PERFORMANCE OF THE LINCOLN F9C RADIOTELETYPE SYSTEM

• Document Type: CREST
• Collection: General CIA Records
• Document Number (FOIA) /ESDN (CREST): CIA-RDP78-03330A004100060024-0
• Release Decision: RIFPUB
• Original Classification: S
• Document Page Count: 25
• Document Creation Date: December 22, 2016
• Document Release Date: February 8, 2012
• Sequence Number: 24
• Publication Date: October 28, 1955
• Content Type: REPORT

Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
? ? ?
RADIOTELGTYPE SYSTEM
?
',1101 e loco]
I I I Ell !"
LINCOLN LABORATORY
mg]
 Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
1 1 Y I 1 1 ? t Y 1 1 1 1
1 t~ 1 1 t 111 ..i t 1 0 1 :. i 1
1 1 1 I= 1 1 I~ Ii JI t 1'
T 1 1 Y 1
1171
Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
 Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
coNHDENTAL
C 41333
This document consists of 24
pages. 1-!2
of 4 450 50 copies. ~f/a
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
LINCOLN LABORATORY
PERFORMANCE OF THE LINCOLN F9C
RADIOTELETYPE SYSTEM
P.E. Green, Jr.
R.S. Berg
C.W. Bergman
W.B. Smith
Technical Report No. 88 28 October 1955
The Lincoln F9C system is a NOMAC radioteletype system. It
transmits a 10-kcpsrnoise-like carrier through jamming for high-
frequency fixed-point circuits. In this report, a brief description 'of
the equipment and its operation leads to a discussion of the tests
on an experimental transcontinental circuit. After time-diversity
was incorporated, F9C performance gave an average anti-jamming
advantage of some 17 db over standard FSK at an error rate of two
wrong characters or less per line.
This document contains information affecting the national defense of
the United States within the meaning of the Espionage Laws, Title 18,
U.S.C. Sections 793 and 794. The transmission or the revelation of its
contents in any manner to an unauthorized person is prohibited by law.
LEXINGTON MASSACHUSETTS
CON
FIDENTIAL
Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
 Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
C N F i D EN11 A L
For several years, communication systems using noise-like carriers (NOMAC sys-
tems) have been under study at Lincoln Laboratory as having possible application against enemy
jamming 1, 2' 3 In July 1953, design and construction was begun on the F9C system, which em-
bodies the basic NOMAC idea in an experimental system for anti-jamming use on high-frequency,
point-to-point radioteletype circuits. Through the cooperation of the Signal Corps, facilities
were provided for field testing the completed system on a transcontinental link between Davis,
California and Deal, New Jersey. The equipment was completed and installed during July 1954,
and tests were run from 12 August to 1 October 1954. At the end of this series of tests, sev-
eral modifications were made. A second series of tests was then run from 1 February to 31 May
1955. As a result of these two series of tests on the experimental system, an operational pro-
totype, the F9C-A, is currently being built for the Signal Corps by the Sylvania Electronic De-
fense Laboratory at Mountain View, California.
This report includes an explanation of the principles of operation of the F9C system,
followed by a discussion of the test results and their significance. The description of system
operation given in Sec. II begins with the equipment as it existed at the beginning of the first
series of tests, and then takes up subsequent changes. Similarly, each of the test results (Sec.III)
is treated chronologically.
II. FUNCTIONAL DESCRIPTION OF THE F9C SYSTEM4
The Lincoln F9C system* is a stored-reference NOMAC** communication system.
In the F9C system, a teletype signal is modulated onto a noise-like carrier at the transmitter
and recovered at the receiver by crosscorrelating the arriving signal with a stored copy of the
noise. The system is called a "stored reference" (as distinguished from a "transmitted ref-
erence") system, since the reference noise-like signal is stored independently at transmitter
and receiver instead of being transmitted.t
For purposes of explanation, it is convenient to compare the F9C NOMAC system
with a conventional Frequency-Shift Keying (FSK) radioteletype system. The difference is por-
trayed in Fig.1, where it is seen that the binary MARK-SPACE teletype data are transmitted
for FSK as a sine wave whose frequency is one of two slightly different values, whereas, for the
F9C system, the sine wave is replaced by a noise.
The upper expressions at the bottom of Fig.1 give the signal-to-jamming power
ratio at the receiver output to the teletype printer for these two systems. It is this (S/J)0 that
specifies the frequency of errors in the resulting teletype copy. (S/J)i is the signal-to-jamming
*UNCLASSIFIED name.
**CONFIDENTIAL name. The word NOMAC stands for NOise Modulation And Correlation.
to transmitted-reference NOMAC equipment called the Lincoln P9D system was built and field-tested in 1952.
This type of system was abandoned in favor of the stored-reference type because of the greater resistance of the
latter to jamming. (See footnote on p.3.)
1
CO N FII E NT I A L
C lr` D C T
~.. CST? T
Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
 Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
SECRET
ratio at the receiver input, W is the effective bandwidth of the signal, and T is the symbol (or
baud) duration. These expressions assume that the enemy is using his jamming power in a sen-
sible way. In the case of FSK, this means that he will concentrate his power at the MARK or
SPACE frequencies. For a stored-reference NOMAC system, it turns out to be substantially
immaterial how he distributes his power in the signal passband; the output signal-to-jamming
ratio will still be TW times the input signal-to-jamming, whether he sends in the receiver band-
width W a sine wave, two sine waves, a noise, etc 5 The only way in which the jamming can be
more effective than this for a given power is for the jamming signal to have a waveform like that
of the noise-like signal. That is, the jamming signal must correlate positively with one or the
other of the possible genuine signals. If the legitimate and jamming signals have precisely the
same waveform, then the output signal-to-jamming ratio is no better than that of FSK. This is
made as difficult as possible by the complexity of the noise-like carrier. Then, if the enemy
wants to interrupt the communication by jamming (or insert a bogus message), he must somehow
duplicate the noise-like carrier waveform. If, in addition, the noise-like signals for MARK and
SPACE cannot be distinguished as such by the enemy, one has the additional advantage of cryp-
tographic security of the message.
us
(7) _ (J )i (mo)o TWJ,)i
FOR FLAT GAUSSIAN NOISE OF DENSITY No: FOR FLAT GAUSSIAN NOISE OF DENSITY No,
(N) = 2TS/No N - 2TS/No
No No
FREQUENCY-SHIFT KEY SYSTEM STORED-REFERENCE NOMAC SYSTEM
Fig. I . Signal spectra of stored-reference NOMAC system and FSK system.
In the F9C system, a bandwidth W of 10 kcps is used, and the symbol duration T is
22 msec (60 words per minute) so that TW = 220, or 23 db. Thus we might anticipate from the
expression in Fig. I that a 23-db advantage of F9C over FSK in the presence of jamming would
be observed. Actually, multipath in the propagating medium deteriorates the performance of
both systems. (The equations given in the figure do not take multipath into account.) As will
appear later in the discussion, this deterioration is greater for F9C than for FSK.
Figure 2 depicts the essential circuit features of the F9C system (the version used
in the first series of tests). At the transmitter, a bandpass noise is translated up in frequency
by either fM or fS through the action of the balanced mixer which is supplied with frequency fM
when a MARK is to be transmitted, and with the slightly different frequency fS for a SPACE.
The filter serves two purposes: first, to eliminate the unwanted sideband resulting from the
conversion, and second, to pass only the center of the frequency-shifted noise band, so that
there remains almost no tell-tale shift of the band edges due to the frequency-shift keying. This
shift, which is shown in Fig. 1, might otherwise allow the message to be discovered.
SECRET
Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
 Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
SECRET
- -- --------~
IDENTICAL
I NOISE-LIKE
SIGNALS
CONVERTER E
ANDPOWER HETERODYNE
AMPLIFIER RECEIVER I
At the receiver, the incoming signal is converted back down from the operating fre-
quency, and heterodyned in a mixer tube with a locally stored copy of the original noise. Since
the difference between the stored signal and that arriving from the transmitter is that the latter
has been frequency-shifted by either fM or fS, then conversely the mixer output must contain a
tone of the same frequency, either fM or fS. From here on, the operation of the receiver is sim-
ilar to that of an FSK system. A pair of narrow-band averaging filters, one tuned to fM and the
other to fS, is used to detect which is present; that is, whether a MARK or SPACE was sent at
the transmitter. The combined operation of heterodyning the received and reference signals and
filtering the difference-frequency tone is equivalent to crosscorrelating them. The comparison
device compares the envelopes of the two filter outputs and actuates the teletype printer accordingly.
BAND
NOISES ~--{ BALA M XERED SIF LTERD MIXER I I DIRCES
SOURCE
I SOURCE
I
fc=150 kcps fc =130 kcps
W =12.7kcps O-~ W=IOkcps
1
OSCILLATOR f OSCILLATOR
fm i fS
TELETYPE
DISTRIBUTOR
AVERAGING
FILTER
fm
fc=150kcps
W=IOkcps
19,908cps 20,092cps
COMPARISON
DEVICE
TELETYPE
PRINTER
Fig.2. Simplified block diagram of the Lincoln F9C system.
SYNC
CIRCUITS
It is clear that the operation of a NOMAC communication system depends on having
the noise-like carrier available at both transmitter and receiver. For a stored-reference sys-
tem, reproducible "noise-like" signals must be generated at transmitter and receiver and then
synchronized * This synchronization is necessary because the output signal-to-noise ratio drops
off seriously if the incoming and stored signals are mistimed by more than about the reciprocal
of the bandwidth (some 50 ?sec for the F9C system). This quantity is the width of the central
peak in the "correlation curve", the plot of output vs desynchronism between incoming and ref-
erence signals.
One obtains the anti-jamming capability described previously only at the expense of
other factors: the system requires simultaneously a large bandwidth allocation and a low data
rate, its efficiency against jamming depending on the ratio of just these two quantities. Further-
more, such a system needs considerable equipment complexity to deal with the two problems of
signal storage and synchronization, as will appear shortly.
*It is clear why transmitting the reference signal largely eliminates the anti -'lamming property. The enemy need
only transmit two coherent signals (e.g., two sine waves) - one in the signal band, and the other in the refer-
ence band. If the frequency difference is properly adjusted, the beat note generated is indistinguishable from
the legitimate one.
SECRET
Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
 Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
SECRET
I00-kcps
FREQUENCY
STANDARD
OUTPUT
128 125 121 117
7-STAGE COUNTER 7-STAGE COUNTER 7-STAGE COUNTER 7-STAGE COUNTER
10
'AND'
.AND'
GATES
10
'AXD'
GATES
'AND'
GATES
Fig.3(a). Generation of the noise-like signal as used in the first series of tests.
SECRET
BANDPASS
LIMITER
BA NDPASS
FILTER
IT]
IE
Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
 Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
SECRET
Because of these factors, the F9C system is envisioned to be held in reserve for
occasions when jamming is severe. If one is not being jammed, there are several conventional
systems that should perform as well, without such an expenditure of bandwidth, equipment, and
specially trained operators. For example, the bottom set of equations of Fig. 1 shows that FSK
and stored-reference NOMAC systems behave identically when the only interference is Gaussian
noise with a flat spectrum. Much non-man-made interference tends to fall in this category.
B. The Noise Source
The synchronization problem requires (1) that the source have an extremely stable
time scale, and (2) that this time scale be quickly readjustable. Therefore, a digital type of sig-
nal generator driven by a primary frequency standard seemed to be most suitable. The standard
provides the stability, and the digital circuitry provides ease of retiming, both being not quite
so convenient with most of the alternative storage schemes (magnetic tape, drums, cathode-ray
tube and storage-tube devices, etc.)
Block diagrams of the generators of the noise-like signals are given in Fig. 3; Fig.3(a)
shows that used in the first series of field tests, and Fig.3(b), that used in the second. In both
cases, a pseudo-random train of positive and negative pulses is derived from the 100-kcps out-
put of the frequency standard. The desired noise-like waveform is the bandpass signal resulting
from shock-exciting the bandpass filter with this pulse train, and then limiting.
100 - kc ps
FREQUENCY
STANDARD
63
6-STAGE COUNTER
64
6-STAGE COUNTER
59
6-STAGE COUNTER
BAND PASS
FILTER
BANDPASS
LIMITER
53
6-STAGE COUNTER
Fig.3(b). Generation of the noise-like signal as used in the second series of tests.
SECRET
Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
 Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
SECRET
In Fig.3(a), the exciting pulse train is derived as follows: Four seven-stage binary
counters are driven simultaneously at the 100-kcps rate. The repetition periods of these counters
are 1 28, 125, 121 and 117, respectively. The four counts being relatively prime to each other,
the over-all repetition period of a given condition of all four counters simultaneously is the
product of the counts times 10 ?sec, or some 40 minutes. The 4 X 7 = 28 counter outputs are
connected to 40 seven-digit AND gates in an arbitrary configuration as shown. At least one of
the seven inputs is derived from each of the four counters. Half the AND gates supply positive
pulses to the filter, and the other half supply negative pulses. For illustration, suppose that
the first AND gate of the upper set of ten is set up so that an output will occur if each input volt-
age is up. Then an output pulse is fed to the bandpass filter every time there is an up output
from the first and fourth stages of the 128 counter, the third stage of the 125 counter, the fifth
and sixth stages of the 121 counter, and the second and seventh stages of the 117 counter. All
the four AND gates are set to respond to independently chosen seven-digit binary numbers.
These numbers, which constitute the "key" from which the noise-like carrier is generated, are
set up on punched cards which are easily replaced when the key is to be changed.
The newer noise source design of Fig. 3(b) is the result of extensive theoretical
work6'7 initiated in the summer of 1954 while the equipment of Fig.3(a) was nearing completion.
The newer system provides an improved correlation curve and greater security. It differs from
the older one in four respects:
(1) Each of the four sets of AND gates is fed individually from a corresponding
set of counters without the crossconnection used previously. (It was convenient to
change the counts to 64, 63, 59 and 53, respectively), thus making the period about
2 minutes.*
(2) The number of AND gates in each set is increased to the point where, at the
output of each set, the voltage is in the up state roughly half the time.
(3) The four outputs of the four sets of AND gates are added modulo-two; that is,
the output voltage of the modulo-two adder is up when none, just two or all four of the
inputs are up, and down for just one or just three inputs up.
(4) Positive or negative pulses excite the filter every 10 ?sec, depending on
whether the modulo-two adder output is up or down, respectively.
Synchronizing the reference noise source with the incoming signal is a considerable
problem. Because the system must operate through jamming, it is impractical to transmit syn-
chronization pulses continually, since they would be vulnerable to imitation techniques. The
system therefore uses no special synchronization information added to the signal transmitted;
the only indication at the receiver of proper synchronism is a sizable output signal from the
averaging filters.
The synchronization procedure has three phases, which are, in order of occurrence:
initial synchronization, searching and tracking. We will consider them in reverse order.
(1) In tracking, the variable delay T of the reference noise is continually and auto-
matically readjusted for coincidence with the incoming signal. This is done by developing a
*In operational F9C systems, e.g., the F9C-A, the period will be made at least one day, and the digital key
will be changed daily.
SECRET
r
Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
 Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
SECRET
RECEIVED
SIGNAL
STORED-
REFERENCE
SIGNAL
AVERAGING
AVERAGING
AVERAGING
FILTER
EN
VELOPE
DETECTOR
ENVELOPE
DETECTOR
Fig.4. Means of derivingerrorsignal A-B with polarity the same sign CIS T, the desynchronism
between stored and received signals. Curves of output vs T are shown at the right.
DC error signal whose polarity corresponds to the direction of any desynchronism (Fig.4). This
error signal causes a small positive or negative shift in the effective clocking frequency.
(2) In searching, the narrow region in T representing synchronism is sought in some
wider range of T by speeding up the reference noise slightly so as to advance it in T, or by re-
tarding it so as to lag. The long response time of the narrow-band averaging filters limits the
search rate to less than one part in 1,000. With a repeat period for the signal of 24 hours, it
would take nearly three years to make a complete search. On the other hand, once the proper
synchronism has been attained, even if the tracking function is disabled for a day or two (as it
would be during a standby condition), the accumulated desynchronism will not amount to more
than 2 or 3 milliseconds because of dissimilar drifts of the two standards, plus 5 or 6 cosec be-
cause of changes in the time of propagation. Therefore in the F9C system, search consisting of
an automatic sweep back and forth in T, 10 msec on either side of the initial value, is considered
sufficient. Smaller sweep ranges are also available. During search, an oscilloscope is used to
display the integrating filter outputs as a function of T. Figure 5 shows two typical patterns
from this "search scope". Because the signal is propagated to the receiver by several modes
having different strengths and times of flight, several correlation peaks are visible in Fig.5(a),
instead of just one. By comparison, Fig.5(b), taken using a complete test transmitter locally
at the receiver, shows only one peak, since multipath effects are absent.
Fig.5. Typical search-scope patterns: (a) transmitted signal received over ionospheric path
(note multipath effect); (b) transmitted signal generated locally.
SECRET
I L
Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
 Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
SECRET
(3) Clearly initial synchronization is required when the system is first set up, and
whenever there is a loss of synchronism, because of a power failure, for example. To perform
the initial synchronization, a surprise tone burst can be transmitted at a prearranged time and
frequency known only to transmitter and receiver. So far as the enemy knows a priori, the
burst might appear anywhere in a frequency range so wide that he cannot jam all of it effectively.
Upon transmission of the burst, all four transmitter noise-source counters are reset to zero,
and upon reception all four receiver counters are likewise reset. Then a brief search for a few
milliseconds of T will reveal the region of synchronism. The method is impervious to repeater
jamming since, by the time the jamming is initiated, it is too late to be effective. In the exper-
imental tests, the tone burst was sent at the operating band center, but any prearranged frequency
would have served as well.
At this point it is necessary to discuss in more detail that portion of the receiver
circuitry in Fig. 2 between the mixer tube and the output to the teletype printer. As shown in the
figure, the fM or fS tones from the mixer may be converted into a two-valued DC output to the
printer by filtering them separately and then forming the difference of the envelopes of the filter
outputs. There are several disadvantages of this scheme as it stands, the principal one being
that the time constant of the averaging filters must be a fraction of the baud length for a clean
rise and fall in the signal to the printer, whereupon the factor T in the expression of Fig. 1 is
reduced accordingly 8
The F9C system employs a variation of this method, the so-called baud-synchronous,
or integrate-and-dump method of demodulation, developed by Coles Signal Laboratory and Collins
Radio Company9 In this method (1) the filter time constant is made several times the baud
length, (2) the difference in filter envelopes is sampled at a time coinciding with the expiration
of each received baud, (3) according to the polarity of the observed sample the printer is sup-
plied a constant DC signal at one of the two levels lasting one baud length, and (4) both filters
are dumped, or deenergized, immediately after sampling and are then free to respond to the
next received baud. Operations (1) and (2) insure that the factor T in the expressions of Fig. 1
is preserved at the full value of the baud duration. Operation (3) eliminates printer errors due
purely to a noisy printer waveform. (The printed characters can still be in error because noise
in the filter outputs causes the sample to have the wrong polarity, but at least the signal to the
printer is a clean square wave.) Operation (4) eliminates any inter-baud overlap effects by re-
moving before each new baud all traces of signal from the previous one.
To use baud-synchronous demodulation, one must sample and dump within about one
millisecond of the expiration of each received baud. In the F9C system the sampling and dump-
ing pulses are derived from the digital circuits of the noise source, which must be already syn-
chronized to the received signal within 50 ?sec. It then remains only to modulate the teletype
information onto the carrier (the switch in Fig. 2), using identical timing derived from the trans-
mitter's noise source.
Each teletype character consists of a start interval followed by the five information-
bearing bauds, followed by a stop pulse. The start and stop pulses are generated internally at
SECRET
Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
 Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
SECRET
the F9C receiver, and those arriving from the transmitter are disregarded, since they are
occasionally in error.
It has already been mentioned that the form of noise source was changed radically
for the second series of field tests. The study leading to the new noise source design also indi-
cated (as did observations reported in Sec. III-A) that greater security could be had by using the
type of modulation and demodulation depicted in Fig.6. Whereas the two binary teletype states
are distinguished by a frequency shift in the previous scheme, the newer method in effect uses
two different noise sources - one for each teletype state. Actually, the MARK and SPACE digital
waveforms need be different only to the extent that they possess a small enough correlation with
each other. This was effected by taking the MARK digital signal from point X of Fig. 3(b) and
complementing every other output pulse in the "complement" circuit of Fig. 6. To prevent a
tell-tale keying transient upon transition from MARK to SPACE or vice versa, actuation of the
switch in Fig. 6 is in effect interleaved between two successive pulses exciting the filter.
DIGITAL
CONVERTER
SUPER-
NOISE
AND POWER
HETERODYNE
CIRCUITS
AMPLIFIER
RECEIVER
FIG.
3(b)
SPACE
MARK
FILTER AND
SIDE AND
FILTER AND
LIMITER
FILTER
MIXER
LIMITER
COMPL
EMENT
SPACEq
FIG. 3(b)
FIG. 3(b)
COMPLEMENT
L - - ----
FILTER
DIGI
TAL
LIMITER R
NOI
SE
FIG. 3(b)
X
CIRC
FIG.
UITS
3(b)
DUMP
MARK SPACE
AVERAGING AVERAGING
FILTER FILTER
(o (o
COMPARISON
DEVICE
TELETYPE
PRINTER
Fig.6. Block diagram of Fig.2, redrawn to show the modulation system used in the second
series of tests.
One of the unique problems in the F9C system is the RF frequency stability required.
It is seen from Figs.2 or 6 that the conversion of the noise-like signal up to the carrier frequency
at the transmitter, and back down again at the receiver, must- involve an accumulated frequency
error smaller than the integrating filter bandwidth, since otherwise the difference frequency tone
will be shifted out of the filter passband. For this reason, the converters made the desired
translations of the signal in several steps in which the injection frequencies for the larger steps
were multiples of the 100-kcps standard frequency (having an observed stability of better than
one part in 108 per day), and those for smaller interpolating steps were taken from separate
crystal oscillators. An over-all frequency stability of one cycle per second resulted.
OSCILLATOR
)o
SECRET
Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
 Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
SECRET
At the transmitting end, an AN/FRT-22 transmitter served as a power amplifier.
One reason for using a limited noise-like signal (Fig.3) was to allow this transmitter to be driven
Class C. During the first series of tests, adjacent channel interference due to spectrum spread-
ing was reported. Accordingly, the FRT-22 was operated linearly (Class B) thereafter, with an
attendant reduction in transmitter power from 40 to 18 kw. During wartime anti-jam operation,
the adjacent channel interference (which is 55 db down one bandwidth from the band edge) 10 may
not matter.
F. Diversity
During the first series of tests it became apparent that multipath effects were seri-
ously deteriorating F9C performance and would have to be countered with some sort of diversity.
The cause of the trouble can be seen by referring back to Fig.5(a).11 If we imagine the receiver
reference noise source to be synchronized at some particular value of T such as T0, the energy
delivered to the receiver at T1, T2, etc., arrives too late or too early to give a correlated out-
put. [This is clear from the narrowness of the correlation curve in Fig.5(b).] These other sig-
nals therefore add their power to any noise or jamming present, with the result that the quantity
(S/J) i in Fig.1 is greatly reduced. Specifically, it becomes the ratio of the power in the signal
arriving at TO to the sum of the jamming power and the powers in signals at T1, Tetc. Be-
cause of the random fading of the amplitudes at the various values of T, occasionally the signal
at T 0 becomes so small compared to the sum of jamming and unsynchronized signal powers that
the 23-db gain given by the product TW is not enough, and errors are printed.*
Two diversity schemes were proposed and tested: T-diversity, and space-diversity.
In T -diversity, the receiving equipment of Figs.2 and 6 connected between the superheterodyne
receiver and the decision circuit is duplicated, with the noise sources synchronized to different
values of T. The correlated outputs are added in some fashion in the comparison device so that
a large output signal-to-noise ratio is preserved, on the supposition that both paths are unlikely
to fade simultaneously. The two noise sources are made to track their respective paths inde-
pendently. (In many of the tests of Sec. III this tracking was omitted, since their duration was
so short that tracking was found to offer negligible improvement.)
Whereas T-diversity uses one receiver and two noise sources, space-diversity em-
ploys, conversely, two receivers whose outputs are correlated simultaneously with the signal
from one noise source. (Referring to Figs. 2 and 6, this means a duplication of all the blocks
preceding the comparison device, except the noise source and associated synchronization cir-
cuits.) By feeding the receivers from antennas spaced far enough apart, one should observe in-
dependent fading of the signal component arriving at a given value of T.
In these tests the two antennas were Signal Corps Type C rhombics spaced 300 feet in
the direction of arrival and 550 feet laterally. The method used for combining the two voltages in
both forms of diversity was to add linearly immediately prior to the sampling operation. (Each
voltage represents a difference in envelopes of MARK and SPACE averaging filters.) A slight im-
provement in performance would probably have been obtained by a suitable nonlinear form of addition.
*The effect of multipath on FSK is of a different sort. So long as the difference in time delays for the various
paths is a small fraction of the band length (which was usually the case in these tests), the only effect is "selec-
tive fading" of the received FSK tone due to phase addition or cancellation of signals arriving via the various
paths. The received signal has no self-jamming introduced as in a NOMAC system.
SECRET
Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
 Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
SECRET
III. TEST RESULTS
A. Noise-Source Performance
Several tests were made of the behavior of the noise sources schematized in Fig.3.
The bandpass noise spectra and correlation curves were measured, and tests were made to learn
whether the MARK-SPACE teletype modulation was audibly perceptible in the transmitted signal.
In all respects, the second noise source with its method of modulation (Sec. II-D) proved better.
Figure 7 compares the spectra resulting from the old and new noise sources, re-
spectively. These plots were taken by substituting a variable-frequency signal generator for
the receiver in Figs .2 and 6, and observing the averaging filter output with the dump input dis-
connected. Since the natural bandwidth of the averaging filters without dumping is about 6 cps,
it was possible to resolve spectral components as little as 6 cps apart. It is seen from the fig-
ures that the many narrow spikes that were present with the old noise source are greatly reduced
with the new one. The spikes are caused by hidden periodicities in the digital exciting wave-
form. A concentration of the signal energy into several strong spectral lines is undesirable,
since it makes the system more vulnerable to jamming by multiple sine waves.*
OLD NOISE SOURCE
[FIG 3(a)]
NOISE SOURCE
[FIG 3(b)]
FREQUENCY (kcps)
Fig.7. Measured noise-source spectra.
Figure 8 shows the correlation curve for the older noise source. For the newer one,
the curve is substantially the same, except that slight humps (such as shown in the figure 1250
?sec from the main peak) are much smaller. These spurious peaks are caused by noise-source
periodicities at approximately the average period of the three counters whose periods are most
nearly equal, namely, the 128, 125 and 121 counters. A spurious correlation peak produces a
vulnerability to repeater jamming having a delay adjusted to arrive in synchronism with the
spurious peak. This weakness is portrayed more vividly in Fig. 9(a), which shows the A-B curve
used for tracking the old noise source. The spurious peak is accentuated here by limiters l2 used
*The data in Fig.7 (light curve) were taken with the limiter in Fig.3(a) bypassed, but inadvertently the limiter
in Fig.3(b) was included for the data of Fig.7 (heavy curve). However, the conclusion that the spectral spikes
are greatly reduced should not be affected by this (see Ref. 10).
SECRET
Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
 Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
SECRET
SECRET
Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
 Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
SECRET
in the tracking circuit (points L, Fig.4) to stabilize the amplitude of the A-B curve with changes
of signal strength. If a strong repeater jamming signal arrives near one of the spurious S-
curves, the tracking can be disrupted. (Because of propagation delays, it is probably not pos-
sible for the jamming signal to arrive soon enough after the main peak to disrupt tracking there.)
Figure 9(b) shows that these spurious peaks have been largely eliminated with the new noise
source.
Vulnerability to repeater jamming is also lessened by the new method of modulation.
With the older frequency-shift scheme, it was possible for the enemy to receive the transmission,
frequency-shift it, and then retransmit it with a suitable delay. Because of the spurious correla-
tion peaks just mentioned, or because of careless synchronization by the operator, this signal
could jam the system effectively. With the newer modulation scheme, MARK can be transformed
into SPACE, or vice versa, not by a simple frequency translation, but only by recovering the
digital waveform at point X of Fig.3(b) and complementing every other digit - a considerably
more difficult operation.
During the first series of tests it was observed that, under certain conditions, one
could detect keying transients, faintly audible in the noise-like transmission. It was concluded
that this perceptibility of the frequency shift was due to the strong lines in the spectrum (Fig.7).
The newer noise source was not tried with frequency-shift modulation, but one should find the
audibility of such keying much reduced. Using the newer form of modulation, extensive listening
tests were made, using keying rates from 6 cps to 100 kcps, and no keying transients were
detected.
B. System Operation Without Jamming
Table I summarizes the performance of the F9C system without jamming, using
various modes of diversity operation, and compares it with that of a standard FSK system using
space-diversity. For each of three frequencies, three conditions of F9C operation were em-
ployed: non-diversity, space-diversity and T-diversity. The test message was the standard
"quick brown fox" text, constituting one 70-character printed line. The table gives the number
of lines of test copy printed and the number of errors made. It also summarizes the error rate
for each frequency, and gives an over-all average error rate for each mode of operation.
It was found that FSK operation was consistently superior to F9C under these no-
jamming conditions, and that T-diversity F9C operation was better than space-diversity, which
in turn was better than using no diversity; T-diversity proved superior to space-diversity even
during periods when there was only a single propagation path. At such times the two values of
T were made to straddle the correlation peak.
Figure 10 shows in more detail the way in which the errors listed in Table I occurred.
The abscissa is the number of erroneous characters occurring in a run, and the ordinate is the
percentage of the runs that were of given length. Data are given for the four modes of transmis-
sion at each of the three frequencies. It is seen that FSK tended to make predominantly single
errors, whereas F9C errors were often made in longer groups. This is to be expected from the
nature of the F9C self-jamming effect in which errors occur when the in-synchronization re-
ceived signal fades. These fades often persist for several characters. The average length of
F9C error runs is seen to be decreased by the use of diversity, and by using higher frequencies.
SECRET
Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
 Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
CONFIDENTIAL
TABLE I
SUMMARY OF PRINTING DATA COMPARING F9C AND FSK WITHOUT INTENTIONAL JAMMING*
F9C
Date
Frequency (kcps)
FSK-S
No diversity
Space-diversity
T -diversity
25 March
17,460
3/226
110/227
46/328
1/150
25 March
8040
3/69
33/43
46/61
15/33
28 March
12,270
6/165
102/157
23/252
0/10
29 March
12,270
7/275
407/349
157/402
28/260
29 March
17,460
4/210
122/208
22/223
10/200
30 March
12,270
0/80
78/96
7/93
4/55
30 March
17,460
9/214
121/237
38/262
21/218
30 March
8040
2/108
193/98
114/110
85/104
31 March
12,270
1/92
62/119
11/125
5/95
31 March
17,460
5/171
66/167
10/171
6/167
1 April
12,270
0/57
30/52
14/122
8/126
1 April
17,460
1/94
8/96
34/108
2/107
Summary by Frequency
12,270
14/669
679/773
212/994
45/546
= 0.021
= 0.88
= 0.21
= 0.082
17,460
22/915
427/930
150/1042
40/842
= 0.024
= 0.46
= 0.15
= 0.048
8040
5/177
226/141
160/171
100/137
= 0.028
= 1.6
= 0.94
= 0.73
Totals
41/1761
1332/1843
522/2207
185/1525
= 0.023
= 0.72
= 0.24
= 0.12
*Shown as number of erroneous characters per 70-character line.
TABLE II
COMPARISON OF F9C AND FSK PRINTING ERROR
RATES IN SEVERE
MULTIPATH WITHOUT JAMMING*
FSK
F9C
F9C
Space-Diversity
Space-Diversity
T-Diversity
4/93
365/165
74/201
=0.043
=2.21
=0.37
*Shown in errors per line.
CONFIDENTIAL
Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
 Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
SECRET
4
I FSK WITH SPACE DIVERSITY
2 2 F9C WITHOUT DIVERSITY
J F9C WITH SPACE DIVERSITY
4 C??.~ F9C WITH - DIVERSITY
8040 kcps
L
TM
0.1
5
-34-2764
?
THEORETIC
AL
of
? OBSERVED
----ONE ERROR PE
R 20 LINES ----
?
OF COPY; 70 C
HARACTERS
PER L
I __,_
INE
L
-14
0
-12 -10 -8
(S/N)I ; NO
Fig. 10. Distribution of errors summarized in Table I Fig. 11. Frequency of character errors vs signal-to-
(unjammed operation). Arrows indicate average length jamming ratio of F9C equipment.
of error run .
The reason for the disappointing performance of space-diversity is that the receiv-
ing antenna spacing used was not enough to make the fading of a given multipath mode at the two
antennas independent. The spacing was sufficient for independent fading of a received FSK tone,
however. It is not known what spacing is required for good space-diversity F9C performance.
In all the experiments summarized in Table I and. Fig. 10, the operating frequency
chosen was that one of the three giving the best performance, i. e., the highest frequency less
than the "maximum usable frequency". At frequencies well below the MUF, where the multi-
path structure is more complicated, the margin of superiority of FSK over F9C widens, as does
that of T-diversity over space-diversity. This is shown by Table II, which gives data taken at
12,270 kcps when 17,460 kcps would have been a much better choice.
C. Local Jamming of a Locally Transmitted Signal
It will be recalled that when multipath effects are absent, a given signal-to-jamming
ratio at the F9C receiver input is improved by a factor TW = 220. A theoretical curve, 3 of the
frequency of erroneous characters vs (S/N)i is given as the solid line of Fig.11. Also shown
SECRET
Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
 Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
SECRET
are data from several experimental determinations made using a locally generated NOMAC sig-
nal to simulate the transmitter, and a bandpass Gaussian noise (of approximately the same spec-
tral shape) for the jamming. Measurement of the signal-to-noise ratio was accurate to about
0.5 db. The theoretical and experimental results show good agreement.
D. Local Jamming of the Received Signal
Several tests were made in which the received signal was jammed by injecting either
a bandpass noise or an FSK signal into the receiving antenna circuits. The noise-spectrum shape
was approximately the same as that of the NOMAC signal, and the MARK and SPACE FSK fre-
quencies were adjusted to coincide with those of the FSK transmitter.
The jamming signal had a constant and measurable level, whereas the received sig-
nal was subject to the usual time-varying propagation effects. Thus the signal-to-jamming ratio
as well as the multipath structure was not constant with time, and the following procedure was
used for comparing FSK and F9C performance. During a 2-1/2 minute interval, FSK signals
were received (using space-diversity) and the jamming power increased to the point where
roughly half the characters were in error. It was possible to make up to five such determina-
tions in the 2-1/2 minute interval. The process was repeated for 2-1/2 minutes of F9C without
diversity, then F9C with space-diversity, then F9C with T-diversity.
Table -III presents the data obtained during a number of such 10-minute cycles. The
figure given as "average anti-jamming margin" is the difference between the average jamming
power used in a 2-1/2 minute F9C interval and that of the preceding FSK interval. It is seen
that all three modes of F9C operation show a significant anti-jam advantage over FSK. Con-
clusions as to the performance of the various F9C modes of operation relative to each other are
probably impossible because of (1) the small number of readings of the margin and their large
dispersion (about 3 db), and (2) the fact that the readings were made at such a high error rate
TABLE III
MARGIN OF ANTI-JAMMING ADVANTAGE OF F9C vs FSK
F9C
No Diversity
Space-Diversity
T-Diversity
Noise Jamming
(23 March 1955, 8040 kcps)
Average db margin
9
8
8
Number of comparisons
4
4
4
FSK Jamming
(23 March and 1 June 1955,
8040, 12,270 and 17,460 kcps)
Average db margin
19
20
21
Number of comparisons
6
12
10
SECRET
J
Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
 Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
SECRET
that the errors were due more to the intentional jamming than to self-jamming by multipath.
This situation, which is not typical of the usual operating condition, tends to obscure the effect
of diversity in reducing the self-jamming.
To imitate an actual operating environment, remote jamming tests were run using as
the jamming signal either a 710-watt transmission at 12,270 kcps from Collins Radio Companyat
Cedar Rapids, Iowa or an 8- to 10-kw transmission at 17,460 kcps from Army Communication
Station ABA in Honolulu. Both jammers used FSK signals with MARK and SPACE frequencies
tuned to those of the "friendly" transmitter at Davis, California. A calibrated attenuator was
used to adjust the Davis transmitter power to
give an error rate at the receiver (Deal, New
Jersey) of about two characters per line, and then
an average of about twenty lines of cony were
f 21PA a single predominant propagation path, space-
diversity; otherwise r-diversity). The data,
which are summarized in Fig.12, show that F9C
exhibited an average anti-jamming advantage of
about 17 db over FSK. Each determination en-
Fig.12. Anti-jam advantage of F9CoverFSK when tered in Fig.12 represents an instance where F9C
remote jamming is used.
and FSK were compared within 5 minutes or less.
? 12,270 kcps, CEDAR RAPIDS
1 17,450 kcps, HONOLULU
Fig.13. Distribution of errors for remotely jammed operation. Arrows
indicate average length of error run.
SECRET
1-
7
1
1
f SPACE-DIVERSITY FSK
2-I
2
DIVERSITY F9C
12,270 kcps
Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
 Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
SECRET
The printed copy produced in the runs summarized in Fig.12 was analyzed for the
length of error runs and tabulated in Fig.13. Comparison with the data of Fig.10 (unjammed
operation) reveals several things. Since the over-all error rate was higher for the jammed
case (about two errors per line compared with the Table I figures), several longer error runs
were observed for both FSK and F9C. With jamming introduced, the FSK error runs tended to
last longer than the F9C runs, whereas the reverse was true for the unjammed case. The F9C
error runs were usually longer for the highest frequency with jamming present.
The F9C system was designed to provide an increase in signal-to-jamming ratio for
high-frequency radioteletype circuits, not by increasing transmitter power or antenna gain, but
by using the NOMAC modulation and demodulation scheme. In'such a system a noise-like wave-
form is the carrier and, with a noise bandwidth of 10 kcps using 60-word-per-minute teletype,
a 23-db improvement in anti-jamming capability over conventional FSK is predicted, if multi-
path is absent. Although superior to FSK with respect to jamming alone, the F9C system with-
out diversity was found to have a very high error rate because of multipath. The inclusion of
T-diversity improved this situation and allowed an average 17 of the 23 db to be attained against
jamming in the presence of multipath.
1. B.L. Basore, "Noise-Like Signals and Their Detection by Correlation," Technical Report No.7,
Lincoln Laboratory, M.I.T. (26 May 1952).
2. P.E. Green, Jr., "Correlation Detection Using Stored Signals,' Technical Report No.33,
Lincoln Laboratory, M.I.T. (4 August 1953).
3. G.L. Turin, "Probability of Error in NOMAC Systems," Technical Report No.57, Lincoln
Laboratory, M.I I.T. (18 January 1954).
4. A more detailed explanation of the equipment used in the two series of tests is given in:
P.E. Green, Jr., "The Lincoln F9C Radioteletype System," Technical Memorandum No.61,
Lincoln Laboratory, M.I.T. (14 May 1954), and Addendum (14 July 1954), and P.E. Green,Jr.,
"Data on the Lincoln F9C-B Radioteletype System," Group Report No.34-36, Lincoln Labo-
ratory, M.I.T. (2 March 1955).
5. B.M. Eisenstadt, P.L. Fleck, Jr., O.G. Selfridge and C.A.Wagner, "Jamming Tests on
NOMAC Systems," Technical Report No.41, Lincoln Laboratory, M.I.T. (25 September 1953).
6. N. Zierler, "Several Binary-Sequence Generators," Technical Report No.95, Lincoln Labo-
ratory, M.I.T. (12 September 1955).
7. B. Gold and B.M. Eisenstadt, Technical Reports in progress.
8. B.L. Basore, op. cit., Sec.VIll-A.
9. M. Doelz, "A Predicted-Wave Radio Teleprinter,' Electronics 27, 166 (1954).
10. Quarterly Progress Report, Division 3, Lincoln Laboratory, M.1 T. (15 January 1955), p.75.
11. P.E. Green, Jr., op. cit. (Ref.2), Chap.V.
12. Quarterly Progress Report, Division 3, Lincoln Laboratory, M.I T. (15 January 1954), p.69.
13. G. L. Turin, 2E.cit.
SECRET
0
Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
 Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
SECRET
Figures A-1 and A-2 show the transmitting and receiving ends of F9C equipment
used in the first field tests beginning in July 1954. The transmitting-end equipment consists of
a complete transmitter and a test receiver, whereas the receiving end is a receiver plus test
transmitter. This duplex arrangement facilitates local testing at both ends of the system.
Briefly 4 the functions of the different units are as follows:
Driver: provides 100 kcps and resetting pulses to drive counters,
Counters: provide cyclic outputs to diode AND gate matrix [see Figs. 3(a)
(3) Matrix: forms from counter and driver outputs two pseudorandom trains
(MARK and SPACE) of pulses,
(4) Modulator: forms the bandpass noise from the digital output of the noise
source, and performs the frequency-shift modulation,
(5) Modulator Timer: retimes incoming teletype signal to synchronism with
the transmitting noise source,
(6) Exciter Converters A and B: converts modulator output to proper frequency
and power to drive transmitter,
(7) Receiver Converters A and B: a stable superheterodyne receiver,
(8) Multiplier: mixes received and "stored" signals to obtain 20 kcps MARK
or SPACE tone,
(9) Demodulator: decides from multiplier output whether MARK or SPACE
was transmitted and actuates printer accordingly (see p.8),
(10) Demodulator Timer: provides synchronous dump, sample and start-stop
pulses for the demodulator (see p.8),
(11) Discriminator: forms S-curve (Fig.4) for automatic synchronization of
received and stored signals,
(1 Z) Synchronizer: advances or delays counters in accordance with discrimi-
nator output,
(13) -r -Register Unit: controls cyclic sweep (in T) of standby noise source to pro-
vide a multipath structure presentation on the Search Monitor Oscilloscope.
The design, construction and testing of the Lincoln F9C system was
the cooperative effort of most of Group 34 at Lincoln Laboratory.
We would like to record our indebtedness to the many Signal Corps
people whose cooperation made possible the field-test portion of the
work. We acknowledge particularly the assistance of Mr. H. F. Meyer,
Mr.L. Manamon and Capt.H.A. Schulke of Coles Signal Labora-
tory; and Maj.C.R.Knoeller and Mr.A.Gorricho of the Sixth Army
(San Francisco).
SECRET
Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
 Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
C O B~6e~tt a
Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
 Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
COM "VIAL
C FO0E SAL
GPcE9
Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
 Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0
Declassified and Approved For Release 2012/07/09: CIA-RDP78-0333OA004100060024-0