Document text
43
ESD-TR 70-93
W
a?
O
o
LU
ESD ACCESSION LIST;
ESTI Call No. _ 69609 _
Copy No. _ [__ of (_ eys.
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
LINCOLN LABORATORY
ESD RECORD COPY
sciENiiric I rfcS,RN T0
ni. euiwiZ°mlm
52 Q - 0019
INTERFERENCE SUPPRESSION PERFORMANCE OF
SEVERAL FM RECEIVERS USING FEEDFORWARD
Ben H Hutchinson, Jr
July 26, 1961
XD7044Q7
The work reported in this document was performed at Lincoln Laboratory,
a center for research operated by Massachusetts Institute of Technology ,
with the joint support of the U S Army, Navy and Air Force under
Air Force Contract AF 19 (604) -7400
LEXINGTON
MASSACHUSETTS
U-13S0*
; fr \ &'* '.. ■': v Y
¥,V-A-»' ■'*.' v j ' i
r< ^.k/>3 h< t>
3 v
XftW
r ' 'V;'.1.
V , N-
•>y w
>/ r. .Vf ’ ,'■' ■ . .
'■/Vl^'vr,' v'-l',' ' - .
jf -fH
'4-i
1
BfKffijRf: -f.
h M '* ■ • ■
s i? y
' . ' ' .. <■
^>Vv * ;.r . V '-rj
i \\ ' V '
■
■uiafiCH i: & i*
- -
^ c ■
f yA
&W
fv '• *
7?«
r; , ? V ; 7 > ’•*• • ,:
, ■> -‘V 'V ' ' -
5 ■Vv-c/'v * v
> ' V- u r y ' r
*H-rv
' <r*v
.\V*
fpv
iS^w/V
fs. *\ v
vr V-Jv
Vk . i * r*
Vr-
SffU $ ! * ' • v
/. V.UV* Al; ;
.vlj *
<
£5F*
'‘L J ^ 1 >» . / ^ r„ /
t v -‘ ;. ^ -t 'r ' •' * /' ■' J •
A .J.- Sl.ir .-•*. . .< hL . ; ’•' * , " .
\ • -■■
iW;
tV j
■> %
lib a
'j.\<
nr'?, i'l-. ~ ~ 1
K,y7-‘ v - .
" V ‘ ,x- -v ‘ -S ... V.
M
, ' . -• *'f V;. •-/
' V , V •;.. <• ; ••*,<
4 cl ,• . >■ j • ■ 1 , A ^ fFQFj :
' f ^ " ; .<:
SjK^ n^ ' 1 ' , . * - ^ k i
V • >‘ ' fV.-. t * i
*••.,. * ““V '
v. ••- .«•.• .”‘ . . . ■ ■
for Mk &m •* hi
unliiiiteiL
r Jfv; ‘
v.- C :v?,‘>-V X. ' >, '*^..^5
rv-^' 1 <•> ' n -v- ^ "v
,\ 1 r *’ ■ • • .T - * » ,
- * -.A ,*»nj s*_\
■BKKrV"
\ / ^ ;i V 3 ,f'
r • v V- y, •
% >>;•? 1 1 s!; ^ .
' r4* ' ,vt*^ •
HK TaffiF^r HBI
&gh$iy
‘ • 'v v ;v : ^ 'KV:;v.'
r^-
;
■K
*$>/
. vj~r. V .■ ^ • •-■, v ' ;r>- '/
ip*. -y*. / j! •HyS/ ^
1 , ,rv . —a r#vi v ■-1;
hJr-
jK'
•*• ' .■ H ,r. - >• -
■:'..■ / ;. » .IV <
•.K>, >," J#, ‘
• . \ hjl
aT4M: * v-
l ; , . «t . *, *
.c.':'sw5''.- '•*’
t'Vv
v< ■» r> ; ’, ■•..■ .. k -■ r ,'
6 -■ . g ;
'Jj-, y .•*'', fcij* rV. S. / \ f' ‘ 1 A’**
Kk^S '&■ V ' V ‘ ' ^
. ' ‘ Cu-V ' " .•’ Vi \ > -
y^... yV ■ . A _ v / ^
'■; (j.
V
INTERFERENCE SUPPRESSION PERFORMANCE OF
SEVERAL FM RECEIVERS USING FEEDFORWARD
by
BEN H, HUTCHINSON JR,
Submitted to the Department of Electrical Engineering on January 16,
1961, in partial fulfillment of the requirements for the degree of
Master of Science in Electrical Engineering
ABSTRACT
The feedforward signal-cancellation technique is based on
subtractively combining the outputs of limiters and linear amplifiers
having a common input. Used in an FM receiver, feedforward pro¬
vides an attractively simple and effective method for suppressing
interference to an FM signal from other co-channel or adjacent-
channel signals which may be either weaker or stronger than the
desired signah The thesis explores theoretically and experimentally
the potential performance and inherent limitations of practical FM
receivers using feedforward. Design criteria are discussed for
various interference conditions and the relative merits of several
practical feedforward circuits are considered,
A laboratory model FM receiver was built and tested with
three different feedforward circuits its performance being measured
under a variety of interference conditions. Significant improvement
in the stronger-signal capture performance of a mediocre FM de¬
modulator was demonstrated. Sinusoidal modulation was recovered
from FM signals between 0o 05 and 0, 9 times the amplitude of an
interfering signal on the same channel, distortion ranging generally
between 8 per cent and 30 per cent for various interference
conditions. Completely intelligible speech modulation was also
recovered from the weaker of two co-channel FM signals. Numerous
suggestions for further work are given
Thesis Supervisor; Elie J, Baghdady
Title Assistant Professor of Electrical Engineering
Accepted for the Air Foroe
Frank.; in C, Hudson
Chief, Lincoln Laboratory Office
This document has been approved for public release and gale;
its distribution is unliaitei
ACKNOWLEDGEMENTS
The author wishes to thank Professor E. J. Baghdady for
supervising the thesis investigation; his basic theoretical work is
the foundation on which the thesis is built. He would also like to
express appreciation to his colleagues in the Research Laboratory
of Electronics for many helpful informal discussions, especially
J. M Gutwein and John Boatwright. Special thanks go to Miss Sheila
Hayes and Mrs. Jean Coombs for typing the manuscript.
TABLE OF CONTENTS
Page
CHAPTER 1 13
Introduction and Background 13
Previous Experimental Work 19
Purpose of Present Investigation 20
CHAPTER 2 23
Problems Encountered in Practical Feedforward
Systems 23
Stronge r -Signal Capture 23
Weaker -Signal Capture 2 5
Capture of the Weaker of Two Co- Channel
Signalsc 26
Practical Requirements for Good Weaker-
Signal Capture 28
(1) Problems of Practical Limiters 28
The pentode limiter 29
The gated-beam limiter '30
The diode limiter 30
(2) Maintaining Accurate Inter¬
ference Cancellation 31
Variations in a 31
Variations in input signal
amplitude 33
Limiter imperfections 3 5
Chapter 2 (continued) Page
Bandpass filter difficulties 36
Ordinary circuit difficulties 37
Degree of precision to which
requirements must be met 37
Difficulties for a^ > 0, 5 37
Difficulties for a^ 0. 5 38
(3) Feedforward Amplifier
Linearity 43
A minor practical problem 43
CHAPTER 3 45
Relative Merits of Various Practical Feedforward
Circuits 45
Summary of Requirements 45
Possible Basic Circuit Configurations 45
The Transformer-Input Feedforward 45
The Transformer-Output Feedforward 49
The Dnver-Limiter Feedforward 49
The Grounded- Grid-Amplifier Feed¬
forward 51
Use of a Split-Load Phase Inverter 55
CHAPTER 4 56
'Description of Experimental Equipment 56
Laboratory Receivers 56
The I* Fo Amplifier 58
Chapter 4 (continued) Page
The Demodulator 61
The Transformer-Input Feedforward 73
Experimental Transformer -Output Feedfor¬
ward 74
Experimental Driver - Limiter Feedforward 76
CHAPTER 5 79
The Experimental Measurements 79
The Experimental Setup 79
Measurement Procedures 82
Audio Filtering 87
Results of Experimental Measurements 90
The Transforme r -Input Feedfor¬
ward 90
The Transformer - Output Feed¬
forward 100
The Driver Limiter Feedforward 105
Capture Plots 105
Varying Modulation Frequencies 111
Spectrum of Captured Weaker-
Signal Modulation 123
Reduced Deviation Tests 127
Speech Intelligibility Tests 128
CHAPTER 6 134
Summary and Conclusions 134
CHAPTER 7
Page
138
Suggestions for Further Work 138
Improved Narrow-Band Limiters 13 8
Improved Bandpass Filters 139
Demodulator Improvements 139
Automatic Control Systems for K 140
Reducing Audio Distortion 143
More Complex Systems Based on the
Feedforward Principle 144
The "Ultimate" Basic Feedforward 145
Feedforward Demodulator 146
REFERENCES 148
LIST OF FIGURES
CHAPTER 1
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
CHAPTER 2
Figure 1
Figure 2
Figure 3
CHAPTER 3
Figure 1
Page
Transfer Characteristic of Ideal
Amplitude Limiter, 14
Frequency Characteristic of
Ideal Bandpass Filter for go > 0o 14
Ideal Narrow- Band Limiter
With Two - Signal Input , 14
Effect of Narrow- Band Limiter
in Reducing Interference Ratio, 16
Considering only the Two Output
Components at the Input Signal
F requencieso
The Basic Feedforward Circuit, 16
An example of stronger - signal
cancellation is shown.
Amplitude of Stronger and Weaker 32
Signal Components at Limiter
Output (A and A as a function
of a
— in
Optimum value of K (Koo) and Outer 40
Boundaries on K for Weaker- Signal
Capture, 0o 001 < a < 0o 5,
Expansion of Figure 2 for a < 0, 1, 41
Basic Transformer Input Feedforward
Circuit,
46
Chapter 3 (continued)
Page
Figure 2
Basic Transformer Output
Feedforward Circuit,
48
Figure 3
Basic Driver -Limiter
Feedforward Circuit,
50
Figure 4
Basic Grounded-Grid Amplifier
Feedforward Circuit.
52
Figure 5
Oscillator Circuit Formed by
Cathode Follower with Tuned
Input Circuit
54
CHAPTER 4
Figure 1
Block Diagram of Laboratory
Model Receiver.
57
Figure 2
Schematic of the 3 -Stage I. F
Amplifier o
59
Figure 3
Frequency Response Curves of
I. F Amplifier.
60
Figure 4
Schematic of the High-Performance
Demodulator with Later-Model
Audio Section.
62
Figure 5
Frequency Response of Bandpass Speech
Filter used in Later-Model Audio
Section,
64
Figure 6
Early-Model Audio Section used in
Demodulator.
65
Figure 7
Frequency Response of Early-Model
Audio Section with 75 and 7 50 p- second
de -emphasis
66
Figure 8
Frequency Response of Early-Model
Audio Section with Low-Pass Filter
67
Chapter 4 (continued)
Pag
Figure 9
Capture Plot of Demodulator
as Originally Constructed, with
320-kc discriminator (Miller
Noo 14641 Used ±n Recommended
Cj rcuxt.o
68
Figure 10
Demodulator Capture Plot
Obtained by Lowering Limiter Time
Constants and Reducing Discrimina¬
tor Time Constant to 3 psec*
69
Figure 1 1
0)
Circuit same as in Figure 10
except 900-kc Discriminator
Transformer (Miller No. 1464-WB)
Substituted for Original Unit*
70
Figure 12
Circuit same as in Figure 1 1
except Discriminator Time Constant
lowered from 3 p-seconds to 1. 4
p-seconds .
71
Figure 13
Circuit, of the Experimental Transformer -
Input Feedforward*
75
Figure 14
Circuit of the Experimental Transformer-
Output Feedforward
77
Figure 15
Circuit of the Experimental Driver-
Limiter Feedforward
78
CHAPTER 5
Figure 1
The Experimental Setup,
80
Figure 2
Block Diagram of the Signal-Generator
Modulator.
83
Figure 3
Method Used to Dynamically Plot
Frequency Response Curves.
84 a
Chapter 5 (continued)
Page
Figure
4
Method Used to Dynamically Plot
Limiter Transfer Characteristics.
84 a
Figure
5
Enhancement of Demodulator Stronger -
Signal Capture Performance by
Transformer-Input Feedforward.
91
Figure
6
Weaker-Signal Capture Performance
of Transformer-Input Feedforward
with Fixed Value of K
93
Figure
7
Weaker-Signal Capture Performance
of Transformer -Input Feedforward; K
Optimized for Each Value of a.
94
Figure
8
Improved Weaker-Signal Capture
Performance with Better Alignment
of Post-Feedforward Filter
95
Figure
9
Improved Capture Performance for
Small a Obtained by Careful Alignment
of Entire Receiver.
96
Figure
10
Best Weaker -Signal Capture Obtained
Using Transformer-Input Feedforward.
97
Figure
1 1
a) System Response Curves for
Solid Percent- Capture Curve of
Figure 10,
98
b) System Response Curves for
Dotted Percent-Capture Curve
of Figure 10 (after re -alignment).
98
c) Appearance of Captured Weaker-
Signal Modulation in Test of
Figure 10.
99
Figure
12
Weaker-Signal Capture Performance
Initially Obtained from Transformer -
Output Feedforward,
101
Figure
13
Improved Performance of Transformer-
Output Feedforward after More Careful
Alignment.
102
Chapter 5 (continued)
Page
103
Figure 14
Figure 15
Figure 16
Figure 17
Figure 18
Figure 19
Figure 20
Figure 21
Figure 22
Figure 23
Performance of Transformer -Output
Feedforward with Bandpass Speech
Filter in the Demodulator Audio.
Improved Transformer- Output 104
Feedforward Performance with
Bandpass Speech Filter and Better
Alignment.
Weaker -Signal Capture Performance 106
of Driver -Limiter Feedforward With
and Without Pre » Limiter
Best Weaker-Signal Capture 108
Performance Obtained, Dnver-
Limiter Feedforward in Receiver.
Appearance of Captured Weaker- 109
Signal Modulation Waveform a = 0 05,
for Test of Figure 17*
Weaker - Signal Capture Performance 110
of Driver- Limiter Feedforward with
K Alone Adjusted for Every a «
Capture of Weaker - Signal Modulation 112
of Arbitrary Frequency; Stronger
Signal Modulated by 1000 -Cycle Sine
Wave
Capture of Weaker - Signal Modulation 113
of Arbitrary Frequency Stronger
Signal Modulated by 400- Cycle Sine
Wave
Capture of Weake r -Signal Modulation 114
of Arbitrary Frequency, Stronger
Signal Modulated by 1000 Cycle Sine
Wave
Capture of Weaker - Signal Modulation 115
of Arbitrary Frequency Stronger Signal
Modulated by Output of Thyratron Random -
Noise Generator Passed Through 3-kc
Low -Pass Filter Simulating Random
Program Modulation.
Chapter 5 (continued)
Figure 24
Figure 25
Figure 26
Figure 27
Figure 28
Figure 29
Figures 30-32
Figures 33-34
Figures 35-36
CHAPTER 7
Figure 1
Capture of Weaker-Signal
Modulation of Arbitrary Frequency;
Stronger Signal Modulated by a
100 -Cycle Sine Wave
Capture of Weaker -Signal Modulation
of Arbitrary Frequency; Stronger
Signal Modulated by 400 -Cycle
Sine Wave.
Capture of Weaker-Signal Modulation
of Arbitrary Frequency, Stronger
Signal Modulated by 1000-Cycle
Sine Wave.
Capture of Weaker -Signal Modulation
of Arbitrary Frequency; Stronger
Signal Modulated by 3000-Cycle
Sine Wave.
Capture of Weaker-Signal Modulation
of Arbitrary Frequency, Stronger
Signal Modulated by 5000-Cycle
Sine Wave
Capture of Weaker -Signal Modulation
of Arbitrary Frequency; Stronger Signal
Modulated by Output of Thyratron
Random-Noise Generator Passed
Through 3-kc Low-Pass Filter,
Simulating Random Program Modulation.
Spectra of Captured Weaker -Signal
Modulation*,
Effect of Reduced Weaker-Signal
Deviation
Effect of Reduced Stronger -Signal
Deviation*,
Block Diagram of a Proposed Control
System to Automatically Maintain
Interference Cancellation
Page
117
118
119
120
121
122
124-
126
129
130
131
132
141
13
CHAPTER 1
INTRODUCTION AND BACKGROUND
The feedforward signal-cancellation technique, applied to an
FM receiver, allows capture of a desired FM signal in the presence
of an interfering signal which may be either weaker or stronger than
the desired signal at the receiver input and whose frequency occupancy
may be quite close to or even within the channel occupied by the
desired signal*
The idea for the technique was conceived by E J0 Baghdady
as a by-product of his theoretical investigation of the FM interference -
suppression properties of narrow-band amplitude limiters* Several
detailed accounts of this work have been puUished ^ 4) hence,
the complex mathematical analysis involved will not be repeated here.
The assumptions involved and the conclusions which suggested the
feedforward idea will be briefly summarized*
In his analysis, Baghdady assumes an ideal amplitude limiter,
defined as a device which delivers a constant output voltage amplitude
so long as the amplitude of its input signal is above a certain minimum,
^thresh" (See Figure 1)0 In operation, the instantaneous amplitude
of the input signal is maintained above thresh ^imeSo An ideal
bandpass filter is assumed to follow the ideal limiter, its bandwidth
being small relative to its center frequency (Figure Z) Two FM signals
Ik
Frequency Characteristic of Ideal Bandpass Filter For CQJ ^ O *
E,
Ideal
Amplitude
Limiter
X
Ideal
Bandpass
Filter
(Figure 1)
(Figure 2)
Figure 3
A o
Ideal Narrow-Band Limiter with Two-Signal Input
15
having normalized amplitudes 1 and a_ln and center frequencies within
the filter passband are assumed at the limiter input The quantity a ^
is, of course, the ratio of weaker-signal amplitude to stronger-signal
amplitude, and, in applications in which the stronger signal is the desired
one, has been called the "input interference ratio"*, For purposes of
(5)
computation, the so-called "quasi- static" analysis is applied; i0 ec ,
the modulation on the FM signals is assumed to be slow enough relative
to their center frequencies and frequency difference so that over several
cycles of the difference frequency the two signals can be treated as two
stationary carriers0 Thus the analysis starts with two carriers of
frequencies p and p + r, r p, and amplitudes 1 and a , ^ 1 ,
fed to the input of an ideal narrow-band limiter (See Figure 3)c The
purpose of the analysis was to determine the character of the signal
at the output of the limiter under various interference conditions and
for various bandwidths of the post-limiter filter0 The filter bandwidth is
for convenience expressed in units of one I F0 bandwidth (BW this
being, of course, the minimum bandwidth necessary for reproducing the
modulation of the desired signal in a practical system*.
The non-linear action of the limiter produces many new frequency
components above and below the two input signal frequencies. These
components are spaced apart by the frequency difference r0 An exact
Fourier analysis of this complex limiter output signal reveals that the
ratio a
— * out'
, amplitude of component at frequency p + r ,
where a , = - t ± -3 - 1 - - — r — n - <--£- -
— out amplitude of component at frequency p
is less than the corresponding input ratio ^ in° The amount of this
1 6
Acr NtK**Li2fe Anptxrvpe
OP Ovrror Cor«r6rt6Nr
Figure 4
Effect of Narrow-Band Limiter in Reducing Interference Ratio,
Considering Only the Two Output Components at the Input Signal Frequencies
The Basic Feedforward Circuit.
An Example of Stronger -Signal Cancellation is Shown.
17
reduction is given in Figure 4. For an a in of 0e 5 or less, is
reduced by a factor of approximately 2, or about 6 db.
If the limiter filter is BW^ wide and the frequency difference r
is greater than BW^/2, only two components will appear at the filter
output, those at the frequencies of the two input signals, The limiter
thus accomplishes a reduction in the amplitude of the weaker signal
relative to that of the stronger with no "side effects" for r^ BW^/20
If r ( BW.^/2, additional components are passed by the filter and the
picture is more complex; however, detailed analysis shows that the net
effect is beneficial reduction in interference from the weaker signal for
a ^ 0, 863. Limiter bandwidth must be increased to obtain beneficial
interference reduction for ^ 0. 863. Theoretical demonstration of
this interference reduction and derivation of minimum allowable limiter
filter bandwidths constituted the main purposes of the analysis „
The idea for the feedforward technique ^ arose from the observa¬
tion that if two FM signals occupied the same channel or adjacent channels,
a narrow-band limiter could easily be arranged such that the instantaneous
frequency difference r would be greater than half the limiter bandwidth
over a significant portion of the modulation cycle, Over this portion of
the cycle, the limiter would have no other effect than reducing the
amplitude of the weaker signal relative to that of the stronger, the amount
of the reduction being given by Figure 4. If the two signals fed to the
limiter are also fed to a linear amplifier (Figure 5), their relative
amplitude in the amplifier output will, of course, be the same as that at
the input. If the outputs of the two parallel channels are combined
18
subtractively with correct relative amplitudes, either the weaker or the
stronger signal can be completely cancelled, leaving a residual output
at the frequency of the other signal* Figure 5 shows an example of
cancellation of the stronger signal* The technique derives its name
from the fact that signals are "fed forward" around the limiter through
the linear amplifier,,
When r is less than BW^/2, the situation is more complex,
since more than two components pass the limiter filter* However, if
the limiter-amplifier combination is followed by a high capture - ratio
FM demodulator, the weaker signal can be captured as long as the
average frequency of the resultant of all of the passed components equals
the frequency of the weaker- signal component. When a number of
extra components are admitted, this condition is, in general, no longer
satisfied and weaker-signal capture fails* Thus, weaker - signal capture
is possible only over part of the modulation cycle* If, however, the
circuit is adjusted for suppression of the weaker signal, the average
frequency of the resultant signal at the output of the feedforward circuit
will always equal the frequency of the stronger signal* Thus, stronger-
signal capture is possible ever the entire modulation cycle
A >
In the analysis, a parameter K is defined as
A E
_ s
klim
in which A
E
Tim
feedforward amplifier gain
input signal amplitude to limiter and
amplifier (stronger - signal amplitude for
two signals)
constant output signal level of amplitude
limiter
19
Physically, K is the ratio of the linear -channel output to the
limiter output for a single unmodulated carrier input* Negative
values of K correspond to 180° phase difference in the two channel
outputs, resulting in subtraction. For values of K near -1, the stronger
signal will be nearly or completely cancelled, allowing the originally
weaker signal to predominate in the output* Values of K in the neighbor¬
hood of -0. 5 to -0. 7 result in suppression of the weaker signal* Reduction
in interference from a weaker signal equivalent to that obtainable from
several stages of narrow-band limiting is possible*
The theoretical analysis by Baghdady outlined above predicts
that a feedforward using a good narrow-band limiter (approaching the
ideal) can cause the weaker of two co-channel FM signals at its input
to predominate at its output as long as sufficiently few additional
components pass the limiter filter* The analysis also predicts that
weaker-signal capture will be lost over part of the modulation cycle,
since part of the time the average frequency of the resultant output
signal will not equal the frequency of the weake r - signal component
Previous Experimental Work
Several experimental investigations of feedforward circuits have
been completed or are presently in progress; however, only two accounts
of such work have so far been published, i e , the S. M theses
recently completed at M. I* T by R. H Small' ^ and R* G Griffin^*
Unfortunately, both of these investigations were concerned with specific
application of the feedforward technique to fairly complex systems at a
20
time when no proven circuit design existed. In order to carry out their
thesis plans, it was necessary in both cases to quickly freeze a circuit
design, build several copies, incorporate them into a complex system,
and make a number of measurements. Time was not available to go
deeply into the workings of the circuit or to optimize its performance
However, Small and Griffin did succeed in demonstrating that their
particular feedforward circuits were capable of improving the stronger-
signal capture capabilities of a demodulator of mediocre performance and
of recovering weaker- signal modulation with quality ranging from poor
to excellent, depending on whether the two signals occupied contiguous
or overlapping channels. They were unable within the limits of their
thesis plans to devote sufficient effort to the problems of operational
feedforward circuitry, to explore the many different ways of realizing
the basic block diagram of Figure 5, to investigate the practical
limitations on the interference suppression performance obtainable
with simple feedforward circuits, including the effects of the performance
of other portions of the receiver or to explore the effects of arbitrarily
varying the modulation frequency on both desired and interfering signals.
Purpose of Present Investigation
There are several ways to realize the basic bxock diagram of
Figure 5 in the laboratory. Different tvpes of limiters are available,
moreover, a cascade of several limiters may be used instead of the
single limiter indicated in Figure 5C
The necessary phase opposition at the two channel outputs may
be obtained by several different combinations of grounded grid, grounded
21
cathode, and cathode follower circuits or by using a center -tapped
transformer or some other phase -splitting means
In the present investigation, an attempt is made to investigate
both theoretically and experimentally several of these different circuits
and to measure their performance under a wide variety of interference
conditions, the primary purposes being: (1) to gain sufficient
understanding of the feedforward technique to be able to formulate a
few general principles to guide the designer of FM receivers using
feedforward; (2) to gain some idea of the sort of signal-capture perform¬
ance potentially available from a feedforward-equipped receiver; (3)
to discover some of the fundamental limitations of the technique, and
some of the problems involved in applying it*
The body of the thesis report consists of six chapters* Chapter 2
is a general discussion of some of the basic problems encountered in
designing an FM receiver using feedforward. Chapter 3 presents the
advantage and disadvantages of several specific types of feedforward
circuits. Chapter 4 contains a description of the design and functioning
of the experimental equipment built for the investigation. Chapter 5
presents the results of experimental interference tests with
interpretations. Chapter 6 contains the over-all conclusions, while
Chapter 7 is devoted to suggestions for further work. The list of
suggestions is quite long because of the exploratory and problem-defining
nature of the study.
Both the theoretical and experimental portions of the study are
"use-oriented" in the sense that these questions are constantly raised:
22
(1) Will the feedforward technique be useful in
an existing or presently conceivable FM system9
(2) Does it offer any net advantages over competitive
signal-processing techniques 9
(3) What basic engineering problems must be solved
in the development of a workable, operational FM receiver using
feedforward ?
23
CHAPTER 2
PROBLEMS ENCOUNTERED IN PRACTICAL FEEDFORWARD SYSTEMS
The basic feedforward system of Figure 5, Chapter 1, may be
used with a good FM demodulator to capture either the weaker or the
stronger of two competing FM signals, The signals may occupy the
same channel (nco-channelM signals) or adjacent, non- overlapping
channels The requirements for optimum performance depend upon the
particular interference situation; the various situations will therefore
be considered separatelya
Stronger-Signal Capture
As shown in Chapter 4, use of a feedforward circuit ahead of a
mediocre FM demodulator can dramatically improve the ability of
the demodulator to reject interference from a co-channel signal only
slightly weaker than the desired signaL Furthermore, the requirements
on the components of the feedforward (limiters, bandpass filters,
amplifiers) are less critical than for the case of weaker-signal capture,
A feedforward will usually outperform a simple limiter in reducing
(7)
weaker-signal interference, if it works at alL Small demonstrated
dramatic improvement in stronger - signal capture performance with
feedforward circuits which had many shortcomings 0
When adjacent- channel interference weaker than the desired
signal is involved, the generalizations given above still hold, provided
24
only that the limiters in the system are sufficiently fast-acting to
cope with the amplitude disturbance associated with the maximum
(3)
frequency difference r and the interference ratio a to be encountered .
The interference need not be weaker than the desired signal at the
receiver input as long as the receiver front end and I. F. amplifier are
selective enough to insure that the interfering adjacent-channel signal
is always weaker at the feedforward inputo
Methods other than feedforward are available for achieving
excellent stronger-signal capture: the wideband approach^ the use
(4)
of cascaded narrow-band limiters , and the use of an oscillating
(13)
limiter . The wideband method is usually so expensive and complicated
as to be obviously inferior to the other schemes. An oscillating limiter
is substantially equivalent to a feedforward in circuit complexity, but
is much more critical in adjustment. A chain of narrow-band limiters
is more straightforward in design, construction, and alignment than a
feedforward, since the problems of maintaining the correct value of K
and of matching phase shift in limiter and amplifier channels are not
involved* Very good capture performance can be obtained from a limiter
chain and a moderately wideband discriminator, as shown in Chapter 4.
In many applications, the potential improvement in capture performance
obtainable by the use of feedforward instead of a limiter chain would
not be worth the extra effort involved in realizing itc
The existence of the competitive alternate solutions discussed
above, the relaxed circuit design requirements compared with those for
weaker-signal capture, and the fact that excellent performance has
already been demonstrated (reference 7 and Chapter 4) combine to make
25
the problem of building better feedforwards for stronger- signal
capture neither very interesting nor very challenging compared with
the problem of weaker - signal capture, except as noted briefly in
Chapter 7 For these reasons, little effort was devoted to the problem
in this investigation, except that mentioned in Chapter 4.
Weaker-Signal Capture
An interfering signal stronger than the desired signal may be
either adjacent-channel or co-channel. If an adjacent-channel signal
is stronger at the receiver input, there are several possibilities.
Arbitrarily good I. F selectivity can reduce the problem to that treated
above,, If, however, the receiver front end and the I F amplifier
are flat over the full frequency range covered by both signals so that the
interference arrives at the feedforward input unattenuated, the feed¬
forward technique can deal with it quite adequately under laboratory
conditions. Demonstration of this fact was the major accomplishment of
(7)
Small's thesis
The L F. amplifier must, however, fully include both signals
in its passband for optimum weaker-signal capture; if the interfering
stronger signal is on the "skirt" or sloping portion of the passband,
its amplitude at the feedforward input will vary dynamically with
modulation as its frequency rides up and down the sloping skirt. This
may cause the interfering signal to be sometimes stronger than the
desired signal and sometimes weaker, making any consistent adjustment
of the feedforward impossible* Even if the I. F characteristic is such
that the interfering signal remains consistently stronger, the inter¬
ference ratio a will vary over the modulation cycle, preventing an
26
optimum adjustment for K, which varies with a, this is explained
later.
Thus, it is necessary to employ an L F filter which passes
the interference without attenuation in order to effectively utilize the
weaker- signal capture capabilities of feedforward against a stronger
adjacent-channel interfering signal,, This technique is not an obvious
choice; it seems a bit strange to make no use at all of I. F. selectivity
to reject adjacent-channel interference, Great pains must be taken
to obtain good weaker - signal capture performance, especially under
field conditions, as explained later in this chapter,, The problems
involved are sufficiently important to raise serious questions as to
whether or not a feedforward plus a wide L F* filter offers any
advantages over a straightforward steep- skirted L F* filter in dealing
with stronger adjacent-channel interference, even if it is necessary to
go to the extreme of using a crystal or mechanical I* F filter.
Capture of the Weaker of Two Co- Channel Signals
The only signal-processing techniques besides feedforward
which allow capture of the weaker of two co-channel signals are the
recently developed ’’dynamic trap" technique^’ ^ and its variation,
the so-called ’’fixed trap” technique^' ^ Both of these ideas are
based on selectively reducing the amplitude of the stronger signal
with a notch filter* In the dynamic trap, the notch dynamically tracks
the stronger signal over the passband, while in a fixed-trap receiver
the stronger signal is ’’frozen” in frequency by a mixing process,
27
allowing the use of a fixed notch filter. Receivers using both techniques
have shown good performance in the laboratory, especially the fixed
trap receiver recently built by J. M. Gutwein^^. Although the
performance of this receiver is superior to that of any feedforward
receiver built to date, feedforward is inherently much simpler than
either trapping scheme.,
Effective application of the feedforward technique to weaker-
signal capture is thus an important problem, because of the attractive
simplicity of the technique as compared with the only alternatives The
idea of weake r - signal capture in general is also interesting, partly
because it was a problem generally considered insoluble until recently.
A practical high-performance receiver capable of capturing either the
weaker or the stronger of two co-channel FM signals would be
extremely useful; it would allow an FM system to continue operation
in the presence of intentional or unintentional interference from other
systems using the same channel, even if the interfering signal was the
stronger. This is an important extension in system capability. Such a
receiver would also allow ,,stuntsM such as multiplexing or simultaneous
two-way transmission on a single channel.
The present investigation is primarily concerned with using
the feedforward technique to capture the weaker of two co -channel FM
signals, since this is both the most interesting and the most difficult
problem connected with feedforward, as explained above.
28
Practical Requirements for Good Weaker -Signal Capture
A practical system patterned after the theoretical block dia¬
gram of Figure 5, Chapter 1, must meet several requirements if it
is to deliver good weaker - signal capture performance:
(1) The limiter must approximate as closely as
possible the action of the ideal narrow-band limiter described in
Chapter 1.
(2) The components in the outputs of the limiter and
amplifier at the frequency of the stronger input signal must be exactly
equal in amplitude and exactly opposite in phase at all times to insure
complete cancellation of the stronger signal*
(3) The feedforward amplifier must be linear*
The implications of these requirements and the problems of satisfying
them in a practical system will be discussed separately*
(1) Problems of Practical Limiters
Three types of amplitude limiters have been used in
previous practical FM systems: the pentode limiter, the diode limiter,
and the gated-beam limiter (usually employing the 6BN6 tube). The
salient characteristics of these three types differ somewhat; none
satisfies completely the requirements for an ideal feedforward limiter.
Good discussions of practical limiter problems are given in references
11 and 14.
29
The pentode limiter . Typical pentode limiters are
shown in Figure 4, Chapter 4. The grid capacitor charges from the
driving source and discharges through the grid resistor, clamping
the positive peak of the R. F. input voltage at zero or at a slight
positive voltage. The tube will conduct only over the portion of the
cycle between zero grid voltage and cutoff, if input voltage is large
enough, this time interval is approximately constant, and average
plate current is essentially independent of input voltage amplitude.
Screen and plate voltages are kept low to lower the grid cutoff voltage.
The pentode limiter is simple, cheap, and has no critical
adjustmentSo It is probably the most widely used type of FM limitero
However, for feedforward use, it has serious disadvantages: the
grid-circuit time constant R^ must be quite small for the limiter
to cope with reasonably large values of a and frequency difference rf
(3)
but it cannot be reduced indefinitely „ The grid capacitor must
remain significantly larger than the tube input capacitance, and the
grid resistor must remain much larger than the forward resistance
of the grid-cathode diode. A low grid resistor also results in low
input impedance, making the limiter hard to drive and hard to use
with tuned circuits. Also, the limiter characteristic of a pentode
limiter is usually gently rounded near the origin, so that its threshold
is high, and has an inescapable slight upward slope instead of being
perfectly flat. These disadvantages usually combine to make the simple
pentode limiter a poor choice for use in a weaker - signal capture
feedforward
30
The gated-beam limiter, Figure 14, Chapter 4, includes
a diagram of a gated-beam limiter stage using the 6BN6 tube0 The
operation of the limiter depends on the internal geometry of the 6BN6
tube, The electron stream is formed into a narrow beam which is
"gated" by the control grid; the tube’s plate current saturates when the
control grid rises a few volts above cutoffo
The 6BN6 limiter has a reasonably low threshold, a higher input
impedance than a low time -constant pentode limiter, and is free from
the time -constant problem of the pentode limiter 0 By careful tube
selection and bias voltage adjustment, an excellent limiter characteristic
can be achieved with the 6BN6,
The outstanding disadvantage of the gated-beam limiter is
that the characteristics of the one available tube, the 6BN6, vary over
quite a wide range from tube to tube. Some tubes are inherently
capable of better limiter performance than others, and observed
performance varies over a wide range, Among the "good" tubes, the
optimum bias voltages are different for each tube, requiring careful
individual adjustment. Moreover, the tube has a nonlinear input
impedance with a nonlinear reactive component, which disrupts tuned
(14)
circuits to which the input is connected, McLaughlin studies the
6BN6 in some detail; Gutwein^ ^ also studied gated-beam limiters.
The diode limiter, A very simple limiter can be made
from two diodes biased to clip symmetrically. Semiconductor diodes are
more convenient than thermionic types, and no bias is necessary if
31
silicon diodes are used because of their 0„ 5 volt threshold before
forward conduction The basic limiter could hardly be simpler, and
there are no time -constant problems and no adjustments., The per¬
formance available depends on diode characteristics, such as forward
and reverse resistance and switching time.,
The disadvantages of the diode limiter are two no gain and
very low input and output impedances 0 It is therefore usually
necessary to use two tubes per limiter stage if tuned interstage filters
are used; both tubes act as amplifiers, with the diode limiter between
the amplifier stages.. The diode limiter should be capable of excellent
performance, if designed carefully and built with high-performance
diodes It has not been widely used in narrow-band applications, and
further investigation is indicated, as noted in Chapter 7„
(2) Maintaining Accurate Interference Cancellation
A host of problems prevent perfect cancellation of the
fundamental stronger-signal component in a practical feedforward
circuito Some would exist even with ideal system components, while
others arise from equipment imperfections., The various disturbing
influences will be discussed separately,,
Variations in a„ The results of Baghdady's analysis
of the output spectrum of a limiter with two-signal input ^ show that
even with constant stronge r - signal amplitude at the limiter input,
the amplitude of the fundamental stronger-signal component at the
limiter output varies with a in a manner shown by Figure lo Therefore,
32
33
even though the stronger signal amplitude at the feedforward
amplifier input remains constant, the feedforward amplifier gain (and
hence K) must vary with a to maintain perfect stronger -signal
cancellation The optimum value of K as a function of a is numerically
equal to the normalized stronger - signal output amplitude (A^) of
Figure 1; this quantity is designated K<x> and plotted by Baghdady^.
Variations in Input Signal Amplitude,, In the simple
basic feedforwar d of Figure 5, Chapter 1, the amplitude of the stronger -
signal component at the amplifier output obviously varies directly with
the stronger-signal amplitude at the feedforward inputo Cancellation
is therefore perfect for only one value of input signal amplitude,
(7 81
In previous experimental investigations of feedforward ’ , a
narrow-band limiter was used ahead of the feedforward proper in an
attempt to hold the feedforward input and hence amplifier output at a
constant level, maintaining proper cancellation despite rapidly
fluctuating stronger-signal amplitude,. This end is achieved at a
price : the threshold value of a below which worthwhile weaker-
signal capture cannot be achieved is raised by about 6 dbo This has been
observed experimentally (see Chapter 5) and can be predicted
theoretically, as follows: a weaker- signal feedforward receiver of
even reasonable performance is capable of capturing the weaker signal
down to at least a = 0o 5, For a < 0, 5, limiter requirements are
not particularly severe (see Reference 3h Therefore, the theoretical
34
reduction in a of 6 db is very closely approached by even a poor
pre-limiter0 This means, of course, that the value of a below
which the system becomes useless is effectively doubled by the use of
a pre-limitero Or, for a specified threshold value of <1, allowable
tolerance on circuit components is effectively halved by use of the
pre-limiter as explained later0 This difficulty is usually not serious,
if a threshold no lower than a = 0o 1 or 0„ 2 is desiredo
A second source of trouble with the pre- limiter system is the
difficulty of building a one -stage limiter of good enough performance
(flat enough limiter characteristic) to maintain a high degree of
constancy in the input signal amplitude to the amplifier, as brought
out earlier in this chapter0 Any irregularities in the performance of the
pre-limiter will adversely affect the capture characteristic of the
system. Therefore, in a high-performance system the pre-limiter
must be designed and constructed very carefully,, Even with an ideal
pre -limiter, the optimum value of K is still a function of a (see
Figure 6 of reference 6)„
When the fluctuations in input signal level are sufficiently
slow, there are two alternative solutions to the problem which avoid
the threshold degradation of a pre-limiter at the cost of increased
complexity. One way is to use a slow-acting pre -limiter; i0 e, , a
limiter with a time constant which is long compared to the slowest
important variation in input signal level due to modulation or inter -
signal interference but fast enough to compensate for '’long-term"
variations in input signal strength,, In ramified form, this suggestion
35
grows into a sophisticated AGC system of high performance but long
time constant* which controls the gain of everything ahead of the
feedforwardo Such a scheme should compensate for slow variations
in input signal level without reducing the interference ratio at the
feedforward input and thereby raising the threshold value of a*
The other method of avoiding the pre - limiter when slow signal
level variations are encountered, also offers potentially better
performance at the price of increased complexity* The idea is to use
an over-all feedback control system which would automatically adjust
the gain of the feedforward amplifier to minimize distortion in the
captured weaker- signal message at the demodulator output* Such a
system would compensate for variations in feedforward limiter output
level and amplifier gain due to supply voltage, temperature
variations and component aging and for the variation in the optimum
value of K for different a and different degrees of modulation.
The range of allowable input signal amplitude variation would be limited
only by the dynamic range of the L Fc amplifier and feedforward limiter
and the range over which the feedforward amplifier gain could be
controlled automatic ally 0 A suggested design for such a system is
given in Chapter 7,
Limiter Imperfections* Any departure of the feedforward
limiter characteristic from the ideal of Figure 1. Chapter 1, will allow
the amplitude of the strong e r~ signal component in the limiter output
to vary with limiter input signal amplitude in addition to the inescapable
variation with a mentioned earlier* contributing to imperfect cancellation.
36
Bandpass Filter Difficulties, Ideally, all portions of a
feedforward receiver ahead of the feedforward proper (front end and
I. Fc amplifier) should have a completely flat amplitude -vs -frequency
characteristic over the entire range of frequency deviation of both
stronger and weaker signals,, Otherwise, the amplitudes of the two
signals at the feedforward input will vary as their frequencies sweep
over the passband, upsetting perfect stronger-signal cancellation over
part of the modulation cycle,, Achieving and maintaining such a flat
I. Fc characteristic while retaining steep skirts for adjacent-channel
interference rejection requires considerable effort,, When an L Fc
stage is overloaded by too much input signal, its grid conducts and
places a heavy, non-linear load on the tuned circuit connected to the
grid, distorting the frequency response of the tuned circuit,, A
practical limiter has a relatively low, non-lmear input impedance, which
loads the tuned circuit to which it is connected similarly distorting its
frequency response,,
It is also necessary, of course, to accurately match the
amplitude and phase characteristics of the parallel limiter and
amplifier channels over the entire signal bandwidth,, In the simpler
feedforward systems* this is easily done by combining the amplifier
and limiter outputs before the limiter filter instead of after it„ This
makes no difference in the basic theory, of course, since the components
which canid each other are unaltered by the ideal filter, making
combination before filtering equivalent to combination after filtering
More complex systems which may include frequency-sensitive
elements in each channel require greater effort to achieve matching over
the passband.
37
Ordinary Circuit Difficulties 0 Familiar design
problems such as component aging, temperature variations, supply
voltage variations, and shock and vibration can be rather troublesome
in a feedforward receiver designed for small values of a, as will be
explained quantitatively in the next section0
Degree of Precision to Which Requirements Must be Met.
The specifications for the various components in a feedforward receiver
have been discussed qualitatively from the standpoint of ideal require¬
ments and the practical considerations which prevent the requirements
from being met exactly0 The next question is, exactly how closely must
practical system components approximate the ideal? What exactly
are the allowable tolerances on various parameters 7 It happens that
the necessary precision varies rather widely with the input interference
ratio a «
It is generally not too difficult to build a practical feedforward
circuit which will capture the weaker co “channel Signal at an a of about
0o 5. As a increases toward 1, or decreases toward zero, weak-signal
capture becomes increasingly difficult The problems when a ^ ^ 0„ 5
are different from those when a ^ 0, 5« There is no real theoretical
significance to the value a = Ck 5; it is merely a convenient "bench mark".
Difficulties for a ) 0. 5o As a becomes larger than
0o 5 and approaches 1, the shape of the passband of that part of the
receiver ahead of the feedforward becomes increasingly important. If the
passband is not flat over the modulation bandwidth, the stronger and
38
weaker signals may exchange roles over a portion of the modulation
cycle. An upper bound on the allowable departure from flatness is
given by lOO(l-a) per cent, since such an error will make the two
signal amplitudes instantaneously equal at one point in the modulation
cycle.
Another effect is equally troublesome for 0, 5 ( a ^ 1: the
(3)
requirements on the feedforward limiter become very stringent
The maximum allowable limiter time constant approaches zero, and the
requirements on limiting threshold and range of input amplitudes over
which the limiter must remain saturated become more severe. Moreover,
feedforward action is degraded because the maximum reduction in a
available from a limiter declines from 6 db at a ( 0, 5 to zero at a = 1.
Since practical limiters have a non-zero threshold, required limiter
drive to maintain saturation quickly becomes unreasonably high. The
net effect of these problems is that above some a , 0„ 5 ( a / 1,
a practical limiter will fail to perform adequately in a feedforward system.
It is usually possible to solve the problems mentioned adequately in
the range 0. 5 { a { 0. 8 or 0o 9» (See experimental results, Chapter 5. )
Difficulties for a ^ 0. 5. One fundamental problem
arises at small values of a which is the source of a host of secondary
difficulties: the maximum allowable variation in the value of K becomes
very small at small a , being approximately equal to 100(a/2) per cent
for a ^ 0, 5, The outer limits on K as a function of a are plotted
by Baghdady^; they are derived as follows. Obviously, if the
39
cancellation of the stronger - signal component in the limiter output
by that in the amplifier output is not perfect (incorrect value of K)f
there will be a residual stronger-signal component in the feedforward
output. If this residual component is equal to, or greater in amplitude
than the residual weaker-signal component, capture of the weaker-
signal component by a stronger - signal demodulator is impossible. The
values of K at which the residual stronger-signal component is equal to
the weaker-signal component are easily derived; these boundary values
are given by
A + A
Lower Bound = 1 = - -
1 + a
A - A
Upper Bound = PUW1 = — - -
1 - a
m which. Aq - normalized amplitude of stronger - signal
component at limiter output
A i = normalized amplitude of weaker - signal
component at limiter output
The values of A and A t as a function of a have been
o-l —
computed (1) and are given in Figure L Computation of an(^
^ LW1 straa§ktforward with a knowledge of A^ and A ^ ; these
boundary values are plotted in Reference 6 as Figure 3* They are
here replotted on a logarithmic scale with expanded abscissa to bring
Ul
ib Z
H
O
O
O
o*
6
c-
tr-
O
vO
<r
O
vn
<r
O
42
out their behavior for very small a, along with ~A^, the value
of K which gives perfect stronger-signal cancellation
Figure 2 shows very clearly the drastic tightening of the outer
boundaries on K as a decreases, Figure 3 is an expansion of the
portion of Figure 2 below a = 0, 1, and shows that the allowable
variation in K (denoted by£) quickly falls from 5 per cent to less
than 1 per cent as a decreases,. This is a pretty stiff requirement in
terms of practical equipment, since it means that limiter output level,
feedforward amplifier gain, and feedforward amplifier input voltage
must all be held to a precision of better than £ per cent over the
entire receiver passband, despite short-term variations in input signal
amplitude, component values, and supply voltages*
Neglecting the familiar perturbations due to such things as
component drift and supply voltage changes, the requirements on K
mean that the limiter characteristic must be flat to better than £ per
1 + *
cent over a range of ^ - : 1 in input voltage0 Fortunately, a
i ■ a
smaller portion of the limiter characteristic is involved at smaller a«
The requirements also mean that all portions of the receiver ahead
of the feedforward must have a frequency characteristic that is flat
within £ per cent over the entire range of frequency deviation of the
input signals, as must the bandpass filter following the feedforward
limiter0 This requirement would not be nearly so difficult were it
not also necessary for the frequency characteristic of the front end
and L Fe to slope off sharply at the band edges to reject adjacent -channel
43
interference and for the feedforward limiter filter to cut off sharply
in order to reject as many of the additional components introduced by
limiting as possible Thus, as a decreases, the required bandpas s
filter shapes required for reasonable performance approach the ideal
rectangular shape very quickly,,
From an engineering point of view, the precision of better than
1 per cent required to capture weaker signals below a = 0, 02 (see
Figure 3) is very difficult to achieve. Thus, a feedforward for the
capture of weaker signals below a = 0* 02, though conceptually very
simple, would be anything but simple to design and construct, and
would be rather unattractive from a practical or economic standpoint.
(3) Feedforward Amplifier Linearity
Linearity of the feedforward amplifier is usually not
too difficult to obtain with reasonable care in design. The amplifier
1+2-
must be linear over a range of — * — - to 1 in input signal amplitude
i *■ a
when no pre-limiter is usedo Since its gam must be controllable,
the amplifier must maintain linearity for all gain control settings,, It
is also necessary to insure that the largest receiver input signal to
be encountered will not overload the amplifier,,
A Minor Practical Problem, In practice, the amplitude
of the residual weaker- signal component in the feedforward output is
usually too low to adequately drive the first limiter in the demodulator,
making it necessary to provide a single-stage amplifier after the simple
feedforward of Figure 5, Chapter 10 For the feedforward to work over a
44
wide range of a, this amplifier must have a reasonably wide dynamic
range, since residual weaker-signal output varies with a_. This point
(7 81
was missed in earlier feedforward designs' * , and no amplifier was
included; this reduced system capabilities considerably.
CHAPTER 3
RELATIVE MERITS OF VARIOUS PRACTICAL
FEEDFORWARD CIRCUITS
Summary of Requirements
The construction of a high-performance feedforward requires an
excellent narrow-band limiter, a linear amplifier of readily controll¬
able gain, bandpass filters which are accurately flat over the modula¬
tion bandwidth, and a circuit arrangement which provides an accurate
180° phase difference at the limiter and amplifier outputs. The quanti¬
tative requirements on these various components were discussed in
Chapter 2.
Possible Basic Circuit Configurations
The Transformer-Input Feedforward
Figure 1 is a basic diagram of one possible feedforward
circuit. The outputs of the limiter and linear amplifier are combined
in phase by simply adding their plate currents in a common plate load
The 180° phase difference is achieved by the U9e of a tuned transformer
with a center-tapped secondary, its primary being fed from the plate
of the preceding stage (the last I. F, stage if no pre-limiter is used).
LIMITER
Figure 1: Basic Transformer -Input Feedforward Circuit
47
The inputs to the limiter and amplifier are connected to opposite ends
of the balanced secondary winding, which are close to 180° apart over
the passband of the tuned transformer.
The limiter and amplifier outputs are combined ahead of the
limiter filter, as explained in Chapter 2, The value of K is controlled
by the bias potentiometer in the cathode of the variable-p amplifier
tube .
One primary advantage of this circuit is its simplicity, there
being only two tubes and four tuned-circuit adjustments in the feedfor¬
ward proper. It is adaptable to almost any frequency at which the
tubes will function well and at which a suitable input transformer can
be built.
The main problem of the circuit is loading of the tuned input
transformer by the limiter0 A limiter input usually presents a low
impedance, non-linear load which can distort the passband shape of the
tuned transformer. The non-linear load on the last I, F stage also
means that the input signal to the feedforward amplifier will be partially
limited, which is undesirable. The bad effects of non-linear loading
may be eliminated by sufficiently lowering the impedance level of the
transformer, but an engineering compromise is necessary, since lower
impedance at this point means less voltage to drive the limiter for a
fixed g^ in the last I. F* stage0 Use of a broadband untuned input
transformer might alleviate the problem somewhat; recent advances in
ferrite core materials and winding techniques have made construction
of such transformers quite feasible for frequencies up to 50 me.
LIMITER
•rA
3
u
w .ij
4-1 7*
u
»r4
c
Figure 2: Basic Transformer -Output Feedforward Circuit
49
The Transformer-Output Feedforward
Figure 2 indicates another possible way in which a
center -tapped transformer may be used to obtain correct phase relation¬
ships in a feedforward circuit. The limiter and amplifier inputs are
fed in phase, their outputs being combined subtractively in the trans¬
former. This circuit arrangement preserves the simplicity of the trans¬
former-input feedforward while avoiding the problem of limiter loading
on the phase-inverting transformer. If used in its simplest form, i e. ,
if fed through a double -tuned circuit from the plate of the last high-gain
I- F. stage, loading of the tuned circuit by the limiter will, of course,
be a problem. The addition of a driver stage with a broadly tuned, low-
impedance output circuit sacrifices the inherent simplicity of the circuit
to solve the loading problem.
The Driver-Limiter Feedforward
The feedforward circuit of Figure 3 avoids the problems
of limiter loading and design of a center-tapped tuned transformer. The
limiter is driven by a linear amplifier with a very low Q single -tuned
plate circuit. The necessary 180° phase difference between channels
is obtained by the use of two stages in the limiter channel and one in the
amplifier. The tuned circuit between the driver amplifier and the
limiter must have a low enough Q to have negligible phase shift over the
passband of interest, and must have a low enough impedance so that the
non-linear load represented by the limiter input will not affect its
characteristics. The two amplifier grids connected to the last tuned
circuit in the I. F. amplifier have a negligible loading effect.
LIMITER
50
a
CO
UQ
rt
(X M
XI V
S3
CQ hi
- If-
nrmwrcn
— I( —
CQ
d
u
«J -H
rt y
■s.2
M ~0
o 5
^ T>
r-l t>
01 o
tn ^
Q0
rt
(Q
+
!•
Figure 3: Basic Driver -Limiter Feedforward Circuit
51
The driver-limiter feedforward has several controls and com¬
plications not shared by the two simpler circuits described earlier
such as the gain control on the driver amplifier and the tuning of the
driver amplifier plate circuit. However, these extra controls provide
advantages as well as complications. The gain of the driver amplifier
may be adjusted to insure that the limiter remains saturated at all
times and that the input signal amplitude variations occupy the flattest
and "best" portion of the limiter characteristic at small a The
interstage tuning control can be varied slightly to adjust the phase
difference in the channel outputs to exactly 180°.
The Grounded-Grid- Amplifier Feedforward
Figure 4 is a diagram of the circuit used in two previous
(1 8 \
investigations of feedforward, ’ The amplifier and limiter are
connected in phase at both input and output; the necessary phase dif¬
ference in the two channels is achieved by using a grounded-grid
amplifier. A cathode follower is necessary to provide a low-impedance
source to match the low input impedance of the grounded-grid amplifier.
The circuit appears reasonably attractive at first glance, but
a second look reveals some serious fundamental difficulties, most of
them connected with the cathode follower. The circuit uses a minimum
of three tubes but offers few if any compensating advantages over the
simpler arrangements using two tubes. The cathode follower is usually
thought of as extremely stable, linear, and trouble-free, but these
generalizations no longer hold when it is used with tuned circuits and at
LIMITER
52
i - L-— If-— J —
+
-
.a
4)
U 4)
4) ts£)
M (4
Figure 4: Basic Grounded -Grid Amplifier Feedforward Circuit
53
frequencies in the megacycle region, such as are commonly used as
intermediate frequencies in FM receivers. Special precautions may be
necessary to prevent oscillation when a tuned circuit is connected to a
cathode -follower grid; the stray capacitances combine to form a Colpitts
oscillator circuit, as shown in Figure 5. Also, the stray capacitance
from cathode to ground of the cathode follower may be quite large,
since it consists of cathode -heate r capacitance plus plate -cathode
capacitance plus the input capacitance of the load, which in the present
case consists of heater-cathode and grid-cathode capacitance of the
tube in the grounded grid stage. The total shunt capacitance may amount
to 50 mmf or more, unless the impedance from cathode to ground is
kept extremely small the shunt capacitance can easily slow the rise
time of the cathode circuit to such an extent that the cathode -followe r
grid will rise quickly to the grid conduction point or fall below the
cutoff point before the cathode voltage can change correspondingly. The
result is non-linearity and clipping* If the cathode impedance is made
low enough to prevent clipping, it becomes difficult to provide enough
voltage at the cathode follower output to drive the limiter adequately
without resorting to tubes of extremely high g^ for the cathode follower,
which in turn aggravates the problem of oscillation
In the versions of this circuit which were actually constructed,
a potentiometer in the cathode follower output was used to control the
feedforward amplifier gain* This is a rather dubious method of gain
control for frequencies in the megacycle region, but there is no obvious
(7, 8)
alternative*
I
X
a>
C
a
4>
X
o
A
nJ
v
i
cn
t: >■
jg* c
o «
U
00
<L>
6
O
O
0>
A
•M
3
a
c:
0)
•M
■M
o o
1_4 f- 4
O 3
s *
• H
o) Z
X
c*I -
nJ
3 Si
5 g.1
S.2§
Om§
O X «T*
P:“o
£TJ3 in
•
c
o
a
cj
u
00
O
cn
co
0)
(X
(X
3
oo
(X
r— 4
a>
Figure 5: Oscillator Circuit Formed by Cathode Follower with Tuned Input Circuit.
55
The problems of this circuit would be somewhat less formidable
at low Io F, frequencies where the cathode follower circuit would be
less troublesome. A step-down transformer may be used to match the
low input impedance of the grounded-grid amplifier; however, this would
require very high primary voltages to provide enough drive at the
secondary to drive the limiter adequately. It is very difficult to see
any advantage to this circuit over the others mentioned in this chapter
at normal FM I F frequencies of several megacycles*
Use of a Split-Load Phase Inverter
The 180° phase difference necessary in a feedforward
circuit might be conceivably obtained from a split-load phase inverter
of the type popular in audio amplifiers, in which outputs are taken
from both plate and cathode of a single tube. However, this approach
would be useful only at rather low I. F 1 s since the phase inverter
circuit has the same troubles as a cathode follower plus a few of its
own when used at high frequencies
56
CHAPTER 4
DESCRIPTION OF EXPERIMENTAL EQUIPMENT
Laboratory Receivers
Figure 1 is a block diagram of the type of laboratory-model
receiver used in the experimental measurements. Since all measure¬
ments were made with signal generators, an R. F, stage and mixer
were unnecessary and were not included, signals being fed in at the
intermediate frequency (10o 7 me). The feedforward circuit was in¬
serted between a more or less conventional L F. amplifier and a high-
capture - ratio demodulator. The entire system was designed around
the standards used in F M broadcasting* 10 7 me I F. and + 75 kc
maximum peak deviation.
In order to compare the performance of different types of feed¬
forward circuits, experimental models were built of the three most
promising circuits of Chapter 3 the transformer -input circuit, the
transformer -output circuit, and the driver-limiter circuit. Each was
tested separately; the same demodulator and I. F. amplifier were used
for all tests. The transformer-input feedforward was built first and
thus included a pre -limiter. The driver-limiter feedforward, con¬
structed next, used a pre-limiter at first and was later modified to
eliminate it. The transformer -output feedforward was built initially
without the pre-limiter.
57
U
o
■M
rH
P
O
6
V
Q
Figure 1
Block Diagram of Laboratory Model Receiver
58
From these basic component circuits, three different ,fmodelsrr
of the feedforward receiver of Figure 1 could be assembled. The
various receiver components will be described separately,,
The I. F- Amplifier
Figure 2 is a schematic of the I. F« amplifier used in the ex¬
perimental receivers. Most of its design features are strictly conven¬
tional, The resistive input network is designed to properly terminate
the two signal generator cables and to provide isolation between gener¬
ators. Multiple bypass capacitors connected to different ground lugs
are used at several points; this arrangement greatly improved the
stability of the amplifier. Shielded power cables plus the isolation
chokes shown in the power leads were necessary to provide isolation
between stages and between the amplifier and other units sharing the
common power supply.
Commercial 10o 7 me I0 F. transformers of the type commonly
employed in F0 M. broadcast receivers were used in the amplifier, for
several reasons. They are compact, inexpensive, readily available,
and well shielded, and their use greatly simplifies the construction of
equipment using tuned circuits. Savings in construction time were
quite valuable, since a considerable amount of hardware had to be built
in the limited time available for the investigation.
Unfortunately, the degree of precision necessary in the I. F«
amplifier response in a feedforward receiver (see Chapter 2) was not
fully appreciated at the beginning of the investigation. Therefore, the
All tubes 6AU6 or 6BA6; all transformers Miller 1463.
59
ft
U
u
•H
r-4
a
6
<
Ui
4)
00
(4
<M
CO
I
ro
4)
u
CtJ
g
4>
X
U
CO
rg
4)
3
00
M
u
6o
40 kc per diviaien
lft kc ft diviaien
25 kc per dlviaion
The 1. F. amplifier frequency -re epeaae curre shown above
waa meaaured at the same time aa the capture plot of Figure 19,
Chapter 5.
The alignment adjustments are the same for all three
picture a; only the frequency-ecale calibration (horiaontal acale) ia
different, aa indicated. The vertical acale ia linearly calibrated in
relative amplitude.
Figure 3
Frequency-Reaponae Curvea of I. F. Amplifier
61
worst disadvantages of the commercial transformers (very poor
stability, relatively high impedance, rounded frequency characteristic)
did not appear too serious to begin with. It later became apparent that
the tuning adjustments of the transformers drifted with time, tempera¬
ture, and vibration to a degree which was quite tolerable in a broadcast
receiver but inacceptable in a precisely adjusted feedforward receiver.
It was therefore constantly necessary to touch up the alignment of the
experimental receiver to obtain best performance. The commercial
transformers were improved considerably by opening their cans and
sliding the coils closer together to increase the coefficient of coupling
and flatten their frequency response. The first transformer used in the
Ic F. amplifier was overcoupled, the second undercoupled slightly, and
the third loaded somewhat and critically coupled to obtain the flat-
topped over-all response of Figure 3,
It was discovered that the use of 6BA6 tubes in the amplifier
resulted in slightly less gain than 6AU6‘s but a wider dynamic range,
i.e. , the response curve preserved its shape over a wider range of
input voltages - an important advantage.
The Demodulator
Figure 4 is a schematic of the receiver demodulator. It
employs four pentode -type limiters; commercial I. F„ transformers
identical to those in the I F. amplifier were used as interstage narrow-
band filters. The discriminator is basically the conventional Foster-
Seeley type. The audio section includes a cathode follower to isolate
62
Figure 4: Schematic of the High-Performance Demodulator with Later Model Audio Section
s (Part 1 of 2 parti)
62a
K
TJ
O
3
C
• H
W
fi
O
u
TT
V
3
00
.«H
The Audio Section
63
the discriminator output circuit from following stages and from test
equipment used to observe the discriminator output directly. This
arrangement keeps the discriminator output capacitance low, insuring
the low discriminator time constant necessary for high capture
(3)
ratio The audio section also includes one conventional voltage
amplifier stage and provision for switching in either a bandpass speech
filter (see Figure 5) or varying amounts of R-C de-emphasis. The
earlier model audio section shown in Figure 6 included provision for
R-C de-emphasis plus low-pass filtering (Figure 7 and 8) and was
used for initial measurements before the design of Figure 4 was
evolved as a much better compromise for the reception of speech-
modulated signals, as explained in Chapter 5*
Figures 9 10, 11, and 12 show the dramatic improvement in
the performance of a practical demodulator which can be obtained by
applying the theoretical principles outlined in References 1 through 4.
Figure 9 shows the capture plot of the demodulator as originally built,
with reasonably low limiter time constants, A commercial broadcast-
type discriminator transformer with 320 kc peak separation was used
in the manufacturer’s recommended circuit. Figure 10 was obtained by
lowering the discriminator time constant to 3 psec and reducing the
limiter time constants somewhat. Figure 11 shows the improvement
obtained by substituting a commercial wideband discriminator trans¬
former (900 kc between peaks) which is sold for high-fidelity tuners;
no other changes were made from the conditions of Figure 10 The
characteristic of Figure 12 was obtained by merely lowering the dis¬
criminator time constant to 1. 4 psec.
64
Ul >
CD O H
O 5 <
W o £
0.0,2
to _
O^o
c* ot *n
V\ o tr.
o
FREQUENCY IN CYCLES PER SECOND
means 1000 ohms*
65
Figure 6: Early-Model Audio Section used in Demodulator
66
?.o ioo icc*>
FREQUENCY IN CYCLES PER SECOND
67
68
o
69
70
900-kc Discriminator Transformer (Miller No. 1464- WB) substituted for original unit.
71
o
except Discriminator Time Constant Lowered from 3 fisec, to 1. 4 psec.
72
Figure 12 is an example of the demodulator performance
obtainable with simple circuitry and noncritical commercially available
parts by careful attention to the really important design factors
Gutwein obtained essentially equivalent performance by using the
same discriminator circuit with three fast-acting 6BN6 limiters instead
of the four pentode limiters used here
The best approach to designing practical high-performance
demodulators seems to be to use the widest bandwidth discriminator
which can be conveniently built} considering requirements on sensi¬
tivity, audio hum and noise, and complexity0 One to four narrow-
band limiters followed by the usual wideband limiter should then be
used ahead of the discriminator to improve capture performance,
The wideband discriminator transformer used in the experi¬
mental demodulator (5 to 6 I F0 bandwidths) is sold as an noff-the-
shelf11 commercial item, demonstrating its practicality Its one
disadvantage is its reduced audio output compared with that available
from a narrow-band unite If this reduced output should necessitate
an additional audio stage in a receiver, design and construction of this
one stage would be far simpler than adding one or more additional
narrow-band limiters in order to obtain equivalent performance with
a narrow-band discriminator
The capture performance shown in Figure 12 is not a great
deal better than that theoretically obtainable from a single wideband
limiter plus a discriminator of the bandwidth used This seems to
indicate that the three narrow-band limiters used are delivering no¬
where near the interfe rence - rejection performance theoretically ob-
73
tainable from three ideal limiters The conclusion is not too sur¬
prising, since the pentode limiters employed have many shortcomings:
the shape of their limiter characteristics is poor; the interstage
tuned circuits leave much to be desired; and their grid time constants
are marginal; though they have been reduced as much as possible.
Probably their worst problem is insufficient drive from one limiter to
maintain saturation in the next for large . Therefore, the per¬
formance of the demodulator could probably be improved somewhat
without adding significantly to its complexity by the rise of better
limiterSo
The Transformer-Input Feedforward
Figure 13 is a schematic of the experimental transformer input
feedforward. It employs pentode limiters quite similar to those in the
modulator of Figure 4, and includes a pentode pre -limiter ahead of the
feedforward proper.
The amplifier following the basic feedforward is included to
raise the low- amplitude residual output signal from the feedforward
proper to a level sufficient to drive the demodulator adequately. Use
of the amplifier stage also makes it possible to use two double -tuned
transformers instead of one in the filter following the feedforward.
Commercial transformers are used as in the I, F. amplifier. The
amplifier design is strictly conventional, using a vanable-p 6BA6 tube
with adjustable cathode bias to provide control of amplifier gain. Screen
voltage is supplied from a voltage divider to minimize its variation with
cathode bias.
74
The feedforward amplifier uses the same basic circuit as the
demodulator driver amplifier. Its gain control provides the operating
control for the parameter Mkrl. This basic amplifier design is used as
a “building block11 in all of the experimental feedforward systems to be
described
The center-tapped feedforward input transformer is a modified
commercial narrow-band discriminator transformer. The modifications
consisted of removing the coupling capacitor, adding tuning capacitance
to the primary, and sliding the coils further apart to decrease the co¬
efficient of coupling. A transformer carefully designed and specially
constructed for the job would undoubtedly have been better; however,
its construction would have also taken a great deal more time than
modifying the commercial unit, and performance of the modified dis¬
criminator transformer seemed adequate to deomonstrate the feasibility
of the circuit.
The A, M, detectors are included to allow frequency response
curves of various parts of the system to be dynamically plotted, as ex¬
plained in Chapter 5. They are designed to cause a minimum of circuit
loading, and are wired in permanently. The basic detector circuit is
also used in the other feedforwards.
Experimental Transformer - Output Feedforward
The transformer -output circuit of Figure 14 employs the same
type of modified discriminator transformer as the circuit of Figure 13,
but in a different manner. Note that no pre -limiter is included in this
circuit and that a 6BN6 gated-beam limiter is used instead of a pentode
75
H)
U)
Figure 13: Circuit of the Experimental Transformer -Input Feedforward
76
limiter. Provision is included for independently adjusting the bias
voltages on all three 6BN6 grids in order to obtain the best possible
limiter characteristic. The amplifier circuits are identical to those
of Figure 13.
Experimental Driver-Limiter Feedforward
The schematic of the experimental driver-limiter feedforward
in its final form is shown in Figure 15, The basic limiter and amplifier
circuits used are the same as those of Figure 14, phase opposition
being obtained in a different manner. The driver -limiter circuit was
first built using two pentode limiters, one as a pre-limiter, it was
later modified to the circuit shown to obtain improved performance.
The only circuit feature not covered in descriptions of the
other feedforward circuits is the plate circuit of the driver amplifier.
The variable inductor (slug-tuned) resonates with circuit capacitance
at the operating frequency of 10 7 me.
The low limiter input impedance loads this low-C tuned circuit
heavily; the 5600-ohm 6BA6 plate load resistor provides additional load¬
ing and swamps the non-linear limiter input impedance to some extent,
since the limiter input impedance varies between about 5000 and 20, 000
ohms depending on the input voltage. In practice, the phase character¬
istic of the low-Q tuned circuit was more important than its amplitude
characteristic, which was quite broad. Detuning of the slug-tuned coil
resulted in excessive phase shift and consequent cancellation of the
stronger signal long before the reduced limiter drive caused any trouble.
77
C>
H
P
Si
u<
Circuit of the Experimental Transformer -Output Feedforward
70
CO
-£>
Tf
P-H'
Ph
V
l\
looooooo.
rvwmir
- U-^r-.
~b
- \
c
M
O
+j
■rt
u
nj
a
rt
u
o
o
o
5
a>
8
x
rn
sO
1 / °*
M 1
UmmmJ
_ O
Figure 15: Circuit of the Experimental Driver -Limiter Feedforward
79
CHAPTER 5
THE EXPERIMENTAL MEASUREMENTS
The Experimental Setup
Figure 1 shows the arrangement used for experimental
measurements on the laboratory model feedforward receivers. Two
FM signal generators provide modulated signals at 10. 7 me with peak
deviations up to + Z40 kc (normally + 75 kc). The nominal output
amplitude of the signal generators is continuously variable between 0. 1
5
and 2X10 microvolts. Both generators can be modulated by internal
audio oscillators at frequencies of 50, 100, 400, 1000 or 5000 cycles;
higher modulating frequencies are also provided but were not used.
The generators can also be modulated by an external low-impedance
audio source; an audio oscillator or a specially built modulator (Figure
2) is used for this purpose. The modulator allows direct speech
modulation or modulation by a thyratron random noise generator or
other source and includes provision for speech clipping and filtering
such as is commonly used in communication transmitters
The audio output from the feedforward receiver was monitored
at all times by an oscilloscope which was extremely useful for initial
adjustments and for determining in detail what was happening to the
modulation waveform, A harmonic wave analyzer, essentially a tunable
filter with a bandwidth of a few cycles, was used to examine the
amplitude of individual frequency components, while total distortion
measurements were made with an ordinary null-type distortion analyzer.
Scop«
80
81
Equipment Used in Setup of Figure 1
Signal Generators!
No. 1: Boonton 202 -B
No. 2: Boonton 202 -C
Both used with 203-C Univorters
Audio Oscillators:
Hewlett-Packard 200- B
Wave Analyzers:
General Radio 73 6- A
Distortion Analyzer:
Hewlett-Packard 305
Oscilloscope:
Either a Du Mont 304-H or a Tektronix 515, the
latter being used for photographs.
82
A tape recorder and speaker amplifier were used for subjective
listening tests.
Figure 3 shows the arrangement used to plot dynamically the
frequency response curves The method is standard, and has been des¬
cribed adequately elsewhere. Such an arrangement is essential when
working with a feedforward receiver because of the importance of the
passband shape of the various filters involved. The method can be used
to plot the response curve of the single filter following a limiter in¬
dependently of the other filters, because the limiter action removes the
effect of the amplitude variations caused by preceding filters.
A similar method was used to plot dynamically the amplitude
characteristics of limiters, as shown in Figure 40 This is a slight
modification of the method used and explained by Gutwein Its
major limitation is inability to plot limiter characteristics over a very
wide range of input amplitudes due to the limited capabilities of the
signal generator modulator, namely, the impossibility of linear 100%
modulation in the negative direction This limitation could be overcome
to some extent by varying the signal generator R F output control to
examine different portions of the limiter characteristic.
Measurement Procedures
Techniques for evaluating the performance of the feedforward
receivers were in part developed as the experimental work proceeded,
since, of course, no standardized procedures yet exist for measuring
the performance of weaker signal capture receivers. Several of the
measurement techniques were improved while the work was in progress;
83
>
d ^
a o
ri 5
r*i ^
<vi
d
c*
Block Diagram of the Signal-Generator Modulator
84
therefore, different procedures were used at different stages of the in¬
vestigation. It is usually possible, however, to obtain some comparison
between early results and those obtained later by slightly different
procedures .
The so-called "capture plot" is one method for portraying
graphically the performance of FM receivers designed to separate co¬
channel signals of different amplitudes,. Figures 9 through 12 of Chapter
4 are examples of capture plots. They are obtained by feeding two
signals of known amplitudes into the receiver input, the two signals
being modulated by sinusoids of different frequencies. At each value
of relative amplitude (interference ratio a), harmonic wave analyzers
are used to measure the amplitudes of the two modulating frequencies
at the demodulator output. The demodulator output amplitude at each
frequency is plotted as a percentage of the value it would have if the
signal in question occupied the channel without interference. This per¬
centage is plotted vertically for each signal and is called "per cent
capture"* The interference ratio a is plotted on the horizontal axis.
The capture plot thus presents a comprehensive picture of how well the
desired signal modulation is captured as a function of a and what fraction
of the modulation from the undesired signal leaks through at each value
of a e Furthermore, both of these measurements are independent of
the type of audio filtering and/or de -emphasis used in the demodulator,
since they are expressed as percentages of audio amplitudes in the
absence of interference.
The simple capture plot described above has one disadvantage,
it gives only an approximate idea of the quality of the captured signal
04a
Figure 3
Method Used to Dynamically Plot Frequency Response Curves
AM
Signal
Generator
I. F.
Amplifier
External
Modulation
Input
Figure 4
Limiter
Under
Test
Tektronix
Model 515
Sc ope
'fy Sawtooth Output
From Scope
Sweep Generator
Method!. Used to Dynamically. Plot Limiter Transfer Characteristics
85
modulatioru If total distortion as measured by a distortion analyzer is
also included on the capture plot as a function of a , this defect is largely
remedied. Total distortion is generally more important than per cent
capture of inte rfering- signal modulation, especially since it was found
possible to reduce interfering-signal modulation to less than 1% of its
undisturbed value, other frequency components being much more
prominent in the output and causing considerable distortion. The per
cent total distortion depends on the kind of audio filtering used, as
amply brought out in the data to follow In obtaining a capture plot, the
following procedure was followed.
(1) Set attenuator dials of signal generators No 1
and No. 2 at some convenient value such as 2000 p volts.
(2) Adjust the generators for equal output and
substantially equal frequency by watching the interference pattern at
the discriminator output with both generators unmodulated and K = 0
Equal output is indicated by maximum amplitude of the spiked inter¬
ference pattern; the spike repetition frequency, being equal to the dif¬
ference frequency between the two signals, is adjusted to a minimum
of a few cycles per second,
(3) Set the attenuator dial of generator No. 2 to a
minimum. Modulate generator No. 1 with the desired modulating
frequency for the stronger signal. Tune one wave analyzer to this
frequency and set its input attenuator for a meter reading of 100%,
corresponding to interference -free reception.
86
(4) Repeat step 3 with generator No. 1 set for
minimum output and generator No. 2 modulated at the desired weaker,
signal modulating frequency, another wave analyzer being tuned to this
frequency and calibrated as above.
(5) Reset the attenuator dial of generator No. 1 to
the reference value of 2000 \x volts. Set up the desired initial value of
a, read from the attenuator dial of generator No. 2 by simple arithmetic;
for example, 800 \± volts correspond to an a of 0„ 4,
(6) Adjust the receiver for best capture performance
and read the two values of per cent capture directly from the meters
of the wave analyzers. Total distortion is measured by the distortion
analyzer.
From this point on, two alternate procedures may be followed
resulting in two different types of capture plots. One method is to
vary the amplitude of the weaker signal generator to obtain various
values of a between zero and unity, recording data at evenly spaced
values of a, the adjustments of the receiver all being left fixed at some
compromise setting. The resulting capture plot indicates how well the
receiver works with varying a for a constant level of stronger signal
input and fixed adjustments.
In the second method, all important receiver adjustments are
carefully reset for optimum performance at each value of a „ A capture
plot made in this way is an indication of the performance which could
be obtained from a receiver in which no detrimental second-order effects
87
occurred due to variations in a and which had a control system capable
of maintaining the proper value of K as a varied
A few plots were made in which only K was readjusted for each
value of a, the other controls being set at some compromise adjustment
These plots indicate the performance that the receiver would actually be
capable of if provided with a control system to optimize K under all
conditions
Initially* capture plots were made '’double - ended1’ so that ^ex¬
tended to values greater than 1 the originally weaker and stronger
signals exchanging roles for a ^ 10 The only advantage that results from
including the region a ^ 1 is that the behavior of the receiver is measured
with changing stronge r - signal amplitude and with an interchange of
modulating frequencies0 The dynamic range of the receiver is of
secondary interest in seeking to determine the basic potential of the.
feedforward technique and dependence of receiver performance on mod¬
ulating frequency, if desired, may be fully investigated separately
Therefore, the double-ended capture plot was soon abandoned and only
values of a less than 1 were considered. The modulating frequencies
were standardized at 400 cps lor the weaker signal, 1000 cps for the
stronge r <,
Audio Filtering
As explained in Chapter 4 the demodulator as originally built
included a 3 kc low-pass filter which could be switched in or out and
R-C de-emphasis networks with time constants of 75 ^seconds and 750
^seconds either of which could be switched in alone or in combination
with the low-pass filter0 These various devices were included merely
88
because they represented standard practice in various types of FM
systems^ The 75 p second de-emphasis is standard in broadcast re¬
ceivers, while the 3 kc low-pass filter is often used in communications
equipment along with a de -emphasis of 6 db/octave, maintained within
3 db over the range 300 cycles to 3 kc„ (750 p second de-emphasis)
Early capture plots were made with 750 p second de -emphasis
plus the 3 kc low-pass filter, since this heavy filtering gave the lowest
distortion figures when capturing 400 cycle weaker - signal modulation
with 1000 cycle modulation on the stronger signal,, As results improved,
later plots were made with 75 p second de-emphasis alone, in hopes of
producing a system of acceptable distortion levels using broadcast-
type standards throughout,, This was not achieved, and 750 p second
de -emphasis was again used in a few plots, this time without the filter,
which hardly made enough difference to justify its use„
Although de -emphasis was used in the receiver audio section
for the initial tests described above, no attempt was made to use pre¬
emphasis in the signal-generator audio section Since only one modu¬
lating frequency was involved at each generator, the use of pre¬
emphasis would have changed only the value of peak deviation of the
signalSo In order to obtain a preliminary indication of feedforward re¬
ceiver performance to be expected, without going to the trouble of set¬
ting the deviation of the two signal generators according to some pre¬
emphasis curve, the first tests were conducted with full nominal peak
deviation (+ 75 kc) on both signals, sinusoidal single -frequency modu¬
lation being used as explained above,,
89
The first test in which weaker-signal modulating frequency was
varied was also made with "flat” response in the signal-generator audio
section, both signals were thus fully modulated at all times0 In practice,
of course, a transmitter with a flat audio response would not be used
with a receiver which included audio de-emphasis, however, the test
was conducted in this way for the sake of convenience in obtaining a pre¬
liminary estimate of the effect of varying weaker-signal modulating
frequency6
The results of this initial test are shown in Figure 20c The dis¬
tortion at the higher weake r - signal modulating frequencies was found to
be considerably worse than at the previously used frequency of 400
cycleSo The reason is apparent: the de-emphasis filter attenuated the
higher modulating frequencies heavily while favoring distortion com¬
ponents at lower frequencies,. It was reasoned that the use of pre-
emphasis in the signal-generator audio would merely reduce the peak
deviation for the lower modulating frequencies, having little effect on
the distortion of the higher modulating frequencies, this conclusion was
later confirmed experimentally, as will be describedo It was tenta¬
tively decided that the use of flat transmitter audio response and a flat-
topped audio bandpass filter in the receiver would provide a better
compromise over the speech band in reducing the distortion on captured
weaker - signal modulation than would the conventional pre -emphasis --
de -emphasis system Accordingly, the audio section of the demodulator
was redesigned around a bandpass filter, as described in Chapter 40
The tentative decision to use a bandpass receiver audio filter and
flat transmitter audio was fully justified by experiment, as shown later
90
in this chapter, all subsequent measurements were then made using the
bandpass filter* However, time was not available for the tedious work
of re -measuring all previously obtained capture plots with the new audio
system; a few points only were measured to provide some basis for de¬
termining what difference the new audio filter would make in distortion
figureso
The account of the evolution of audio filtering techniques is in¬
cluded to explain the use of different audio filters in obtaining data
taken at different timeSo It is important to note that "per cent capture"
figures are unaffected by changes in audio filtering, and provide an
excellent criterion for comparing systems whose performance was
measured with different types of audio filters*
Results of Experimental Measurements
The Transformer -Input Feedforward
Figures 5 through 10 are capture plots obtained using
the transformer -input feedforwardc All were made, with pentode
limiters in the feedforward proper and a pentode pre-limiter* Since
they were the first capture plots made for this investigation, distortion
measurements were usually not included*
Figure 5 shows the improvement in stronger -signal capture
obtainable with the feedforward* The inner dotted curves are the curves
of Figure 10 of Chapter 4 and represent the performance of the demodu¬
lator alone in an early stage of its development* The outer curves were
obtained by inserting the feedforward between the I* F. amplifier and
the demodulator and adjusting K for best capture, a single value of K
being used for the entire plot*
91
o
Enhancement of Demodulator Stronger-Signal Capture Performance by Transformer
Input Feedforward
92
Figure 6 shows receiver performance with K set for optimum
capture of the weaker signal at an a of 0„ 5o Performance of course
falls off on either side of a = Ch 5, note that the deterioration occurs
much more rapidly for a ( 0C 5 than for a ^ 0o 5 because of the more
stringent requirements on K for small a, as explained in Chapter 20
The plot is asymmetrical because of amplifier overload and poor
limiter performance at the large input signals which occur for a ^ 1„
Figure 7 was obtained by re -aligning the receiver and
optimizing the value of K for each a . The improved performance at
high input signal levels was obtained at the expense of that for smaller
signals*.
Figure 8 shows somewhat improved weak-signal capture per¬
formance over that of Figure 7, obtained by careful re-alignment of the
post-feedforward filter to obtain a flatter curve* Figure 9 indicates
further improvement obtained by a more careful alignment of the en¬
tire receiver,, In both of these plots* K was optimized for every a.
Figure 10 represents the best capture performance obtained
from the transformer- input feedfoTward although the symmetry of the
plot is none too good0 Figure 11 shows the response curves of various
portions of the system at the time the plot was made, and indicates the
importance of flatness in the curves at small values of a* Note the
very slight improvement in the flatness of the curves of Figure 11-B
over those of Figure 11 -A and the resulting improvement in capture
performance shown in Figure 10o
Figure 11»C shows the appearance of the recovered 400 cycle
weaker-signal modulation with various types of audio filtenng0 The
K set for optimum weaker -signal
capture with a = 0.5.
93
U
X
o
o
Weaker-Signal Capture Performance of Transformer-Input Feedforward with Fixed Value of K
ure
94
o
f o
Weaker Signal Capture Performance of Transformer -Input Feedforward;
K Optimized for Each Value of a
ure
95
o
Improved Weaker -Signal Capture Performance with Better Alignment of Post-Feedforward Filter
96
97
Best Weaker-Signal Capture Obtained Using Transformer
* * .... 1 — ? .J , ' *• ■ -* . j . j
98
I. F.
Amplifier
Feedforward
Input
Transformer
Post-
Feedforward
Filter
The vertical scale is linearly calibrated in relative amplitude. The
horizontal scale is frequency, 40 kc. per division.
Figure 1 1 -A
System Response Curves For Solid Percent-Capture Curve of Figure 10
I. F.
Amplifier
Feedforward
Input
Transformer
Poet-
Feedforward
Filter
Scale* lime a* above.
Figure 1 1-B
System Response Curves For Dotted Per cent-Capture Curve of Figure 10
(After Re-alignment)
99
75 |x second de -emphasis alone.
Distortion approximately 15%
750 p second de-emphasis alone.
Distortion approximately 10%
750 p second de •emphasis plus 3 kc.
low-pass filter.
Pictures taken at a ■ 0. 7, 400 cycle weaker -signal modulation shown,
at 90% capture. Stronger signal modulated at 1000 eyeless 9% capture
of 1000 cycle modulation. Both signals deviated +_ 75 kc. with center
frequencies within a few cycles of each other (co-channel signals).
Figure 11-C
Appearance of Captured Weaker-Signal Modulation In Test of Figure 10
100
photos were made at the same time as the plot of Figure 10 at an a of
0. 7, the point of best performance. The top waveform represents a
distortion of 15% with 75 p second de -emphasis.
The capture plots of Figures 5 through 11 illustrate the importance
of flat frequency-response characteristics in the filters of a feedforward
receiver and the necessity for exact alignment. They show that a very
definite improvement in performance can be obtained by only a very
slight re-alignment. The improvement thus obtained is most marked
for small a , as predicted in Chapter 2.
The Transformer -Output Feedforward
Figure 12 is the first capture plot obtained with the
transformer -output feedforward of Figure 14 in Chapter 4C The distortion
figures, rather low for 0. 2 a <^0o8, were obtained with 750
H-sec de -emphasis and low-frequency rolloff, which rather favors the
400 cycle modulation used. Figure 13 shows improved results obtained
with more careful alignment.
Figure 14 was obtained with the bandpass speech filter in the
audio instead of the heavy de«emphasis. Although Figures 12 and 14
were measured at different times, their per cent capture curves are
substantially the same, allowing comparison of the bandpass speech
filter and the heavy de-emphasis under similar conditions. Use of the
bandpass filter instead of de-emphasis increases 400 cycle distortion
from 10% to about 25% in the middle range of a however, it represents
a better compromise over the entire speech band, as brought out later.
Figures 13 and 15 allow the same comparison of audio filtering methods,
101
102
Figure 1 4
103
104
105
they show slightly improved performance due to more precise alignment,
All of the capture plots shown for the transformer output feedforward
were obtained by optimizing adjustments at each value of a „ No pre-
limiter was used
The Driver Limiter Feedforward
The feedforward circuit of Figure 14, Chapter 4, was
found capable of better over-all performance than either of the other
two circuits built, both in ability to capture a weaker signal at very
small values of a and in amount of distortion and per cent capture in the
region around a = 0o 50 Hence, more extensive measurements of its
performance were made than for the other two circuits* In addition to
capture plots, tests were made at an a of 0o 5 in which the modulating
frequencies and degrees of modulation of the two signals were varied.
The spectrum of the captured weaker - signal modulation and its
accompanying distortion was measured under several different con¬
ditions.
Capture Plots
Figure 16 shows the capture performance obtained from
the driver-limiter feedforward with and without a pre-limiter* Notice
that the receiver is capable of a given per cent capture at about a 6 db
lower value of a without the pre -limiter, as predicted by theory* This
plot also clearly indicates that more drastic filtering than the broad¬
cast-type 75 p second de-emphasis is in general necessary to reduce
audio distortion to tolerable levels*
106
107
Figure 17 shows the best over-all performance obtained from
any feedforward receiver to date Note that capture of the weaker -
signal modulation is better than 80% over the range 0* 06 a <( 0C 9 and
that the residual stronger - signal modulation is less than 3% over most
of this range and never more than 5%, Total distortion is, of course,
rather low because of the heavy de -emphasis used: judging by the results
of measurements on the transformer -output feedforward, values of
distortion would probably range between 20% and 30% over the range of
80% capture with the more realistic bandpass speech filter in the audio
system in place of de -emphasis* Figure 18 shows the appearance of the
captured weaker - signal modulation waveform at an a of 0* 05*
It is important to interpret Figure 17 correctly* It does not rep¬
resent the performance of an operational receiver, since a number of
adjustments had to be carefully optimized at each value of a to obtain
the performance shown. The plot does indicate something of the poten¬
tial performance of which the feedforward technique is capable at each
value of a * In order to realize this performance in an operational re¬
ceiver under field conditions, careful design would be necessary, as
outlined in Chapter 2C
Figure 19 provides a slightly more realistic picture of how well
an operational feedforward receiver might perform if equipped with a
control system capable of maintaining the optimum value of K The plot
was made by varyirg only the value of K for different a all other ad¬
justments remaining at some compromise value* Again, the distortion
curve would probably lie closer to 25% than 10% if the bandpass filter
had been used instead of the heavy de -emphasis as explained earlier*
108
109
K * 0. 100% capture of 1000 cyclo
stronger »«igtial modulation.
Ks-1, 75% capture of 400 cycle weak«r-
signal modulation, a ■ 0. 05. 15% dis¬
tortion with 750 pane. de-emphasis.
About 1% capture of 1000 cycle stronger-
signal modulation.
■ Same weaker-signal modulation as
above • with different scope sweep ad¬
justment to show details of distortion.
Co-channel signals. ^ 75 kc deviation. Only one adjustment (feed¬
forward amplifier gain) was moved between top and bottom photos.
Figure 18
Appearance of Captured Weaker-Signal Modulation Waveform, a « 0. 05.
For Test of Figure 17
no
Ill
Varying Modulation Frequencies
One of the major purposes of the experimental investi¬
gation was to determine how the weaker - signal capture performance
of a feedforward receiver changed as the modulating frequencies of the
two input signals were varied arbitrarily over the audio range,, Accord¬
ingly, the receiver was set up with the driver -limiter feedforward and
adjusted for optimum performance at an a of 0o 5, a single modulating
frequency being selected for the stronger signal* The frequency of the
weaker - signal modulation was varied over the audio band, distortion
and per cent capture being measured at a number of points * The re¬
sults are plotted as Figures 20 through 29> which show per cent
capture and total distortion as a function of weaker-signal modulating
frequency for various conditions*
The significance of Figure 20 is explained in the introductory
portion of this chapter* The results of Figures 21 through 23 were ob¬
tained with a pre -emphasis of 6 db per octave in the signal- generator
audio section; the receiver audio used a 6 db/octave de -emphasis over
the range 300 to 3000 cycles as well as the bandpass filter of Figure 5,
Chapter 4* The distortion curves confirm the tentative conclusion
derived from Figure 20; use of the standard pre-emphasis - de-emphasis
technique results in excessive distortion in captured weaker-signal
modulation at the higher modulating frequencies* This is rather
serious for a speech channel* since the higher frequencies are known
to be the most important for intelligibility*
In light of the results described above, the bandpass filter of
750 (-L second de-emphasis in demodulator audio. Flat response in signal-generator audio.
112
£
<d
£
CO
a)
■ — i
u
>*
O
rO
’"0
<1)
+J
a)
r-H
2
x)
o
s
’CJO
VH
CO
<D
W)
o
4-»
CO
WEAKER- SIGNAL MODULATING ff?£CUENCY IN CYCLES PER SECOND
113
0
O
n4
FRCQUF NC y UV CYCLES PER SECOND
115
0
116
Figure 5, Chapter 4 was tried alone in the receiver audio, flat audio
response being used in the signal generator modulator* The bandpass
filter characteristic has 3 db points at 300 and 3000 cycles, this char¬
acteristic was chosen because it is the response curve generally used
as a design goal for high -intelligibility speech channels It was hoped
that a receiver filter which was flat over the passband ol greatest
interest and fell off sharply outside would provide a better compromise
for reducing audio interference which could fall anywhere in the audio
spectrum than would the standard pre-emphasis - de-emphasis
technique.,
Use of the flat receiver audio filter with flat signal- generator
audio proved as predicted to be an excellent compromise, as shown by
the plots of Figures 24 through 29, all measured with the bandpass
filter in the audio section of the demodulator* The plots are the same
type as those of Figures 20 through 23* Each was measured with a
different modulating frequency on the stronger signal* In Figure 24,
the stronger signal modulating frequency is below the passband of the
audio filter; in Figure 25, it is inside the lower edge; in Figure 26, it
is near the geometric center of the filter passband in Figure 27, it is
at the upper edge and in Figure 28 above the passband
It seems safe to conclude from the figures that per cent capture
is essentially independent of the weake r - signal modulating frequency for
the high deviation ratio used (25)* Distortion is surprisingly low when
the stronger - signal modulating frequency is just inside the upper edge
of the audio filter passband or above the passband altogether* When
the stronger-signal modulating frequency is within or below the filter
ny
o
y/eaxeh- signal mobu la ting FREQUENCY IN CYCLES PER SECOND
118
8
I too 10OO
WEAKER-SIGNAL. MODULATING FREQUENCY IN CYCLES PER SECOND
119
1R -SIGNAL M0DULATD1G FREQUENCY IN CYCLES PER SECOND
5s
121
00 l
122
WEAKER-SIGNAL MODULATING FREQUENCY IN CYCLES PER SECOND
123
passband the distortion level should allow fair but not good quality
speech transmission on the weaker signah In all cases, distortion
rises rapidly when the weaker - signal modulating frequency begins to
move out of the filter passband because the modulation is then attenuated
with respect to the distortion within the passbando
Spectrum of Captured Weaker-Signal Modulation
The spectrum of the inherent distortion involved in
weaker - signal capture seems to follow a simple rule when the modulation
on both weaker and stronger signals is sinusoidal: components appear
at frequencies that equal the highest common factor of the two modu¬
lating frequencies and all of its harmonicSo For example, if the
stronger signal is modulated by a 1000 cps signal and the weaker by a
300 cps signal, the distortion accompanying the recovered weaker-
signal modulation will have components at 100 cycles and all of its
harmonics, since 100 is the highest common factor of 300 and 1000u
An actual spectrum measured for this case is shown in Figure 30o The
strongest components are those corresponding to lower-order inter¬
action, such as 1000 + 300 cycles and 2(1000) 300 cycles0 The inter¬
action between signals does seem to be of a fairly high order, however,
since all of the harmonics of the 100 cycle HCF frequency are measur¬
able up to 5 kc and beyondo Physically, 100 cycles is basic repetition
frequency of the interference pattern of the 300 cycle and 1000 cycle
sine waveSo Distortion is produced whenever the two FM signals cross
in frequency, and the pattern of the frequency crossings repeats at a
loo
124
H - h
O Q
f'o rx
o
0
too }£>00 YOOQG ZOQOQ
FREQUENCY IN CYCLES FEFt SECOND
125
o
126
e
127
100 cycle rate, producing the 100 cycle fundamental and its harmonics
in the demodulator output*
Figures 31 and 32 are further examples of the spectrum of
weaker-signal modulation plus distortion* Note the clear predominance
of the weaker-signal modulation component the relative unimportance
of the fundamental stronger-signal modulating frequency compared to
other interference components, and the adherence of the component
frequencies to the ’’highest common factor” rule*
Reduced Deviation Tests
Tests were conducted at an a of 0* 5 with fixed modulating
frequencies on both signals to determine how weake r - signal capture
performance would be affected by reducing the frequency deviation on
each signal
Figure 33 shows the effect of reducing the deviation of the weaker
signal, stronge r- signal deviation remaining at the full + 7 5kc* Note
that distortion decreases with deviation down to a point, then increases
rapidly* The per cent capture is actually greater than 100% for small
weaker -signal deviations; this means that the fundamental weaker -
signal modulation component at the demodulator output is stronger in
the presence of interference* The reason for this effect is not known,
nor was any reason found for the dip in the per cent capture curve with
400 cycle modulation on the stronger signal*
Figure 34 indicates that some improvement in distortion can be
obtained by readjusting the feedforward for smaller values of weaker-
signal deviation.
128
The effects of reduced stronger-signal deviation are shown in
Figures 35 and 36c Note that performance deteriorates significantly
for reduced deviation* then improves for very small values of stronger-
signal deviation Also* a significant improvement in performance is
obtainable at small values of deviation by careful adjustment These
facts seem to suggest that the feedforward circuit can reject unmodulated
or low-deviation signals but that it works in a somewhat different way
than for fully- modulated interference,,
Speech Intelligibility Tests
A few brief tests were conducted to obtain a subjective
evaluation of the quality and degree of intelligibility of speech modulation
recovered from the weaker of two co-channel signals The stronger
co-channel signal was modulated successively at 100, 400, 1000, and
5000 cycles; deviation was varied from zero up to the full 75 kc„ The
weaker co-channel signal (a = 0o 5) was modulated with a voice signal
from a microphone* and the output of the feedforward receiver was
recorded on an ordinary tape recorder for later evaluation by ear„
Some tests were conducted using a speech clipper and low-pass filter
in the signal- generator speech modulator The bandpass speech filter
was* of course, used in the receiver audio section,,
The speech recorded on the tape by the method just described
was surprisingly intelligible «, There was, as expected, background
noise consisting of a complex audio tone whose amplitude and quality
varied with the voice modulation; it was generally loudest in the absence
of voice modulation,, As might be predicted from the distortion plots,
129
3
n'
t* : }
■ . i-
i<H
4
Stronger Signal 1000 cps, Weaker 400 cps.
- Stronger Signal 400 cps, Weaker 1000 cps. 1
'I 1 ~ !' 1:Tl!niTjriin:‘!lli-.in = :|::;;|in!|j!!:i ; : I’ : •'! !!!; I ; . , ;i : : :ii'inn ii nn1 ii ■ i i •
HI
: , } ' i 5 ■ * *
111 ™
rrd
Wf
rrrrr Tir~
0. 5; Stronger-Signal Deviation, 75 kc.
:l.i: i.
::!i
tjii
'fit
;u:
rm
* f.trrii4
ui
liiiMM
ui.
Ml
HI
Figure 33: Effect of Reduced
Weaker-Signal Deviation
Weaker-Signal Deviation, Kc.
lo
— i —
40
— r~
30
— i —
+o
i:*
t • »
!lji
iililli' :
i-L.L: .
H
tr
tj}{
;ij;
in:
li; hill
to:
! t ’ ■ 1 • : H
Ml
50
6o
70
90
130
131
132
133
speech quality was excellent with 5000 cycle modulation on the stronger
signal, the background noise was more like a hiss than a tone0 For other
modulating frequencies on the stronger signal, there was noticeable
distortion of the speech when compared with speech transmitted over
the same system without interfe rence0 The speech quality was sur¬
prisingly good for low stronger - signal deviations and even with an un¬
modulated stronger signal; this was somewhat unexpected, considering
the results of the quantitative distortion measurements with reduced
stronger-signal deviation
The use of speech clipping and filtering in the signal-generator
speech modulator produced pretty much the expected results the re¬
covered speech sounded fuller and more powerful but less natural The
clipping seemed to help somewhat against the background interference,
but the difference was not very greata
In a situation in which both the desired weaker signal and the
interfering stronger co-channel signal have the same nominal peak
deviation and the stronger signal is fully modulated, as in the test, the
reduction in distortion of captured weaker - signal modulation for reduced
weaker-signal deviation acts to favor speech modulation Since a speech
waveform consists of high peaks with a generally lower average level
(R0 Mo So value approximately 1 / 3 peak value) the deviation of the
speech-modulated weaker- signal transmitter is less than its maximum
value a sizable percentage of the time, resulting in reduced distortion.
In summary the recovered voice modulation is probably best
described as usable fully intelligible speech of fair but not particularly
good quality
134
CHAPTER 6
SUMMARY AND CONCLUSIONS
The feedforward technique can be used to enhance the perfor¬
mance of an FM receiver in rejecting interference either weaker or
stronger than a desired signal* Its usefulness in stronger-signal
capture must be examined in a given circumstance in the light of
alternative stronger-signal capture techniques^ Feedforward can
deliver excellent performance, but the improvement in performance
over an alternate technique of comparable complexity may not be
worth the additional design problems that go with feedf orwardo
However, the flexibility and conceptual simplicity of feedforward may
be quite attractive in particular stronger- signal capture applications,
especially in a receiver used by a skilled radio operator*
The problem of weaker-signal capture is both more complex
and more interesting; the feedforward technique is conceptually
simpler and more straightforward than any other existing weaker-
signal capture technique* but its simplicity may often be offset by the
close tolerances on component circuits necessary for high performance.
Precision of at least 100(a /2) per cent is required for a ( 0o 5 and
at least lOO(l-a) per cent for a ) 0, 5 for weaker -signal capture.
Three basic types of practical feedforward circuits were
built as laboratory models All demonstrated various degrees of
135
weaker-signal capture and some improvement in stronger - signal
capture. The best of these experimental results indicate that
reasonably good weaker-signal capture can be achieved under labora¬
tory conditions with a feedforward receiver for 0 06 ^ a 0. 9;
this corresponds to tolerances of about 3 pe r cent on all important
system parameters, Extension of performance to smaller values
of a is dependent on attaining better than 3 per cent accuracy in
bandpass filter responses, limiter characteristics, and amplifier
gain over all parts of the modulation cycle.
Extensive tests were made of the distortion encountered in
the recovered weake r - signal modulation under a wide variety of
modulation conditions for an a^ of 0. 5, at which the feedforward
technique delivers its best weaker - signal capture performance. The
results indicate that in general considerable distortion of recovered
weaker-signal modulation is an inherent feature of the feedforward
technique, and that the use of large deviation ratios and receiver
audio filters flat over the passband of interest and falling off sharply
outside the passband are generally necessary to reduce distortion to
tolerable levels*
The best experimental feedforward receiver built was designed
for + 75 kc frequency deviation on both weaker and stronger signals,
and used an audio filter whose response was flat from 300 to 3000 cycles
and fell off sharply outside this band. At an a^ of 0. 5, the measured
distortion on recovered weaker - signal modulation varied between 10
per cent and 30 per cent, depending on the two modulating frequencies
136
involved and the peak deviation of the two signals*, Reduced weaker-
signal deviation resulted in reduced distortion down to one -half or
one-third full deviation* Reduced stronger-signal deviation with
full deviation on the weaker signal resulted generally in increased
distortion. Distortion was fairly uniformly distributed as a function of
weaker-signal modulating frequency over the filter passband. The
distortion conditions were such as to allow usuable and completely
mtelligibile (but not high-quality) speech transmission on the weaker
signal at an a near 0* 5 with the experimental feedforward receiver
used for reception*
A feedforward receiver designed for weaker-signal capture
would be at its best in a situation in which it was required to capture
a weaker signal with a near 0 5 whose deviation was about half
that of the interfering stronger co-channel signal and whose modulation
consisted of one or more narrow-band audio signals such as teletype,
remote control, or low-rate digital signals, allowing the use of high
deviation ratios and narrow audio filters* The receiver would also
be useful for communication-quality reception of a speech-modulated
weaker signal, but would not do well with high-quality program
modulation, such as music*
The major engineering problems in designing and building a
feedforward receiver are: design of limiters and bandpass filters
whose parameters remain within the tolerances required for small a ,
design of a high-performance demodulator to use following the feedforward
137
and design of an auxiliary circuit which will maintain complete
interference cancellation in spite of variations in input signal amplitude,
interference ratio, and circuit parameters.
If vacuum tubes are to be used in a feedforward receiver, the
transformer -output circuit of Figure 14, Chapter 4, is probably the
simplest and least complicated circuit available, it is capable of
good performance if the output transformer is designed carefully
and the preceding stage has enough output to insure adequate limiter
drive and is not adversely affected by limiter loading The driver-
limiter circuit (Figure 15, Chapter 4) seems to be the basic feedforward
design capable of best performance, though it uses a minimum of three
tubes instead of two. One untried idea of great promise is the use of
two amplifier stages in the upper channel with two very fast silicon
computer diodes between the amplifiers as a limiter
138
CHAPTER 7
SUGGESTIONS FOR FURTHER WORK
Because of the newness of the feedforward technique, there are
many unsolved problems connected with the design of practical feed¬
forward receivers. A number of ideas for future investigations arose
out of the present study because of its exploratory and problem-defining
nature.
Improved Narrow-Band Limiters
So far, no one has designed or built a narrow-band limiter which
has a low threshold, high output, flat limiting characteristic, freedom
from precise adjustments and selected components, a high enough input
and output impedance to use easily with tuned circuits, and is simple
and economical, Ideally, a single stage should suffice A few two-stage
limiters have exhibited fairly good perfomance at the cost of com¬
plexity.
One possibility is the use of a special gated-beam tube of the
6BN6 type intended solely for limiter application, with its parameters
controlled tightly enough in manufacture to insure uniformly flat limiter
characteristics and low thresholds without special bias adjustments or
tube selection. This obviously involves considerable design effort.
The use of diodes in a narrow-band limiter is worth investigating.
Most previous diode limiters for FM receivers were wideband and used
139
diodes which are now obsolete* Recently developed semiconductor de¬
vices such as Zener diodes, tunnel diodes, and silicon computer diodes
deserve study as possible limiter components0 An interesting possi¬
bility is the Microwave Associates IN903 silicon diode, which has ex¬
tremely fast switching time, uniform characteristics, and the inherent
CL 5 volt gap before forward conduction found in silicon diodes which
makes bias unnecessary Two diodes connected in parallel with opposite
polarities constitute a limiter which saturates at 1 volt peak to peak*
The problem with diode limiters is to reconcile their low impedance
with the necessity for reasonably high-impedance tuned circuits and the
high voltage output necessary to adequately drive the next limiter0
Improved Bandpass Filters
In designing feedforward systems, special attention must be given
to the bandpass filters used (see Chapter 2)0 Double-tuned circuits can
be used in many cases if properly designed with adequate thought given
to stability, ease of adjustment, and freedom from loading effects0 One
attractive untried possibility is the use of potted toroidal inductors,
silver mica capacitors and mutual-capacitance coupling, insuring
great stability and ease of adjusting coupling coefficients0
Demodulator Improvements
As mentioned in Chapter 4, better limiters would have improved
the performance of the demodulator described there* An improvement
in demodulator capture performance is reflected in improved feedfor¬
ward receiver performance* However, demodulator performance im-
140
provements become less and less important as the capture ratio is
pushed toward lc 0, since the range of a over which distortion occurs
becomes so small Therefore, work on improved demodulators is
valuable up to a point, beyond which it is not worth the effort
Automatic Control Systems For K
As mentioned in Chapter 2, a feedback system which would
automatically adjust K for best weaker - signal capture as determined
by the quality of the demodulator output would be an extremely valuable
addition to a feedforward receiver,, Such a system would be somewhat
complex, but w^uld enhance performance materially,,
One difficulty with the scheme is the problem of building a
system which can accurately determine the point of correct adjustment*
The most obvious scheme is to transmit a pilot tone on the desired
signal, outside the audio bandwidth used* Maximum amplitude of
this pilot tone at the demodulator output indicates the optimum value
of K, the tone being separated from the message modulation by a
narrow filter,. The outstanding advantage of this idea is that if the
control system is designed to sweep slowly from K = 0 to slightly
below K = 1, stopping at the point of maximum pilot tone output, it
would adjust K for optimum desi red- signal capture regardless of
whether the interference was stronger or weaker.
Figure 1 shows a block diagram of a proposed control system
embodying the above ideas* The sawtooth generator is a phantastron
sweep circuit of the type used in "search" type radar AFC^^. It
generates a negative -going sawtooth waveform until a negative Stopping11
l4l
*1
O
U
P
00
U<
Block Diagram of a Proposed Control System to Automatically Maintain Interference Cancellation
142
bias is applied to its control terminal, at which time the circuit becomes
a DC amplifier, its steady output level being directly proportional to
the "stopping" bias, The output of the pilot-tone filter is rectified to
provide this stopping biasc The sweep generator output is applied as
bias to the feedforward amplifier grid causing the gain of the amplifier
to sweep from a maximum down to zero and thereby varying K from be¬
low -1 up to zeroo
The graph in Figure 1 shows how an equilibrium point would be
reached slightly below the optimum value of K An extraneous in¬
fluence acting to push K toward K would result in greater output from
the pilot tone filter and increased negative bias on the sweep circuit
(now behaving as an inverting DC amplifier) raising its output level
and decreasing feedforward amplifier bias to increase its gain and move
K upward to compensate for the disturbance An extraneous drift of K
in the other direction would be similarly compensated.
If K suddenly moves past ^max faster than the system can
compensate or the pilot tone output falls below the triggering level due
to worsened interference the sweep generator will start moving K a-
way from the equilibrium pointo Presumably, the system will lock in
on the next sweep cycle at a new value of amplifier gain, The demodu¬
lator output level at which the sweep actually stops is subject to so
many extraneous influences that it is probably best provided as a front
panel control which could be adjusted until the system barely locked m0
The system as it stands has the disadvantage that it. will not hold
K at its exact optimum value but at some nearby value This might be
gotten around in a more refined system which could determine the point
143
of best adjustment exactly, either by the pilot-tone method or by some
other technique*,
Reducing Audio Distortion
The results of Chapter 5 show clearly that sinusoidal modulation
on the weaker of two co-channel signals can be recovered with practically
its full amplitude0 However, it will inherently be accompanied by a
fairly sizable amount of distortion, no matter how good the R. F* portions
of the feedforward receiver,, Therefore, methods of reducing the audio
distortion would be very valuable.
Since the captured modulation waveform is intact over a part of
the cycle and "broken up" over discrete segments of the waveform, the
(15)
"speech repair" techniques used by Aiguimbau, et ah in their
transatlantic FM experiments might be of value. Briefly, their technique
consisted of replacing the violently disturbed portions of the waveform
with linear approximations based on the value of the modulation wave¬
form and its derivative just before onset of the disturbance.
In a system in which a strong FM signal with speech or music
modulation occupied a channel all or part of the time, it might be possible
to effectively utilize a weaker signal on the same channel to transmit
remote control signals, teletype signals, or low-rate digital informa¬
tion with good results, since such signals allow the use of narrow audio
filters following the weak-signal capture receiver, minimizing the in¬
evitable audio distortion By careful choice of audio frequencies, it
should be possible to transmit several such narrow-band signals on a
single carrier without undue crosstalk. Investigation of this possibility
144
would not be difficult, given a reasonably good feedforward receiver
capable of weaker-signal capture.
More Complex Systems Based on the Feedforward Pnnciple
It has been suggested previously that if the simple narrow-
band limiter of Figure 5> Chapter 1 were replaced by a more effective
device for reducing the interference ratio of two signals at its input,
the amplitude differential at the point of subtraction would be larger,
increasing the amplitude of the residual weaker-signal component in the
output and hopefully increasing the fraction of the modulation cycle over
which weak signal capture could be achievedo
One suggested replacement for the single limiter is another
entire feedforward circuit arranged to improve the predominance of the
stronger signaL The phasing problem involved in such a scheme would
be its most unattractive feature0 It would also be advisable to determine
by theoretical analysis whether or not the potential improvement in the
amplitude of the residual weaker-signal component would actually allow
weaker -signal capture over a significantly larger fraction of the modu¬
lation cycle before going to the trouble of building such a device, if not,
it would offer no improvement over the simpler system The feedfor¬
ward used to replace the limiter could be the pre-limiter type, since
the pre-limiter offers no disadvantage in a stronge r- signal feedforward*
Another possible replacement for the single limiter is a
cascade of two or more narrow-band hmiters* Phasing problems are
troublesome in this scheme also. If two limiters are used with a
145
double -tuned inductively coupled circuit between as a bandpass filter,
a 90° phase lead must be introduced into the amplifier channel because of
the 90° phase lead introduced by the double -tuned circuit over its pass-
band. All sorts of increasingly complex schemes can be worked out
which will yield the proper phase relationships 0 The next most
complicated possibility uses three limiters in the upper channel with
doublt-tuned inductively-coupled circuits between, resulting in zero net
phase shift and allowing the use of a single stage feedforward amplifier*
If too many limiters are included in the upper channel, however, envelope
delay will become troublesome, the ongin and effects of this trouble
are described by Gutwein^^l
Theoretical evaluation of the performance to be expected from
the cascaded-limiter feedforward awaits a detailed analysis of the
spectrum at the output of a cascade of two or more narrow-band
limiters with two-signal input for the case in which more than the
original two components are included within the filter passbands0
The "Ultimate" Basic Feedforward
From a knowledge of the circuit problems usually encountered and
the requirements on the basic components (see Chapter 2), it is possible
to outline the salient design features of a practical feedforward receiver
whose performance should approach that of an ideal system.
The L F amplifier would employ a crystal or mechanical filter
in order to approximate as closely as possible the desired rectangular
passband shape and be free of alignment adjustments 0 The filter would
be isolated from the feedforward limiter by at least one linear amplifier
146
stage with a broadband, low-impedance output circuit so that the I. F0
passband would be unaffected by limiter loading* The limiter would be
of the "improved" type specified earlier in this chapter, and would be
followed by a second mechanical or crystal filter of nearly rectangular
passband shape, well isolated as was the first* The demodulator should
have a capture ratio as high as possible, three or four properly designed
narrow-band limiters plus a wideband limiter and discriminator of 5 or
6 I. F* bandwidths should do the job0 An automatic control system would
be included capable of maintaining the optimum value of K for any
usuable input signal strength and for all values of a above some small
, including a_^> 1, when a pilot tone is usedc It should be possible
to automatically maintain a manually preset value of K independent of
input signal variations when no pilot tone is present on either input
signal*
Construction of such a receiver, using either tubes or transistors,
is well within the state of the art but would require considerable develop¬
ment effort and would be rather costly0 Its performance capabilities,
however, should be quite useful, as well as interesting.,
Feedforward Demodulator
A basic feedforward for stronger - signal capture could be built
with three tubes a pre-limiter, a feedforward limiter, and a feedfor¬
ward amplifier* The output of the feedforward could be fed to a wide¬
band limiter -discriminator combination or, conceivably* directly to an
amplitude -insensitive FM detector such as a ratio detector or gated-
beam discriminator. The combination would constitute a demodulator
with a reasonably good capture ratio* It would be very interesting to
147
compare the performance of this "feedforward demodulator" with the
performance of a conventional chain of narrow-band limiters plus
wideband limiter and discriminator which used the same number of tubes.
It seems possible that the feedforward might offer better performance
than the straightforward chain of limiters using the same number of
tubes; if so, its greater complexity would be justified. An experimental
investigation to settle this question would not be difficult and its results
would be quite valuable*
REFERENCES
E, Jo Baghdady, "Interference Rejection in FM
Receivers", Technical Report No 252, Research
Laboratory of Electronics, September 24, 1956*
E. Jo Baghdady, "Frequency Modulation Interference
Rejection With Narrow-Band Limiters", Proceedings
of the L RoEo , Vol 43, Noc 1, (January, 19551 p. 5T
E0 Jo Baghdady, "FM Demodulator Time - Constant
Requirements for Interference Rejection", Proceedings
of the Io R Eo , Volo 46, No0 2 (February,
pp 432- 440 0
E Jo Baghdady, "Theory of Stronger - Signal Capture
in FM Reception", Proceedings of the L R E , VoL 46,
Noo 4, (April, 1958T "
Eo Jo Baghdady, "Theory of Low- Distortion Reproduction
of FM Signals in Linear Systems", I R E Transactions
on Circuit Theory, VoL CT-5, (September, 1958)
PP 202-214
Eo Jo Baghdady, "New Developments in FM Reception
and Their Application to the Realization of a System of
’Power Division1 Multiplexing", I0 R E, Transactions
on Communications Systems, VoL CS-7, Nou Tl
(September, 1959) ppo 147- 1 6 1 0
Ro Ho Small, "A Single -Channel, Two- Carrier FM
Multiplex System", So M0 Thesis, Department of
Electrical Engineering Me I0 T0 , June 3, 1958c
Ro Go Griffin, "A Single Channel Multi Carrier FM
Multiplex System", S0 Mo Thesis Department of
Electrical Engineering, Mc I T* , August, 1959o
G Jo Rubissow, "Dynamic Trap for Capture of Weaker
Signal in FM Interference", S M Thesis, Department
of Electrical Engineering, M, I T0 , January 31 19580
149
(10) Fo L Sheftman, nA Fixed-Trap System for Capturing
the Weaker of Two Co-Channel FM Signals'1, S„ M.
Thesis, Department of Electrical Engineering, M. I. T. ,
September 5, 1958,
(11) Joseph M Gutwein "The Capture of the Weaker of
Two Co-Channel Frequency-Modulated Signals",
S, M, Thesis, Department of Electrical Engineering,
M, L To , May 21, I960.
(12) J. Granlund, "Interference in Frequency-Modulation
Reception", Technical Report No* 42, Research
Laboratory of Electronics, M IT. , January 20, 1949.
(13) E Jo Baghdady, "FM Interference and Noise-Suppression
Properties of the Oscillating Limiter", L R. Et Transactions
on Vehicular Communications, Volc PGVC - 13 ,
(September, 1959)<>
(14) Ro J McLaughlin, "A Study of FM Capture Effects",
S M Thesis, Department of Electrical Engineering,
Mol Tc , June 1958*
(15) Lc Be Argiumbauj et al , "Transatlantic Frequency-
Modulation Experiments ", Technical Report No. 278,
Research Laboratory of Electronics, MIT,
September 20, 1954
Microwave Receivers, Voh 23, Radiation Laboratory
Series, Section 30 13, p 64 ffc
(16)
UNCLASSIFIED
Security Classification
151
DOCUMENT CONTROL DATA - R&D
(Security ctmeeifi cation of tit/m, body of mbatrmct end indexing annotation muat be entered when the overall report ia c laaaifted)
1. ORIGINATING ACTIVITY (Corporate aasthor)
Lincoln Laboratory, M. l.T.
2a. REPORT SECURITY CLASSIFICATION
Unclassified
2b. GROUP
None
3. REPORT TITLE
Interference Suppression Performance of Several FM Receivers Using Feedforward
4. DESCRIPTIVE NOTES (Type of report and in ctuaiv a date a)
Group Report
5. AUTHOR(S) (La at name, fire t name, initial)
Hutchinson, Ben H., Jr.
6. REPORT DATE
26 July 1961
7a. TOTAL NO. OF PAGES
156
76. NO. OF REFS
16
8a. CONTRACT OR GRANT NO. AF 19(604)-7400
b. PROJECT NO 649L
c.
d.
9a. ORIGINATOR’S REPORT NUMBER(S)
52G-0019
9 b. OTHER REPORT NO(S) (Any other numbera that may be
aaai&ned this report)
ESD-TR -70-93
10. AVAILABILITY/LIMITATION NOTICES
This document has been approved for public release and sale; its distribution is unlimited.
M. SUPPLEMENTARY NOTES
12. SPONSORING MILITARY ACTIVITY
None
U.S. Army, Navy and Air Force
13. ABSTRACT
The feedforward signal -cancellation technique is based on subtractively combining the outputs
of limiters and linear amplifiers having a common input. Used in an FM receiver, feedforward pro¬
vides an attractively simple and effective method for suppressing interference to an FM signal from
other co-channel or adjacent-channel signals which may be either weaker or stronger than the desired
signal. The thesis explores theoretically and experimentally the potential performance and inherent
limitations of practical FM receivers using feedforward. Design criteria are discussed for various
interference conditions and the relative merits of several practical feedforward circuits are considered.
A laboratory model FM receiver was built and tested with three different feedforward circuits ,
its performance being measured under a variety of interference conditions. Significant improvement
in the stronger -signal capture performance of a mediocre FM demodulator was demonstrated. Sinusoi¬
dal modulation was recovered from FM signals between 0.05 and 0.9 times the amplitude of an inter¬
fering signal on the same channel, distortion ranging generally between 8 per cent and 30 per cent for
various interference conditions. Completely intelligible speech modulation was also recovered from
the weaker of two co-channel FM signals. Numerous suggestions for further work are given.
14. KEY WORDS
interference FM receivers feedforward demodulation
UNCLASSIFIED
ptf-1800
Security Classification