Sampling & Pulse Modulation Lecture Notes in “Digital Communications”

by: Dr. Mohammed AlMahamdy Electrical Engineering | University of Anbar

1.6 PULSE CODE MODULATION (PCM)

After sampling an analog signal, it is possible to code the samples using discrete symbols, such
as binary. The communication through binary alphabets is more efficient than the analog
transmission. If the value of a sample is sent using only two possible elements, the reception
becomes more reliable. It is easier for the receiver to discriminate the reception into two
possible signals than estimating the value of a continuous signal. This significantly helps in
reducing the effects of distortion and of additive noise.

In PCM, the sample values are rounded to the values of certain levels. The rounding off
operation is known as “quantization”. Then each sample is coded into a binary number which
is equivalent to the index of the quantization level that is closest to the sample value.

1.6.1 Quantization
When an information signal is pulse amplitude modulated, it becomes discrete in time only. It
remains analogue in amplitudes since all the values within the specified range are allowed. PAM
signal is said to be quantized when each pulse of the PAM signal is adjusted in amplitude to
coincide with the nearest level within a finite set.

It is clear from the figure below, quantization error (noise) can be reduced by increasing the
number of quantization levels (𝐿), i.e. decreasing the intervals (𝑞) between the levels.

-11-
Lecture Notes in “Digital Communications” Sampling & Pulse Modulation
Electrical Engineering | University of Anbar by: Dr. Mohammed AlMahamdy

1.6.2 Encoding
The quantized samples are now to be coded with 𝑙 bits per sample (𝑙 = log 2 𝐿). One of the most
popular quantizer/encoder circuits is the parallel quantizer, which requires 𝐿 − 1
comparators. The following is a 2-bit PCM coder.

1.6.3 Decoding
The transmitter sends binary codes to the receiver via a channel. The receiver must decode the
bit sequence back to a time function. This is done by associating each group of bits with the
corresponding quantization level; thereby reconstructing the quantized waveform by LPF. In
this simple 2-bit PCM decoder, S1 & S2 are received binary, S1=MSB and S2=LSB.
-12-
Sampling & Pulse Modulation Lecture Notes in “Digital Communications”
by: Dr. Mohammed AlMahamdy Electrical Engineering | University of Anbar

R/4

Quantized
- samples Analog
S2 R LPF
+ output
S1 R/2

The PCM technique is considered as an Analog to Digital Converter (ADC) at the transmitter
and Digital to Analog Converter (DAC) at the receiver. The following figure illustrates the ADC

-13-
Lecture Notes in “Digital Communications” Sampling & Pulse Modulation
Electrical Engineering | University of Anbar by: Dr. Mohammed AlMahamdy

The complete block diagram of PCM system is:

Modulator:

Anti- Parallel
Analog Xs L-Levels Xq Digital
Aliasing Sampler to PCM
Input Quantizer Encoder
Filter Serial

fs

Demodulator:

Serial
Decoder Xq Analog
PCM to LPF
(D/A) output
Parallel

1.6.4 Non-Uniform Quantization

UNEQUAL STEP-SIZE
Uniform quantization assumes that the information signal has uniform PDF, i.e. all quantization
levels are used equally. For most signals, it is not the case.

If the PDF of the information signal is not uniform (nevertheless known and constant with
time), then we can optimize the locations of the quantization levels to obtain minimum
quantization noise introduced.

As an illustration, let the normalized signal 𝑥(𝑡) has the typical probability function 𝑃𝑥 (𝑥):

Px(x)

x
-1 0 1

The shape of 𝑃𝑥 (𝑥) means |𝑥(𝑡)| ≪ 1 most of the time. Therefore, we can use non-uniform
quantization as indicated by the dashed lines. The quantization lines are located here close to
each other near 𝑥 = 0 . They are sparse for large values of |𝑥(𝑡)| , as large |𝑥(𝑡)| occurs
infrequently.

The depictions below show the uniform and the non-uniform distribution of the quantization
levels.

-14-
Sampling & Pulse Modulation Lecture Notes in “Digital Communications”
by: Dr. Mohammed AlMahamdy Electrical Engineering | University of Anbar
Output
Output

Input
Input

Uniform Non-Uniform
Quantization Quantization

Practically, such optimization is a difficult procedure because it requires prior knowledge of the
signal PDF. Problems arise if the information signal has an unknown PDF or if its PDF changes
with time. However, they are similar in some signals. For example, in the case of voice signals,
the PDF shape of different speakers is usually similar, but the gross level can vary widely
between speakers, e.g. man is shouting and woman softly spoken. Therefore, the approach
taken in practice is to use uniform quantizing after non-linear compression (the companding).

COMPANDING (COMPRESSING-EXPANDING)
This is the process of compressing the information signal prior to linear quantization at
transmission. The compression is achieved via a non-linear amplitude characteristic circuit.

Information Non-Linear Uniform Non-Uniform

signal Operation Quantizing Quantizing

The receiver expands the reconstructed signal with the inverse characteristic to restore the
original waveform.
Output Output

Compression

No Compression

Input Input

Compression Expansion
Characteristic Characteristic
H(x) H-1(x)

So, PCM with companding system will be:

Input Signal Compression Uniform CHANNEL PCM Expansion Output Signal

x(t) H(x) PCM Decoder H-1(x) x’(t)

-15-
Lecture Notes in “Digital Communications” Sampling & Pulse Modulation
Electrical Engineering | University of Anbar by: Dr. Mohammed AlMahamdy

Companding Methods

There are two types of companders that are practically and widely used for speech coding:

(1) The µ-law compander (used in the US, Canada and Japan). The parameter µ controls the
amount of compression and expansion. The standard compressor uses 𝜇 = 255
followed by a uniform quantizer with 128 levels (7 bits per sample).

log(1 + 𝜇|𝑥|)
𝐻(𝑥) =
log(1 + 𝜇)

(2) The A-law compander (used in most countries). A is chosen to be 87.56.

𝐴|𝑥| 1
for 0 ≤ |𝑥| ≤
1 + ln 𝐴 𝐴
𝐻(𝑥) =
1 + ln(𝐴|𝑥|) 1
{ 1 + ln 𝐴 for ≤ |𝑥| ≤ 1
𝐴

-16-