ost people assume that the trend in communications has been from

analog to digital. Actually, digital communication came first. The first

practical means of electrical communication was by means of Morse code,
which is digital. lt is not binary, however, as it uses three elements (clot, dash.
Morse Code: The First and space) rather than two.
Practical Digital
Eommunication
5gstem

7"X ieltrod*ati*n
Many of tire signals used in modern communication are digital (for example. . ,
codes for alphanumeric characters and the binary data used in computer p: ,
grams). In addition. digital techniques are often used in the transmission of r, _-
log signals. Digitizing a signai often results in improved transmission qual .
with a reduction in distortion and an improvement in signal-to-noise ratic. r.
chapter looks mainly at the digital transmission of analog signals such as ro-_.
and video signals. Dara cornmunication wlli be studied in the Chapter 9.
Fi-gure 7.1 shows severai possible types of signal transmission. In Fis-:,
7.1(a) an analog signal is sent over a channel with no modulation. A typical ..
ample is an ordintry public-address system consisting of a uricrophone, an anl:.
fler. and a speaker and using twisted-pair wire as a channel.
Figure 7. l(b) shows ;rnalog transmission using modulation and democlu -
tion. Broadcast radio and television are good examples"
Figure 7. I (c) and 7.1(d) srart with a digital signal (for example. a data file 1i -
acomputer). In(c.t,thelinkcanhandlesomekindofdigitalpulsesignaldirecth _-
(d), the channel cannot transmit pulses directly (a radio channel, for exam:..
requires a modulation process, and an ordinary telephone connection cannor p*,
dc). In these cases the digital signal has to be modulated onto a carier at one i- _

and demodulated at the other. The moclem shown is a combination modula. -.

demodulator. Transmission of digital data will be discussed in the next chapter
Figures 7.1(e) and 7.1(f) show an analog signal that is digitized at the rr;_.-
mitter and converted back to analog fbrm at the receiver. The difference benr;.-
these two systems is that in (e) the transmission is digitai, while in (f) the transn' .-
sion channel cannot cary pulses, so modulation ancl demodulation are requirec
Analog transmission of analog signals, such as voice, seems to make sens; -
is certainly simpler than convertin-q the slgnal to digital form and back ag,
Similarly, it seems obvious that signals that begin as di-eital, such as the conte.
of computer memories. should be kept in digital form as much as possible. \\-. -
seems awkward at first glance is the idea of converting analog signals to dig:._
form for transmission. Actually the use of digitai techniques with analog siur,*
is one of the fastest growing areas in communications, for several good reasor..
Throughout this book, beginning in Chapter 1, we have looked at the etir,
of noise and distortion on analog signals. once noise and distortion are pre:.:
there is usually no way to remove them. Furthermore. the effects of these r-
pairments are cumuiative. Noise is added in the transmitter. the channel. antj .-.
zAO
 Section 7.1 S lntroduction

been from
it. The first I _--) Baseband Channel i___--
orse code, Anaiog Source Analog
idot, dash, Destination
(a) Analog signal and baseband transmission

Analog Source Modulator Demodulator Analog

(Transmitter) (Receiver) Destination
(b) Analog transmission using modulation and demodulation

i F__ r
- I Digital Channel
Digrtal Source Coder Decoder Digital
Destination
(c) Digital signal transmitted on digital channel
:rample, the
mputer pro-
sion of ana-
;ion quality, Analog Channel
Digital Source Modem Modem Digital
: ratio. This Destination
Lch as voice (d) Digital signal transmitted by modem
.o
. In Figure
typical ex- Digital Channel
:. an ampli- Analog Source A,/D Conversion Decoding and Analog
and Coding D/A Conversion Destination
(e) Analog signal transmitted digitally
demodula-

tafile from
drectly. In
r example,
Analog Source A/D Conversion Modem Channel Modem Decoding and Analog
and Coding D/A Conversion Destin"ation
annot pass (f) Analog signal digitized and iransmitted by modem
at one end )
nodulator- Figure 7,! Analog and digital communication
'hapter.
the trans- :ceiver. If the communication system involves several trips through arnpliliers
e between .nd channels, as in a long-distance telephone system, the signal-to-noise ratio
transmis- .-, gradually decrease with increasing distance from the source.
rl1
:quired. Digital systems are not irnmune from noise and distortion, but it is possible to
e sense. It :educe their effect. Consider the simple digitat signal shown in Figure 7.2. Sup-
ck again. .l
rose that a transmitter generates v for a binary one and 0 v for a binary zero.
) contents rhe receiver examines the signal in the middle of the pulse and has a decision
5le. What .hreshold at 0.5 v; that is, it considers any signal with an amplitude greater than
to digital i).5 v to be a one and any aniplitude less than that to represent azero. Figure
,g signals -.2(a) shows the signal as it emerges from the
transmitter. and Figure 7.2(b)
reasons. .hows it after its passage through a channel that adds noise and distorts the pulse.
re effects ln spite of the noise and distortion. the receiver has no difiiculty deciding cor-
present, fectly whether the signai is a zero or a one. Since the binary value of the pulse is
hese im- the only information in the signal, the distortion has had no effect on the trans-
. and the mission of information.
 Ehapter 7 ,if Digital fommunications
t

i:I15::r-+: ?."..i:li Removal of noise and

distortion from digital signal

S 1v

Time
(a) Digital signal as transmitted

Threshold

Time
(b) Received signal with added noise and distortion

5i*;""
Distorted
3,"ff1,"""
Receiver Transmitter
(c) Digital repeater

The perfectly received signal of Figurei.2(b) could now be used to gener:.:

a new pulse train to send fumher down the channel. This receiver-transmitter cor:-
bination, called a regenerative repeater and illustrated in Figure 7.2(c),has n_
only avoided adding any distorlion of its own, but it has also removed the effe;:,
ofnoise and distortion that were added by the channel preceding the repeater. Tl:
elusive goal of distortionless transmission seems to have been achieved!
Unfortunately, as in most areas of life, absolute perfection is impossibi.
Since noise is random, a noise pulse can have any amplitude, including one th.
will cause a transition to the wrong level. Similarly, extreme distorlion of puls;,
can cause errors. Figure 7.3 demonstrates these problems. Errors can never b.
eliminated completely, but by judicious choice of such parameters as signal lei e.,
and bit rates, it is possible to reduce the probability of error to a very small valu.
There are even techniques to detect and correct some of the errors.
The other major source of error in the digital transmission of analog signa_.
appears in the conversion of the inflnitely variable analog signal to digital fonr
This conversion inevitably results in the loss of some information and the crearic,
of a ceftain amount of noise and distortion. once again, however, it is possible i-
predict quite accurately the amount of error that will be introduced and reduce :.
to any required value.
other advantages of digital communication include convenience in muit.-
plexing and switching. Time-division multiplexing (TDM) is quite easy with dis-
ital signals, and different types of signals (for example, voice and data) can b.;
multiplexed together on the same channel.
The only real disadvantages of digital systems are their greater complexir,
and the larger transmission bandwidth they require. Large-scale, low-cost di_ri-
tal integrated circuits are reducing the difficulty and expense of constructin;
complex circuitry, and ingenious data-compression techniques, combined wirr
 Section 7-l & lntroduEtion

FEqcsre ?"H Excessive noise on a

digital signal
010011101101

Time
a)Digital signal as transmitted

0)
b0
,J 1v
Threshold'--?---- --

Time
bl Received signal with excessive noise and distortion

ar-atec

are beginning to decrease the

-:r-bandwidth media like Iiber-optic cable,
- :s idth penalty. In general, the advantages outweigh the disadvantages'

hannels and Information CaPacitY

J0ir1- they pass frequen-
. :ractical communication channels are band-limited; either
.: nrri
]:Iiomdctosomeupperlimit,ortheyhavebothloweranduppercutofffre.
thermal noise' and
-:r.ies. In addition, alichannels are noisy' There is always
of noise)'
-
:i- be other kinds of noise as well (see chapter 1 for a discussion
111o]
be sent along
There are theoretical limits to the rate at which information can
.lble ratio. of course, there is no
_ irnnel with a given and signal-to-noise
bandwidth
: ih.. time to
to the amount of information that can be sent if there is sufficient
--'::

:r be
. j lt. The relationship between time, information capacity' and channel band-

:ih is given by a simple equation called Hartley's Law:

I : ktB O.L)

:n"]. ..re 1 : amount of information to be sent

tttll]^. k: aconstant
/: time available E
1- -
B: channel bandwidth d
t.-3 -:
='n1ey,s law tells uS that the information rate (that is. the amount of information
..'..unbesentinagiventime)isproportionaltothebandwidthofthechannel.It
lr - - codins used
:t - : rlso propoftional to the factor ft, which depends on the type of data " (
, i the signal-to-noise ratio ofthe channel' Lower Center UPPer
If the noise is assumed to have the same flat spectrum and random ampli- Cutoff FrequencYCutoff
is of the type called FrequencY FrequencY
-le as thermal noise (the amplitude probability function shape shown
:.;ttssian.) and if the bandwidth of the channel has the square-sided FrequencY
c_,:-- given channel can be
, Figure 7 .4, thenthe maximum theoretical data rate for a
- t--,: ::q;..:;;1q,: .::-.:= Ideal bandpass channei
'i: ,-: -rnd using two simPle equations.
?44 Ehapter 7 s trigital Eommunications

First, ignoring noise, there is a limit to the amount of data that can be sent l:,
a given bandwidth. This limit is given by the Shannon-Harlley theorem:

C - 28 logzM (7.:

where C : information capacity in bits per second

B: the channel bandwidth inhertz
M: number of levels transmitted

In the above equation, C and B should be easy to understand. M is a little 1e.,

obvious. The idea is perhaps best explained by looking at a low-pass (rather th-
a bandpass) channel. Suppose that the channel can pass all frequencies from ze:.
to some maximum frequency B. Then, of course, the highest frequency that c-
be transmitted is B. Suppose that a simple binary signal consisting of altern;:.
ones and zeros is transmitted through the channel. This time let a logic 1 be 1 ''
and a logic 0 be -1V The input signal will look like Figure 7.5(a): it wili be -
square wave with a frequency one-half the bit rate (since there are two bits, a o:.
and a zero, for each cycle). Since it is a square wave, this signal has harmonics -
all odd multiples of its fundamental frequency, with declining amplitude as r.
fiequency increases. At very low bit rates, the output signal after passage throus-
the channel will be similar to the input, but as the bit rate increases, the frequen; ,

of the square wave also increases and more of its harmonics are filtered ou:
Therefore, the output will become more and more distorted. Finally, for a bit ra:.
of 28, the frequency of the input signal will be B, and only the fundamental of rL.
square wave will pass through the channel, as shown in Figure 7.5(b). Noneth=-
less, the receiver will still be able to distinguish a one from azero, and the info:-
mation rvill be transmitted. Thus, with binary information, the channel capaci -
will be

C :28
Now suppose that instead of only two possible levels, several different ler -
els, each corresponding to a different number, are possible. For instance, eac.,
level could represent one of four possibilities. The possible levels could b.
-1V, -0.5 V, +0.5 V, and *1V. Figure 7.6 shows this four-level code. S-ir-:.
this code, measuring the voltage level once at the receiver would actuall..
provide two bits of information, since it takes two bits to express four differer,.

0)
bD b0
c! cd

-1V

Time Time
(a) Input square-wave signal (b) Output signal at maximum rate

Figure 7.5 Digital transmission through a Iow-pass channel

 SeEtion 7-l s lntroduEtion za5

sent m

(:7.2) o
b0

-1V

le less fime
:r than
n zero gure 7-6 Four-1eve1 code
at can
ernate
rc1V
lbea ..ibilities. However, the maximum frequency of the signal would not change.
a one - have, it seems, managed to transmit twice as much information in the same
ucs at
-
-:Ju'idth. This idea can be expanded to any number of levels, in order to give
as the - -.rtion (7.2).
rough It might seem that any desired amount of infonnation could be transmitted in
channel by simply increasing the number of levels. This is not possible,
,1ency
i out.
' -.\en
.,. ever, because of noise. The more levels there are, the closer together they are,

it rate j rhe more like1y it is that noise will cause the receiver to mistake one level for
of the
- rher (this can also be seen from Figure 7.6). Therefore, there will be, for a
tethe- -. :n noise level, a maximum data rate that cannot be exceeded without errors,
infor- :natter how elaborately the data is coded. This maximum rate is called the
raci$ ' ;nnon limit:
C:Blog2(1 +S/N) (7.3)

:re C: information capacity in bits per second

B: bandwidth in hertz
each S/I/ : signal-to-noise ratio (as a power ratio, not in decibels)
dbe
\\ irh There is no contradiction between Equations (.7.2) and (7.3). Each represents
uailr' - :-arirnum rate. so the one that applies in a given situation is the one that gives
-: lorver data rate. Remember also that these rates are theoretical maxima, and
3rent
:.-. practical equipment is unlikely ever to reach these limits.
-\t this point, a small reminder may be in order. Not every scientific calcula-
: rs capable of linding logs to the base 2, but they can all find logs to the base 1 0.
- irindy way to find logs to the base 2 is to use the following equation:
1o s , ,,ly'
(7.4)
log 112

EXAMPLE 7.1 bandwidth of 3.2kHz and a signal-

A telephone line has a
:o-noise ratio of 35 dB. A signal is transmitted down this line using a four-
.eve1 code. What is the maximum theoretical data rate?
.i ea5 Ehapter 7 * tligital Eommunications
I

Solution First, we use the Shannon-Hafiley theorem to find the maximLi:'

data rate for a four-level code in the available bandwidth, ignoring noise.
From Equation (1.2),

C:28log2M
:2(.3.2 x 103)Xlog24
: 12.8 x 103 b/s
: 12.8 kb/s

Next, we use the Shannon lirnit to find the maximum data rate for any cod;
given the bandwidth and signal-to-noise ratio. Remember that S/N is requi:. -
as a power ratio.

S/N : antilogl6 r35/l0t

: 3162

From Equation (7 .3')

,= B 1og2 (i + S/N)
(3.2 x 103) 1og2 (1 + 3162)
31.2kbls

Since both results are maxima, we take the lesser of the two, 12.8 kb/s. Th.
means that it would be possible to increase the data rate over this channel b'
using more levels.

SCTE'FSEru ?"8 ffiffiE/$ffiq# The Shannon-Hartley theorem shows that bandwidth and information rate are
ffiAJffiSru'HgBru interchangeable, at least in theory. That is, a large quantity of information can be
transmitted in a small bandwidth by using more time. Give an example of the
application of this theorem. Is it always possible to take advantage of the
equivalence between bandwidth and time?

t:.'i.,,..=,:.,:..-
li.,j. rilr1.1i;L-,1 : =, l:::

To transmit an analog signal by digital means, it is Iirst necessary to sampl; - ,,

signal at intervals. The amplitude of each sample can then be expressed as a bi: --
number for transmission. At the receiver, the samples can be reconstituted '- -
used to form a replica of the original signal.

Sampling
.
In I 928, Har:ry Nyquist showed mathematically that it is possible to reconstru - -
band-limited analog signal from periodic samples, as long as the sampling rr:-
at least twice the frequency of the highest frequency component of the si,c- -
This assumes that an ideal low-pass fi1ter prevents higher frequencies from er:.-
ing the sampler. In practice, the sampling frequency should be consider::
greater than twice the maximum frequency to be transmitted. For exampie :
telephony, a sample rate of 8 kHz is used for a maximum audio frequencl'
3.4kI12, and compact disc systems have a 44.lkHz sampling rate and & n1:'
mum audio frequency of 20 kHz.

-
 Sertion 7-Z * Pulse Modulation

rimum
se.

Samples Samples remain at

follow signal's level at start
signal of sample.
level.
code,
required a) Natural samPling (b) Flat-topped samPling

F iqure 7"7 Natural and flat-topped sampling

Figure 7.7(a) demonstrates the simplest type of .sampling, called natural

,ampling. The incoming analog signal is low-pass flltered and then multiplied by
:ulse train. If we assume that the pulses have an amplitude of 1 V at the input to
then the output voltage will Lre equal to the arnplitude of the analog
: multiplier,
_:nal at the time of the pulse. Since a real
pulse has a linite duration, the ampli-
-:e of the pulse varies during the length of time it is present. In some situations
s. This tere this is unclesirable, a sample-and-hold circuit can be used to keep the
nnel by ,.-\e amplitude constant for the duration of the pulse. This variation, which is
'. rrr
n in Figure 1 .7(b), is called flat'topped sampling.
The penalty for a sampling rate that is too low is called aliasing or foldover
jistortion. In this form of distortion, frequencies are translated downward.
. :ure 7.8 shows how aliasing develops. In Figure 7.8(a), the sampling rate is
i:e t3quate and the signal can be reconstructed. In Figure 7.8(b), however, the rate
l,in be , too low and the attempt to reconstruct the original signal results in a lower-
lie lquency signal. Once aliasing is present, it cannot be removed'
The frequency of the interf'erence generated by aliasing is easier to see by
-rking at the frequency domain. Recall from Chapter 3 that multiplying a mod-
----------------{
- rting signal and a sine-wave carrier produces a double-sideband suppressed-
-,rrier signal. We might expect similar results when the carrier is a pulse train.
_:t us find out. For simplicity, we shall assume that the modulating signal is a
. :re wave.
:.imple the
rs a binary €u, : E,, sin co,r,/ (7.s)
:ituted and
,.nere e- : rhe instantaneous modulating-signal voltage
E* : the peak modulating-signal voltage
<o,, : the radian frequency of the modulating signal
.-onstruct a
1in-e rate is -he carrier signal is a pulse train like the one in Figure 1.9.Let us assume for
the signal. ,implicity that the pulses have an amplitude of 1 V. The spectrum for such a signal
rom enter- . qiven in Chapter 1 as:
rnsiderably
2 rtlT
rarnple, in ,,: i't + 2 ir (#
/ sin ttlT
cos .u,r + sin2*ilr cos 2o,/
:quency of (7.6)
nd a maxi- sin 3nt/f
CoS J t'J.I -
3nrlT )
 Ehapter 7 S Digital EommuniEataons

Sampling fimes Sampling Times

Original signal Samples Reconstructed signal
(a) Satisfactory sampling rate

nlltrlil

1ilil

,utif,

rti[

Sampling Times Sampling fimes ,L{!il

Original signal Samples Reconstructed signal iiiltf,

(b) Sampling rate too low ,itr

i.:i::::,.-:-: - J',,ti Aliasing

ln!

iW

k-T------4 where es the instantaneous voltage of the sampling pulse ,rll]ll

T the duration of a pulse

Voltage T the period of the pulse train
o)s the radian frequency of the pulse train

Multiplying the two signals together gives the following output:

T
: E*7sin r / sin r.r/T
v(t) "'r |\ trtT sin t,l.-I cos t,l^r
a*t + 2E*- (7.7)

sin2rr. lT sin3rrlT
lE^ sin <rr./ cos 2osl + E^ sin oar cos 3o)rl +
2m/T )
As is often the case with equations of -this type, we do not have to o'solve"
anything. We simply need to study the equation to see what is happening. In par-
ticular, look at the first term. It is just the modulating signal multiplied by a con-
stant. This tells us that applying a low-pass filter to the pulse signal can recover
the original modulation, provided that there are no other components of the signal
occupying the baseband frequency range.
Next, look at the second term, which contains the product of sin <o-l and cos
o"/. We have seen many terms of this sort before, and we should remember that
they represent sum and difference frequencies. In this case the frequencies
produced will be f, - f* and f" + f*. (We have changed from radian notation
 Section 7-Z * Pulse Modulation za9

Pr

f* f"* f*
Frequency
a) Satisfactory sampling rate (b) sampling rate too ,o*u'"nt""'
.::"::: Aliasing in the frequency domain
=...:i..:

::.ause while radian notation simplifles equations, most practical work is done
!h hefiz, not radians per second.)
Figure 7.10 shows sketches of the frequency spectrum for the first two terms
. Equation (1 .1). The subsequent terms produce still higher frequency compo-
-.rts and are not of interest in this discussion. In Figure 7.10(a), the baseband fre-
"f, - f*,
, -3ncy is lower than the component at so a low-pass filter can recover the
-'.eband signal. In Figure 7.10(b), however, the difference term has a lower fie-
- -3ncy than the original baseband signal, and the two cannot be separated by a
'.\ -pass filter. This is a frequency-domain representation of the aliasing that we
..'.r earlier in the time domain.
It appears from Figure 7.10 that aliasing will take place ifl, is greater than
- .f-.If "[ : 2f,,,, then f" - f,, : f*, and we are just on the edge of aliasing. In
:Jer to separate the signals with a real low-pass filter, the carrier frequency
'. hich is also the sampling rate) must be greater than 2f,,,, as stated earlier. If

-.Lasing does take place, the interfering component will be at a frequency

f,: f, - f^ (7.8)

.'. here the frequency of the aliasing distorlion

.ls - the sampling rate
.lnt - the modulating (baseband) frequency
(7.7)

't
EXAI\,IPLE tl.m An attempt is made to transmit a baseband frequency of
-10 kHz using a digital audio system with a sampling rate of 44.1 kHz. What
solve"
audible frequency would result?
n par-
a con- Solution From Equation (7.8),
)cover
signai .la - Js Jnt
: 44.1kHz - 30 kHz
rd cos : l4.l kHz
:r that
:ncies In addition, of course, the original baseband signal of 30 kHz would be present.
tation
 Ehdpter 7 e trigital Cornmunications

Analog Pulse-Modulation Techniques

Sampling alone is not a digital technique. The immediate result of sampling :; -
pulse-amplitude modulation (PAM) signal like the one in Figure 7.1l(b). p{l".
is an analog scheme in which the amplitude of each pulse is proporlional to --.:
amplitude of the signal at the instant at which it is sampled. The exampie -
Figure 7.1 1 uses flat-topped sampling (described earlier), in which a sample-ar,:-
hold circuit is used to keep the amplitude constant during each pulse.
For our pulposes, the main use of PAM is as an intermediate step; bef,::.
being transmitted, the PAM signal has to be digitized. Similarly, at the receir.:
the digital signal must be converted back to pAM as part of the demodular:::
process. The original signal can then be recovered using a low-pass filter.
Before looking at digital schemes, a quick glance at two other analog pui..,
modulation techniques is in order. Pulse-duration modulation (pDM), which .,
shown in Figure 7. 1 1(c), uses pulses that all have the same amplitude. The dur.-
tion of each pulse depends on the amplitude of the signal at the time it is sar:,-
pled. PDM has its communications uses; for instance, it is often used in the hig:-
powered audio amplifiers used to modulate AM transmitters. It has also be;-
used for telemetry systems. Though still an analog mode, it is more robust th-
PAM because it is insensitive to amplitude changes due to noise and distonio:,
Like PAM, PDM can be demodulated using a low-pass firter. Sometimes the ter.'
pulse-width modulation (PWM) is used for this method.

0)
o

q q
i
{

fime Time
(a) Original signal (b) Pulse-amplitude modulation (PAM)

C)
0)

p. g
J

Time fime
(c) Pulse-duration modulation (PDM) (d) Pulse-position modulation (PPM)

Figure 7.It Analog pulse modulation

 seEtion 7-3 s Pulse'Eode Modulation z9t

in Figure 7'11(d)' is. closely

Pulse-position modulation (PPM), shown
nglsa .:edtoPDM.Allpulseshavethesameamplitudeandduration'buttheirtiming
sees some use in
I. PAM .:s with the amplitude of the original signal' PPM also
Ito the :netry systems.
ryle in
te-and-
that can be sEETIIIN 7-E REVIEW
\\'hy is there always a definite upper iimit to the baseband frequency
before QUEsTION
,ransmitted in a digital communication system?
ceiver,
ulation

pulse-
hich is -,3 Fulse'Code l\{odulation
: dura- FtedAudio"
s sam- ?ulse-codemodulation(PCM)isthemostcommonlyuseddigitalmodulation
_ .eIne. In pcM rhe av;ilable range of
signal voltages is divided into levels.
t hiCh- represented by the
l been J each is assigned a binary nu*L"t' Each sample is then
- '.ary number representing tire level closest to its amplitude' and this number is
il than
ortion. ..nsrnittedinserialform."InlinearPCM,levelsareseparatedbyequalvoltage
e term -' :dations.
- -ih"
of bits used to express
;r-ber of levels available depends on the number
: sample value. The number of levels is given by
(7.e)
N:2'
here number of levels
m: number of bits Per samPle

EXAMPLET.3Calculatethenumberoflevelsifthenumberofbitsper
sample is:

(a) B (as in telePhonY)

1b) 16 (as in compact disc audio systems)

Solution
from Equation (7'9),
ia) The number of levels with 8 bits per sample is,

N :2^
^8
:iso
(b) The number of levels with l6 bits per sample is.
irom the same equation'

N:2,,
: 2t6
: 65,536
 fhapter 7 :l lligital EBmmuniEataons

a a

o
Fl FI
b0 b0

N N

Cd cd
5
@ @

Time Time
(a) Samples before quantizing (b) Samples after quantizing
{
i':i=::::-::r--:,1-:-.i:l* Quantizing emof I L LiLIUUI]il

,it
This process is called quantizing. Since the original analog signal can h. .
an infinite number of signal levels, the quantizing process will produce er.:
called quantizing errors or often quantizing noise.
Figure 7.12 shows how quantizing errors arise. The largest possible effor
one-half the difference between levels. Thus the en'or is proportionately grea::-
for small signals. This means that the signal-to-noise ratio varies with the sig:-
level and is greatest for large signals. The level of quantizing noise can be ;:,
creased by increasing the number of levels, which also increases the number '
bits that must be used per sample. q
The dynamic range of a system is the ratio of the strongest possible sigr--
that can be transmitted and the weakest discernible signal. For a linear PCM s1,,, , :iu
tem, the maximum dynamic range in decibels is given approximately by
r i,i: ;.llu!

DR : (1.j6 + 6.02m) dB (7.ltl ,,. ilu

.ilm
where DR : dynamic range in decibels
la : number of bits per sample . :nmll:

This equation ignores any noise contributed by the analog portion of the system ' _iilr

r tht

rd&- rf,

Hx.&rujl$$*il y-'d$ Find the maximum dynamic range for a linear PCM
system using I 6-bit quantizing.

Solution From Equation (7.10)

DR: 1.76 + 6.02mdB

: 1.76 + 6.02 x 16
: 98.08 dB

Increasing the number of bits per sample increases the data rate, which i.
given very approximately by

D:f,* (7.11'
 SeEtiBn 7-3 3 Pulse-Esde Modulation 293

where D: datarate in bits per second

f, : sample rate in samPles Per second
m: namber of bitspersample

Extra bits arenneeded to detect and correct errors. A few bits, called framing bits,
are also needed to ensure that the transmitter and receiver agree on which bits
constitute one sample. The actual bit rate will be somewhat higher than calculated
above.

EX,q&t$3*-X f"X Calculate the minimum data rate needed to transmit

audio with a sampling rate of 40 kHz and 14 bits per sample'

Solution From Equation (7.11)

can have
Lce elTOrS D:f,*
:40x103x14
e error is : 560 x 103 b/s
y greater : 560 kbis
he signal
urbe de-
rmber of
Companding
le signal
The transmission bandwidth varies directly with the bit rate. In order to keep the
CM sys-
bit rate and thus the required bandwidth low, companding is often used. Com-
panding involves using a compressor amplifler at the input, with greater gain for
(7.10) low-level than for high-level signals. The compressor reduces the quantizing elror
for small signals. The effect of compression on the signal can be reversed by
usiag expansion at the receiver, with a gain characteristic that is the inverse of that
at the transmitter.
It is necessary to follow the same standards at both ends of the circuit so that
,.\ stem the dynamics of the ou@ut signal are the same as at the input. The system used in
the North American telephone system uses a characteristic known as the mu (y")
law, whichhas the following equation for the compressor:

V.ln(l -f pv;lVi)
V^: +
(7.12)
ln (1 p.)

where r/o : ouq)ut voltage from the compressor

V, : maximum output voltage
v; : actual inPut voltage
%: maximum input voltage
p: a parameter that defines the amount of compression
(contemporary systems use p : 255)

European telephone systems use a similar but not identical scheme called A-law
rlch is
compression.
Figure 7.13 shows the p.-255 curve. The curve is a transfer function for the
compressor and relates the input and output levels. It has been normalized, that is,
t7.11)
v i lV i and v o I V o are plotted, rather than v ; and v o-
 fhapter 7 e Digital fornmunications

Figure 7.13 Mu-law compression. Output Level (Normalized)

Input Level
(Normalized)

liLrLrr rililu

Iiitl]lt' ;tr

;:rrriluli

,,i]}lM

EXAMPI-E 7.8 A signal at the input to a mu-law compressor is positive.

with its voltage one-half the maximum value. What proportion of the
maximum output voltage is produced?

Solution From Equation (7 .12)

+ p"u/V;)
V" ln (1
uo
ln(l + p)
liu

ml
V,,ln(l+255x0.5)
1n (l + 255)
:0.816V.
This problem can also be solved graphically, as shown in Figure 7.13.

Digital companding is also possible. The method is to quantize a signal using

a greater number of bits thanwill be transmitted, and then perform arithmetic on
the samples to reduce the number of bits. This is the way companding is done in
most modern telephone equipment. This type of companding is part of the codinS
and decoding process, which is the topic of the next section.

Coding and Decoding

The process of converting an analog signal into a PCM signal is called coding:
and the inverse operation, converting back from digital to analog, is known as
decoding. Both procedures are often accomplished in a single integrated-circuit
device called a codec. t
 Section 7.3 s Pulse-fode Modulation 295

t;;-] l---------------l
A""trs l3-l I l-l T- l I Digital
Inpui Low-Pass Filter Sample-and-Hold A/D Converter (and ParalleVSerial Serial
Circuit Compression if Conversion Output
Required)

Fiqure 7-14 PCM coding

Figure 7.14 is a block diagram showing the steps for converting an analog
signal into a PCM code. The first block is a low-pass filter, required to prevent
aliasing. As shown in section 7 .2, the filter must block all frequency components
above one-half the sampling rate. This requires a high-order filter.
The next step is to sample the incoming waveform using a sample-and-hold
circuit. There are many such circuits; a simple one is shown in Figure 7.15. The
field-effect transistor Q turns on during the sample time, allowing the capacitor to
charge to the amplitude of the incoming signal. The transistor then turns off, and
the capacitor stores the signal value until the analog-to-digital converter has
had time to convert the sample to digital form. The two operational amplifiers,
connected as voltage followers, isolate the circuit from the other stages. The low
output impedance of the first stage ensures that the capacitor quickly charges or
discharges to the value of the incoming signal when the transistor conducts.
The samples must now be coded as binary numbers. If we are using linear
FCM, all that is required is a standard analog to digital (A/D) converter. Com-
pression, if required, can be applied to the analog signal, but it is more common to
incorporate the compression into the coding process.
The codecs used in telephony generally accomplish compression by using a
piecewise-linear approximation to the mu-law curve shown earlier in Figure 7.13.
The positive- and negative-going parts of the curve are each divided into eight
segments, with an additional segment centered around zero, resulting in a total of
sixteen segments. Figure 7.16 shows the segmented curve. Segments 0 and t have
the same slope and do not compress the segment. For each higher-numbered
segment, the step size is double that of the previous segment. Each segment has
sixteen steps. The result is a close approximation to the actual curve.
The binary number produced by the codec in a telephone system has eight
bits. The first is a sign bit, which is a one for a positive voltage and a zero for neg-
ative. Bits 2,3, and 4 represent the segment number, from zero to seven. The last
four bits determine the step within the segment. If we normalize the signal, that is,
set the maximum input level equal to one volt, the step sizes can easily be calcu-
lated as follows: let the step size for the 0 and 1 segments be r mV. Then segment

Output

ng.
:: Sampling Pulse
L.:
Figure 7.t5 Sample-and-holdcircuit
296 Ehapter 7 S Digital fornrnunications

6/ 2 has a step size of 2x, segment 3 a step size of 4r, and so on. Since each segmer
9^ has 16 steps, the value ofx can be found as follows.
Eb
2r
E4 l6(x + x I 2x t 4x t 8x * 16x * 32x + 64x) : l000mV
9.
tso x : 0.488 mV
Br
a- The relationship between input voltage and segment is shown in Table 7.1.
1
0
Input Voltage

F5g-g:-ss* ?-€S Segmented mu-law Mu-Law compressed PCM coding

curve (Positive half)
Segment Voltage Range (mV) Step Size (m\ t

0 0-7.8 0.488
1 7.8-15.6 0.488
2 15.6-3 r.25 0.9772
J 31.2s-62.5 1.953
1 62.5-125 3.906
5 125-250 7.813
6 .250-500 15.625
7 500- 1000 31.25

il
EXAMPLE 7"7
Code a positive-going signal with amplitude 30Vo of ,-ll
the maximum allowed as a PCM sample. .:
,rS!

:[,u] |

Solution The signal is positive, so the first bit is a one. On the nornalized i":il:llflL

voltage scale, the amplitude is 300 mV. A glance at Table 7.1 shows that the
signal is in segment 6. That means the next three bits are I l0 (6 in binary).
This segment starts at 250 mV and increases 15.625 mV per step. The signal
voltage is 50 mV above the lower limit, which translates into 50115.625 :
3.2 steps. This is less than halfway from step 3 to step 4, so it will be
quantized as step 3, making the last four bits 0011 (3 in binary). Therefore
.,i;.trLil
the code representing this sample is 1110001 1.

!'ryn I

In operation, many modern codecs achieve compression by first encoding

the signal using a 12-bit linear PCM code, then converting the 12-bit linear code
into an S-bit compressed code by discarding some of the bits. This is a simple "nffi
,,,il][
example of digital signal processing (DSP). Once an analog signal has been
drgitized., it can be manipulated in a great many ways simply by performing illll

arithmetic with the bits that make up each sample. In the case of the 12-to-8 bit 1

conyersion described here, some precision will be lost for large-amplitude sam- til

-
ples, but the data rate needed to transmit the information will be much less than :ll|
,NI
for 12-bit PCM. Since most of the samples in an audio signal have amplitudes
much less than the maximum, there is a gain in accuracy compared with 8-bit
linear PCM.
Briefly, the conversion.works as follows. The 12-bit PCM sample begirs
with a sign bit, which is retained. The other eleven bits describe the amplitude of
 seEtion 7-3 * Pulse-Eode Modulation ?97

n segment

Digital Analog
Serial Serial/Parallel D/A Converter (and Sample-and-Hold Low-pass Filter Output
Input Conversion Expansion if Circuit
Required)

ble 7.1. F-x3xr= ?-17 PCM decoding

(mV)
: sample, with the most signiflcant bit first. For low-level samples, the last few
:e
: :. and the sign bit may be the only non-zero bits. The segment number for the
,58 '-rii code can be determined by subtracting the number of leading zeros (not
s8 - :nting the sign bit) in the 12-bit code from 7. The next four bits after the ltrst
-t2 .r: counted as the level number within the segment. Any remaining bits are
- -: "'arded.

EXAMPLE 7.4 Convert the 12-bit sample 100110100100 into an S-bit

.-ompressed code.

Solutisn Copy the sign bit to the S-bit code. Next count the leading zeros
of I ) and subtract from 7 to get 5 (101 in binary). The first four bits of the S-bit
;ode are thus 1 101. Now copy the next four bits after the first 1 (not counting
ihe sign bit) to the 8-bit code. Thus the next four bits are 1010. Discard the
red
rest. The corresponding 8-bit code is 11011010.
he
l.

nal

)
The decoding process is the reverse of coding. It is illustrated in the biock
rirgram in Figure 7 .17 .The expansion process follows an algorithm analogous to
.rat used in the compressor. The low-pass filter at the output removes the high-
-:equency components in the PAM signal that exits from the digital-to-analog
,-rrnveftef.

iding
;rtde
rple Differential PCM
been Instead of coding the entire sample amplitude for each sample, it is possible to
iring ;ode and transmit only the difference between the amplitude of the cuffent sam-
: bit ple and that of the previous sample. Since successive samples often have simiiar
Jm- rrnplitudes, it should be possible to use fewer bits to encode the changes. The
:han most common (and most extreme) example of this process is delta modulation,
,-t- uhich is discussed in the next section.
Lt-
-LI]L

:ln
I trf
s
Explain the difference between normal and differential PCM. SEilE*f,}ru 7.3 SE€er*XfiW
*aJ**Ygg3:H
 Chapter 7 * Oigital Communications

(iii) Rapid
'-1 Variation
7.4 Detrta N{odulation
o
I In delta modulation. only one bit is transmitted per sample. That bit is a one if r:.
- current sample is more positive than the previous sample, zero if the current sa: -
ple is more negative. Since only a sm.rll amount of information about each samp..
is transmitted, delta modulation requires a much higher sampling rate than PCl,l
for equal quality of reproduction. Nyquist did not say that transmitting samples _'
Time
(a) Input sigrral
trvice the maximurr signal frequency would always give undistorted results. or. .
o
that it could, provided the samples were transrnitted accurately.
d
Figure 7.18 shows how delta moduiation generates emors. In region (i). r:.
signal is not varying at all; the transmitter can only send ones and zeros, ho,., ,
ever, so the output wavefbrm has a triangular shape, producing a noise sicr-*
0
called granular noise. On the other hand, the signal in region (iii) changes mc-,
(b)
rapidly than the system can follow, creating an error in the output called slop=
Transmitted digital signal
overload.
(iii) SIope
o
.d Overload

A
Adaptive Delta Modulation
Adaptive delta rnodulation. in which the step size varies according to previo-.
values, is more efllcient. Figure 7.19 shows how it works. After a number of ste;. ',,1

Time in the same direction. the step size increases. A well-designed adaptive delta mo.-
(c) Reconstructed output signal ulation scheme can transrnit voice at about half the bit rate of a PCM svstem. \\ 1: -
equivalent quality.
Figure 7.18 Delta modulation

SEITION 7.4 REVTEW Why is it necessary to use a greater sampling rate tbr deita modulation than for
OUE5TTTIN PCM?

Figure 7.t9 Adaptive delta

modulation Variation -.- " r

o
(i) Constant : (iii) Rapid
Amplitude r Variation
a
5

Time
(a) Input signal (b) Transmitted digrtal signal

(iii)
-a
a
(i) Granular
Slope
Overload
i Reduced
a Noise
E

Time .,,d
(c) Reconstructed output signal
 SeEtion 7-5 ,*r Line Eodes 299

7.5 Line Codes

,rne if
the So far, our discussion of digital signals has been in terms of zeros and ones. It is
r-nt sam- now time to convert these binary numbers into voltage or current levels on a line'
h sample This can be done in many ways. Probably the simplest line code is to use the
ran PCM presence of a current or voltage to represent one logic state, while its absence
Lmples at indicates the other logic state. For example, TTL voltage levels could be used.
:1ts. only Figure 7.20 shows a binary sequence sent using TTL levels. Positive logic, in
which a high level represents a logical one, was used for the example, but nega'
1 1i), the tive logic could be used just as well.
rs. how- This example requires a line with dc continuity, since the current, when pres-
e signal ent, always flows in the same direction. Codes of this sort are called unipolar
es more I{RZ codes. The term unipolar means that the voltage or current polarity is
:d slope always the same, and NRZ (non-retum-to-zero) means that there is no require-
ment for the signal level to return to zero at the end of each element. For example,
the message 111 requires the voltage or current to go to the high state and stay
there for three bit periods.
sometimes, the condition of dc continuity cannot be met (for instance, when
there are transformers or ac-coupled amplifiers on the line)' Codes that have zero
,revious average dc content have been developed for ac-coupled lines. In the long run, the
rl steps bipolar NRZ code shown in Figure 7.2t will accomplish this, provided the
ta mod- message contains equal numbers of ones and zeros. On the other hand, a long
n. with string of ones or zeros will result in a component at very low frequency, and for
some systems there is no guarantee that there will be equal numbefs of ones and
zeros.
Low-frequency ac components and dc components can be eliminated com-
pletely with bipolar RZ (return-to-zero) codes. Figure 7 .22 provides two exam-
ples. Figure 7.22(a) shows a system used in telephony; it is called AMI, for
altemate mark inversion. A binary zero is coded as 0 Y and binary ones are
recorded alternately by positive and negative voltages. This signal will have no
dc or low-frequency ac content, even with long strings of ones or zeros. On the
other hand, long strings of zeros must be avoided when synchronous communica-
tion takes place with this code. A string of zeros can cause the signal to disappear
and the timing to be lost. There is some elror detection built into this code: any
time two consecutive pulses with the same polarity are received, an error must
have occurred.

Logic 1 = 4V

Logic 0 = 0V
Time

Egerre F"ffiB Unipolar NRZ code using TTL 1evels

 3t]t] fhapter 7 @ Digital fommuniEdtians
'I i

Logic 1 = +V

h0
0

Logic 0 = -V nl!lx[o

l]iulllllu

Time tl
FiEure ?.E! Bipolar NRZ code

Figure '7 .22(b) shows the Manchester code, which is a type of biphase code
Every bit has a levei transition in the center of the bit period. For a one there i.
an upward transition; for a zero, a downward one. There is no dc or 1o\\ -
frequency energy regardless of the proportion of zeros and ones in the signa,
The Manchester code also provides strong timing information regardless of th;
pattem of ones and zeros. Its disadvantage is that it requires more bandwidr:,
than the AMI code.

ltfitr"ri][

sEfTtt N 7.5 REVIEW Which of the line codes discussed in this section have a dc component, and under Iillllfll

SUEsTIEN what circumstances does this.pose a problem?

uffnfiilq

$ltum

?.6 Tirae-DivisionS'Iultiplexing
There are two basic types of multiplexing. We looked at frequency-division mu1-
tiplexing (FDM) in the previous chapter. Time-division multiplexing (TDM) is
used mainly for digital communication. In TDM, each information signal r:
allowed to use all the available bandwidth, but only for part of the time. Frorr-
Hartley's Law (Equation 7 .l'), it can be seen that the amount of infbrmation trans-
mitted is proportional to both bandwidth and time. Therefore, at least in theory. i: fllru{

is equally possible to divide the bandwidth or the time among the users of a chan-
nel. Continuously varying signals, such as analog audio, are not well adapted tc

C) o
bo b!
cB

*o o

Time Time
(a) AMI code (b) Manchester code

Figure 7,?t RZ codes

 Sertion 7.5 s Time-Eivision Multiplexing 301

IDM because the signal is present at all times. On the other hand, sampled audio
b very suitable for TDM, as it is possible to transmit one sample from each of
xveral sources sequentially, then send the next sample from each source, and so
m- As already mentioned, sampling itself does not imply digital transmission, but
in practice sampling and digitizing usually go together.
Many signals can be sent on one channel by sending a sample from each
rignal in rotation. Time-division multiplexing requires that the total bit rate be
mrltiplied by the number of chamels multiplexed. This means that the bandwidth
requirement is also multiplied by the number of signals.

fDM in Telephony
IDM is used extensively in telephony. There are many different standards for
IDM. One commonly used arrangement is the DS-1 signal, which consists of
!4 PCM voice charmels, multiplexed using TDM. Each channel is sampled at
3 COde. tkllz, with 8bits persample. This gives abitrateof 8 k X 8: 64 kb/s foreach
here is roice channel.
r low- The DS-1 signal consists of frames, each of which contains the bits repre-
signal. $nting one sample from each of the 24 channels. One extra bit, called the fram-
of the ing bit, is added to each frame to help synchronize the transmitter and receiver.
lwidth Eachframe contains 24 x 8 + 1 : 193 bits.
The samples must be transmitted at the same rate as they were obtained in - 125 Microseconds
193 Bits
order for the signal to be reconstructed at the receiver without delay. This requires
the multiplexed signal to be sent at arate of 8000 frames per second. Thus the bit
*_
<:
er rate is 193 x 8000bls:1.544 Mb/s. Figure 7.23 contains an illustration of a
24PCM Framingbit
frame of a DS-l signal. When this signal is transmitted over twisted-pair line
----{ using the AMI line code just discussed, the whole system is known as a T1 carrier.
Samples
8 Bits Each
fhat is, the signal includes only the coding into ones and zeros, while the caruier
dso includes the voltage levels used. Figure 7.23 One tiame of a DS-1
signal
The framing bits are used to enable the receiver to determine which sample
mul- md which bit in that sample are being received at a given time. In addition, the
!1 ) l\
receiver must distinguish between frames in order to decode the signaling infor-
ll is mation that is sent along with the signal. In one frame out of every six, each of the
rtolt-l
hast significant bits in the 24 samples is used for signaling information rather
.JNS-
rhen as part of the PCM signal. This results in a very slight degradation of signal
rr. it quality-for instance, the signal-to-noise ratio is degraded by about two decibels.
han- The signaling information is different on the sixth and twelfth frames, known as
dto the A and B frames, in a sequence. In effect, the receiver is required to count
frames up to twelve. A group of twelve frames is called a superframe.
To allow the receiver to accomplish this, the framing bit alternates between
two sequences, 100011 and 011100. The underlined bits indicate the A and B
signaling frames, respectively.
The "stolen" signaling bits can be used to indicate such conditions as on-
hook and off-hook, ringing, and busy signals.

Digital Signal Hierarchy

The DS-1 signal and T-1 carrier described above represent the lowest level in a
hierarchy of TDM signals with higher bit rates. All of these signals contain PCM
audio signals, each sampled 8,000 times per second. As the number of multiplexed
voice signals increases, so does the bit rate. This requires the channel to have a
wider frequency response and variations of time delay with frequency to be held
 Ehapter 7 * Digital fomrnunications

Digital signal hierarchy , tXlt

it ditr
Voice Bit Rate
l

Carrier Signal Channels (MbA) Typical Medium

;.lilll

T1 DS-1 24 1.544 twisted pair l1' .r t ::'

T1C DS-1C 48 3.t52 twisted pair ... : ll

T2 DS-2 96 6.312 low-capacitan;. LlIl r [,iil

twisted patr. ri:l
microwave
ll ,ll l,",m

T3 DS.3 672 14.736 coax. micror.r a'. .

. itLl]. lll- lrilll,
T4 DS-4 4032 2'74.t76 coax, fiber opt,:-
T5 DS-5 8064 560.16 fiber optics
": :[
tL, , ., lLlll

'll llLhilm

to a low 1eve1. Twisted-pair lines can be specially conditioned for use as T1 --

T2 carriers, but higher data rates require channels with greater bandwidth, suc- - ,iLlr, llllllfil

coaxial cable, microwave radio, or optical fiber. See Table 1 .2 for more detai'. ;

r ilr'ti rLllllli

ll i ll irlll]

SEITIIIIU 7.6 *fEV$EtV Explain the difference between TDM and FDM. I i iiillL,

GUESTIAru tLrl liliilf,

,] , iiiltrl

1
T l:,, l\{l

? "
E c>{:,}* +,E :EEE* "L} *,ti€ i:-l +r:= E==
e"-,s= i t-; :.:
= I tLt: iLnLlillll

In a previous section we looked at companding, which we noticed consistec

compression at the transmitter and expansion at the receiver. E,ssentially this i. -- l,urllflrufl
analog technique for improving dynamic range by increasing the signal-to-n, . -,
ratio for low-level signals although, as we saw, it can be implemented using c-: I{l
tal signal processing. We now turn our attention to the bits that result from --, iriililtilililll

analog-to-digital conversionjust discussed and consider whether there is an\ .i. - ti:r ti lllut lllll

to reduce the number of bits that have to be transmitted per second. This reduc:- - llllliiln I
is also called compression, but it is really a completely different process from -:, , i1 liltllll

one just described. We shall call it data compression to emphasize the differei--. lllLllllllrlJll{

Generally, without data compression more bandwidth is required to tran:.- .L,lllrili$llli;

an analog signal in digital form. For instance, analog telephony requires less t: : rl uu ' il

4kHz per channel with single-sideband AM transmission. Digital telephonr ;. --

*ritfl]t[trtr

ventionally operates at 64 kb/s. The exact bandwidth requirement for this depe- -
on the modulation scheme but is likely to be much more than 4 kHz unless '--,. ill]jitl]0

channel has a very high signal-to-noise ratio and an elaborate modulation sche , rurlilrt

is used. To use digital techniques in wireless communication. it is very desir": , llfl' I

to reduce the bandwidth to no more, and preferably less, than that needed for ;.'--
,\ill0

log transmission. tiLlLlnll'

rr ilt

Lossy and Lossless Compression ilffin1iltr

There are two main categories of data compression. Lossless compression -' ufrllflltrI

volves transmitting all of the data in the original signal but using fewer bits. Lo., flr{
compression, on the other hand, allows for some reduction in the quality oi:-. rxilullu

transmitted signal. Obviously there has to be some limit on the loss in qualitr. ;.' tfillM

pending on the application. For instance, up until now the expectation of r.' -.
quality has been less for a mobile telephone than for a wireline telephone. L ilnutiu

expectation is now changing as wireless telephones become more comm,. - i Illllr

 seEtion 7-7 * Yoco,ders and Data Eompression

.:ople are no longer impressed with the fact that wireless telephony works at all;
:-\' want it to work as well as a fixed telephone.
Iedium Lossless compression schemes generally look for redundancies in the data.
: rr instance, a string of zeros can be replaced with a code that tells the receiver the
.:gth ofthe string. This technique is called run-length encoding. It is very use-
.ii '-i in some applications: facsimile (fax) transmission, for instance, where it is un-
:iance
-:--essary to transmit as much data for white space on the paper as for the message.
!air.
J\C In voice transmission it is possible to greatly reduce the bit rate, or even stop
Ii\\ i\.e .nstritting altogether, during time periods in which there is no speech. For ex-
- L)prics
-.rp1e, during a typical conversation each person generally talks for less than half
-: time. Taking advantage of this to increase the bandwidth for transmission in
:.i1 time requires there to be more than one signal multiplexed. When the tech-
-.que is applied to a radio system, it also allows battery-powered transmitters to
- .,nserve power by shutting off or reducing power during pauses in speech.
T1 and
Lossy compression can involve reducing the number of bits per sample or re-
:uch aS
tails.
--icing the sampling rate. As we have seen, the lirst reduces the signal-to-noise
:,io and the second limits the high-frequency response of the signal, so there are
nits to both methods. Other lossy compression methods rely on knowledge of
.,e type of signal, and o1Ien, on knowledge of human perception. This means that
rice, music, and video signals would have to be treated differently. These more
---{ .jranced methods often involve the need for quite extensive digital signal pro-
-:ssing. Because of this, they have only recently become practical for real-time
-:e with portable equipment. A couple of brief examples will show the sort of
lng that is possible.
!ted of
.. is an
\-ocoders
'-nol se
g digi- 1 \'ocoder (voice coder)is an example of lossy compression applied to human
)ir the :leech. A typical vocoder tries to reduce the amount of data that needs to be
\ \\'a] .-rnsmitted by constructing a model for the human vocal system. Human sounds
u,-rion ,:e produced by emitting air from the lungs at an adjustable rate. For voiced
rrn the :,-rurdS this air causes the vocal cords to vibrate at an adjustable frequency; for
-ni oiced sounds the air passes the vocal cords without vibrating them. In either
nsmit -rse, the sound passes through the larynx and mouth, which act as f,lters, chang-
s than rg the frequency response ofthe system at frequent intervals. Typically there are
.'rln - .:om three to six resonant peaks in the frequency response of the vocai tract.
rends Vocoders can imitate the human voice with an electronic system. Modern
!S the '.
ocoders start with the vocal-tract model above. There is an excitation function,
heme -ol1owed by a multi-pole bandpass filter. Parameters for the excitation and the fll-
;.rble :-r response must be transmitted at intervals of about 20 ms, depending on the
'ina- )\'stem. Vocoders of this type are known as lineor preclictive coders because of
.he mathematical process used to generate the lilter parameters from an analysis
rl the voice signal.
The first step in transmitting a signal using a vocoder is to digitize it in the
usual way, using PCM, generally at 64 kb/s. Then the signal is analyzed and the
1tn- necessary excitation and fllter parameters extracted. Only these parameters need
, :s]
ro be sent to the receiver where the signal is reconstructed. The transmitted data
i rhe
rate is typically in the range of about 2.4 to 9.6 kb/s, allowing a much smaller
. rl--
transmission bandwidth than would be required for the original 64 kb/s rate.
,-,i i-
There are two main ways of generating the excitation signal in a linear pre-
Tirs
dictive vocoder. In pulse excited linear predictive (PELP or sometimes RPEL|
l'tn ior regular pttlse excited linear preclicrlue) vocoders, a white noise generator is
ffi S{FE rI
5
Select the created model by using the File-> Open men.'
To demonstrate the operation of option to select and open the file "codec.mdl".
-t
both a mu-law coder and decoder Run the simulation by selecting the Simulation->
in SIMULINK. Start menu option in the model window.
& f+, :illlfl
lfl *Pi Background:
. Double-click on the scope to view the input,
compressed and expanded signals.
In this section, SIMULINK is used to illustrate the non-
'.x
: .',e r- linear sampling that occurs when a mu-law coder is Experiment with other input wave types and frequencre- .. -.u]lll
used, as well as the corresponding inverse process re- from the function generator.
f,r'-u[
d,'+;. quired to restore a mu-law compressed signal. *'lrlffi,

Implementation:
Stafi MATLAB and then type simulink at the command tiLtlnr.im

prompt to start the SIMULINK program. ''uiiiuu{

nwiili
>> simufink ififl,Il
,r1

",fl|!

;: i
iJU.L&.II1J, i]EFEE }ECBDEE SIMUL4TItrH

- t, t: t,uiluf,

ltL, r l
:.,1 * lil

l' r "l

rllfi

llilIll
m u-lJr{ rru- lttq
e;;p:nder ,iL r lll

tit ,ililMn

fr

ul ,llfl

used lbr unvoiced sounds, and a variable-frequenc-v pulse generator produce: :

voiced sounds. The pLrlse generator creates a tone rich in harmonics, &S i: -
,, i,,,L.

r, r'}rll0
sound produced by human vocal cords. Both sources have variable amplituc.
it, i,iiLill0
Figure 7.22i illustrates the process at the receiver.
Resicluol excited linear pretlictitte (RELP) vocoders, on the other hand, ap:
the inverse of the filter that will be used at the receiver to the voice sisnal. T

Figure 7.E4 PELP vocoder Electronic Switch

[_]---_]
Pulse Generator '
, ___+
1_____, _ _i Output
-Audio
1 Filter
L-[--
Noise Genelator i*dl

3D4
 n idea of the effectiveness of data compression can be gained by com-
ln nlenu paring conventional compact-disc audio with the MP3 format popular
on the internet. CDs use sampling a\ 44.1 kHz to allow a maximum informa- n.."iF&,&F tr*Sffie&W
ation-> tion-signal frequency of 20kHz. The coding is linear PCM using 16 bits per
sample. Since there are two stereo channels, the raur data rate, before error fBs and MPSs
correction, is

fa:2x 44.1 x 103 x 16: 1.41 Mb/s

lencles
Compact discs tend to have quite a few errors due to the manufacturing
p!'ocess, so a robust error correction scheme is used which fut'ther increases
the bit rate.
By comparison, audio formats commonly used on the lnternet, such as MP3
and RealAudio, achieve good results with much lower data rates. Many people
have difficulty distinguishing MP3 at 128 kbls from CD audio. This result is
achieved by removing redundant data from the '1 .41 Mb/s data stream de-
scribed earlier, and by ignoring any components that would be inaudible to a
human ear. For instance, very quiet sounds can be masked by louder sounds.
Vocoders wouid not be usable with these formats because they are required
to reproduce all kinds of sounds and not just the human voice.

utput of this hlter is a signal that. when applied to the receiver fiiter. will repro-
-uce the original signal exactly. Figure 7.25 shows how this process works at the
:lnsmitter. The residual signal is too complex to transmil eractly with the avail-
,,ble bit rate, so it must be represented in a more economical way. One method is
.,-r compare it with values in a table, called a cotlebook, and transmit the number
,i the closest codebook entry. The receiver looks up the codebook entry, gener-
Lies the comesponding signal. and uses lt instead of the pulse and noise generators
.hown in Figure 3.13. Many other vocoder variations are possible as well.
Reasonable quality can be achieved with vocodei's using data rates much
..ruer than those required for PCM. So far. the quality is not quite as good as tbr
: Iraightforward PCM. howet,er.

It should be obvious that vocoders are intended lbr use with voice only';
..r'hereas. the PCM system described above can be used to send any 6,1 kbis
i) the
:: the Jata stream. including music. fax, or computer liles. None of these v;ili work
udes. ;roperly with a vocoder. Vocoders even tend to give a somewhat unnatural
.luality to human speech. Still. the gain in bit rate and hcnce bandwidth, com-
.ppl]' rared to PCM, is so great that vocoders are very conrmon in di-eitai wireless
. The voice communication.

Expiain the dift'erence betrveen data compression and the use of vocoders for 5ECTION 7.7 REVIEW
voice signals. QUESTTON

Inverse
\roice Signal
Filter Residual Signal Excitation Signal

Figure 7,P5 Generation of excitation signal using codebook

3t]5
3t]5 Ehapter 7 # Digital fommunications

I illlilLrLll;;t il
. 1"'

sL!s€ruc&ffiY Here are the main points to remember from this chapter. iLlllilut.;lJ t

:
1. Modern communication systems are often a mixture of analog and dr-i''-
Iiii...,.,,.
sources and transmission techniques. The trend is toward digital systems.
illlllur,. rm
2. Modern digital systems have better performance and use less bandwidth :-"-: ..1 1

equivalent analog systems. '.1

3. An analog signal that is to be transmitted digitally must be sampled at 1; ,., ,,
iltilillllflr r r11[

twice per cycle of its highest-frequency component. Failure to do so cre"": r,1iil]ll][lnmi@

undesirable aliasing. ltil

4. PCM requires that the ampiitude of each sample of a signal be converted .- - Lltuililr ril(
binary number. The more bits used for the number, the greater the accur.-
but the greater the bit rate required. ilululII]:llllllu&l

5. Delta modulation transmits only one bit per sample, indicating whether --:. ,ll
signal 1evel is increasing or decreasing, but it needs a higher sampling ::.'
than PCM for equivalent results. ilfi ilLllllllllll Ilulfl /l

6. The signal-to-noise ratio for either PCM or delta modulation signals can o-..: rlI

,,fru".
be improved by using companding. im

7. Lossless compression eliminates redundant data bits, thereby reducing the :

rate with no effect on signal quality.
l|lilrfilrlil|imTn{

ill

8. Lossy compression compromises signal quality in order to reduce the bit r.:.
lllllUl llll I ilxl{l'n{
For voice transmissions, vocoders are often used to achieve great reduction. .

bit rate.
Illlililllli{11i11"4

r
gsLiAs!ffiru I-r5Y
r .Liilflll0

.titi[|[l

I-ktB (7.r) f,,:f,-f, (7.5 ILllillllrlllllllii&

C : 28 logzM (7.2) N :2,,, (7.e r lllI

C:B1og2(1 +S/N) (7.3) DR : 1.16 + 6.02m dB (7.1ti

Eh . E6 sin rr/T D:f,m (7.|t

v(n: -i sin or,t + ' sin(co, - co7,)1
T"trt (7.7) V" 1n(1 * ;ru;/Vi) (7.1:
E6 sin rrlT yo
ln(l + p)
'l t l
l

ffi**ffiffi&&mv !:+,:5:fl

aliasing Distortion created by using too low a sampling rate companding Combination of compression at the transmu..-
when coding an analog signal for digital transmission. and expansion at the receiver of a communications system.
bipolar code A data code that uses both polarities of voltage or compression Amplification of a signal in such a way that the:.
current. is less gain for higher-level input signals than lbr lower-ler.
codec (coder-decoder) Device that converts sampled analog input signals.
signal to and from its PCM or delta modulation equivalent. decoding Conversion of a PCM or delta modulation bitstrea-:
coding Conversion of a sampled analog signal into a PCM or 1o analog samples.

delta modulation bitstream.

 Ouestions

=. ,:==.,.5-iK
:.:a modulation A coding scheme that records the change in pulse-position modulation (PPM) A series of pulses, in
- .1 leve1 since the previous sample. which the timing of each pulse represents the amplitude of the
.ital signal processing (DSP) Filtering of signals by con- information signal at a given time.
rd digitat : --rs them to digital form, performing arithmetic operations on pulse-width modulation (PWM) See pulse-duration modu-
ems. ' :;ta bits, then converting back to analog form. Iation (PDM).
-=:-topped sampling Sampling of an analog signal using a quantizing Representation of a continuously varying quantity
idth than
.:1e-and-ho1d circuit, such that the sample has the same am- as one of a number of discrete values.
-ie for its whole duration. quantizing errors Inaccuracies caused by the representation of
J at least
.,lor.er distortion See aliasing. a continuously varying quantity as one of a number of discrete
c creates values.
-rming bitsbits added to a digital signal to help the receiver
-:rect the beginning and end of data frames. quantizing noise See quantizing errors.
'ned to a
:. code A system for translating logic ones and zeros into regenerative repeater A device that decodes and recodes a
.!-curacy,
rqe or curent levels for transmission. digital signal as well as amplifying it.
'.:ural sampling Sampling of an analog signal, so that the run-length encoding Method of data compression by encoding
:iher the --..:1e amplitude follows that of the original signal for the dura- the length of a string of ones or zeros instead of transmitting all
ing rate -. .-,f the sample. the one or zero bits individually.
-
=:ative logic A logic system in which a low level represents RZ (return-to-zero) code A line code in which the voltage or
dn often : .' one and a high level represents logic zero. current retums to zeto at the end of each bit period.
''. RZ (non-return-to-zero) code A data line code in which the sample-and-hold circuit A device that detects the amplitude
: the bit . -lEe or curent does not necessarily return to zero between bits. of an input signal at a particular time called the sampling time
,.,sitive logic A logic system in which a high level represents and maintains its output at or near that amplitude until the next
one and a low level represents logic zero. sampling time.
bit rate. --,-
: -ilse-amplitude modulation (PAM) A series of pulses in slope overload In delta modulation, an error condition that
tions in
--rh the amplitude of each pulse represents the amplitude of the occurs when the analog signal to be digitized varies too quickly
- ,nnation signal at a given time. for the system to follow.

- ulse-code modulation (PCM) A series of pulses, in which unipolar code A line code in which the polarity of the voltage
: lmplitude of the intbrmation signal at a given time is coded as or the direction of the current remains the same at a1l times.
:-nary number. vocoder Circuit for digitizing voice at a low data rate by using
knowledge ofthe way in which voice sounds are produced.
,^ :ulse-duration modulation (PDNI)
-::h the duration of
A series of pulses, in
each pulse represents the amplitude of the
o.e) : rmation signal at a given time.

O.to) .' :q. i"

Q.1t)
Give four advantages and one disadvantage of using digitai 6. (a) Briefly explain what is meant by companding.
rrather than analog) techniques for the transmission ofvoice (b) What advantage does companded PCM have over linear
A.tz1 telephone signals. PCM for telephone communications'l
-. What factors limit the theoretical maximum information rate 7. Explain why the sample rate must be greater for delta modu-
on a channel? lation than for PCM.
".. What happens when a signal is sampled at less than the 8. What are the two functions of a codec? Where in a tele-
Nvquist rate? phone system is it usually located?
- (a) List three types of anaiog pulse modulation. 9. Explain briefly how mu-lau,'compression is implemented in
mitter lb) Which pulse modulation scheme is used as an interme- a typical codec.
diate step in the creation of PCM? 10. What advantages does the AMI bipolar code have over
there (c) Which pulse modulation scheme also flnds use in audio unipolar NRZ coding?
-ier el amplifiers and motor speed-control systems? 1 1. Compare the bipolar NRZ and Manchester codes:
5. For a PCM signal, describe the effects of: (a) Which of these codes is more suitable for synchronous
.ream
(a) increasing the samplin_e rate communication? Why?
(b) increasing the number of bits per sample (b) Which of these codes has more energy at 1ow
frequencies? Illustrate your answer with a bit pattem
3trEl fhapter 7 S lligital EommuniEations

that contains a lot of low-frequency energy with 14. How is signaling information incorporated into a DS-1
this code. signal?
(c) Which of these codes requires the greater bandwidth? 15 How is framing information incorporated into a DS-1
Illustrate your answef with a bit pattern that requires the signai?
greatest bandwidth with this code. 16. Explain the difference between lossy and lossless compr--
t2 How many voice signals are combined into a DS-1 signal? sion.
How is this done?
t3 Distinguish between a DS-1 signal and a T1 carrier.

q
pseffiffiLffi&*s
ffi
Section 7.1 and coded using linear PCM with sixteen bits per
17. A broadcast television channel has a bandwidth of 6 MHz. sample. Find:

(a) Calculate the maximum data rate that could be

(a) the maximum audio frequency that can be recorded
(assuming ideal filters)
carried in a TV channel using a 16-1evel code. *iil
Ignore noise. (b) the maximum dynamic range in decibels
(b) What would be the minimum permissible signal-to- (c) the bit rate, ignoring error coffection and framing bit
noise ratio, in dB, for the data rate calculated above? (d) the number ofquantizing levels t5
18. How many levels would be required to communicate at 9600 23. Suppose an input signal to a pJaw compressor has a
bits per second over a channel with a bandwidth of 4 kHz positive voltage and an amplitude 25Vo of the maximum 5
and a signal-to-noise ratio of 40 dB? possible. Calculate the output voltage as a percentage of 1fl,T1]'

the maximum output. iillillulll

Section 7.? 24. A PCM/TDM system multiplexes 24 voice band i,rl:iluM

19. A 1 kHz sine wave with a peak value of I volt and no dc channels. Each sample is encoded into seven bits, and a ,uuNl

offset is sampled every 250 microseconds. Assume the first frame consists of one sample from each channel. A framir: i I rttillllllll

sample is taken as the voltage crosses zero in the upward bit is added to each frame. The sampling rate is 9000 sam-
direction. Sketch the results over 1 ms using: ples per second. What is the line speed in bits per second.'
!5r
(a) PAM with all pulses in the positive direction l[ ]il
25. Suppose composite video signal with a baseband
a
i I Ltlllilllllx
(b) PDM frequency range fiom dc to 4 MHz is transmitted by
rL]:rffiiliillll

(c) PPM linear PCM, using eight bits per sample and a sampling
tiruUflS
rate of 10 MHz.
20. It is necessary to transmit the human voice using a ilrlnllllllfl

frequency range from 300 Hz to 3.5 kHz using a digital

(a) How many quantization levels are there?
system. (b) Calculate the bit rate, ignoring overhead. 5t
iurffiiltr
(a) What is the minimum required sampling rate, according (c) What would be the maximum signal-to-noise ratio, ir .[' x0
to theory? decibels?

O) Why would a practical system need a higher rate than (d) What type of noise determines the answer to part (b): rM
the one you calculated in part (a)?
Section 7-5 .T
21. T\e human voice actually has a spectrum that extends to
26. Code the following data using each of the codes
much higher frequencies than are necessary for communica- ill]fiil[
specified. The data is to be sent in time as it reads, from
tion. Suppose a frequency of 5 kHz was present in a sampler 'Iltm,!
left to right.
that sampled at 8 kHz.
01 10001 11
(a) What would happen?
(b) How can the problem (a) unipolar NRZ
described in part (a) be
prevented? (b) AMr
22. The compact disc system of digital audio uses two (c) Manchester code
channels with TDM. Each channel is sampled at 44.lkIIz
 Answers to Review Ouestions 3Og

a DS-1 Figure 7.26 shows a signal in AMI code. Decode it into ones 29. T\e basic European TDM telephone signal combines 30
and zeros. voice channels with 2 signaling channels that have the same
DS-1 datarute as the voice channels. The sampling rate is 8 kIIz
and there are 8 bits per sample for each voice channel.
Sampling Times Calculate the total bit rate for this signal.
s compres_

Section 7.7
o 30. A typical PCS system using a vocoder operates at
h0
6 9600 b/s. By what factor has the amount of data
required been reduced, compared with standard digital te-
lephony?

Sgstems
3 1. A certain cable television system has a usable bandwidth of
300 MHz. How many television signals could be transmitted
Figure ?.EE
in this bandwidth
liection 7.5 FDM and standard NTSC television channels,
(a) using

18. (a) Calculate the bandwidth required to transmit a DS-1 6 MHz wide

rrded signal using AMI code. (b) using MPEG-2 with 3 Mb/s required per channel, time-
(b) Suppose Manchester code were used instead. division multiplexed and transmitted using an ordinary
unipolar NRZ code
What would be the required bandwidth.
ng bits

4ru51ry8ffiS T'tr ffiEt/5EW ffilJES?'Ed3ruS

lUm 5ection 7.1 deal of difference between sample amplitudes. Therefore, the
eof number of samples must be higher than for PCM, in which
rere are many applications of this theorem. For instance, a
. .rnd or video signal whose bandwidth exceeds the channel several bits are used to encode the complete amplitude informa-
tion for each sample.
, -r.icity can be transmitted in a greater time than normal. This is
da :,t al1owab1e where real-time communication is required. as in a
raming
liection 7.5
.-ephone conversation.
sam- The unipolar NRZ code has a dc component and may not pass
ond? 5ection 7.2 through transformers and many amplifiers.
- :e fact that a signal must always be sampled befole being liection 7.6
--,rrtized, and that the sample rate must be at least twice the In FDM each information signal is assigned part of the available
tg :-.lrimum baseband frequency, places a stnct upper limit on the
- -r:eband bandwidth for digital communication that is not
bandwidth on a full-time basis; in TDM the whole bandwidth is
given to each information signal for part of the time.
. ',r'avs present in analog systems.
lSectlon 7-7
Section 7.3
Data compression acts to reduce redundant data in already-
r. 1l1 \rrrmol PCM requires that the con-rplete amplitude information
coded signals. Vocoders rely on knowledge of the characteristics
:: encoded for each sample. In ditl'erential PCM, only the differ-
of the human voice to avoid the necessity of conventional
b,t ::ce between the arnplitudes of the previous and current samples
sampling at or above the Nyquist rate.
.. encoded.

5ection 7.4
Since only one bit of information is transmitted fol each sample,
"le signal cannot be represented accurately if there is a great