastman Kodak Com-
pany are promising applications of the PT
method.
SUMMARY OF PRIVACY METHODS
Several kinds of coding units have been
tested which could be built in a volume of 1
Cu ft and in mechanical form suitable for
military service.
Synchronizing and phasing will be critical
problems in radio transmission.
The B system and combinations will not
function over radio circuits with severe multi-
path distortion.
Methods TDS, PR, and VS will not be seri-
ously affected by multipath distortion.
Methods A, B, and TDS can be used
ultimately for facsimile and voice coding.
Methods PR and VS are special to facsimile
and would be of questionable effectiveness for
voice coding.
CONCLUSIONS
In general, privacy systems which are useful
for speech may or may not be suitable for
facsimile. This is because the original facsimile
signal has a regular periodicity which tends to
reveal the code that has been applied.
Of the three existing speech privacy systems,
A-3 gives no privacy for facsimile transmission
whereas TDS and Myopia Mark I are reason-
ably effective.
The privacy of several variations and com-
binations of the six basic systems has been
estimated in terms of the time required to
decode them and will be found in the final
report37 together with proposed methods of
decoding and lists of the necessary apparatus.
Much greater privacy is obtained with a
tandem combination of two systems than with
either system alone.
The best single system so far evaluated (VS)
requires three hours for 80 per cent decoding
of the picture. It is continuously variable drum
speed wherein variations of not less than ?-5
per cent occur within each scanning line. ?
The most effective tandem combination of
two systems will apparently require 10 hours
to decode 80 per cent of the picture. This is
the combination of continuously variable drum
speed and D-Specification TDS. Application of
CVS with the continuously varying code TDS
(C-50) should still further increase the time
for decoding. This combination merits investi-
gation.4
Pretransmission systems have been proposed
during this survey, and have been evaluated.
Such systems include coding of the subject
copy prior to transmission by normal facsimile
means, and decoding of the record copy after
it has been received. In general, these methods
can effectively code the subject copy but there
appear to be technical difficulties in the decod-
ing process.
6.3 PRETRANSMISSION FACSIMILE
PRIVACY
Project 13.3-97 comprised an investigation
of the feasibility of automatically enciphering
and deciphering graphic material such as maps,
photographs, drawings, etc., by strobotronic
photography. The material is scrambled by
transposing elementary sections of the graphic
original before transmission.
BASIC METHOD UTILIZED
Basically the pretransmission scrambling
method investigated under this project is
equivalent to cutting the copy into vertical
strips which are transposed and remounted so
that although the information within the strips
is in proper order vertically, the picture is
mutilated horizontally. After remounting the
vertical strips, the resulting copy is cut into
strips horizontally and after transposition of
the elements, the copy is again assembled. Now,
of course, the original information is hidden
to a degree depending upon the smallness of
the individual elements and the manner in
which they are scrambled. Although the fea-
ture was not incorporated in the model pro-
duced under the instant contract, some of the
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
118
FACSIMILE PRIVACY SYSTEMS
elements might be upside down as well as out
of Order; it is also conceivable that some' of
the elements might be reversed.
The final product of the several transposi-
tions can be sent to the receiver by courier, by
mail, or by wire or radio facsimile or photo-
graphic transmission methods.
THE ACTUAL MECHANISM
In the machine developed and the model
produced, the following sequence of events
occurs. The copy to be scrambled is wrapped
upon a cylindrical drum and rotated by a
motor at a speed which is not critical. A single
strip of this copy is illuminated and its image
is projected upon a similarly mounted sheet
of photographic paper which rotates, not con-
tinuously, but in steps.
After the first portion or "frame" of the
photographic paper is exposed, the drum carry-
ing the sensitized paper moves one step bring-
ing another frame of the paper into readiness
for the second exposure. In the meantime the
subject has moved to a new location, according
to a key or code, and at the proper time this
new portion of the original is photographed
upon the sensitive paper. When all portions of
the subject strip have been photographed, a
new strip of the copy, adjacent to or otherwise
with respect to the first strip, would be photo-
graphed. In this manner all the copy would
be impressed upon the photographic paper but
with the elements out of order.
After development, the negative image with
vertical elements transposed would be used as
an original and a new print made, this time
with the scrambling performed in a direction
at right angles to the first encipherment. The
second print would be a positive like the
original material but with all elements out of
order.
Increased privacy could be secured by per-
forming another mutilation this time encipher-
ing the transposed copy which is placed on the
drum displaced, say, a half-strip in width.
The deciphering is electrically and mechani-
cally the inverse of the operation which
scrambles the copy. According to the proper
key, portions are photographed upon a sheet
of paper which is developed to a negative image.
As many copying operations must take place
as occurred in scrambling the material in the
first place and, of course, the proper key must
be used.
THEORY OF OPERATION
Optically it is possible to "stop" the motion
of a rapidly moving object at various points in
its line of travel if it is illuminated by flashes
of light of such short durati.on that for the
per4od of the flash no appreciable movement
of the object occurs.
In the case of the graphic privacy system
developed under this project, the moving object
is the message to be scrambled. The flashes of
light are furnished by the discharge of elec-
trically charged capacitors through gas-filled
strobotron tubes. Photographic means are used
to record the results.
The original material is scanned through
an aperture and by its means and that of a lens
system, 1/2x2-in. sections of the copy are photo-
graphed on the sensitive paper rotated by the
drum on which it is mounted.
Naturally, the flashes of light must be
properly timed in accordance with the relative
positions of the two drums and must have the
proper intensity. Timing is effected by elec-
tronic circuits described in the final report on
the project.
Sixteen exposures result in transferring a
column 2 in. wide by 8 in. high to the photo-
graphic paper in coded random 1/2x2-in. sec-
tions. Thereupon the optical assembly is moved
axially and another 2-in, strip is photographed.
Naturally, the consecutive sections of the
material are not photographed in consecutive
position on the sensitive paper, but in some
other order determined by the key.
The duration of the flash is about 40 Rsec
so that the continuously rotated copy is, in
effect, stationary during the photographic
exposure. The overall effect of two cycles of
scrambling is to produce a new positive picture
made up of 1/2-in, squares completely out of
place with respect to each other and with
respect to the original matter.
MECHANICAL DETAILS
The apparatus constructed under the project
SECRET
Declassified and Approved For Release 2013/12/11: CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
PRETRANSMISSION FACSIMILE PRIVACY
119
was composed of the two drums approximately
4 in. in diameter and 9 in. long, one being
rotated at approximately 140 rpm by an electric
motor through a 25 to 1 reduction gear and
the other rotated in steps by a plunger-type
solenoid magnet through a ratchet and pawl
mechanism. The optical assembly was composed
of two strobotron tubes (Sylvania R-4215) and
a projection lens and moved along guide rails
between and parallel to the axis of the drums.
In the model, coding was accomplished by a
switchboard composed of 64 flexible electric
cords with phone tip plugs and a like number
of tip jacks. Eastman Kodak Aero Enlarging
Paper (mapping paper) single weight, No. 2,
was used since this material is specially treated
to minimize dimensional changes in processing.
The copy drum steps once for every two
revolutions of the subject drum. During the
alternate revolutions, switching and other
mechanical actions take place. In deciphering,
the enciphered copy is placed upon the stepping
drum, the optical assembly is reversed so that
the same key may be used to photographically
rearrange the elements in their original order.
CONCLUSIONS
Operating tests of the Model GPM-X1 ma-
chine proved the validity of the principles
involved and indicated: (1) that in a new model
much higher drum speeds could be employed,
(2) titat it would be advisable to complete the
encipherment in a single photographic opera-
tion to avoid cumulative errors, (3) that it
would be feasible to build a machine that would
automatically encipher a clear message* in a
two-dimensional randomized copy consisting of
squares at least as small as 0.2 in. on a side,
and (4) that a newly proposed plan for chang-
ing keys should be utilized. Such a new machine
would provide a substantial term of privacy
for graphic copy.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Chapter 7
MISCELLANEOUS PROJECTS
THE FINAL TWO PROJECTS considered in this
volume are related in a general way to the
subject matter that has gone before. Under
Project C-52,a Division 13 developed a crypto-
graphic rotor which had certain advantages
over the similar rotor in use at the time.
Project C-7P was set up to study the radio
transmissions of German submarines in an
endeavor to determine if frequency-modulation
were being used in addition to the normal
amplitude modulation. Methods were developed
for recording and studying these signals, the
recording apparatus being somewhat similar to
the schemes described in the earlier portions
of this volume.
7'1 ROTOR FOR CRYPTOGRAPHIC USE
Project C-52 was concerned with the design,
development, and test of a set of rotors for
multiple transposition coding when used on a
suitable printer typewriter. Decoding was
accomplished by use of a reversing switch.
STATE OF THE ART
At the time this project was started the
Signal Corps had in use a typewriter type of
coding and decoding machine which would
automatically perform a quadruple transposi-
tion ciphering each time a key was struck.
This was accomplished by means of four disks
(actually rotary switches), each disk having
input and output terminals for each letter of
the alphabet. The interconnecting wires be-
tween these terminals gave the particular
transposition for that disk. Four such disks
were used in series. After each letter was
coded, one or more of these disks was rotated
to change the code for the next letter. The
machines in use by the Signal Corps had two
disadvantages. First, it was quite difficult to
a Project C-52, Contract OEMsr-542, Fournier In-
stitute.
b Project C-71, Contract OEMsr-880, Western Elec-
tric Company, Inc.
120
rotate the _disks since heavy pressures had to
be kept on the contact fingers in order that
good electrical connections could be made.
Second, the construction of the contacts in the
disks was such that many troubles were en-
coui4ered with arcing and tracking across the
insulating sectors between contact points.
These two difficulties prevented the machines
from being used more widely.
OBJECT OF THE INVESTIGATION
The widespread use of codes and ciphers by
the military forces in time of war makes it
essential that the process of coding and decod-
ing messages be made as simple as possible.
This is usually accomplished by either 'Mechani-
cal or electrical devices or a combination of
both. An ideal solution of this problem is a
machine with a standard typewriter keyboard
upon which a message may be written in plain
language resulting in a printed version of this
message being coded automatically by the
machine and delivered instantaneously. The so-
called transposition cipher is most commonly
used, in which the letter "a" for instance, when
struck on the keyboard, causes some other letter
to replace it in the coded message. For purposes
of additional security, it is arranged so that
the next time the letter "a" is used a different
letter represents it than that one which was
originally used. A wide variety of multiple
transpositions may be used to obtain the neces-
sary security.
Such a transposition can be obtained by
having an electrical connection made when a
key on the keyboard is depressed. This electrical
connection is made to a contact finger which
touches one of twenty-six contacts on an insu-
lated rotary disk. On the other side of this
disk there are also twenty-six contacts, one for
each letter of the alphabet. An electrical con-
nection is permanently made by wires from
contacts on one side of the disk to contacts on
the other side in a predetermined manner to
give the alphabetical transposition which is
desired. Twenty-six outgoing contacts bear on
SECRET
Declassified and Approved For Release 2013/12/11: CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
to
at
e.
le
n-
s.
BS
it
3-
e.
a
-cl
in
is
ie
o-
ly
?s
Tt
ts
le
s-
)3.7
a
al
is
1-
)11
Lo
is
in
ROTOR FOR CRYPTOGRAPHIC USE
121
each of the rotor contacts and are connected
to a printer circuit which causes a letter to be
printed corresponding to the contacts through
which the electrical circuit was made. Such an
arrangement with one rotor disk gives a single
alphabetical transposition. A more secure code
can be obtained if several transposition disks
are used in series, and are rotated in some
predetermined fashion to make the breaking
of the coded messages even more difficult.
sary to use a rather strong spring on the con-
tact fingers. This made it very difficult to turn
the rotors, an operation which was necessary
for each new letter which was transmitted.
Since these devices were to be used in advanced
locations where electric power was not avail-
able, it was not possible always to provide some
strong solenoid action to force this rotation
against the necessary friction. To accomplish
tis by a straight mechanical linkage to the
FIGURE 1. General view of testing equipment for cryptographic rotor, rotors in place ready for use.
It is also possible to design an electrical
reversing mechanism which enables the same
machine to be used in decoding messages. The
coded message is transcribed on the keyboard
as it was originally received, and the plain test
version appears automatically decoded on the
printer tape from the machine.
Machines of this description were in use by
the Signal Corps at the time of the project,
but two difficulties had been encountered. The
first difficulty was that to make a reliable
contact with the existing designs, it was neces-
keyboard would have necessitated a very special
keyboard with a long distance of travel for
each key, and would have required the use of
considerable force when the key was struck.
A second difficulty was that the existing de-
signs used a bakelite disk in which were molded
brass inserts which were turned until they were
flush with the bakelite surface. As the disks
were rotated after each letter, the contact
finger rubbed across the bakelite in its travel
from one rotor contact to the adjacent one. This
resulted in a track being made across the bake-
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
122
MISCELLANEOUS PROJECTS
lite surface, and ultimately a low-resistance
path was built up, which led to electrical break-
down and sparking across this path. When this
happened the rotor had to be discarded and a
new one substituted.
? The object of ?this investigation was to
develop a rotor which could be used in such a
device without the difficulties of voltage break-
down between the contacts and excessive
design would be satisfactory. Two actual rotors
were designed and the second which was
smaller and less complex than the first was
actually constructed and submitted to the
Signal Corps. Most of the time and effort were
spent on test equipment. After rotors and test
equipment had been designed and constructed,
life tests were run and demonstrations were
made.
FIGURE 2. Back view showing details of eight rotor disks.
mechanical friction during rotation. It was
necessary that any proposed rotor be readily
manufacturable, and if possible, it was desired
to make the connection between the contacts
on each side of the rotor sufficiently flexible so
that they could be changed readily in the field.
PROCEDURE
There were two principal problems involved
in attacking the project, (1) to design and
construct the rotors and contact and (2) to
devise proper test equipment to insure that the
The coding disks and rotors constructed were
quite easy to rotate while still giving reliable
electrical contact as proved by the life tests.
Substantially complete freedom from leakage
and tracking between terminals and contacts
was achieved. The final model should be readily
and easily manufactured in large quantities.
So far as is known no use was made of the
designs or models.
The final report38 on the project gives com-
plete mechanical description including working
drawings.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
RADIO RECORDING
123
7.2
RADIO RECORDING
Project C-71 comprised an investigation of
the possible existence of frequency modulation
in certain telegraph transmissions and develop-
ment of means for recording the transmissions.
SUMMARY OF THE PROJECT
The work under this project was carried out
at the request of the Navy Department as a
means of identifying German naval vessels by
the characteristics of their telegraph trans-
mitters. Such identification, by recordings,
would supplement the direction-finder net set
up to watch for enemy signals.
Under the project more than a thousand
transmissions were recorded by means of
apparatus developed to detect the frequency-
modulation in the transmissions and were
transmitted to the Navy.
ale final report39 of the project gives the
means used and the results obtained.
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
BIBLIOGRAPHY
Numbers such as Div 13-302-M3 indicate that the document listed has been microfilmed and that its
title appears in the microfilm index printed in a separate volume. For access to the index volume and to
the microfilm, consult the Army or Navy agency listed on the reverse of the half-title page.
1. Speech Privacy Systems, Interception, Diagnosis,
Decoding, Evaluation, OSRD 4573A, Final Report
Project C-43, W. Koenig and C. H. G. Gray, BTL,
Oct. 12, 1944. Div. 13-302-M3
2. A Coding Arrangement for C-50 A-3 Privacy Sys-
tem, J. M. Fraser, OSRD 4573B, Preliminary Re-
port 21 of Part II, Final Report Project C-43,
BTL, July 31, 1943. Div. 13-302-M4
3. Methods for the Automatic Scrambling of Speech,
G. Guanella, OSRD 4573B, Preliminary Report 5
of Part II, Final Report Project C-43, December
1941. Div. 13-302-M4
4. Methods for the Automatic Scrambling of Speech,
W. Koenig and P. W. Blye, OSRD 4573B, Pre-
liminary Report 5 of Part II, Final Report
Project C-43, BTL, December 1941.
Div. 13-302-M4
5. Analysis of Brown Boveri Two-Dimensional Speech
Scrambling System, W. Koenig, OSRD 4573B,
Preliminary Report 9 of Part II, Final Report
Project C-43, BTL, Sept. 25, 1942. Div. 13-302-M4
6. "The Carrier Nature of Speech," Homer Dudley,
Bell System Technical Journal, Vol. 19, October
1940, p. 495.
7. "Remaking Speech," Homer Dudley, Journal
Acoustical Society of America, Vol. 11, October
1939, p. 169.
8. Speech Privacy Development, R. K. Potter,
OSRD 201, Final Report Project C-1, BTL, June
4, 1941. Div. 13-301.3-M2
9. Privacy Considerations in Connection with Present
and Future TDS Unit Designs and Recommenda-
tions for New Models of TDS Privacy Equipment,
R. K. Potter, OSRD 196, Final Report Project
C-IA, BTL, May 20, 1941. Div. 13-301.3-M1
10. Continuously Coded TDS, Speech Privacy Equip-
ment, Eugene B. Mechling, OSRD 2068, Final Re-
port Project C-50, BTL, Nov. 1, 1943.
Div. 13-301.31-M1
11. Frequency Time Division Speech Privacy System,
L. Schott, OSRD 1725, Final Report Project C-66,
BTL, May 29, 1943. Div. 13-301.3-M4
12. Code Changing Attachment for TDS Speech
PriVacy Units, C. W. Carter, OSRD 1541, Final
Report Project C-65, BTL, Apr. 30, 1943.
Div. 13-301.3-M3
13. Telegraphy Applied to TDS Speech Privacy Sys-
tems, C. W. Carter, OSRD 1047, Final Report
Project C-55, BTL, Oct. 3, 1942. Div. 13-304.3-M1
14. RCA Speech Secrecy Research, Project C-54, RCA
Laboratories.
Part I. Proposed Portable Speech Privacy Unit
with High Security, A. V. Bedford, OSRD 1882,
Feb. 2, 1943. Div. 13-301.1-M1
Part II. Description of Speech Privacy Unit
Model RCAL-1 and Related Information, A. V.
Bedford, OSRD 3107, Oct. 30, 1943.
Div. 13-301.1-M2
Part III. Modification of Speech Privacy Units,
4adio Tests and Conclusions, A. V. Bedford,
OSRD 3395, Feb. 25, 1944. Div. 13-301.1-M3
15. Speech Secrecy System Development, 0. M. Dun-
ning, OSRD 207, Final Report Project C-15,
Hazeltine Service Corp., Oct. 28, 1941.
Div. 13-301.2-M1
16. Variable Band Shift Filter, D. F. Hoth, OSRD
4573B, Preliminary Report 11 of Part II, Final
Report Project C-43; BTL, Nov. 25, 1942.
Div. 13-302-M4
17. Large Spectrograms Made on Variable Area
Pattern Machine, W. Koenig, OSRD 4573B, Pre-
liminary Report 13 of Part II, Final Report
Project C-43, BTL, Jan. 13, 1943. Div. 13-302-M4
18. Method of Finding the Decoding Permutations in
Brown Boveri Two-Dimensional Scrambling Sys-
tem, W. Koenig, OSRD 4573B, Preliminary Report
10 of Part II, Final Report Project C-43, BTL,
Oct. 31, 1942. Div. 13-302-M4
19. Mechanical and Numerical Aids for Cracking Re-
peated Code TDS, A. D. Fowler and W. Koenig,
OSRD 4573B, Preliminary Report 14 of Part II,
Final Report Project C-43, BTL, Jan. 25, 1943.
Div. 13-302-M4
20. Experiments in Cracking Two-Dimensional
Scrambles, W. Koenig, OSRD 4573B, Preliminary
Report 22 of Part II, Final Report Project C-43,
BTL, July 23, 1943. Div. 13-302-M4
21. Evaluation of the Privacy Afforded. for Non-
Repeated Code TDS Systems (C-50 TDS), W.
Koenig, OSRD 4573B, Preliminary Report 26 of
Part II, Final Report Project C-43, BTL, Nov.
30, 1943. Div. 13-302-M4
22. Variable Area Speech Patterns, W. Koenig, OSRD
4573B, Preliminary Report 1 of Part II, Final
Report Project C-43, BTL, May 21, 1942.
Div. 13-302-M4
23. Variable Area Pattern Machine, D. F. Hoth, OSRD
4573B, Preliminary Report 7 of Part II, Final
Report Project C-43, BTL, Sept. 14, 1942.
Div. 13-302-M4
24. Playback Machine for Variable Area Speech
Patterns, W. Koenig, OSRD 4573B, Preliminary
Report 12 of Part II, Final Report Project C-43,
BTL, Jan. 6, 1943. Div. 13-302-M4
25. Privacy Evaluation of Repeated Code TDS and
A-3 Privacy Equipment in Tandem, W. Koenig,
SECRET 125
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
126
BIBLIOGRAPHY
OSRD 4573B, Preliminary Report 19 of Part II, 34.
Final Report Project C-43, BTL, May 27, 1943.
Div. 13-302-M4
26. Use of Monotone or Whispered Speech on Privacy 35.
Systems, W. Koenig, OSRD 4573B, Preliminary
Report 16 of Part II, Final Report Project C-43,
BTL, Feb. 25, 1943. Div. 13-302-M4 36.
27. Evaluation of the Security Afforded by the RCA-
Bedford Speech Privacy System, W. Koenig, OSRD
4573B, Preliminary Report 18 of Part II, Final
Report Project C-43, BTL, May 26, 1943.
Div. 13-302-M4 37.
28. Decoding Equipment for TDS,0Split Band, and
Two-Dimensional Speech Scrambling System, W.
Koenig, OSRD 4573B, Preliminary Report 15 of
Part II, Final Report Project C-43, BTL, Jan. 25, 38.
1943. Div. 13-302-M4
29. Playback Devices for Spectrograms, L. Y. Lacy,
OSRD 4573B, Preliminary Report 17 of Part II,
Final Report Project C-43, BTL, Mar. 30, 1943. 39.
Div. 13-302-M4
30. Analysis of Generalized TDS Codes, A. D. Fowler, 40.
OSRD 4573B, Preliminary Report 3 of Part II,
Final Report Project C-43, BTL, June 16, 1942.
Div. 13-302-M4
31. Analysis of Self-Converse TDS Codes, A. D. 41.
Fowler, OSRD 4573B, Preliminary Report 6 of
Part II, Final Report Project C-43, BTL, Aug.
19, 1942. Div. 13-302-M4
32. TDS Field Decoding Test at Camp Coles, W. 42.
Koenig and A. E. Ruppel, OSRD 4573B, Pre-
liminary Report 24 of Part II, Final Report
Project C-43, BTL, Oct. 19, 1943. Div. 13-302-M4 43.
33. Part II, Final Report on Project C-43, containing
all the preliminary reports, most of which are in-
cluded in this bibliography. Those reports not so
included are: 44.
No. 2. Results of Experimental Intercept Work
at Holmdel, New Jersey and Point Reyes, Cali-
fornia, P. W. Blye and L. Y. Lacy, OSRD 4573B, 45.
BTL, May 29, 1942. Div. 13-302-M4
No. 4 Sonovox Secrecy System, L. Y. Lacy, OSRD
4573B, BTL, July 2, 1942. Div. 13-302-M4
No. 8. Speech Scrambling System Proposed by
Frank D. Lewis, W. Koenig and P. W. Blye, OSRD 46.
4573B, BTL, Aug. 29, 1942. Div. 13-302-M4
No. 20. Speech Privacy and Synchronizing Sys-
tem Devised by Captain Henry P. Hutchinson,
D. 0. Slater, OSRD 4573B, BTL, July 31, 1943.
Div. 13-302-M4 47.
No. 23. Intercept Work at Point Reyes, California,
H. Kahl, OSRD 4573B, BTL, July 26, 1944.
Div. 13-302-M4
No. 25. Interception of Enemy Radiotelephone 48.
Communications Employing Privacy Systems;
Problems, Procedures and Tools, W. Koenig, OSRD
4573B, BTL, Oct. 12, 1943. Div. 13-302-M4
SECRET
Speech Privacy Problems, A. D. Fowler, OSRD
5686, Final Report Project 13-106, BTL, Aug. 18,
1945. Div. 13-300-M1
Spectrographs for Field Decoding Work, C. H. G.
Gray, OSRD 3824, Final Report on Project 13.3-86,
BTL, May 31, 1944. Div. 13-302.1-M3
Facsimile Privacy, H. H. Beverage, OSRD 1881,
Final Report Project C-73, RCA Laboratories,
June 7, 1943. Div. 13-303-M1
Abridged Report, OSRD 2005, Oct. 15, 1943.
Div. 13-303-M2
Investigation of Pre-Transmission Facsimile
Privacy Methods, J. V. L. Hogan, OSRD 6346,
Final Report Project 13.3-97, Faximile, Inc., June
30, 1945. Div. 13-303-M3
Development of a Rotor for Cryptographic Use,
Harner Selvidge, OSRD 1607, Final Report Project
C-52, Fournier Institute, May 1, 1943.
Div. 13-304.1-M1
Radio Recording, A. M. Curtis, Final Report
Project C-71, BTL, June 15, 1943. Div. 13-304.2-M1
Control Circuits for the Sound Spectrograph, W.
Koenig, OSRD 4573B, Preliminary Report 27 of
Part II, Final Report Project C-43, BTL, Apr.
29, 1944. Div. 13-302-M4
The Sound Spectrograph, A Time-Frequency-
Intensity Analyzer, W. Koenig, Preliminary Re-
port Project C-43, BTL, Oct. 1, 1943.
Div. 13-302.1-M2
Operating Notes for Spectrograph Model Nos. 2
and 3, R. G. McCurdy, Project C-43, BTL, Jan.
30, 1943. Div. 13-302.1-M1
Notes on Description and Operation of Special
Speech Privacy Decoding Equipment Used at Point
Reyes, California, 0. M. Akey and H. Kahl,
Project C-43, BTL, June 30, 1944. Div. 13-302-M2
Speech Privacy Decoding, OSRD 386, Final Re-
port Project C-32, BTL, Jan. 31, 1942.
Div. 13-302-M1
Interception of Enemy Radiotelephone Cammuni-
cations Employing Privacy Systems; Problems,
Procedures and Tools, W. Koenig, OSRD 4573B,
Preliminary Report 25 of Part II, Final Report
Project C-43, BTL, Oct. 12, 1943. Div. 13-302-M4
Results of Experimental Intercept Work at
Holmdel, New Jersey and Point Reyes, California,
P. W. Blye and L. Y. Lacy, OSRD 4573B, Pre-
liminary Report 2 of Part II, Final Report Project
C-43, May 29, 1942. Div. 13-302-M4
Intercept Work at Point Reyes, California, H.
Kahl, OSRD 4573B, Preliminary Report 23 of
Part II, Final Report Project C-43, BTL, July 26,
1944. Div. 13-302-M4
Memo of Changes in Speech Privacy Units as Re-
leased to the Army by RCA Laboratories, A. V.
Bedford, Project C-54, RCA, Oct. 19, 1944.
Div. 13-301.1-M4
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
OSRD APPOINTEES
DIVISION 13
Chief
C. B. JOLLIFFE (December 1942 to May 1945)
HARADEN PRATT (May 1945 to May 1946)
J. L. ALLISON
C. F. DALZIEL
K. C. BLACK
0. E. BUCKLEY
J. H. DELLINGER
W. L. EVERITT
G. C. Floc
R. H. GEORGE
C. H. G. GRAY
A. HAZELTINE
Section 13.1
Section 13.2
Section 13.3
Section 13.4
Section 13.5
Section 13.6
4
Deputy Chief
K. C. BLACK
Technical Aides
Members
J. F. MCCLEAN
A. F. MURRAY
J. A. HUTCHINSoN
C. M. JANSKY
L. F. JONES
D. G. LITTLE
R. K. POTTER
H. PRATT
C. A. PRIEST
F. M. RYAN
Section Heads
Direction Finding
Radio Propagation Problems
Speech Secrecy
Special Communications Problems
Precipitation Static
Miscellaneous Projects
L. V. BERKNER
H. H. BEVERAGE
Consultants
R. K. POI'TER
L.
J.
R.
C.
D.
F. JONES
H. DELLINGER
K. POTTER
A. PRIEST
. PRATT
G. LITTLE
E. D. BLODGETT
D. G. LITTLE
Interference Reduction Committee
K. C. BLACK
H. D. DooLirrLE
R. G. FLUHARTY
A. HAZELTINE
0. W. TOWNER
SECRET
J. C. R. LICKLIDER
C. T. MORGAN
A. F. MURRAY
S. S. STEVENS
127
nedassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
CONTRACT NUMBERS, CONTRACTORS, AND SUBJECT OF CONTRACTS FOR DIVISION 13
Contract
Number
Name and Address of Contractor
Subject
Refer to
Chapters
NDCrc-125
NDCrc-196
NDCrc-139
OEMsr-230
OEMsr-435
OEMsr-490
OEMsr-542
OEMsr-592
OEM sr-628
OEMsr-782
OEMsr-795
OEMsr-880
OEMsr-837
OEMsr-1110
OEMsr-1202
0 EM sr-1440
Western Electric Company, Inc.
New York, N. Y.
Western Electric Company, Inc.
New York, N. Y.
Hazeltine Service Corporation
New York, N. Y.
Western Electric Company, Inc.
New York, N. Y.
Western Electric Company, Inc.
New York, N. Y.
Western Electric Company, Inc.
New York, N. Y.
Fournier Institute
Lemont, III.
Radio Corporation of America
Camden, N. J.
Western Electric Company, Inc.
New York, N. Y.
Western Electric Company, Inc.
New York, N. Y.
Western Electric Company, Inc.
New York, N. Y.
Western Electric Company, Inc.
New York, N. Y.
Radio Corporation of America
Camden, N. J.
Western Electric Company, Inc.
New York, N. Y.
Faximile, Inc.
New York, N. Y.
Western Electric Company, Inc.
New York, N. Y.
Speech secrecy system.
Continuation of speech secrecy system.
Hazeltine speech secrecy system.
Decoding speech Ves.
Continuation of decoding speech codes.
Improved speech secrecy.
Development of rotor for cryptographic use.
Speech secrecy research.
Telegraphy applied to TDS speech secrecy system.
Code changing attachment for TDS speech privacy
unit.
Frequency-time division speech analyzer.
Radio recording.
Facsimile secrecy.
Spectrographs for field decoding work.
Graphic privacy system.
Speech privacy problems.
2
2
3
4
1, 3, 4
2, 6
6
3, 5, 6
2
2
6
6
5
6
5
128
SECRET
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
SERVICE PROJECT NUMBERS
The projects listed below were transmitted to the Executive
Secretary, NDRC, from the War or Navy Department through
either the War Department Liaison Officer for NDRC or the
Office of Research and Inventions (formerly the Coordinator of
Research and Development), Navy Department.
Service
Project
Number
Subject
Refer to
Chapters
SC-12 Speech secrecy system. 2
SC-12 Continuation of speech secrecy system. 2
SC-12 Hazeltine speech secrecy system. 3
SC-12 Improved speech secrecy. 2, 6
SC-12 Speech secrecy research. 3, 5, 6
SC-12 Code changing attachment for TDS speech secrecy system. 2
SC-12 Frequency-time division speech analyzer. 2
SC-19 Telegraphy applied to TDS speech secrecy system. 2
SC-25 Development of rotor for cryptographic use. 6
SC-28 Decoding speech codes. 4
SC-28 Continuation of decoding speech codes. 1, 3, 4
SC-43 Facsimile secrecy. 6
NS-130 Radio recording. 6
NS-134 Graphic privacy system. 6
NS-349 Speech privacy problems. 5
SECRET
129
neclassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
INDEX
The subject indexes of all STR volumes are combined in a master index printed in a separate volume. For
access to the index volume consult the Army or Navy Agency listed on the reverse of the half-title page.
A2 speech scrambling system, 41
A3 facsimile privacy system, 107
A3 speech scrambling system, 21
A4 phase reversal system, speech
scrambling
cryptographic decoding, 57
noncryptographic decoding, 42
spectrograms, 88
A5 split-phase system, speech
scrambling, 41, 89
"Aided tracking" for interception
of speech privacy systems, 40
Amplitude representation for
spectrographic decoding of
speech, 64-65
Articulation tests for TDS speech
scrambling systems, 21
Audio frequency code-waves, RCA-
Bedford speech privacy sys-
tem, 25
Automatic code changing, 17-18
Automatic decoding of speech, 44-48
frequency band methods, 44
multiplication systems, 46
parallel-automatic method, 46
summary, 47
wobble band method, 45
Automatic trial systems for speech
decoding, 83
B1 facsimile privacy system, 107
B2 band displacement system,
speech scrambling, 44
B2 facsimile privacy system, 107
B3 facsimile privacy system, 107
B3 wobble band displacement sys-
tem, speech scrambling, 40
B4 double modulation system,
speech scrambling, 57
B4 facsimile privacy system, 107
B5 facsimile privacy system, 107
B6 facsimile privacy system, 107
B7 facsimile privacy system, 107
Band displacement speech scram-
bling systems
B2; 44
decoding, 42
Hazeltine, 33-34
spectrograms, 89
superposition, 43
Band splitting speech scrambling
systems
Dl; 41, 57
D2; 43, 57
description, 3-4, 73
Bedford type speech scrambling
systems
see RCA-Bedford speech privacy
system
Bell Telephone Laboratories
D-specification model, bacsimile
privacy system, 115
sound spectrograph, 35
TDS facsimile privacy systems,
107-108
British modulator type 2C speech
scrambling system, 100-101
British two-dimensional speech pri-
vacy system, 101, 103
Cl re-entrant inversion system,
speech scrambling, 48
C2 continuously varied re-entrant
displacement system, speech
scrambling, 89
C2 triple modulation system, speech
scrambling, 57
Channel mixing systems, speech
scrambling
Ll, L2, L3; 59, 89
multiple transmission paths, 11-
12
noncryptographic decoding, 43
Code waves, RCA-Bedford speech
privacy system, 28-31
Code-changing attachment for
portable TDS speech scram-
bling, 22-23
Code-changing method, RCA-Bed-
ford speech privacy system,
29
Coding systems
see Speech privacy systems
Compandor for RCA-Bedford
speech privacy system, 27, 29
Continuously coded TDS systems,
speech scrambling, 17-20
automatic code changing, 17-18
equipment, 19
frequency-band shift system, 20
number of sequences, 20
ten-element system, 18
Continuously varied re-entrant dis-
placement system, speech
scrambling, 88
"Cracking" speech scrambling sys-
tems
SECRET
frequency-TDS, 21
portable TDS, 14
RCA-Bedford, 32
time measurements, 84
Cryptographic decoding of speech,
48-61, 120-122
applications, 57-59
message, 59
oscillographic traces, 54
playback, 59-61
rectification, 56-57
"rotor," 120-122
spectrograms, 48-51
summary, 48
variable-area patterns, 52-54
visual, 54-56
CVS1 (continuously-variable), fac-
simile privacy system, 108,
112
Cyclic speech coding system, 85
D1 band splitting system, speech
scrambling, 41, 57
D2 band splitting system, speech
scrambling, 43, 57
Decoding methods, speech communi-
cation, 35-99, 104
automatic methods, 44-48, 83
cryptographic, 48-61
equipment, 104
evaluation, 84-86
history, 35-36
instantaneous speech patterns, 78
interception, 36-40
noncryptographic, 40-48
partial matching system, 76-78
sound spectrograph, 61-83
unknown signals, 86-89
Delay networks for speech privacy
system, 28-31
Directive antennas for intercepted
signals, 37
Double modulation speech scram-
bling system
automatic decoding, 45-46
B4; 57
spectrogram, 88
Double-sideband radio transmis-
sion, 38
Eastman Kodak Company, pre-
transmission facsimile pri-
vacy system, 112, 117
131
Declassified and Approved For Release 2013/12/11: CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
132
INDEX
Enemy privacy systems, captured
sets, 40
Equalizer unit for RCA-Bedford
speech privacy system, 29
Fl speech scrambling system, de-
layed subbands, 41
F2 TDS system, repeated code,
speech scrambling, 57
F3 TDS system, nonrepeated code,
speech scrambling, 57, 89
F4 speed variations system, speech
scrambling, 45, 58
Facsimile privacy systems, 105-119
see also under name of system
background, 105-106
CVS-1 system, 108, 111
frequency band
method, 106,
methods, 106-117
polarity reversal
108, 110-111
pretransmission, 107,
112, 116-119
RCAL-1 system, 113
TDS system, 106-108, 110, 115
variable speed system, 107, 108,
113, 116
Western Electric
107
Faximile, Inc., facsimile privacy
system, 117
Filters for spectrographs, 66-69
Fixed-code speech scrambling
tems, 100-101
Foreign language
speech decoding, 83
Frequency multiplication
simile privacy system
applications, 113
description, 106
Myopia Mark I; 109-110
nomenclature, 107-108
Frequency substitution
speech? scrambling
band-splitting, 3-4, 33-34, 73
double modulation, 2
single modulation, 1-2
time division multiplex
4-5
triple modulation, 2-3
Frequency-band shift speech scram-
bling systems, 7, 20, 44
Frequency-band transposition
facsimile privacy systems
Al, A2, A3; 107
applications, 113
description, 106
samples, 109
transposition
107, 109, 113
systems, 106,
108, 111-
system (A3),
records
in
sys-
for
fac-
systems,
(TDM),
in
Frequency-time division systems,
speech scrambling, 21-22
G1 tape plus modulation system,
speech scrambling, 58, 89
G2 tape plus modulation system,
speech scrambling, 58
G3 tape plus modulation system,
speech scrambling, 58
G4 tape plus modulation system,
speech scrambling, 58
G5 tape plus modulation ystem,
speech scrambling, 89
G6 tape plus modulation system,
speech scrambling, 89
Gabrilovitch, L. E., speech privacy
systems, 102
German vessel detection by tele-
graph transmitter identifica-
tion, 123
GPM-XI, pretransmission facsimile
privacy system, 119
Graphic copy, scrambling systems
see Facsimile privacy systems
Graphic patterns in spectrographic
speech decoding, 69-71
H1 wave form distortion system,
speech scrambling
automatic decoding, 46
cryptographic decoding, 58-59
spectrograms, 88
H2 wave form distortion system,
speech scrambling, 44, 59
H3 wave form distortion system,
/speech scrambling, 44, 59
Hazeltine band displacement sys-
tem, speech scrambling, 33-
34
High-power sig-nal interception, 36
High-security speech privacy sys-
tems, 32
Inflection effects in diagnosing pri-
vacy systems, 63
Interception of speech privacy sys-
tems, 36-40
decoding devices, 39-40
receiver sets, 37-38
recording methods, 39
signal quality, 36-37
types of radio systems, 36, 38
"Interlace," TDS speech scrambling
system, 6-7
Inversion systems, single modula-
tion speech scrambling, 1,
71, 73
J1 masking system, speech scram-
bling, 59
SECRET
J2 masking system, speech scram-
bling, 44
J3 tone sequence system, speech
scrambling, 41
K1 voeoder system,
bling, 59, 89
K2 vocoder system,
bling, 59, 89
K3 vocoder system,
bling, 59, 89
K4 vocoder system,
bling, 59, 89
speech
speech
speech
speech
scram-
scram-
scram-
scram-
L1 channel mixing system, speech
scrambling, 43, 59
L2 channel mixing system, speech
scrambling, 43, 59
L3 channel mixing system, speech
scrambling, 59, 89
Level compression in spectrographic
decoding of speech, 63-64
Level modulation systems, speech
scrambling, 42, 44
Limiter, speech decoding, 42
Linguaphone, speech decoding, 83
Low-power signal interception, 36
Magnetic tape speech recording
systems
see Time division speech scram-
bling (TDS) system
Masking systems, speech scram-
bling
J1; 59
J2; 44
summary, 9-10, 103
Military strategy, decoding speech,
85-86
Model RCAL-1, privacy unit, 29-31,
113
Modulation systems, speech scram-
bling, 1-3
double, 2
single, 1-2
triple, 2-3
Modulator Type 2C speech r.ecord-
ing, 100-101
Multiplication system (H1) speech
scrambling
automatic decoding, 46
cryptograph decoding, 58-59
spectrograms, 88
Multivibrator for RCA-Bedford
speech privacy system, 28
Musical effects in spectrographic
speech decoding, 68
"Myopia Mark I" for facsimile
privacy system, 107, 109
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
INDEX
133
Networks for code-waves of RCA-
Bedford speech privacy sys-
tem, 28, 31
New Zealand switched-band speech
privacy system, 101-102
Noncryptographic decoding of
speech, 40-48, 83
automatic, 44-48, 83
captured set, 40
compromise methods, 41-44
Nonrepeated code systems, speech
scrambling, 57, 84-85, 89
Nonscrambled speech, Hazeltine
band displacement system,
34
Oscillograms for cryptographic de-
coding of speech, 54
Parallel-automatic speech decoding
method, 46
Peak chopper for speech decoding,
42
Phase varied inverter-distorter
speech secrecy set, 102 :
Phase-reversal speech scrambling
system
cryptographic decoding, 57
noncryptographic decoding, 43
spectrograms, 88
Phonographic recordings for speech
scrambling systems, 103-104
Playback, cryptographic decoding
of speech, 59-61
Polarity reversal (PR) system,
facsimile privacy
description, 106, 108
samples, 110-111
Portable TDS systems speech
scrambling, 13-17/22-23
PR-1 facsimile privacy system, 108
PR-2 facsimile privacy system, 108
PR-3 facsimile privacy system, 108
Pretransmission facsimile privacy
system, 107, 108, 111-112
equipment, 118-119
evaluation, 119
method, 117-118
operation, 118-119
Privacy systems
see Facsimile privacy systems;
Speech privacy systems
"Private" coded facsimile subjects,
105
Radio recordings of telegraph
transmissions, 123
Radio transmission, interception
see Interception of speech pri-
vacy systems
RCA-Bedford speech privacy sys-
tem, 25-33, 103
code waves generated by delay
network, 28, 29-31
code-changing method, 29
compandor, 27
decoding, 32, 103
present status, 32-33
principles, 25
synchronization ?of delay net-
work, 29
wave multiplier, 25
RCAL-1 facsimile privasy system,
113
Receiving sets for code intercep-
tion, 37-38
Recording methods, speech scram-
bling systems, 39, 103
Rectification methods for speech de-
coding, 42, 56-57
Re-entrant inversion system, speech
scrambling, 48
Repeated code system, speech
scrambling, 57
Repeated code waves of RCA-Bed-
ford speech privacy system,
29-31
Rotating commutator for "inter-
lace" speech scrambling sys-
tem, 6
Rotors for cryptography, 120-122
Scanning filters for cryptographic
decoding of speech, 51-52,
66-69
Scrambling systems for speech
privacy
see Speech privacy systems
Secrecy systems
see Facsimile privacy systems;
Speech privacy systems
"Secret" coded facsimile subjects,
105
Self-decoding time division scram-
bling system, 6
Signal quality, interception meth-
ods, 36-37
Single modulation speech scram-
bling systems, 1-2
Sound spectrograms, 87-99
alternate inversion, 91
backwards, 98
channel-mixing, 99
fixed displacement, 91
modulation sidebands, 90
multiplication, 98
re-eiffrant inversion, 92
scrambled speech, 88
simple inversion, 90
SECRET
speed wobble, 97
split band scramble, 93, 94
subbands delayed, 95
TDS combined with split-band
scramble, 96
time and frequency measure-
ments, 88
time division multiplex, 94, 95
two dimensional scramble, 97
wobbled displacement, 92
Sound spectrograph, 61-83
amplitude, 64-65
applications, 66-83
description, 65-66
diagnosis, 88-89
history, 61
improvements, 64
level compression, 63-64
measurements, 87
operation, 62-63
Speech patterns for scrambling
systems, 71-74, 78
Speech privacy systems
see also under name of system
band displacements, 40, 42-44, 89
band-splitting, 3-4, 33-34, 73
British systems, 100-101, 103
channel-mixing, 11-12, 43, 59, 89
continuously coded, 17-20
diagnosis of unknown systems,
86-99
double modulation, 2, 45-46, 57,
89 ,
evaluation, 84-86
Hazeltine band displacement, 33-
34
interception, 36-40
inversion systems, 1, 48, 71, 73
masking, 9-10, 44, 59, 103
modulation systems, 1-3
multiplication systems, 42, 46, 88
phase reversal, 42, 57, 88
portable systems, 13-17, 22-23
RCA-Bedford, 25-33, 103
recording methods, 39, 103
repeated code, 57
single modulation, 1-2
speed variations, 45, 58
split phase, 41, 89
summary list, 47
switched band, 101-102
tape recording, 5-6, 13-24, 52-56,
74-78
time division multiplex (TDM),
4-5
time division scrambling (TDS),
5-6, 13-24, 52-56,34-78
tone sequence, 41
triple modulation, 2-3
two-dimensional, 101, 103
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
134
INDEX
unrepeated code, 57, 84-85, 89
vocoder, 10-11, 45, 54, 59, 89
wave for modification, 8, 58
Speed variations speech scrambling
system, 45, 5/8
Spill-over effects in cryptographic
decoding of speech, 51, 64
Split-band systems, speech scram-
bling
Dl; 41, 57
D2; 43, 57
description, 3-4, '73
Split-phase speech scrambling sys-
tem, 42, 88
Spread-band radio transmission, 38
Stepped displacement speech scram-
bling systems
B2; 45
decoding, 42
Hazeltine, 33-34
spectrograms, 89
superposition, 43
Subcarrier frequency modulation
systems, facsimile privacy,
108
Superposition for speech decoding,
43-44
"Suppressed carrier" radio trans-
mission, 38
Switched band speech privacy sys-
tem, 101-102
Synchronizing blanking circuits for
RCA-Bedford speech privacy
system, 29
Synthetic speech in Vocoder sys-
tems, 10-11
Tandem transmission in facsimile
privacy systems, 116, 117
Tape plus modulation systems,
speech scrambling, 58, 89
Tape recording systems
see Time diviyion speech scram-
bling system (TDS)
TDM (time division multiplex)
speech privacy system, 4-5
TDS facsimile privacy systems
see Time delay system (TDS),
facsimile privacy
TDS speech scrambling system
see Time division speech scram-
bling (TDS) system
Telegraphy transmission in speech
privacy systems, 23-24, 39
Ten-element TDS speech scram-
bling system, 18
Time delay system (TDS), fac-
simile privacy
applications, 113-116
description, 106
nomenclature, 107-108
samples, 110
Time division multiplex (TDM)
system, speech scrambling,
4-5
Time division
speech scrambling
(TDS) systems
code-changing attachment for
portable TDS, 22-23
continuously coded TDS, 17-20
decoding, 74-78
"interlace," 6-7
matching variable ? area pat-
terns, 52-54
portable TDS, 13-17, 22-23
summary, 5-6, 13
telegraphy, 23-24
visual methods, 54-56
Times Telephote Equipment, Inc.,
facsimile privacy methods,
111
Tone sequence system (J3) speech
scrambling, 41
Transposition coding of speech,
120-122
Transposition of frequency bands
in facsimile privacy sys-
tems, 107, 109
Triple modulation speech scram-
bling system, 2-3
"Twin-channel" radio transmission,
38
Two-channel speech privacy system,
2
SECRET
Two-dimensional speech privacy re-
cording system, 7, 100-101,
103
Unknown systems, diagnostic
methods, 86-89
illustrations, 88-89
introduction, 86-87
spectrograms, 87
Unscrambling methods, speech sys-
tems
see Decoding methods, speech
communication
Variable speed (VS) system, fac-
simile privacy
applications, 116-117
description, 107
nomenclature, 108
samples, 110-112
Variable-area patterns for crypto-
graphic decoding of speech,
52-54, 78
Visual methods, cryptographic de-
coding of speech, 54-56, 61
Vocoder speech scrambling sys-
tems, 10-11
automatic decoding, 45
description, 10-11
Ki, K2, K3 and K4 systems, 59
oscillographic traces, 54
spectrograms, 89
Voice scrambling systems
see Speech privacy systems
VS facsimile privacy systems
see Variable speed (VS) sys-
tem, facsimile privacy
Wave form modification systems,
speech scrambling, 8, 58
Wave form traces for decoding
TDS, 76
Wave multiplier for RCA-Bedford
speech privacy system, 25-26
Western Electric Company, A-3
facsimile privacy system, 107
Wobble band displacement system,
speech scrambling, 40, 45-46
Declassified and Approved For Release 2013/12/11 : CIA-RDP10T02279R000100180001-2
..1.4 ?
i..;!!!!-?
Declassified and Approved For Release 2013/12/11: dIA-RDP10T02279R06010.0180001-2 .
4101.121