204 IRE TRANSACTIONS ON SPACE ELECTRONICS AND TELEMIIETRY September

misconieeptions often arise through the use of inconsistent a clear and complete definition of the significance of the
terminology and incomplete definitions of terms. For term can minimize the confusion of misinterpretation.
example, if the term "delay time" is used in a general
ACKNOWLEDGMENT
sense in one case, used to mean carrier delay time in
another case, or used in the context of envelope delay The author wishes to express appreciation for the
many
time in another instance, misunderstandings are certain helpful eontributions of MI. G.
Pawley,
T. B. Jackson,
to arise. Therefore, a mutual understanding between filter and E. Hygh in the preparation of this report. Acknowl-
designers and filter users regarding fundamental terms and edgment is also due C. Waltner for his careful work in
definitions of terms is highly desirable. Even though a conducting experimental measurements and assembling
particular delay characteristics may be improperly named, measurement data.
A Telemetering System by Code Modulation
- X-
M
Modulation*
H. INOSEt, MEMBER, IRE, Y. YASUDAt, AND J.
MURAKAMIIt
Summary-A communication system by code modulation is the input signal before it enters the modulator so as to
described which incorporates an integration process in the original
generate output pulses carrying the
infoimation
corre
delta modulation system and is named delta-sigma modulation
after its modulation mechanism. It has an advantage over delta
sponding to the amplitude of the input signal. The delta-
modulation in dc level transmission and stability of performance, sigma modulation (A-EM) system is a realization of this
although both require essentially an equal bandwidth and complexity principle.
of circuitry. An experimental telemetering system employing delta-
sigma modulation is also described. THE PRINCIPLE OF THE A-IM SYSTEM
Fig.
1
is the block diagram which shows the principle
INTRODUCTION
of this system. It is essentially a feedback system composed
p ULSE CODE MODULATION (PCMI) and delta of a pulse
generator,
a pulse modulator, an integrator and
inodulation (AM) have been known as communica- a difference circuit. The output pulses p(t) are fed back
tion systems by code modulation. Although PCMI to the input and subtracted from the input signal s(t)
is considered to be the most efficient among the existing which varies sufficiently slower than the sampling pulses.
communication systems, the circuitry used in modulation Then, the difference signal d(t) = s(t) - p(t) is integrated
and demodulation is quite complicated and expensive. to produce e(t)
=
f d(t) dt, and enters the pulse modulator.
On the other hand, AM requires wider bandwidth than The pulse modulator compares the amplitude of the inte-
PCM, but its circuitry is much simpler. grated difference signal E(t) with a predetermined reference
However, in a AM system, pulses which are sent over level and opens the gate to pass a pulse from the pulse
a transmission line from the sending end carry the in- generator when the polarity of e(t) is positive, and closes
formation corresponding to the differentiation of the the gate to inhibit the pulse when it is negative. Through
amplitude of the input signal, and in the receiving end this negative feedback procedure, the integrated dif-
these pulses are integrated to obtain the original wave- ference signal is always kept in the vicinity of the reference
form. Hence transmission disturbances such as noise or level of the pulse modulator, provided that the input
hit result in an accumulative error upon the demodulated signal is not too large. Hence, if the amplitude of the
signal. input signal becomes large, the output pulses appear
This is a drawback of the AM
system and,
in
spite of more frequently. In other words, the output pulses
its simplicity of circuitry, has confined its use to the
transmwission of an audio signal or others which need not XBln
transmit the dc level of an input si g n a
l.l peP1..r
This drawback of AM can be made up for by integrating
*Received May 6, 1962.
t)(j-t;c)=d/t
t Faculty of Engineering, University of Tokyo, Japan.l
t Tokyo Shibaura Electric Co., Japan. Formerly with Faculty
of Engineering, University of Tokyo. Fig. 1-The block diagram of a-:~modulation system.

1962 Inose, Yasuda and Murakanii: Telemetering by Code Modulation 205
carry the information corresponding to the input signal principle and to investigate its features.
amplitude. The equipment is designed for the input signal with the
Demodulation in the receiving end is carried out by frequency range of dc to 50 cps, and the repetition fre-
reshaping the received pulses and passing them through quency is chosen as 1 kc.
a low-pass filter and no integration procedure is in- Fig. 2 and Fig. 3 show, respectively, the block diagram
volved. Therefore, no accumulative error due to trans- and waveforms of the equipment.
mission disturbances results in the demodulated signal.
input signal s(t)
AN EXPERIMENTAL TELEMETERING EQUIPMENT
sampling pulse cQ)
UTSING A-:1MODULATION
An
experimental telemetering equipment is constructed output of addition circuit
and tested in order to demonstrate the realizability of X .A _ output of integration circuit
the communication system based on the above-mentioned
Schmitt trigger output
flg(t)
sampling,tl
t output
of
phase
inverter
pulse
.geneurator.
d(t) -Itr-q(t) c(t) r d(t'dt l I I | 0 | t t
i c(t) gated by g(t)
Sit) o{|Milr Sc bhmitt | -
in
interator triggerge p
=l c(t) gated by -g(t)
signal .. p(t) modulated
s 4 < Vrlp
(t 1,
trset output of flip flop
de odula:ed reset output of flip flop
q)
_e_t_qtfilter
demodulated output
Fig. 2-The block diagram of the telemetering system. Fig. 3-Waveforms of the telemetering system.
1 fror y6-
2</ sampling pulse
from
wave
Isquare wave
H
k i Hi ll
generator /ooo 0/I
I 14 ; 11 '
sampling pulse generator -le I
output
V.-- . -- ~~~~~~~~~~~~ -~~~~~:50pulse
Fig. 4 The schematic
/Odiagr
of thetelemete-
pulse shaper
3K
V
FK62 itr -
-sr +6r - -
49A

206 IRE TRANSACTIONS ON SPACE ELECTRONICS AND TELEMETRY September
The sampling pulse generator generates a series of Therefore, the dynamic range of the input signal should
pulses c(t) with the repetition frequency of 1 kc, the be given as follows:
amplitude of 6 v and the pulse width of 5 p,sec.
The series of pulses c(t) is gated by g(t) and -g(t), L D < S < (I- D.
(3)
which are the Schmitt trigger outputs of opposite polarity, fr fr
and becomes, respectively, the set input p(t) and the
On the other hand, even if fd > JG, the pattern of the out-
reset input -5(t) of the flip flop. resetlputpt ftef lp
put pulses may contain the frequency component which
The flip flop reshapes and widens the gated pulse to
pu lses
mar
cnti
thel
f.req eny
c p
e
which
havetheampltud of 3.5v an tohavethewidt ofpasses
through the low-pass filter. When
this iS
the case,
have the amplitude of ---3.5 v and to have the width
ofthcopntadsotedmdutdotutsa the component adds to the demodulated output as a
one sampling period. The reset output -q(t) of the flip
periodic noise.
flop is fed back to the input, added to the input signal
Next, as the result of the theoretical calculation de-
s(t) in the addition circuit composed of resistances and
scribed in the Appendix, the signal-to-quantizing noise
enters the Miller integrator. The output of the Miller r i
integrator is directly coupled to the following Schmitt
trigger. When the output of the integrated difference S3/2
signal E(t) is larger than the trigger level, the output
N=
M
4r(4)
g(t)
of the Schmitt trigger becomes positive and opens
the gate and passes a pulse c(t) to the set input of the where M! is the ratio of the signal amplitude to the maxi-
flip flop. mum amplitude that does not overload the modulator.
On the other hand, when it is smaller than the trigger In a delta modulation system SNR is given as follows.
level, g(t) becomes negative and inhibits the pulse c(t).
3/2
In this case, as the output of the phase inverter
-g(t)
s
l3/2
()
becomes positive, it passes a sampling pulse to the reset fc f(5
input of the flip flop. Thus the output of the integrator where
f,
is the signal frequency.
is always kept in the vicinity of the trigger level of the Therefore, both the systems have the same relation
Schmitt trigger. The set input p(t) of the flip flop is fed as to the pulse repetition frequency, while, in contrast to
to the transmitter where it is shaped to match the charac-
the delta modulation system, the signal frequency has no
teristic of the transmission line.
issrelation to the SNR in the A-2 modulation system.
In the
experimental equipment
the transmitter is
For the purpose of comparison, the SNR of the pulse
omitted and the set output of the flip flop is directly fed
number modulation is given as
follows:
into an emitter-follower and demodulated by means of a
low-pass filter with the cutoff frequency of 50 cps. S Jr
The schematic diagram of the equipment is shown in
N
1/2Jl/2 (6)
Fig. 4. The entire equipment is built with semiconductor
components. where
fr
is the
pulse repetition frequency
and
f
is the
sampling frequency. Reference to (4)-(6) indicates that A
NOISE IN THE A-:I MODULATION SYSTEM
and A-2 modulation systems have an advantage over the
pulse number modulation in the degree of improvement,
The system suffers two sorts of noise, one of which is with regard to the pulse repetition frequency. The reason
periodic and the other is random or quantizing noise is that in the former, the pulse repetition frequency is
which generally accompanies in any coding process. With equal to the sampling frequency and as it becomes larger
regar(d to the former, the number of the output pulses the spectra of the quantizing noise extends to the higher-
in every second, that is, the density
fd
is proportional to frequency range, and its component which passes through
the amplitudes S of the input signal. If we assume that the low-pass filter is reduced; while in the latter, as the
the amplitude of the flip flop is D and the repetition sampling frequency is constant, the frequency composition
frequency of the sampling pulses is r, the following of the quantizing noise remains unchanged.
equation results:
THE STABILITY OF THE A-2 MODULATION SYSTEM
Jd = Jr. (1) The variation of the reference level of the comparator
is generally the most serious problem in an analog-to-
Then provided that digital conversion device when it has to convert the de
level of an input signal exactly into the digital form.
f <f (2) ~~~~Inthe
1A-: modulation system an input signal is com-
Jd
c
~ 2
pared with the reference level after it is integrated.
where
JC
is the cutoff frequency of the
low--pass
filter, the 1FdeJgr")etmouainaehdofPMrns
periodic
compon1ent
Of the output pulses passes through
mis3sion
using the
o>ne ulnit
code," Phi!lip8 lies. Repts., vol. 7, p.
the low-pass filter, so that this condition should be avoided. 442; 1952.

1962 Inose, Yasuda and Murakamli: Telemetering by Code Modulation 207
Therefore, the trigger voltage of the comparator be- Fixing input signal frequency at 30 cps, the relative
comes noise power is obtained from the demodulated signal
through a 30 cps rejection amplifier composed of a twin
e(t)
= k d(t) dt = K j
{s(t)
-
p(t)} dt. (7) T network. While changing the input signal frequency
to pass through the rejection amplifier and fixing its
Now, if we assume that the reference level of the amplitude constant, the component of the signal power
comparator has had a variation of E' volt during the in- is obtained from the demodulated signal.
terval sufficiently longer than the time constant of the
integrator, the equivalent variation of the input of the ---
integrator becomnes d' volt, where
e' =K d' dt
=
KTd' -li- _ i
or -2 -. e l_ |_
d_ E(8)
K'T
Then the output pulses appear to compensate for this
excursion. As K . 1/RC and -r . ARC in the Miller
4 -2 0
2 4
integrator, K-T A. Therefore, provided that the gain Input (V)
is increased, d' can be made neglgibly small.
Fig. 5-The input-output
characteristics of the telemetering
svstem
iS increased, d' can be made negligibly
small.
(de input).
In the experimental telemetering equipment, as the
gain A is approximately 100, an error of only 0.01 volt
results in the signal, with a variation of the level of 1 volt. TABLE I
THE INPUT-OUTPUT CHARACTERISTICS OF THE TELEMETERING
In this way, the variation of the level after integration IMODULATOR
(DC INPUT)
scarcely effects the system performance. Therefore, only
one component of the circuitry which must be well stabi- Input
lized is the flip flop which feeds the pulse back to the Level I II III Average
input. The amplitude of the pulses is easily regulated by -4 V
29391 29391 29395 2939/see
means of zener diode.-
-3
25073 25076
25077
2508
CHARACTERISTICS OF THE EXPERIMENTAL -2 20875 20881 208S4 2088
rELEMETERING EQUIPMENT -1 16883 16877 16876 1689
The Input-Output Characteristics 0 12701 12708 12710 1271
Fig. 5 shows the relation between a de input level and 1 | 893 8593 8590 859
the output level of the low-pass filter where the repetition 2 4718 4738 4725 473
frequency fr = 3 ke and the time constant of the integrator 3 548 547 543 54.6
X = 10 msec.
Table I indicates the mean numbers of the output 4 0 0 0 0
pulses per second which are counted three times for each -
during 10 seconds.
These results reveal that the linearity of the equipment
- - | - - - - - l
is satisfactory. O0.2; = =1 = - - -
The Frequency Characteristic
Iig. 6 shows the relation between the input signal
D - - - - -
frequency and the output of the low-pass filter where
the amplitude of the input signal is fixed constant. As
the characteristic iS exactly the same to the frequency ,
-
characteristic of the low-pass filter, it is known that the Frequency of input signal (c/s)
modulation system itself has the characteristic which is Fi. 6-h frqec.hrceitc ftetlmtrn ytm
constant all over the signal frequency range.
The Signal-to-Noise Ratio 3/oBa
The measurement of the SNR is carried out with the ln u
method showxn in fFig. 7. Fia. 7-The method of measuring SNR.

208 IRE TRANSACTIONS ON SPACE ELECTRONICS AND TELEI1IETRY September
Fig. 8 shows the theoretical and experimental SNR APPENDIX
with the input signal amplitude fixed at 2 volts, where CALCULATION OF SNR
f, is taken as a parameter. Fig. 9 shows the input and The SNR of the A-: modulation system
can
be esti-
output waveforms with various amplitude and frequency mated as follows, where the symbols mean
ranges which are recorded on a pen recorder.
S((_ ___ ~ input signal w < c =
27rfc
5"F - 1
I I
I/P(tw)l:
output pulses
240 L(w) } ~~~~~~~~~~~~~~difference
signal
- | 4 5t E(t)t ~
integrated differenee signal
s(t)
,~.
demodulated output
H(w) frequency characteristic of the low-
;' 20 -5 - pass filter.
, I t l !~~~~~~1 IwI<@c~ o
l~~~~~~~~~~~~~~~~~~~~~o
< W>co
/0
~~~~~~~~~~~~~~~~~0,
>
w
l_______ ____ e(t) does not exceed the amplitude A of the integrated
0 _
,
-
pulse as long as the input signal is kept in the nonsaturating
/ 23 5 7
Repetition frequency of sampling pulse (k() range.
Fig. 8-Measured SNR. Therefore, for the sake of convenience, we assume that
e(t)
is a
sequence
of
rectangular pulses
shown in
Fig.
10.
-, ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~T'
-4
__sec_._X_____.
-s
-4
O
/ 2
3 4 5
N
sr= Ilicys r= IKCIS Fig.
10-The waveform of
e(t)
assumed for calculation.
Then E(t) is expressed as follows:
h=3S.c/s l l fr-i6
n e(t) = E ?74(t - nT) (9)
f_\~f~ / \ <_| |where
(a) (b)
(c)
+I e(t) > O
IFig. 9=Input and output waveforms recorded bv a pen recorder. 71n
(a) Sinusoidal input (1 cps). (b) Sinusoidal input (10 cps). (c) V-1 e(t) < 0.
Staircase input.
Now, the Fourier transform F(w) of
f(t)
is
CONCLUSION ji~F(w) =AS(j-) (10)
Although this system is originally derived from the
delta modulation system, it may be considered to be a where T =
11fr,
the period of the sampling pulses, and
new communication system by code modulation in view S,(x)
= sin x/x.
of the characteristics discussed above. It has the following Hence, the Fourier transform E,(w) of the portion of
major features: e(t) between number -N and N is given as follows:
1) It transmits a dc level of the input signal. This is EN(w) A/T E' S(--)ei'\ ]1
an advantage over A modulation. w n=-N
e T (11
2) The stability of performance is satisfactory as com- On the other hand
pared with other code modulation systems.
3) It requires smaller bandwvidth than pulse number ddE
-
st-p()(2

1962 Inose, Yasuda and MlIurakami: Telemetering by Code Modulation
209
hence, In this equation, expanding
S2(wT12)
in series
and
jwE(w) = S(c) - P(c). (13) leaving its first term
Multiplying H(w) to the both sides,
N = 8 27r A'T (24)
3
jwE(w)H(w)
= S(w)H(w)
- P(w)H(w). (14)
Then, by the previous assumption on H(co) the following
that is
equations result:
N2=!2rf3/2 AT'1/2 (25)
S(w)H(w)
=
S(co)
(15)
N
-r3/
,/2(5
P()H(w)
=
0(U).
(16)
On the other hand, when the pulse width of p(t) is
Eq. (14) can be rewritten as follows: equal to the sampling period, the dynamic range of the
jC0E(w,o)H(wv)
=
S(w)
-
0(wo). (17)
input signal
is
approximately
as follows:
This is equal to the Fourier transform of the difference D = 2 7' (26)
signal L(w). Tlhen, assuming that the part of the difference
signal between number
-
N and N is LN(w),
Therefore, defining that the ratio of peak-to-peak
LN(w) = jwEN(w)H(co). (18) amplitude of the sinusoidal input signal to the dynamic
Therefore, the power spectrum of the difference signal range is
M,
becomes as follows:
IA
S = I -. (27)
WV(W)
=
lim -,(co)L* ()
(2N -H I)T L~)~w
Hence,
the
SNR
is
=
--28
Z >i: ?ltl,?e Hi(w)H*(w). (19)
N 4M
-
7)
f(28)
Then, defining K(w) as This is obtained from the assumption that E(t) is com-
posed
of
rectangular pulses
with the constant amplitude A.
K(w) =- lim ZZ
nm?e
(nEm)e1 m
T
(20) However, in actual cases, the amplitude of e(t) never
Noo
2N
+
I exceeds A.
the n
z
m terms disappear and (2N + 1) n = m terms Assuming that the amplitude of E(t) takes the value
remain so that
between zero and
A
with equal probability, the noise
power
is increased by
the following value:
K(w)
= 1
(21)
hence, yY
_ 1
i
(w) =
A
Tw2S,(-})H()Hf*(w)a
(22)
Therefore, the SNR can be rewritten as (4).
The quantizing noise power Nwhichpassesthrou
The SNR of the pulse
number modulation can be also
The quantizing noise power N which passes through caclae in a .iiapoeue
the low-pass filter is given as follows:
-\A2
=
A2 f
2s2(WT1) dw (23)
ACKNOWLEDGMENT
c7r 2
~ '
The authors are deeply indebted to Prof. T. Sakamoto
where the assumption that H(co) = 1 is used. for his suggestions and encouragements.