WAVEFORM CODING TECHNIQUES

As shown in figure 3.1(c), amplitudes of the signal x(t)

lie in the range (- DO YOU KNOW?
m,, m,) which is partitioned into L
Modern communication systems
intervals, each of magnitude Av =
Now, each sample is are often a mixture of analog and
digital sources and transmission
approximated rounded off to the nearest
or
quantized level1 techniques. The trend is toward
as shown in figure. Since each sample is now approximated
to one of the Lnumbers therefore the information is
digital systems.
digitized.
Important Point: The quantized signal is an approximation of the original one. We can
improve the accuracy of the quantized signal to any desired degree simply by increasing the
number of levels L.

3.4. PULSE CODE MODULATION (PCM)*

1. Definition
Pulse-code modulation is known as a digital pulse modulation technique. In fact, the pulse-code
modulation (PCM) is quite complex compared to the analog pulse modulation techniques ie, PAM,
PWM and PPM) in the sense that the message signal is subjected to a great number of operations.

2. Elements of a PCM system

3.2 shows the basic elements of PCM system. It consists of three main parts i.e.,
a
Figure
The essential operations in the transmitter of a
transmitter, transmission path and receiver. discussed earlier,
a s shown in figure 3.2. As
PCM system a r e sampling, quantizing and encoding is according
analog (i.e., continuous-time) signal sampled
sampling is the operation in which a ndiscrete-time
in a signal. The quantizing and encoding operations
to the sampling theorem resulting
in the s a m e circuit which is
known as a n analog-to-digital converter (ADC).
are usually performed of impaired signals, decoding
in the receiver a r e regeneration
Also, the essential operations a r e usually performed
in
demodulation of the train of quantized samples. These operations
and converter (DAC).
known a s a digital-to-analog
the s a m e circuit which is
Analog PCM
message Low Pass Quantizer Encoder
Sampler signal
O Filter
signal
(a)

Distorted Regene Regenerated

Regene- rative PCM signal
PCM
rative repeater
signal repeater
(b)

Recons- Destina-
Regene truction
Input Decoder tion
ration filter
O-
circuit

(c) Receiver
Transmission path (c)
Transmitter (b)
of a PCM system (a)
elements
Fig. 3.2. The basic

obtained from
a the class of baseband signals
 Purther, at intermediate points, along the transmission route from the
the transr
receiver, regenerative repeaters are used to reconstruct (i.e., regenerate) the transmitte transmitter to the
lntigure
diserete
in
of coded in order to combat the accumulated effects equence
pulses signal distortion of noise As
discussed in article 3.3, the quantization refers to the use of a finite set of amplitud 11se. andd eguence is
ixred
dis
evels.l tt
the selection of a level nearest to a partieular sample value of the message signal as therepreers and
orerror 1
for it. In fact, this operation combined with sampling, permits the use of coded pulsesfor renroation
calledx(n
,
the message signal. Thus, it is the combined use of quantizing and coding that
pulse code modulation from analog modulation techniques. ishes convertsit
encoderis
3. Few Important Points
Now, let us summarize PCM in the form of few points as under: a Imp
)PCM is a type of pulse modulation like PAM, PWM or PPM but there is an imnortand binary
difference between them PAM, PWM or PPM are analog pulse modulation systems serialh
PCM is a digital pulse modulation system.
whereas
a shift
(i) This means that the PCM output is in the coded digital form. It is in the form of of bina
pulses of constant amplitude, width and position.
digital
(iii) The information is transmitted in the form of code words. A PCM system consists of a Alsc
PCM encoder (transmitter) and a PCM decoder (receiver). Convert
and enc
(iv) The essential operations in the PCM transmitter are sampling, quantizing and
encoding.
(v) All the operations are usually performed in the same circuit called as analog-to-digital 3.6. P
converter.
(vi) It should be understood that the PCM is not modulation in the conventional sense. The pa
(vii) Because in modulation, one of the characteristics of the carrier is
varied in proportion Over
with the amplitude of the modulating signal. Nothing of that sort happen in PCM. trans
most
3.5. A PCM GENERATOR OR TRANSMITTER Contr
In the last article, Wave
we had
overview of the elements of a PCM system (i.e.,
an
by
transmission-path and receiver). In this section, we shall discuss the PCM generator (i.e.,transmitUer
from a practical point of view. transmiutet n f
Figure 3.3 shows a practical block diagram of a PCM genera each
In PCM generator of
figure 3.3, the signal x(t) is first passed through the low peri
cut-off frequency fm Hz. This -pass
low-pass filter blocks all the frequency components
Wnic and
lying above fm Hz.* ro
Digitally encoded
Bandlimiting
Low pass signal
x(t)| filter with A Sample A Binary
Continuous cut-off freq
and hold nTs q-level
g-level X(nT) encoder
time (S/H) circuit PAM quantizer Quantized (digitizer)
messagee signal PAM
signal
v digits
Parallel PCM
Timer to serial Outpu
22m converter|r=uf,
Fig. 3.3. A practical PCM
This means that
generator h o l d
c i r c u i t
now the signal x(t) is bandlimited to f Hz. The
then samples this s1gnal at the rate of sample
f. Sampling frequency f. is selected iently
ciently abov
aD
nyquist rate to aviod aliasing i.e. suu
 WAVEFORM CODING
f2f x(nT). This signal x(nT)1s
and hold circuit is denoted by
the output of sample
with
3.3, input x(nT,)
In figure A q-level quantizer compares
time a n d c o n t i n u o u s in amplitude. with its fixed digital
discrete in level to x(nT.)
levels. It assigns any o n e of the digital minimum distort1on
its fixed digital one of digital
the level to x(nT.) which results
in
a digital level
levels. It then ass1gns any of quantizer is
e r r o r . Thus, output
e r r o r is called quantization
encoder
encoder. This
error. This to binary
Now, the quantized signal level x.(nT) is given 'v' binary bits.
This
called x,(nT). is converted to
to 'v digits binary word. Thus x.(nT.)
converts input signal
also k n o w n a s digitizer.
encoder is
each bit of the
possible to transmit
It may be noted that it is not converted to
a Important Point: 'v' binary digits are
transmission line. Therefore usually
separately on to serial converter,
binary word baseband signal. In a parallel
to generate single baseband signal
serial bit stream is thus a single
this The output of PCM generator
job.
a shift register does
of binary bits. to serial
hold circuit and parallel
clocks for sample and and hold, quantizer
oscillator generates the
Also, a n discussed above, sample
code m o d u l a t i o n generator
converter. In the pulse converter (ADC).
form a n analog to digital
and encoder combinely
TRANSMISSION PATH
3.6. PCM
receiver DO YOU KNOW?
t r a n s m i t t e r and PCM
the PCM
The path between PCM
travel, is called a s that the amplitude
which the PCM signal The PCM requires
over
shown in figure 3.4.
path and it is
as
of a signal be
transmission
lies in its ability to of each sample
feature of PCM system binary
number.
most important when the PCM converted to a
distortion and noise bits used for
the
control the effects of this capacity The more
on the
channel. PCM accomplishes the accuracy,
w a v e travels as shown number, the greater
a chain of
regenerative repeaters the bit rate
by m e a n s of using a r e spaced
close enough to but the greater
Such repeaters
in figure 3.4. The r e g e n e r a t i v e required.
t r a n s m i s s i o n path.
each other on the equalization, timing POCM signal
free
operations namely the clean noise
performs three basic reproduces of PCM
each repeater actually the performance
and decision making. Hence, channel noise.
This improves
distorted by the
from the PCM signal
in presence of noise.
Regenerative Regenerated
Regenerative repeater PCM wave
Distorted. repeater
PCM
LM o
o L
PCM+Noise
oClean PCM
P M+Noise
Clean PCM
transmission path
Fig. 3.4. PCM
3.6.1. Block Diagram of
a Repeater
repeater.
of a regenerative the effects
the block diagram compensate for
Figure 3.5 shows d i s t o r t e d PCM
w a v e so a s to
which is
shapes the pulse train
The amplitude equalizer circuit produces a periodic
d i s t o r t i o n s . The timing the decision making
is then applied to
OT amplitude and phase PCM pulses.
This pulse train
equalized PCM pulses.
derived from the input train for sampling the
u s e s this pulse maximum. The
making device to noise ratio is
evice. The decision out at the
instants where the signal has a 0
carried PCM w a v e at its input
he sampling is decision about
whether the equalized
decision device makes a
value or 1 value at the instant
of sampling. Such a Amplitude Decision
decision 1s PCM wave making
made by comparing equalized
equalizer
device Regenerate
PCM wave
PCM with a reference level
called decision threshold as LM Timing
circuit
illustrated in figure 3.6. At the
output of the decision device, Fig. 3.5. Block diagram of a
regenerative repeater
we get a clean PCM signal without any trace of noise.*
3.7. PCM RECEIVER NMAAA w
Distorted
Decision
In this section, we shall discuss a PCM PCM wave
.bOundary
receiver from practical point of view. ada orthreshold
Figure 3.7(a) shows the block diagram Sampling
of PCM receiver and figure 3.7 (b) instants
LHH
decision
shows the reconstructed signal. The
regenerator at the start of PCM Output
receiver reshapes the pulse and Hegenerated
removes the noise. This signal is then PCM)
converted to parallel digital words for Fig. 3.6. Waveforms of a regenerative repeater
each sample.
v digits
Incoming Clean
PCM+Noise A A serial
PCM AA digital to A sample Low pass
Signal
Regene- to parallel:analog (D/A)
converter
and hold filterwith
rator converter (S/H) circuit
Sync Timer
(a)
5/q x(t)
3/q
x(KT)
1/q
KTs
(b)
Fig. 3.7. (a) PCM receiver (b) Reconstructed waveform
Now, the digital word is converted to its analog value denoted as x.(t) with the help of a s
hold circuit. This signal, at the output of
masS
sample and hold circuit, is allowed to paSs throg
reconstruction filter to get the appropriate original message
signal denoted as x(). ssible to
Important Note: As shown in reconstructed signal of figure 3.7 (b), it is 1u i n t r o d u c e d
reconstruct exact original signal x(t) because of
permanent quantization error dnced
rror
educed by
during quantization at the transmitter. In fact, this quantization error can De
 WAVEFORM CODING TECHNIQUES
95
increasing the binary levels. This is equivalent to increasing
But increasing bits 'v increases the binary digits (bits) per sample.
signalling rate
have observed in last article. Therefore the choice of
as well as transmission bandwidth as we
these parameters is made, in such a
manner that noise due to
quantization error (i.e., also called asquantization noise) is in
tolerable limits.
3.8. QUANTIZER (WBUT, Kolkata, Sem. Exam., 2004-05) (08 marks)
As discussed earlier, a q-level quantizer compares the discrete-time
input x(nT) with its fixed
digital levels. It assigns any one of the digital level to x(n'T) with its fixed digital levels. It then
assigns any one of the digital level to x(nT) which results in minimum distortion or error. This
error is called quantization error. Thus, the output of a quantizer is a
digital level called x,(nT,.
3.8.1. Ciassification of Quantization Processs
Basically, quantization process may be classified as follows
Quantization
Uniform Quantization Nonuniform Quantization
Midtread type Midrise type
The quantization process can be classified into two types as under:
(i) Uniform quantization
11) Non-uniform quantization.
This classification is based on the step size as defined earlier.
(6) Uniform Quantizer
the 'step size' remains throughout
A uniform quantizer is that type of quantizer in which
same
the input range.
(ii) Nonuniform Quantizer
varies according
of quantizer in which the 'step-size'
A non-uniform quantizer is that type
to the input signal values.
3.8.2. A Uniform Quantizer
remains
called a s a n uniform quantizer if the step size
As discussed quantizer is
earlier, a
constant throughout the input range.
Types of Uniform Quantizer
3.8.2.1.
a s under:
quantizer
There are two types of uniform
of the midtread type
(1 Symmetric quantizer
of the midrise type the
11) Symmetric quantizer uniform o r nonuniform type.
In a uniform quantizer,
can be of a is nonuniform. Now,
let us
Basically, quantizers otherwise, the quantizer
uniformly spaced; on. The quantizer
epresentation levels a r e
nonuniform quantizer
shall be considered later characteristic
quantizers, input-output
Consider only uniform midtread o r midrise type.
Figure 3.8(a) shows the lies in the
because the origin
Characteristic c a n also be type, which is so called input-output
of the midtread the corresponding
Oa uniform quantizer graph. Figure 3.8(b) shows lies in the
middle
in which the origin
staircaselike
tread of the
ddle of a of the midrise type,
laracteristic ofa uniform quantizer
of a rising part of the staircaselike graph. It may be noted that both the midtread and midriso
symmetric about the origin.*
ypes of uniform quantizers illustrated in figure 3.8 are
Output level
Output level
4 1 2 3
Input
4 level -3 -2 10 1 2 3 Input
4
level
. -1
-2
-3
-4
(a) (b)
Fig. 3.8. Two types of Uniform quantization: (a) Midtread, and (b) Midrise
3.9. WORKING PRINCIPLE OF QUANTIZER
In this section, let us see how uniform
consider uniform quantizer of midrise type.
quantization takes place. For this purpose, we shall1
uniform
Figure 3.9(a) shows the transfer characteristics of a
quantizer of midrise type. In figure 3.9(a), let us assume that
x(nT) varies from 4 A to + 4 A. This means that the peak to peak valuethe input to the quantizer
of x(nT) will be between
-
4A to44. Here, 'A' is the step si1ze.
+
Thus, input x(n'T) can take any value between 4 A
-
to + 4 A. Now, the fixed digital levels
are available E
at
AA and tA. These levels are available at quantizer because ot
its characteristics.
Hence, according to figure 3.9(a), we have
If
x(nT) =
4 A, then xq (nT) A =
and if
x(nT) =
-4 , then
x, (nT) =-5A
Thus, it may be observed from
figure 3.9 (b) that
maximum
From above, we conclude
that
quantization error would De 2
quantization error may be
here 'e exn) -x(nT) expressed as
represents the quantization error
..3.1)
*To increase the
be increased. To
resolution of a digitizing system
increase the number requires
o f q u a n t i z a t i o nl e v e l s
of that the number
digits representing each
voltage level bequantization levels requires that the
e r o fb i n a r y
increased. nu
.
 WAVEFORM CODING TECHNIQUES
Now, when x(nT,)= 0, quantizer will assign
either A/2 or A/2. If A2 is assigned, then
-
any one of the nearest binary levels i.e.,
quantization error will be,
Quantizer output x(nT)
Transfer characteristics
passing through zero
7A/2
Max. -ve input 5/2
3A/2
Max. + Ve input
4A -3A -24 -A A/2,
Quantizer
-/2A 2A 3A 4A input x(nT
-3/2
-5/2
-7N2
Peak to Peak excursion of input
(a)
Quantization error (e)
Maximum = IA/2
quantization error
/2
Input
x(nTs)
(b)
Transfer characteristic of a quantizer (b) Variation of quantization
Fig. 3.9. (a)
= A/2-0 = A/2
with input e =x,(nT,) x(nT)
-
error
From figure 3.9(a), it may also be observed that DO YOU KNOW?
3
= With a uniform quantizer, weak
for A<x (nT) <24, x,(nT) signals would experience a
40-dB
poorer SNR than that of strong
- A < x(nT)<- 2A, x(nT) signals. The standard telephone
=
or
error will technique of handling the large
This means that the maximum quantization input signal
range of possible
be t A/2. is to logarithmic
use a
quantization e r r o r is given by
levels instead of
maximum
In other words, compressed quantizer
a uniform one.
..(3.2)
'.
max
CHARACTERISTICS
QUANTIZATION
INCORRECT
QUANTIZER WITH
10. A UNIFORM (Important)
step the size remains same
quantizer since
known as uniform then quantizer is
known a s
last section is to the input,
#Luantizer discussed in varies according
step size reduce the
the input range, Also, if etc. is to
levels at+A/2,t A...
ghout
for taking the digital
lorm quantizer. The reason
 DIGITAL COMMUNICATIONS
quantization error. This has been ilustrated in figure 3.10. Hence, there ara
as shown in
figure 3.10 (a). That is one characteristie 'A' with thick line two
with thin line. It may be
observed that for characteristic 'A', we
have
: possiblecharcharacte
econd
Quantizer output x(nT
Transfer characteristics
not passing through zero
44
3A
-Characteristic 'A'
2A -Characteristic 'B
A/2
-4-3A -2 -A
A 2A 3A 4 Quantizer
input x(nT
(a)
Quantization error (e)
Maximum
quantization error =A
-4A-3A -2A-A NN 0 A 2A 3A 4A
Input
x(nT)
()
Fig. 3.10. (a) Incorrect
If 0<
xnT) <
A; then
quantization
characteristic (b) Increased
or
2A< x (nT) output x (nT) A
<3 A; then
= quantization error
Therefore, the output x, (nT) 3A
Similarly maximum =
for
shows actual characteristic 'B maximum quantization error
will be
equal to A as shown 3.106
figure 3.10(a).transfer characteristic whichquantization error is otted
through originequal
to -
On the other hand,
passes
for
A. shown
it does not
pass charactern
levels are taken
through characteristic of figure the diga
at 3.10(a). n
quantization error.*
tA/2,tA etc. It provides correct
quantization«characteristic
redu
and
1on
3.11.
TRANSMISSION BANDwIDTH IN A PCM
In this section, we
SYSTEM
shall evaluate 2005D
that the the
quantizerbe
levels that
use 'v
number transmission (BPTU, Orissa, Sem. B x a m .
assus
represented by 'v'binary digits tobandwidth for PCM sys the
may of Let u s
tem. h en u m b e r
Another digits will be, represent each level.
way of improving
the
Tn
accuracy of the
quantized signal is to increaff
e sample
rate
 ..(3.3)
q = 2
For example, if v =
4
levels of a q-level quantizer.
Here 'q represents total number of digital
number of levels will be,
bits, the total q 24 = 16 levels
Each sample is converted to 'v' binary bits. i.e.,
Number of bits per sample v.
=
We know that, number
of samples per second f =
second is expressed as
Therefore, Number of bits per x (Number of samples per
second)
(Number of bits per samples) ..3.4)
(Number of bits per second)
=
second
v bits per sample
x
f samples per
=
r a t e of PCM and
second is known a s signaling
As a fact, the number of bits per
matter of
is denoted by 'r' i.e., .(3.5)
Signaling rate in PCM,
r =
v f,
where f22 fm given by half of
the signaling rate
since bandwidth needed
for PCM transmission is
Also, Bandwidth in PCM,
Transmission
therefore, we have.
.(3.6)
BW2T
But r = vfs
..(3.7)
Therefore, BW2
Again, since f2 2 fm .(3.8)
BW 2v fm
Hence, bandwidth of a PCM system.
expression for
This is the required
NOISE/ERROR IN PCM
QUANTIZATION
3.12. Sem. Exam. 2006-07)
(JNTU, Hyderabad,
PCM system
noise (i.e., error) in a
for quantization
shall derive a n expression Because of quantization,
inherent e r r o r s a r e
In this section, w e quantization. quantization
for linear quantization
o r uniform
quantization error.*
As defined earlier, the
e r r o r is called
introduced in the signal. This
error is given as
...(3.9)
x g (nT)-x(nT, quantizer has
continuous amplitude
linear o r uniform
Let us assume that the input x(nT,) to a
in the range - xmax to max of input x(nT,) is mapped
total excursion
observed that the
it may be
From figure 3.9(a),
output isA and when input
vertical axis. This
means that when input is 44,
into 'q' levels o n
1 - A . Therefore,
represents
and -
x max represents 2
A. Thus, +:xmax 2
is-44,output is
max
the total amplitude range becomes,
= X,max-(-Xmax2 Xmax then the step size'A
Total amplitude range
range is
divided into 'q levels of quantizer,
amplitude
Now, if this total
number.
will be, be converted to a binary
sample of a signal
amplitude of each the greater the bit
rate required.
*PCM requires that the mher the greater the accuracy, but
 00 DIGITAL COMMUNICATIONS
nax 2xXmax
step size' A = max
Again, now if signal x(t) is normalized to minimum and maximum values equal to 1
3.10)
we have then
Amax =l
max 1
Therefore, step side would be, 3.11)
A (for normalized signal)
q .3.12)
Now, if step size 'A' is considered as sufficiently small, then it may be assumed that
tho
quantization error'e' will be an uniformly distributed random variable. We know thatthemari
um
quantization error is given as,
max 2 .3.13)
1.e.,
.3.14)
Hence, overtheinterval- A2
quantization error may be assumed as b- 1
an uniformly distributed random /2--2
variable.
Figure 3.11(a) shows an
uniformly
distributed random variable X over an -A2 2
interval (a, b). Recall that the PDF of Fig. 3.11. (a) A Uniform distribution (b) A Uniform
uniformly distributed random variable distribution for quantization error
X is given as
0 for x S a
b-a tora x Sb12tor linear quantization.
0 for x > b ..(3.15)
Thus, with the help of equation (3.15), the probability density function (PDF) for quantization
error 'e may be defined as
0 for es-
e)= for
A
0 for e> ..(3.16)
Also, from figure 3.11(b), it may be observed that verage
value. In other words, the mean 'm of the quantization quantization
error 'E
has zerd ot the
error is zero. Further, we know u
signal to quantization noise ratio of the quantizer is defined as,
S Signal power (normalized) .(3.17)
Noise power (normalized)
 WAVEFORM CODING TECHNIQUES
If type of signal at input ie., x(t) is known, then it is possible to calculate signal power. The
noise power 1s expressed as,
Noise power = VRoise
noise .(3.18)
R
noise is defined
Here, Vnoise 1S taken as the mean square value of noise voltage. Since, here
varible 'E, and PDF f, (e)therefore, its mean square value is given
by random
Mean square value = E[e2] = e = v2 ...(3.19)
noise
square value of random variable X is expressed as,
We know that the mean a
= EX = xtx(x) dx ..(3.20)
...(3.21)
Here ElE=t,e)de
Using equation (3.16), above equation may be written as,
E lE]
A/2
-A/2 A3 J-a/2
74/2
, 3
Simplifying, we get
.. (3.22)
El 12 would be
square value of noise voltage
Now, using equation (3. 19), the
mean
42
square value
=
mean
Vnoise 12
ohm, then the noise power is normalized i.e.,
Also, if load resistance, R= 1
= noise (putting R =1 in equation (3.18)
Noise power (normalized)
Ah2 1 12
Hence, finally, we write
Normalized noise power
or Quantization noise power
..(3.23)
or Quantization error (in terms of power)= 1
RATIO FOR LINEAR QUANTIZATION
3.13. SIGNAL TO QUANTIZATION NoISE
marks)
(C.G.U.T. Bhilai, Sem. Exam., 2005-06)(08
ratio is
PCM system for linear quantization
the signal to quantization noise
e know that in a
given as,
Here, we are as8umine that x(t) is a voltage signal.
 Normalized signal power
Normalized noise power
But, normalized noise power has been calculated as -12
Therefore, SNormalized signal power
N (A12)
We know that the number of bits 'v and
quantization levels are related
as,
2
Let us assume that
range
input x(nT) to a linear
quantizer has continuous
Xmax to+Xmax Therefore, total amplitude amplitude inmtthe
range
*max
maxFmax
Now, the step size will be
ZXmax
A max
Here, substituting the value of q from ..3.26
equation (3.25) in equation (3.26), we get
A = max
Now 2
substituting this value in
equation (3.24), we
Normalized signal power get,
Let
2a 2
DO YOU KNOW?
normalized signal power be denoted The transmission bandwidth
as P. varies
Then, S
P directly with the bit rate.
42 2
3P
22v
In order to keep the bit rate and
thus the
max v required
This is the
a PCM system.
2v 12
required relation for signal
max
companding is usedbandwidth low,
to
Hence, signal to quantization quantization noise ratio for
noise ratio: linear
quantizatio
22v
This max
expression shows that signal
with increasing bits to
per sample. noise power
Now, if we assume that ratio of ..(3.27)
input x(t) is
normalized, i.e.,
quantizer increases
max 1 exponentially
exp
Then, signal to quantization noise ratio will
be,
N 3 x 23 xP .(3.28)
Also, if the destination signal power "P is
PS 1 normalized, i.e.
.(3.29)
Then, the signal to noise ratio will be
given as
8x22
N
.(3.30)
.(3.31)
 WAVEFORM CODING TECHNIQUES 103
Because Xmax 1 and P < 1, the signal to noise ratio given by equation (3.31) is said to be
normalized. Expressing the signal to noise ratio in decibels, we get
aB 10logs0B s10log,, 13
= x 23]s(4.8+6 ) dB
Thus, signal to quantization noise ratio for normalized values of power P" and amplitude of
input x(t), will be
dB(4.8 +6v) dB ..3.32)
3.14. INFLUENCE OF NOISE ON THE PCM SYSTEM (Important)
To illustrate the influence of noise on the transmitted pulses, let us consider figure 3.12. First,
let us look at figure 3.12(a). Due to the noise superimposed on the pulse, only the PAM system
shall be affected. However, the PWM, PPM and PCM systems will remain unaffected. The
regeneration of the pulses is achieved by using a clipper circuit with reference levels A andB.
Now, let us observe figure 3.12 (b). Here, the sides of the transmitted pulse are not perfectly
vertical. In practice, the transmitted pulses usually have slightly sloping sides (edges). As the
noise is superimposed on them, the width and the position of the regenerated pulses is changed.
Now, this is going to distort the information contents in the PWM and PPM signals.*
Noise- -Noise~
AAA
Clipping A A--
levels B B -----*--{ -**-" ---**-***-f
AAAA
Regenerated
Regenerated pulses
pulses
No change Change in width
in width or position and position
(a) (b)
3.12. Effect of noise on PCM
Fig.
a Important Point: However, PCM is still unaffected as it does not contain any information
in the width or the position of the pulses. Hence, PCM has much better noise immunity as
compared to PAM, PWM and PPM systems.
3.15. VARIOUS IMPORTANT ASPECTS RELATED TO PCM
.15.1. Advantages of Digital Representation of a Signal: Salient Features
The digital representation of a signal has following advantages
interference.
Immunity to transmission noise and
along the transmission path.
() It is possible to regenerate the coded signal
Communication can be kept private and
secured by the use of encryption technique.
Mo use less bandwidth than equivalent analog
ern digital systems have better performance and
systems.
 DIGITAL COMMUNICATIONS
v) The possibility of uniform format for different kinds of baseband signals.
1 s possible to store the signal and process it whenever required.

3.15.2. Drawbacks
The advantages listed above are attained at the cost offollowing factors:
Increased transmission bandwidth.
i) Increased system complexity.
(11) PCM belongs to a class of signal coders known as waveform coders.
iv) The name waveform coders is given since in PCM, an analog signal is usually approxima
by mimicking the amplitude - versus time waveform. imated
3.15.3. Difference Between Waveform Coding and Source Coding
) Basically, the waveform coders are generally designed to be signal independent.
i) The waveform coders are different from the source coders (1.e., linear predictive coder:l
The source coders depend on
parameterization of the analog signal in accordance with an
appropriate model for the generation of the signal.
3.15.4. Encoding in PCM Binary
We know that encoding is the
ata010 1|100
process that follows the sampling unipolar
NAZ
and quantization. Encoding
process converts the quantized
samples into codewords. In a
binary code, each symbol may Polar
have either a 0 value or a 1 value. NRZ
There are various formats (i.e.,
waveforms) to represent the
binary sequence. They are known
as line codes. Figure 3.13 shows Fig. 3.13. Two binary formats
two of such formats called
unipolar NRZ and polar NRZ, where the NRZ stands for non-return to
zero.

3.15.5. Multiplexing in PCM Systems

It is possible to multiplex the PCM
signals using the time division multiplexing (TDM) princ1ple.
With increase in number of independent
message sources, the time interval allotted to each
has to be reduced to accommodate all the sources. source
This reduces the duration of each binary bit in a
PCM codeword. This increases the bandwidth
short then the impairments in the transmissionrequirement the system. If the pulses become very
of
medium start to interfere with the
of the system. Therefore, in practice, it is essential
to restrict the number
proper operauo
of message sources.
3.15.6. Synchronization in PCM
For a PCM system with TDM, it is essential to synchronize the transmitter and receivet for
proper operation of the system. For
synchronization, it is essential to the
transmitter and receiver. synchronize the clocks at
3.16. PERFORMANCE EVALUATION OF PCM: PROBABILITY OF ERROR CM
FOR P
To evaluate the performance of the PCM system, we have to
under
consider two maior sources of noie
) channel noise
(i) quantization noise.
 WAVEFORM CODING TECHNIQUES 10
) Channel Noise gets introduced anywhere along the transmission path. It is also known as
decoding noise.
(i1) Quantization Noise is introduced at the transmitter and is carried along to the receiver
output.
Both of them are present simultaneously but we shall consider them one by one to find their
efffect on the PCM system.
A. Channel Noise and its Effect
The major effect of channel noise is the introduction of transmission errors at the receiver
when the PCM signal is being reconstructed. Due to such errors, the receiver will make mistake in
making the decision about whether a 0 was received or a 1 was received. A 0 may be mistaken as 1
and a 1 may be mistaken as 0. Such errors must be minimized so as to improve the fidelity of PCM
system.
B. Probability of Error (P)
The fidelity of a PCM system in presence of channel noise is measured in terms of error or
probability of error. The probability of error or error rate is the probability that the symbol at the
receiver output differs from that transmitted.
Expression for Probability for Error (P) PCM Matched Decision Output
filter device
To obtain the expression for probabilityavG Sample
for error (P), let us use the matched filter att=Tp
shown in figure 3.14 and assume that the type Noise Threshold
of noise is AWGN, i.e., additive white Fig. 3.14. Matched filter receiver for PCM
gaussian noise*.
in chapter 5.
The derivation for the probability of error for a PCM system (P) has been given
as under:
Here, let us only state the expression for the error probability (P) ofPCM system
1 max .(3.33)
P erte
2 No
where Emax max
Peak signal energy N= Noise spectral density
We can substitute Emar Pmax Th where Pmax is
the maximum or peak signal power and T, is
terms of power is as under:
the bit duration. Hence, the error probability in

1 Pmax X1 . (3.34)
P,-erte No 2 N/T,
the average noise
****"***** ****

The ratioN,/ Th can be seen as 10

transmission bandwidth equal to
the
power contained in a viewed as the peak
bitrate (1/T). Hence, Em/N may be 10
signal to noise power ratio.
****
a***e***

10
C. Observations
e r r o r in PCM
receiver
) The average probability of **********a*************"

10
energy
depends only on the ratio peak signalmeasured at
of max
to the noise power spectral density No 70o ***
******

the receiver input.

function erfc is a
The complementary e r r o r
function. Hence, erfc 10 15 20
monotonically decreasing 10
PCM
Fig. 3.15. Probability of error in
a

- ita norfomance bas been discussed _ystem

 108 DIGITAL COMMUNICATIONS

But, WN =B/2
B
Therefore, R= 2xlog, M
R
B log M
or
M from equation (3.38), we get,
Substituting the expression of

R= Blog2|1NB bits/sec

From Shannon's theorem, for acceptable probability of error, we have

CR
This means that the lowest value will be CR. Using this identity, we get

12 P
C R=B log2|1
This is the required result.
NB
The ideal system is described by Shannon's channel capacity theorem. It sto
capacity C is given by, states the channe
C Blog, 1 NB
Now, let us compare equations (3.41) and (3.42). If they are to be equal, then
transmitted power of the PCM system should be increased by a factor k</12 as
comnaro
ideal system. ote
Trade off between Bandwidth and SNR
By Shannon-Hartley theorem, we get the channel capacity as under:

C Blo821
N
Now, let us find out the maximum possible value of C. From the equation for C, it is evident
that it depends on two factors, which are the bandwidth B and
the S/N ratio. Let us find their effet
on C one by one.

Effect of S/N on C
Ifthe communication channel is noiseless, then, N =0. Therefore, (S/N) - o and, so, Calso w
tend to Thus, the noiseless channel will have an infinite
.
capacity.
Effect of Bandwidth B on C
Now, let us consider that some white
gaussian noise is present, hence, (S/N) is PCM
8
not infinite.
Now, as the bandwidth Ideal
approaches infinity, the channel capacity system -M=8
does not become infinite since N N, B =

will also increase with the bandwidth B.

This will reduce the value of (S/N) with /6 dB -M=4

increase in B, assuming the signal power

S to be constant. Thus, the ideal system 2 -M=2
operates under two constraints namely,
limited bandwidth B and limited power 40 PN,Bdecis

P. But, the M-ray PCM operates under -20 -10 0 10 20 30 idealsysten

an additional constraint of a finite Fig.3.17. Comparison of M-ary PCM with tn

dB vers
number ofamplitude levelsM. Figure 3.17 is plotted as input signal to noise PN,
ratio P
noise ratio
B
 WAVEFORM CODING TECHNIQUES 109
handwidth efficiency R/B bits per second per hertz. In figure 3.17, we have substituted
R C and for an M-ray PCM system for different vallues of M, k has been substituted equal to7.
into saturation when thee
The constraint on M, in M-ray PCM tends to drive the M-ray PCM system
bandwidth efficiency satisfies the following condition:
R
s2
B log,M
that with increase in the value of M, the saturation level goes
on
This expression tells us

When PCM system enters into saturations the error probability reaches its limiting
increasing. a

value of 1.

3.17. COMPARISON OF PCM AND ANALOG MODULATION

Sem. Exam, 2006-07)
(Anna University, Chennai,
of analog modulation methods such as FM
or
The threshold effect in PCM is similar to a property the threshold
reduce the wideband noise above
PPM. The property is that, these systems tend to threshold
noise reduction if it is operated above its
levels. The PCM also provides the wideband
which is given by,

3S,
where q=2" for binary PCM and q M for M-ary PCM.
=

is close to the Nyquist rate and BW

=
Nf,m Hz. Then,
We a s s u m e that the sampling frequency
bandwidth ratio.
q M =Mb
where b BW/f is known as the
=

Therefore, we have

3 x (M)S,
=
3 x
M2S, =3M2b s

destination
Signal to noise ratio at the
Here, =

D
destination
Signal power at the to M2b which is much
S =

to noise ratio (S/N)p is proportional

that the signal b or b>. Hence, PCM
Here, it may be noted wideband FM which is proportional to only
of the of various modulation types
as a
higher than the (S/N), 3.18 shows the performance
FM. Figure 1/2. The dots indicate the
performs better than 3.18 have been plotted for S,
=

in figure
function of y. All the c u r v e s 2 and v = b.
been drawn for M
=

PCM c u r v e s have
threshold points. The

Conclusions 3.18 may be listed as

under
observations from figure (corresponding
Some of the important the threshold value yth
then increase in y beyond
the flat PCM c u r v e s
) For PCM if b is constant, does not increase (S/N), at all. Let us observe
to the threshold point) above the threshold.
PCM must be operated just with the s a m e
in figure 3.18. Hence, s o m e advantages
o v e r FM and PPM,
offer
the PCM does
(i) Near threshold, and expensive
values of b and (S/N)p. of m o r e complicated
is gained at the expense
advantage
(111) However, this
with increase in the value
of y and becomes
circuitry. increases linearly
and PPM
(iv) The (S/N), for FM values of y.
PCM for the higher
better than that of
 Expressing signal to noise power ratio in
WAVEFORM CODING TECHNIQUES
111
dB, we get
dB 10log1o =
10 log1o (1.5 22)
x

Or
dB 10 log (1.5) + 10 log10 (24)
or dB 1.76 +2v x 10 x 0.3
Therefore, we have

dB in PCMI=- aB =1.8+ 6v (for sinusoidal signal) Ans.

N
EXAMPLE 3.2. A Television signal
having a bandwidth of 4.2 MHz is transmitted using binary
POM system. Given that the number of quantization levels is 512. Determine:
V(i) Code word length
(ii) Transmission bandwidth
(iii) Final bit rate
(iv) Output signal to quantization noise ratio.
Solution: Given that the bandwidth is 4.2 MHz. This means that highest frequency component will have
frequency of 4.2 MHz ie.,
fm 4.2 MHz
Also, given that Quantization levels, q= 512
) We know that the number of bits and quantization levels are related in binary PCM as under:
2
1.e., 512 2
or log10 512 = v log1o 2

or V =
log10 512
log102
Simplifying, we get, v = 9 bits

Hence, the code word length is 9 bits. Ans.

i ) We know that the transmission channel bandwidth is given as,
BW2 vfm29x 4.2 x 106 Hz 2 87.8 MHz Ans.
(ii) The final bit rate is equal to signaling rate.
We know that the signaling rate is given as,
r = vtg
.6)
Here, sampling frequency is given as f, 2 2fm
f 2 2 x 4.2 MHz since f,m = 4.2 MHz
Thus, m

or f28.4 MHz
Substituting this value of 'f in equation (i) for signaling rate, we get
r = 9 x 8.4 x 105 bits/sec = 75.6 x 106 bits/sec Ans.
or
The transmission bandwidth may also be obtained as,

BW 2 75.6 x10 bits/sec

or BW2 37.8 MHz which is same as the value obtained earlier.
(av) The output signal to noise ratio is expressed
as

dB (4.8 +6v) dB
But v 9
 112 DIGITAL COMMUNICATIONS

Therefore, dB s 4.8 +6X9

or dB 58.8 dB Ans.

EXAMPLE 3.3. The bandwidth of an input signal to the PCM is restricted to 4 kHz, T
S1gnal varies in amplitude from - 3.8 V to +3.8 V and has the average power of 30 ut
required signal to noise ratio is given as 20 dB. The PCM modulator produces binar The
Assuming uniform quantization,
(i) Find the number of bits required per sample.
Gi)
Outputs of such PCM codersaretimemultiplexed. What would be the minimum
30
transmission bandwidth for this multiplexed signal?
ra..
red
(PTU, Jalandhar, Sem, Exam,
Solution: The given value of signal to noise ratio is 20 dB. 2006-07n
This means that.

dB =
10log1o 2 0 dB

Hence, = 100

i) We know that the

signal to quantization noise ratio is given as,

S 3P 22v
N 2max
Here, we are given, xmax 3.8 V
P 30 mW

and = 100
N
3x 30x10- x22v
Therefore, 100
(3.8)
Solving, we get v =
6.98 bits =
7 bits Ans.
(ii) The maximum frequency is given as

f 4 kHz
We know that the transmission bandwidth is
BW2 vfm
expressed as,
Since there are 30 PCM coders which
time multiplexed, the
are
BW 2 30 xvx transmission bandwidth must be,
We also know that the fm2 30 x 7 x 4 kHz 2
840 kHz Ans.
signaling rate is two times the
Signaling rate, r = 840 x 2 bits/sec transmission bandwidth, i.e.
=1680 bits/sec. Ans.
EXAMPLE 3.4. The information in an analog
/pCM svstem with an accuracy of+ 0.1% (full signal voltage waveform is to be transm smitted
over a

of 100 Hz and an amplitude range of- 10 toscale). The analog dwidth

+ 10 voltage waveform has a Da

volts.
G) Find the minimum sampling rate
required.
ii) Find the number of bits in each PCM word.
iii) Find minimum bit rate required in the PCM
iv) Find the minimum absoute channel signal the
bandwidth required
PCM signal. for the transmisS104
Solution: Here an accuracy 1s given as** *o. his means that(WBUT, Kolkata, error Exam.,
Sem. must
the auantization be *
**14% or
must be + 0.1%.
the maximum quantization error
 WAVEFORM CODING TECHNIQUES
Thus, Emax +0.1%+0.001
We know that the maximum
quantization error for an uniform quantizer is expressed as,

max

or 0.001
Therefore, Step size A = 2 x 0.001 0.002
We know that the step size, number of
related as
quantization levels and maximum value of the signal are

A = 2 Xmax
)
Here, given| xmax l = 10 volts

Substituting, values of A and xax in equation (i), we get

0.002 2X10

or
20
= 10,000
0.002
Hence, the number of levels are 10,000.
) The maximum frequency in the signal is given as 100 Hz, i.e.,
fm 100 Hz
By sampling theorem minimum sampling frequency should be,
f 2 2 fm2 2 x 100 2 200 Hz Ans.
(ii) We know that minimum 10,000 levels should be used to quantize the signal. If binary PCM is
used, then number of bits for each samples may be calculated as under, i.e.
= 2
Here, q= number of levels
v = bits in PCM,

Thus, 10,000 2
log10 10,000 =
v
log10 2
or V
log1010,000
13.288 14 bits Ans.
log102
(ii) The bit rate or signaling rate is expressed as,
r2 vf 214 x 200 2 2800 bits per second. Ans.
(iv) The transmission bandwidth for PCM is expressed as,

BW. x 2800 2 1400 Hz Ans.

EXAMPLE 3.5. Twenty four voice signals are sampled uniformly and then have to be time division
multiplexed. The highest frequeney component for each voice signal is equal to 3.4 kHz. Now
(i) If the signals are pulse amplitude modulated using Nyquist rate sampling, what would
be the minimum channel bandwidth required.
(i) Ifthe signals are pulse code modulated with an 8bit encoder, what would be the sampling
rate? The bit rate of system is given as 1.5 x 10° bits/sec.
olution: (i) As a matter of fact, if N channels are time divis1on multiplexed, then minimum transmission
bandwidth is expressed as,
BW N fm
Here, f is the maximum frequency in the signals
Given, 3.4 kHz
 DIGITAL COMMUNICATIONS

Therefere BW = 24 x 3.4 kHz = 81.6 kHz

(11) The signaling rate of the system is given as8, Ans
r = 1.5 x 105 bits/sec
Since there are 24 channels, the bit rate of an individual channel is,

1.5 x 10
r (one channel) =E = 62500 bits/sec
24
Further, since each sample is encoded using 8 bits, the samples per second will be,

Sample/sec = r (one channel) bits/sec

bits/sample
Note that the samples per seconds is nothing but sampling frequecy.

62500 bits/sec
Thus, we have
8bits/sample
Solving, we get, f 7812.5 Hz or samples per second Ans.

EXAMPLE 3.6. A PCM system uses a uniform

quantizer followed by a 7-bit binary encoder. The
bit rate of the system is equal to 50 x 105 bits/sec.
(i) What is the maximum message signal bandwidth for which the
system operates
satisfactorily?
(ii) Calculate the out ut signal to quantization noise ratio when a full load
sinusoidal
modulating wave of frequency 1 MHz is applied to the input.
(RTU, Kota, Semester Exam. 2005-2006)
Solution: (i) Let us assume that the message bandwidth be
be,
fm Hz. Therefore sampling frequency should
f 2 fm
The number of bits given as v = 7 bits
We know that the signaling rate is given as,
r2 v f,
or r2 7x 2 fm
Substituting value for r, we get
50x1062 14fm
or
fm3.57 MHz Ans.
Thus, the maximum message bandwidth is 3.57 MHz.
(1) The modulating wave is sinusoidal. For such signal, the ed
as,
signal to quantization noise ratio is expre

d B = 1.8 + 6v

Substituting the value of v, we get

dB =
1.8 + 6 x 7 =43.8 dB Ans.

EXAMPLE 3.7. The information in an analog waveform with maximum kHzis to

betransmitted over an M-level PCM system where the number of frequency Tm M 16.
The quantization distortion is specified not to exeed quantization leveis
1% of peak to peak analog signa
(i) What would be the maximum number of bits per this n
sample that should be u
PCM system?
(ii) What is the minimum
sampling rate and what is the resulting bit transmissio
solution: (1) Since the number of quantization levels given here are M 16, =

= M= 16
We know that the bits and levels in
binary PCM are related as,
2

=number ot bits in a codeword

Here, v
 Important Note: Therefore, such large crest factor signals (i.e.
use non-uniform quantization to overcome the problem just disepeech and
reduces at low power levels (P << 1) just now we have
observed . ed. Signal
means that at low signal levels, signal to noise ratio reduces i.e. equatio to musinOtsee shndarai
the quantization noise is directly related to step size. Therefore. ats e increa (85
noise can be kept low by keeping step size low. This means that at low signal leo
noise ratio can be inereased by decreasing step size 'A'. This means tnal levelt vels (P<<
be varied according to the signal level to keep signal to noise ratiotep size s
2e
This is nothing but nonuniform quantIzauon. INow letS see how non inicequire
is achieved through companding in the section to follow.
Nalue
3.20. NONUNIFORM QUANTIZATION*
antnatin
(VTU, Bangalore, Sem.
Exam ., 2005-A8
If the quantizer characteristics is nonlinear and the step size is not eanot
variable. dependent on the amplitude of input signal then the quantization ie eadit
is instead if iti
uniform quantization. In non-uniform quantization, the step size is reduced
reduced withknown
witht as no
in signal level. For weak signals (P < < 1), the step Size is small, therefore the Tedueta
noise reduces, to improve the signal to quantization noise ratio for weak
weak signals.Ttz
is thus varied according signal level to keep the signal to noise ratio adeauatlstep s
the
signals. The
This is non uniform quantization. The non-uniform quantization is practicallyachieveddtthrough
a process called companding. We shall discuss companding in the next section.

3.21. COMPANDING (i.e., cOMPANDED PCM)

As a matter of fact, companding is non uniform quantization. It is required to be implementeit
We know that the quantization
improve the signal to quantization noise ratio of weak signals.
noise is given by

N 12
the step size is fixed, the quantizatm
This shows that in the uniform quantization,
once

power is not
constant. It is propornon
noise power remains constant. However, the signal
will be small for weak signals, but qua
the square of signal amplitude. Hence signal power
to quantization noise
for the weak sigdid
noise power is constant. Therefore, the signal
of signal. The remedy is to use
companding. Compa
poor. This will affect the quality
term derived from two words i.e., compression
and expansion as under
tnown

CompandingCompressing
= + Expanding
non-uniform quantization becausmethod us
is
1s method

In pratice, it is difficult to implement the

applyingtthe
anplying
p a r t i c u l a r

advance about the changes in the signal level. Therefore, a particu

before
d e s i t

in
ock that pr
t h a tp r o v i d e s
attenuated

and strong signals a r e

The weak signals a r e amplified
compression and ne
quantizer. This process is called
as T h e circuitused
to a uniform
mitter

is called as a compressor,
followed which is called expans at th nding

At the receiver exactly opposite is

o c e s so fcompana

is called a s a n expander.
The compression or sig
providing e x p a n s i o n called companding. Iu
the receiver is combined to be
and expansion at
E x p a n d e r

Uniform
in the form
has been shown Input Compressor quantizer
quantizer
block diagram in figure ignat
of a
A companding
mode

3.20. Fig. 3.20. proportionalt 7redominantw s

n
he made
 WAVEFORM CODING TECHNIQUES

3.21.1. Compressor Characteristic

Figure 3.21 (a) shows the compressor characteristics. As shown in figure 3.21 (a),
the compressor
input signals. Thus,
provides a higher gain to the weak signals and smaller gain to the strongnoise ratio. lt may be
weak signals a r e artificially boosted to improve the signal to quantization
signal but
noted that this compressor characteristics has been shown only for the positive input
we can draw
it even for the negative input signals using the some principle. In fact, the compressor
is included at the PCM transmitter.

3.21.2. Expander Characteristics

Figure 3.21 (6) shows the expander characteristics. This characteristics is exactly the inverse
boosted signals by the
of the compressor characteristics. This ensures that all the artificially
are brought back to their original amplitudes at the receiver
end.
compressor
Expander output
Compressor output

Gain for weak

input signals is high
but that for strong
input signal, it is low

Compressor input Expander Input

characteristics
characteristics Fig. 3.21. (b) Expander
Fig. 3.21. (a) Compressor
3.22. COMPANDER CHARACTERISTIC

characteristics which is the combination of the compressor

Figure 3.22 shows the compander Output voltage

Due to the inverse
and expander characteristics.
the overall
nature of c o m p r e s s o r and expander,
characteristics of the compander
is a straight line **************

indicates that all Compression

(dotted line in figure 3.22. This
back to their original -Expansion
the boosted signals a r e brought
amplitudes.*
0 Input voltage
3.23. DIFFERENT TYPES OF Expansion
COMPRESSOR CHARACTERISTICS

2006-07) (10
A
*************s**ansann"
Compression

(GGSIPU, Delhi, Sem. Exam.,

marks)
Linear
characteristics characteristics
ldeally, we need a linear compressor
curves for PCM system
for small amplitudes of the input signal anda Fig. 3.22. Companding
elsewhere. In pratice, this
logarithmic characteristic
18 achieved by using following two methods:

() -law companding
i) A-law companding
constant for all signals within the
used to make the SNR a

o n iniform quantization can be

input range.
 3.23.1. p-law Companding
In -law companding, the compressor is continu
linear for
the for
characteristic
ious. It
smaller values input high
levels and logarithmie approxin
of ut levels. is input
compre-ssor characteristic is mathematically expressed as under:
(+lx|/x,mas
The
= (sgn x) n
z(x)
In(1+)
where 0sIxl/xmax 1.
Here, z(x) represents the output and x is the input to the compressor. Al.
the normalized
value of input with respect to the maximum valuex
represents + 1 i.e., positive and negative values of input and output Furtner,max(sesen
characteristics for different values of have been shown in figure
3.23(a) Taw
sgn
value of u is 255.It may be noted that the characteristic com
he practically
to
corresponding u =o
uniform quantization. The u-law
companding 1s used for speech and music sie corresponds
PCMtelephone systems in United States, Canada and Japan. Figure 3.23(h nals. It is useto the
of signal to quantization noise ratio with to
respect signal level, with and withss the the varmation
is obvious that SNR is almost constant at all the varir
signal levels when
Normalized output z() companding is used. mpanding.t
(S/N), dB

0.8 H 255
50t
= 5
0.6 40
0 with companding
0.4 30

0.2
20
without compandingg
0.4 0.8
40 -30 -20 -10 Signal
0 leve! (dB)
Normalized input,I * max
(a) Compressor characteristics of a
p-law compressor (b) PCM performance with -law companding
Fig. 3.23.
3.23.2. A-law Companding Normalized output
1.0
(VTU, Bangalore, Sem. Exam.,
In the A-law 2004-2005) 0.8 A 87.56
companding, the A =2/
characteristic is piecewise, made up
Compressorr
compressor
for of a linear
low level inputs and a 0.6
level inputs. Figure 3.24 logar+thmic segment segment
for high A =1
shows the A-law
characteristics for different
values of
compressor 0.4
to A = we
A.
1, observe that the Corresponding
to
characteristic is
a
which corresponds linear o2
uniform The
practically used value
of Ais 87.56. Thequantization.
A-law
is used for companding 0.8 1
0.4 0.6
PCM telephone systems in
segment of the characteristics is Europe. The linear 0.2 Normalized
lnpu

whereas the logarithmic segments isfor low level

for high level inputs 3.24.
It is mathematically expressed as input*. Fig. of A-law compressor
o
under: ompress
*For voice signals, the typlcal 1nput signal
dynamic ramge is 40 decibels (db).
 WAVEFORM CODING EGHINI
.
Alx|/x,ma: for 0sLs1
1+ loge A Kmax ...(3.54a)
z(x)
Xmax 1+log [AIx|/%max for s ls1
1+loge A A A max
of Compandingg
3.23.3. Effect
us consider figure 3.25 (SNR), dB
Let
the curves of output
which shows
signal to noise ratio versus input
power for
the uniform and With u law
signal The curve companding (u = 255)
nonuniform quantizers.
nonuniform quantizer 40
for
corresponds to the lawv
255. The DynamiC
companding with u = 30t range
3.25 have been
curves in figure
plotted with the- following
assumptions:
20t
) Value of Q is 256. Go
Uniform
(ti) Parameter u 255
=
10t quantization
random
(iii) Let X denote the 20 log1o lo/madB
variable representing the
-40 -30 -20 -10
input. Let it has the -60 -50
uniform and nonuniform
Laplacian distribution Fig. 3.25. Output NR characteristics for
quantization
which is represented
mathematically as under.
fx) 20, e-21x|/oxx
where, o is the variance of X.
iv) The input signal ranges between -Xmax to Xmax
3.23.4. Dynamic Range
on the
of uniform quantizer is highly dependent
Figure 3.25 indicates that the performance dB (from 15 dB to -45 dB.)
has a dynamic range of about 30
-
Input, whereas the u law compander of input over which the output SNR remains within 3 dB of
Dynamic range is defined a s the range can also be defined as
the ratio of largest possible
he maximum value of 38 dB. The dynamic range the
of the signal, that can be decoded successfully by
magnitude to the smallest possible magnitudeis defined as under:
range
aecoder. Mathematically, the dynamic
DR = max
Vmin
3.23.5. Coding Efficiency
a PCM code is utilized. It is
defined a s the
indication of how efficiently
Coding efficiency is an
certain dynamic range to the actual number
of
to achieve a
0rat minimum number of bits required efficiency is given by,
ES
required. Mathematically, the coding
Minimum number of bits (including sign bits)100
x100
Coding efficiency Actual number of bits (including sign bits)