Communication Systems Lecture-6:

Pulse Code Modulation

Chadi Abou-Rjeily

Department of Electrical and Computer Engineering

Lebanese American University
chadi.abourjeily@lau.edu.lb

September 15, 2017

Chadi Abou-Rjeily Communication Systems Lecture-6: Pulse Code Modulation

PCM (1)

Pulse Code Modulation (PCM): is essentially Analog-to-Digital

conversion (A/D) of a special type where the information
contained in the instantaneous samples of an analog signal is
represented by digital words in a serial bit stream.
The PCM signal is generated by carrying out three basic
operations:
Sampling: where the analog signal is transformed into a
flat-top PAM signal. In other words, the analog signal is
represented by a certain number of samples.
Quantization: the samples of the analog signal can take an
infinite number of values. The signal at the output of the
quantizer can take M possible values. In this case, each sample
is replaced by the closest level (among the M possible levels).
Encoding: each level is mapped into a sequence of n bits. The
output of the decoder is a binary codeword having a length n
where M = 2n .

Chadi Abou-Rjeily Communication Systems Lecture-6: Pulse Code Modulation

PCM (2)

Chadi Abou-Rjeily Communication Systems Lecture-6: Pulse Code Modulation

PCM (3)
In the next figure M = 8:

Chadi Abou-Rjeily Communication Systems Lecture-6: Pulse Code Modulation

PCM (4)
It is up to the system designer to specify the exact codeword
that will represent a particular quantized level.
The next example shows a Gray code with M = 8 (n = 3 bits):

For the Gray codes:

The first bit is the sign bit.
Consecutive levels differ by only one bit. Consequently, noise
that moves the signal sample from one quantization level to an
adjacent level results in a single bit error in the received PCM
codeword.
Chadi Abou-Rjeily Communication Systems Lecture-6: Pulse Code Modulation
Bandwidth of PCM signals (1)

Due to the nonlinear operations introduced by PCM, the

spectrum of PCM signals is not directly related to the
spectrum of the input analog signal (as with PAM).
The bandwidth depends on:
1 Bit rate.
2 Waveform pulse shape.
Since a sample is generated every Ts seconds and since every
sample is encoded into n bits, the bit rate of PCM systems is:
n
R= = nfs
Ts
For no aliasing, it is required that fs ≥ 2B resulting in
R ≥ 2nB.

Chadi Abou-Rjeily Communication Systems Lecture-6: Pulse Code Modulation

Bandwidth of PCM signals (2)

It will be shown that the bandwidth of the binary encoded

PCM waveform is bounded by:
1 1
BPCM ≥ R = nfs
2 2

The minimum bandwidth of 12 R is obtained only when

sinc-type pulses are used.
Usually, rectangular-type pulses are used. In this case, the
bandwidth of the PCM waveform will be larger than this
minimum.

Chadi Abou-Rjeily Communication Systems Lecture-6: Pulse Code Modulation

Effects of noise (1)
Two main effects produce noise in PCM systems:
Quantizing noise that results from the M-level approximation
of the samples.
This can be reduced by increasing the binary length of each
codeword (in other words by increasing M = 2n ).
Channel noise that results in bit errors in the recovered PCM
signal.
It can be shown that the ratio of the average power of the
recovered signal to the average noise power is:
M2
 
S
=
N out 1 + 4(M 2 − 1)Pe
where Pe is the probability of bit error in the recovered binary
PCM signal at the receiver before it is converted back into an
analog signal.
Note that the above equation is valid under the assumption
that the signal has a uniform distribution between −V and V
where V can take any positive value.
Chadi Abou-Rjeily Communication Systems Lecture-6: Pulse Code Modulation
Effects of noise (2)

To consider the quantizing noise alone, we set Pe = 0

resulting in:  
S
= M2
N out

In decibels:
 
S
= 10 log10 (M 2 ) = 10 log10 (22n )
N out,dB
= 10 log10 (4n ) ≈ 6.02n

This is called the 6-dB rule where a 6-dB improvement in the

SNR is obtained when an additional bit is added to each PCM
codeword.

Chadi Abou-Rjeily Communication Systems Lecture-6: Pulse Code Modulation

Example: PCM for telephone systems
An analog audio telephone signal occupies a band from 300
Hz to 3.4 kHz.
The minimum sampling rate is fs = 2 × 3.4 = 6.8 kHz.
In order to allow the use of practical low-pass filters with
reasonable transition bands, the signal is oversampled with a
sampling frequency of 8 ksamples/s.
Each sample is represented by 8 bits and the data rate is:
R = nfs
= [8 (bits/sample) ] × [8 (ksamples/s) ] = 64 (kbits/s)

In this case, the minimum absolute bandwidth is:

1
Bmin = R = 32 kHz
2
The average signal-to-quantizing-noise ratio is:
 
S
≈ 6n = 48 dB
N out,dB
Chadi Abou-Rjeily Communication Systems Lecture-6: Pulse Code Modulation
Nonuniform Quantizing (1)

Uniform quantization presents the following limitations:

Voice analog signals are more likely to have amplitude values
near zero than at the extreme peak values allowed.
When uniform quantizing is used, the level of quantizing noise
is independent of the signal amplitude. Consequently, small
amplitudes suffer more from quantizing effects with respect to
large amplitudes.

Chadi Abou-Rjeily Communication Systems Lecture-6: Pulse Code Modulation

Nonuniform Quantizing (2)
One way for solving this problem is by applying nonuniform
quantization. In this case:
Quantizing steps are made small for small amplitude levels.
Quantizing steps are made large for large amplitude levels.
This quantization can be achieved by the use of a variable step
size.

Chadi Abou-Rjeily Communication Systems Lecture-6: Pulse Code Modulation

Nonuniform Quantizing (3)
The effect of nonuniform quantization can be achieved by:
First passing the analog signal through a compression
(nonlinear) amplifier.
Then applying uniform quantization.
At the receiver side, the inverse operation (expanding) must
be performed.

w (t)
w1 (t) = ⇒ |w1 (t)| ≤ 1
max |w (t)|
Chadi Abou-Rjeily Communication Systems Lecture-6: Pulse Code Modulation
Nonuniform Quantizing (4)

µ-law: The µ-law type of compression is defined by:

ln (1 + µ |w1 (t)|)
|w2 (t)| =
ln (1 + µ)

In the United States, Canada and Japan, the telephone

companies use a µ = 255 compression characteristic in their
PCM systems.
A-law: The A-law type of compression is defined by:
(
A|w1 (t)|
1+ln A , 0 ≤ |w1 (t)| ≤ A1 ;
|w2 (t)| = 1+ln(A|w1 (t)|)
1+ln A , A1 ≤ |w1 (t)| ≤ 1.

The A-law is used mainly in Europe. A typical value of A is

A = 87.6.

Chadi Abou-Rjeily Communication Systems Lecture-6: Pulse Code Modulation

Nonuniform Quantizing (5)

Chadi Abou-Rjeily Communication Systems Lecture-6: Pulse Code Modulation

Nonuniform Quantizing (6)
In practice, the nonlinear characteristics of the µ-law and
A-law are approximated by piecewise linear chords.
For example, the telephone companies in the United states
use the µ = 255 compression for generating 8-bit PCM
codewords. For this 8-bit quantization:
8 chords are used to approximate the µ-law compression curve
for the positive portion of the signal.
Each chord is approximated by a uniform quantizer with 16
steps.
The length of the steps depend on the chord number. For
example, chord 1 has 16 steps having a length of ∆ each.
Chord 2 has 16 steps having a length of 2∆ each. In general,
the step size in the n-th chord is 2n−1 ∆.
The same procedure is applied on the negative part of the
signal.
Consequently, the total number of steps is:
Nsteps = 2 × 8 × 16 = 256
Chadi Abou-Rjeily Communication Systems Lecture-6: Pulse Code Modulation
Nonuniform Quantizing (7)

Chadi Abou-Rjeily Communication Systems Lecture-6: Pulse Code Modulation

Nonuniform Quantizing (8)
As can be seen from the previous figure, the 8-bit PCM
codeword consists of:
1 sign bit that denotes a positive or negative input voltage.
3 chord bits that denote the segment number.
4 bits that denote the particular step within a segment.
A segmenting technique similar to that used with the µ-law is
applied with the A-law.
The total length of the segments in the positive part of the
signal is given by:

16 [1 + 2 + 4 + · · · + 128] ∆ = 16[255]∆ = 4080∆

Suppose that the analog signal is limited between −V and V .

The step size ∆ is chosen to verify:
V
∆=
4080
Chadi Abou-Rjeily Communication Systems Lecture-6: Pulse Code Modulation
Nonuniform Quantizing (9)
Once again, it can be shown the output SNR follows the 6-dB rule:
 
S
= 6.02n + α
N dB
where n is the number of bits per PCM codeword.
For uniform quantization:
 
V
α = 4.77 − 20 log10
xrms
where:
V is the peak value of the input analog signal.
xrms is the root-mean-square value of the input analog signal.
For the µ-law companding with sufficiently large input levels:
α ≈ 4.77 − 20 log10 [ln(1 + µ)]
For the A-law companding with sufficiently large input levels:
α ≈ 4.77 − 20 log10 [1 + ln(A)]
Chadi Abou-Rjeily Communication Systems Lecture-6: Pulse Code Modulation
Nonuniform Quantizing (10)
The ratio V /xrms is called the loading factor.
Notice that the output SNR is a function of the input level for
the case of uniform quantizing, but it is relatively insensitive
to the input level for µ-law and A-law companding.

Chadi Abou-Rjeily Communication Systems Lecture-6: Pulse Code Modulation

Nonuniform Quantizing (11)
For uniform quantizing with a uniformly distributed input signal,
α = 0 and the output SNR takes the simple value given at the
beginning of this lecture:
 
S
= 6.02n
N dB

Proof:
Consider the signal w (t) that is uniformly distributed between
−V and V .
The probability density function of w (t) is given by:

1

2V , −V ≤ x ≤ V ;
fw (x) =
0, elsewhere.

Chadi Abou-Rjeily Communication Systems Lecture-6: Pulse Code Modulation

Nonuniform Quantizing (12)
The mean value of w (t) is given by:
Z +∞
x̄ = xfw (x)dx
−∞
+V
1
Z
= xdx = 0
2V −V

The rms value can be calculated from:

Z +∞
2
xrms = (x − x̄)2 fw (x)dx
−∞
+V
1 V2
Z
= x 2 dx = · · · =
2V −V 3
Consequently, for uniform quantization:
 
V
α = 4.77 − 20 log10
xrms
√
= 4.77 − 20 log10 3 = 0
Chadi Abou-Rjeily Communication Systems Lecture-6: Pulse Code Modulation
Shannon’s Theorem

The capacity of a channel is given by (in bits/s):

 
S
C = B log2 1 +
N

where:
B is the bandwidth of the channel.
S
N is the signal-to-noise ratio.
Shannon’s theorem: a data rate R can be achieved over a
communication channel with an arbitrarily small probability of
error (Pe ) if and only if:

R≤C

Chadi Abou-Rjeily Communication Systems Lecture-6: Pulse Code Modulation

Example: V.90 56-kb/s PCM computer modem (1)

This modem employs µ-law PCM with µ = 255.

The codewords have a length of 7 bits: 1 sign bit, 3 chord bits
and 3 bits to specify the step in a chord.
The modem’s clock is synchronized to the PCM 8-ksamples/s
clock of the telephone company.
This results in a data rate of:

R = nfs = [7 (bits/sample) ]×[8 (ksamples/s) ] = 56 (kbits/s)

The SNR required to achieve this rate with an arbitrarily small

probability of error can be calculated from Shannon’s theorem:
 
S
R ≤ C = B log2 1 +
N
S
⇒ ≥ 2R/B − 1
N
Chadi Abou-Rjeily Communication Systems Lecture-6: Pulse Code Modulation
Example: V.90 56-kb/s PCM computer modem (2)

Assuming that the filtering circuits limit the useful bandwidth

of the telephone line to B = 3.3 kHz:
 
S h i
≥ 10 log10 256/3.3 − 1 = 51.1 dB
N dB

If the SNR is less than 51.1 dB, the data rate of 56 kb/s can
not be achieved with Pe ≈ 0.
In this case, the modem will fall back to a lower speed, such
as 33.3 kb/s, 28.8 kb/s or 24 kb/s.

Chadi Abou-Rjeily Communication Systems Lecture-6: Pulse Code Modulation