Chapter 2

On the Transmission Capacity

of the "Ether" and Wire in
Electrocommunications
V. A. Kotel'nikov
Translated by V. E. Katsnelson

Material for the First All-Union Conference

on Questions of Communication, Izd. Red.
Upr. Svyazi RKKA, Moscow, 1933.

To broadcast a program by radio or by wire one needs not a single fre-

quency but a whole range of frequencies. Because of this only a limited
number of radio stations (which broadcast different programs) may work
simultaneously. It is also possible to transmit simultaneously no more than
a limited number of programs by one pair of wires because it is impossible
for the frequency band of a program to overlap with the frequency band of
another program. Such overlapping may cause mutual interferences.
To increase the channel capacity of the "ether" and of the wire (this
would be of the utmost importance, especially in connection with the rapid
growth of the radio engineering and of the telecasting) it is necessary either
to decrease somehow the frequency range which is required for the trans-
mission of a program (but in such a way as not to deteriorate the quality
of the program) or to invent a method of the separation of programs with
the overlapping frequency bands. May be one needs even to use a method
of separation of programs not on the basis of frequencies but on another
basis. 1
Until the present time no contrivances in these directions allow (even
theoretically) to increase the channel capacity of the ether and of the wire
more than it may be done using the single side band transmission. Therefore
the question arises: is it possible to realize this at all. May be, all attempts
in this direction are equivalent to attempts to build the perpetuum mobile?
At the present time this is a topical question in radio engineering because
the "crowding of the ether" is growing year after year. Now, in connection

lSometimes it is possible to do this using directional antennas, but now we consider

only the case where it is impossible to do this for some reasons.

J. J. Benedetto et al. (eds.), Modern Sampling Theory

© Birkhäuser Boston 2001
28 V. A. Kotel'nikov

with planning of scientific researches, it is extremely important to investi-

gate this problem, because to direct efforts in the right directions we need
to know what is possible and what is impossible at all.
In the present paper we investigate this problem and prove that for the
television program and image transmission (with all semi-tones) as well as
for telephone program transmission there exists certain completely deter-
mined minimal frequency band. This frequency band can not be decreased
by any means without deterioration of the transmission quality and low-
ering of the speed of transmission. We also prove that for such programs
it is impossible to increase the channel capacity neither of the ether nor of
the wire by means of non-frequency selections of various kind, as well as by
means of any other methods (except of course the directional selection by
means of the directional antennas). The maximal possible channel capacity
for these programs may be attained by the transmission on the "single side
band", which is in principle realizable at present.
For the telegraph or for the television program and image transmission
without semi-tones, etc., when transmitted values may vary not continu-
ously but taking certain amplitudes known in advance, we show that the
frequency band which is needed for the transmission may be decreased as
many times as we wont, without deterioration of the transmission quality
and without lowering the speed of transmission. However this may be done
only at the cost of the complexity of the equipment and of the increase of
power. One of the methods for such a decrease of the frequency band is
shown in this paper. We show also which power increase is necessary for
this goal. Thus, there is no theoretical limit for the channel capacity of the
"ether" and of the wire by the transmission of messages of this kind. The
problem is only with the technical implementation.
In this paper the proof of these concepts is carried out independently
of a transmission method on the following basis. By an electrical trans-
mission of any kind the transmitter may transmit and the receiver may
receive a function of time which can not be absolutely arbitrary because
the frequencies, of which it consists and in which it may be expanded, have
to be contained within certain limits. First, very low and very high fre-
quencies can not reach the receiver because of the propagation conditions,
and second, the frequencies which do not belong to a certain narrow range
are usually suppressed intentionally in order to not interfere with other
messages.
We show below that the necessity to transmit by means of functions, con-
sisting of frequencies from a restricted range, results in a certain restriction
on the channel capacity.
To prove the above stated concepts, we will study functions which consist
of frequencies from a certain range.
 Chapter 2. On the Transmission Capacity of the "Ether" 29

Functions Which Consist of Frequencies Between 0

and !I
Theorem I. Any function F(t) which consists of frequencies between 0 and
h is representable by the series

F(t) = f Dk Sinwl (t
t -211
-
~ ik) (1)
-00

where k is an integer number, Wl = 27rh, Dk are constants depending on F(t).

Conversely, any function F(t), represented by the series (1), consists only of
frequencies between 0 and h.

Proof: Any function F(t), which satisfies the Dirichlet conditions (a finite
number of maxima, a finite number of minima and a finite number of discontinuity
points within every finite interval) and is integrable from -00 to 00 (which is
always fulfilled in electrical engineering) ,2 is representable by the Fourier integral 3

00 00

F(t) = / C(w) Coswtdw +/ S(w) Sinwtdw, (2)

a a
i.e., as a sum of infinitely many harmonic oscillations with frequencies from 0 to
00, and amplitudes C(w) dw and S(w) dw, depending on the frequencies.
Moreover,
00

C(w) = *" / F(t) Coswtdt

-00

00
(3)

S(w) = *" / F(t) Sinwtdt

-00

In our case, where the function consists of frequencies between 0 and h, it is

clear that
C(w) = 0, S(w) =0 for w > Wl = 27rh-
Therefore, F(t) may be represented, according to the equation (2), in the follow-

lwl
ing way:

F(t)= a l W1
C(w)Coswtdw+ a S(w)Sinwtdw (4)

As any other function, the functions C (w) and S (w) are representable by Fourier
series on the interval 0 < w < Wl. Moreover, these series may be taken to consist of

2In what follows we consider only functions satisfying the Dirichlet conditions.
3See for example Smirnov, Course of Higher Mathematics, vol. II , 1931, p.427.
30 v. A. Kotel'nikov

Figure 2.1

cosines only (or of sines only, as an option), by considering the periodic extension
to an interval of length 2WI. 4 Thus

(5a)

and
S(W) = fBk Sin ~kw. (5b)
o 2WI

Let us introduce the following notations:

} (6)

Thus, the formulas (5a) and (5b) may be rewritten as

00

C(w) = LDkCos~kw
WI
-00

(7)
00

S(w) = LDkSin~kw
WI
-00

Now, substituting the expressions (7) into the formula (4), we obtain the equation
(1) after some transforms (See Appendix 1) and prove the first statement of
Theorem I.
To prove the second statement of the theorem, we consider the special case of
F(t), in which the frequency spectrum is supported within the limits between 0
and h and is expressed by the equalities (7), in which all Dk, except one, are
equal to zero. It is clear that such F(t) consists of only one term of the series (1).
Thus, the converse assertion is true: if F(t) consists of one arbitrary term of the

4See Smirnov, Course of Higher Mathematics, vol. II , 1931, p.385.

 Chapter 2. On the Transmission Capacity of the "Ether" 31

series (1), then its frequency spectrum is contained within the limits between 0
and h. Therefore any sum composed of terms of the series (1), and hence the
sum of the series itself, consists of the frequencies, which are contained within
the limits between 0 and h. D

All terms of the series (1) are of the same form and differ only in the
factors Dk and in time. One of the terms, corresponding to the index k, is
plotted in Figure 2.1. It reaches the maximum at t = 2}" and its amplitude
decreases progressively on both sides.

Theorem II. Any function F(t), which consists of the frequencies between 0
and h, may be transmitted continuously with any accuracy by means of numbers
which follow one another at intervals of 2}, seconds.

Proof: Indeed, measuring the value F(t) at t 211 (n is an integer), we

FC;J
obtain:
= DnWl. (8)

All terms of the series (1) vanish for this value of t, except the term with k = n,
which equals DnWl. Thus, after each 2}, second we can determine the next Dk.
Transmitting these Dk in turn after each 2}, second, we may, according to the
equality (1), recover F(t) from them term by term with any accuracy. D

Theorem III. It is possible to transmit arbitrary numbers Dk continuously

and uniformly, with the speed of N numbers per second, by means of function
F(t), whose summands are arbitrary small for frequencies larger than h = ~.

Proof: Indeed, as soon as we received a number Dk, we construct the function

Fk(t) such that

for t < --"'-

2h - T

for 2;' - T < t < 2;' + T (9)

for t > 2;' + T

and transmit their sum F(t). If we would have T = 00, then the obtained function
F(t) would consist only of the frequencies smaller than h because we would
have obtained the series (1). Unfortunately, it is impossible to construct such
infinite series, and we have to restrict ourselves to finite T. Let us prove, than the
larger is T, the smaller are the amplitudes on frequencies f > h, and that these
amplitudes may be made arbitrary small. To do it, let us find the amplitudes
C(w) and S(w) for the function (9), substituting it into the equation (3). We
32 V. A. Kotel'nikov

Six

Figure 2.2

obtain
1/
C(w) = -
271 +T
Dk
Sinwi (t - 2~1) Coswtdt
k
7r k
211- T t -2h
-
(10)

1
S(w) = -
/27 1
+T
Dk
Sinwi (t - 2~
k 1
)
Sinwtdt
7r k
2h -
T t -2h
-
After the integration (see Appendix II) we obtain:

C(w) = I?rk Cos wit [SiT (w + WI) - Si T (w - wI)] }

I?f Sin w ik
(11)
S(w) = lSi T (w + WI) - Si T (w - wd]

In this expression Si denotes the integral sine, i.e. the function

x
·
SIX= jSinYd
- - y. (12)
Y
o

This function is tabulated, its values may found in tables 5 It is plotted in Fig-
ure 2.2. As it is seen from the figure, Si x tends to ~ as x -+ 00. Let us consider
the value of the square parentheses in expression (11). In Figure 2.3a this is plot-
ted for T = 2~1' in Figure 2.3b for T = 2~1' in Figure 2.3c for T = 22~, and in
Figure 2.3d for T = 00. As it is evident from these plots, the value within square
parentheses tends to the limit depicted in Figure 2.3d as T increases, i.e.

for w > WI [ ] = 0

for w < WI [ ] = 7r.

5See, for example, E. Jahnke und F. Emde "Funktionentafeln mit Formeln. und Kur-
ven".
Chapter 2. On the Transmission Capacity of the "Ether" 33

0)
T- 3
- 2/

CD
T- 6
- 2/

II
(0
T- 24
- 2/

W @

T=oo
i
7r

J w

Figure 2.3
34 V. A. Kotel'nikov

This is also evident directly from the expression (11), since with the increase of
T the scale of W is as if increasing and Si becomes rapidly damped. Thus the
resulting sum of Fk(t) has arbitrary small amplitudes on the frequencies larger
than f > h, for T big enough. From the received function F(t) it is easy to
recover the transmitted values Dk. Since all terms except one, with k = n, vanish
at t = {k, and the term with k = n equals Dnw, then

Thus, measuring the value of our function at t = 2~1 ' we are able to obtain the
value of a new Dk after each 2} second. In one second we receive N = 2h
~m~rn. 1 D

Functions Which Consist of Frequencies Between 11

and 12
Let us prove

Theorem IV. Any function which consist of frequencies between hand 12,
may be represented in the form

(13)

where

and FI (t), F2 (t) are functions which consist of the frequencies from 0 to f =
12-h
~.

Conversely, if in the equality (13) FI (t) and F2(t) are any functions, which consist
of the frequencies between 0 and f = ¥, then f(t) consists of frequencies
between hand 12.

Proof: If F(t) consists only of frequencies between hand 12, then C(w) and
S (w) for this function may be represented as

C(w) = S(w) = 0 for w > W2 or w < WI,

C(w)

for WI < w < W2,

S(w)
 Chapter 2. On the Transmission Capacity of the "Ether" 35

or, introducing once again the notations

}
we obtain:

C(W)

(14a)
S(W)

for

and
S(W)=C(W)=O} (14b)
for W > W2 or W < Wl .

Substituting the equalities (14) into the equality (2), we obtain after integration
and some transformations (see Appendix III):

C Wl +W2
os 2 t +

S · Wl + W2 t (15)
m 2 '

or, denoting

(16)

(17)
36 V. A. Kate} 'nikav

and taking into account that, according to Theorem I, H (t) and F2(t) must
consist of frequencies from 0 to f = h; 12, we may consider the first part of
Theorem IV as proved.
Since, according to Theorem I, it is possible to represent any function consisting
offrequencies from 0 to 12; h , by the series (16) and (17), and since no conditions
are posed on the coefficients D k , which appear in these series, the second part of
Theorem IV holds as well. D

Let us prove two theorems which are generalizations of Theorems II and

III.

Theorem V. Any function F(t) which consists of the frequencies from h

to 12 may be transmitted continuously, with any accuracy, by means of numbers
which follow one another at intervals

1
2(12 - fI)

seconds.

Proof: Indeed, at t = 12!h (k is an integer) we obtain according to formula

(13):

F (-k)
- -F (k)
h+h -
I
h+h'
-- (18)

since at this value of t cosine equals 1 and sine equals O. At t = f~:t we obtain

for analogous reasons. Thus, after each h ~ 12 seconds, we may determine the
values FI (t) and F2 (t) one by one. From these values we may recover the functions
FI (t) and F2 (t) themselves, because according to Theorem II from the numbers
which follow so frequently one may recover functions which consist of frequencies
from 0 to f = 12 ~ h , whereas the functions FI (t) and F2 (t) consist of frequencies
between 0 and f = 12; h only.
According to Theorem II, each of the two functions FI(t) and F2(t) may be
transmitted by means of numbers which follow one another at intervals of 12~h
seconds, and these two functions simultaneously may be transmitted by means of
numbers which follow one other at intervals 2(12~h) seconds. From these numbers
we recover first FI(t) and F2(t). Further, according to the formula (13), we may
recover the function F(t) itself from them. D

Theorem VI. If
N = 2(12 - h), (19)
 Chapter 2. On the Transmission Capacity of the "Ether" 37

then it is possible to transmit arbitrary numbers Dk continuously and uniformly,

with the speed of N numbers per second, by means of the function F(t), with
summands that are arbitrary small (i.e. practically vanishing) for frequencies f >
12 and f < h·

Proof: Indeed, according to Theorem III, we may transmit N numbers per

second by means of two functions FI (t) and F 2 (t), moreover for each of them its
summands are arbitrary small on the frequencies higher than ¥.
The same functions may be transmitted uniformly by means of the function
F(t), with summands that are arbitrary small on the frequencies f > 12 and
f < h· Indeed, according to the equality (13), we construct the function F(t)
from FI(t) and F2(t). Transmitting the former, we may recover FI(t) and F2(t)
from it, and finally recover the transmitted numbers. D

The last theorem says that it is impossible to transmit too much by

functions, which consist of frequencies occupying some restricted range. To
prove the theorem, we prove the following lemma:

Lemma 2.1. It is impossible to transmit M numbers by means of N numbers,

if
M<N (20)

Proof: Assume that it is possible. It is clear that the M numbers mi ... mM

can be regarded as some functions of N numbers ni ... n N , i. e.:

(21)

Knowing only M numbers mi ... mM and of course the functions i.f!1 . . . i.f!M'
we have to recover the numbers ni ... nN from them.
However, this is equivalent to solving M equations (21) with N variables. And
it is impossible to do this if the number of the equations is less than the number
of the variables, i.e. if the inequality (20) holds. D

Theorem VII. It is possible to continuously transmit arbitrary numbers,

following one another uniformly, with the speed of N numbers per second, and
M arbitrary functions FI (t) . . . F M (t) with the frequency range of the widths
D..h, ... , D..fM' by means of numbers, following one another uniformly, with
the speed of N' numbers per second, and of M' functions F{(t) ... F~, (t), with
the frequency range of the widths D..f{, ... , D..f~, if

(22)
38 V. A. Kotel'nikov

This can not be done in any way, if

M M'
N +2 L I::.fk > N ' +2 L I::.f£· (23)

The first part of this theorem is proved on the basis of Theorem V and
Theorem VI.
Indeed, according to Theorem V, we may transmit our N numbers per
second and M curves by means of P numbers per second, if

M
p= N +2 Lb./k (24)

These P numbers per second may be transmitted partly by means of N'

numbers per second and partly, according to Theorem VI, by the curves
F{(t) ... F~, (t), if the equality (22) holds.
The second part of the theorem is proved by reductio ad absurdum, on
the basis of the lemma.
Assume that we need to transmit P arbitrary numbers per second. Ac-
cording to Theorem VI, it is possible to do it, transmitting N number-
s per second and functions Fl (t) . . . F M (t) with the frequency ranges
b.h, ... , b.fM' if the equality (24) holds. If the second part of the the-
orem would be false, these number and functions could have been trans-
mitted by means of functions F{ (t) . . . F'M ,( t) and of N' numbers per
second. According Theorem V, the latter numbers and functions may be
transmitted by means of pi numbers per second, if

M'
pi = N ' +2 L b.f~. (25)
1

In other words, we would be able to transmit continuously P numbers per

second by means of P' numbers per second, despite the fact that according
to the equalities (24) and (25) and to the inequality (23),

P> P'.

Thus the assumption that the second part of Theorem VII is false leads us
to a result which contradicts the lemma.

Channel Capacity by Telephone Transmission

The speech, the music and other subjects of the telephone transmission are
arbitrary functions of time, which consist of the frequency spectrum whose
 Chapter 2. On the Transmission Capacity of the "Ether" 39

width is completely determined and depends on the quality of the sound

transmission.
Transmitting this function by wire or by radio, we transform it into
another function of time, which properly speaking we transmit. According
to Theorem VII, by the continuous transmission the bandwidth of the latter
function must be not narrower than the bandwidth of the function which
we would like to transmit.
Thus, a continuous telephone program cannot use a frequency range
smaller than the spectral width of the sound frequencies needed for this
program. This is true independently of the method of transmission. It is
impossible to invent any method which would allow to use a narrower
bandwidth.
As it is known, at present such a minimal frequency spectrum can be
obtained by means of the single side band transmission.
The condition "by continuous transmission" is very important, because
transmitting some sounds, for example music, with interruptions, it is pos-
sible to use the frequency range which is smaller than the width of the
sound spectrum which we will obtain in the process. To do this, it is e-
nough first to record the transmitted music onto gramophone records, and
then transmit from these records, turning them, say, twice as slower as in
recording. Then all frequencies become twice as small as the normal ones,
and we can use the half-width frequency range. It is possible to recover
such a message by means of gramophone as well. It is clear that such a
method of transmission can not increase the channel capacity, because by
such a transmission the ether or the wire are busy all the time, but the
transmission go on with interruptions.
This does not contradict Theorem VII, because it is stated there that it
is impossible to transmit "any function" and "continuously". However by
this kind of transmission we may transmit continuously either an arbitrary
function with interruptions, or a function not quite arbitrary but with
known breaks.
From Theorem VII it also follows that it is impossible to increase the
channel capacity by means of any non-frequency selection (except for di-
rectional antennas) or by means of something similar.
Indeed, if it would be possible to do this, then applying these methods
it would be possible to transmit from one place to another one n telephone
programs simultaneously, say, with the frequency spectra width 6.1 for
each one of them, using frequency range smaller than n6.f.
However with such a way of receiving the field intensity (or the current
intensity in the wire) from different programs would be mixed together into
a single function of time with frequency spectrum smaller than n6.1, which
would be received by the receivers. It would result, that we have been able
to transmit n functions of time with the frequency range widths 6.1 by
means of one function with the frequency range width smaller than n6.f.
According to Theorem VII, this is absolutely impossible.
40 V. A. Kate} 'nikav

From the aforesaid it is clear that to increase the channel capacity of

the ether for the telephone it is only possible either applying directional
antennas or extending the exploited frequency range by means of ultrashort
waves.

Transmission of Images and Telecasting With All

Semi-Tones
In the transmission of the images and in the telecasting we need to trans-
mit the degree of blackness of some N elements per second, and this is
equivalent to the transmission of arbitrary numbers with the speed of N
numbers per second. If we would like to do it by means of a function of
time (as it is usually done) than according to Theorem VII it must use the
frequency range not smaller than 1fperiods per second. Thus it is seen
immediately that here it is also impossible to decrease the frequency band
more that it can be done by using single side band transmission. Howev-
er, realizing this, we may encounter serious technical difficulties because of
phase interferences.
It is also impossible to decrease the frequency band by means of any
"group scanning of the images" (scanning multiple cells) because with such
a scanning one has to transmit, though by another method, the blackness
degree of the same N cells, i.e., N arbitrary numbers per second. t is
impossible to do this within a decreased frequency range.
Non-frequency selection methods can not help here for the same reason
that in the telephone transmission.

Telegraph Transmission and Transmission of Images

Without Semi-Tones or With a Limited Number of
Them
In the telegraph transmission and in the transmission of images without
semi-tones we deal with a transmission of some N elements per second
again. This is equivalent to the transmission of N numbers per second.
However, these numbers can not be quite arbitrary but must take preas-
signed values, known in advance.
Therefore, the theorems derived above are not applicable to the trans-
mission of such messages, because these theorems are concerned with the
transmission of arbitrary numbers unknown in advance.
Indeed, for the transmission of such messages the needed frequency range
can be decreased as much as one wants. Hence, the channel capacity can
be increased as much as one wants, at least theoretically. For this purpose
 Chapter 2. On the Transmission Capacity of the "Ether" 41

'"'1

Figure 2.4

one can act as follows. We want, say, to transmit with the speed of N
elements per second elements which can take the values 0 or 1 and to
use frequency range of !f only (instead of If, as in Theorem VII). For
this purpose we transmit two such elements by means of one element (or
number), for example, according to the following table, where the values of
the first element are given in the first column, the value of the second one
- in the second column, and the value of the element by means of which
we want to transmit them - in the third column.

I II III

o o o
1 o 1
1 1 2

o 1 3

Thus, we can transmit N two-valued elements by means of If four-valued

elements which can be transmitted over the frequency range !f. In practice
such a replacement of two elements by one can be realized by a circuit
diagram plotted on Figure 2.4, where <P1 and <P2 are either two photo-
electric cells or two telegraph apparata. The photo-electric cell <P1 activates
the modulator M1 that sends a unit amplitude to the line. <P2 activates the
modulator M2 sending an amplitude equal to three. By joint work of <P1 and
<P2 both modulators are activated. Because they are connected opposite to
each other, the amplitude equal to two is sent. The received signal reaches
three receivers. The first one of them, R 1 , works starting with amplitude 1,
the second R2 - starting with amplitude two, the third one - starting with
amplitude three. The first receiver sets L 1 , the second - L 2 , the third one
closes the access from R1 to L1 upon receiving amplitude three. By means
of such a circuit design we obtain the above decrease of the bandwidth.
In such a transmission one has to distinguish four instead of two levels
of the received signal. In view of this, it is necessary to increase the power
of the transmitter by the factor of 32 = 9, in comparison with the usual
transmission.
42 V. A. Kotel'nikov

It is possible to decrease the frequency band by the factor of n in an

analogous way, transmitting n two-valued elements by means of one 2n_
valued element (2n is the number of all possible combination of n two-
valued elements). However, for such a transmission one needs increasing of
the power by the factor of (2n - 1) 2 •
In the transmission of images with several semi-tones each element may
take several, say m, values (in this case m > 2). To decrease the bandwidth
by a factor n in such a transmission, it is possible to replace n transmitted
elements by one which must take mn values (mn is the number of all possible
combinations of n elements each one of them taking m possible values). It
is clear that one has to increase the power by the factor of

Obviously, such a decrease in bandwidth requires a huge power increase.

Moreover, the methods given above are very bad for the transmission
over the short waves because of fading.
For the wire communications the decrease of the frequency band by
means of this method may be of practical importance just now because
the power is small there and there are no fast variations of the receiving
intensity.

Appendix I
Substituting the expression (7) into the equality (4), we obtain

F(t) = l o
Wl 00

LDkCos~kw . Coswtdw +
-00
WI

+l
Wl 00

L Dk Sin ~ kw . Sinwtdw =
o -00
WI

f
-00
Dk
ior
' (cos ~ kw . Cos wt + Sin ~ kw . Sin wt) dw
WI WI

Integrating and replacing WI with 27l' h in the parentheses, we obtain

F(t)
00 Sin WI (t - 2~J
LDk k'
-00 t--
2h
 Chapter 2. On the Transmission Capacity of the "Ether" 43

Appendix II
Let us substitute
k
t = u + -I' dt = du,
2 I
into

1 !2~1 +T
C(w) = - Dk Cos wt dt
7f -A..-T
211

Then
C(w) = -1 jT Dk Sin WI u Cos W
( U + -I
k ) d
u
7f -T U 2 I

= -
1 jT Dk Sin WI u . Cos wu C k d
os w -I U +
7f -T U 2 I

+ -
1 jT Dk Sin WI U. Sin wu S.
lllW-
k
1
d
u.
7f -T U 2 I
The function under the second integral is even. Because of this the second
integral vanishes. The function under the first integral does not changed
by replacing u with -u. Because of this we may take this integral between
o and T, multiplying it by two. Hence,
C(w) = 2Dk cos~ (Sin WIU Coswu du,
7f 2h 10 u
or

C(w) ~k Cos 2;' [ [ Sin(w~+ w)u du - [Sin(W ,~ w,)u du 1

Replacing in the first integral
(WI + w) u = y,
and in the second
(W - WI) u = YI ,

we obtain:
Dk k
C(w) = - Cos-
[l(W+Wd T -
Sin Y
- dy -
l(W-Wd T -
Sin YI 1
- dYI .
7f 2h 0 Y 0 YI

The integrals in the parentheses are functions of its upper limits. It is

conventional to call them integral sines. Introducing this notion, we obtain

C(W) = ~k cos;l~ [SiT(w+wd - SiT(W-W d].

Doing the same with S(w), we obtain the equality (11).
44 V. A. Kotel'nikov

Appendix III
Substituting the equalities (14) into the equality (2), we obtain

+ l
WI
W2

L Dk
00

-00
Sin 7r
k
W2 - Wl
(w - wd Sinwt dw.

The limits are taken between Wl and W2 because of

C(w) = S(w) = 0

for

After trigonometric transformations

Replacing the difference of the sines with the product and simplifying, we
obtain

F(t)
00
Cos ( '" : W, t_H Sin [ W2 ::' (t - ~ J,) ) 1
k 2(j,
-00
2(j2-11)·

Or, regrouping the terms with the even and with the odd k together, we
obtain the equality (15).
 Chapter 2. On the Transmission Capacity of the "Ether" 45

Conclusions
1. In view of the "crowding of the ether" which is seen already by now
and in connection with the rapid growth of the radio engineering, es-
pecially in connection with the development of the short wave broad-
casting and of the telecasting, the research institutes must be con-
fronted with the acute problem of searching for methods of increasing
the channel capacity of the ether.
The problem of increasing the channel capacity of the wire is of great
economic importance as well. Because of this, this problem should be
also studied.

2. Because of the impossibility to increase the channel capacity of the

ether or of the wire by any method (for example, by means of the
overlapping of the frequencies of two different programs) more than
allowed by the single side band transmission, all attempts in this
direction should be given up as nonrealizable.

3. For the telegraph transmission and for the transmission of the im-
ages without the semi-tones or with a limited number of them, the
channel capacity may be increased without limits, at least theoret-
ically. However, such an increase involves a huge power increase as
well as a considerably more complicated equipment. Such a decrease
of the bandwidth may be applied in the wire communications in the
immediate future. This problem should be worked upon.
4. For first class programs (the telephone transmission and the semi-tone
video transmission) all efforts should be directed towards the elabo-
ration of methods of the single side band receiving and of the single
side band transmission as the methods allow a maximal exploitation
of the ether and of the wire.

5. It is necessary to study the problem of the increase of the channel

capacity of the ether by means of the directional antennas (both for
receiving and for transmitting).

6. It is necessary to enlarge the range of the exploited frequencies in the

ether by means of the ultrashort waves whenever is possible, as well
as by the study of this frequency range.

7. It is necessary to study the problem of the increase of the stability of

the frequencies of the radio stations, since it allows to compress the
ether.