Principles of Communication

Baseband
Digital System (2)

LC 3-3, 7-7

Lecture 14, 2008-10-24

 Contents
„Pulse Code Modulation
„Sampling, Quantizing, and Encoding
„Practical PCM Circuits
„Effect of Noise
„Nonuniform Quantizing

2
 Pulse Code Modulation
„ Pulse code modulation (PCM) is essentially analog-to-digital conversion
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.
„ Relatively inexpensive digital circuit may be used extensively in the
system.
„ PCM signals derived from all types of analog sources (audio, video, etc.)
may be merged with data signals (e.g., from computers) and transmitted
over a common high-speed digital communication system. TDM
„ In long-distance transmission systems, a clean PCM waveform can be
regenerated at the output of each repeater, where the input consists of
a noisy PCM waveform. However, the noise at the input may cause bit
errors in the regenerated PCM output signal.
„ The noise performance of a digital system can be superior to that of an
analog system.
„ Much wider bandwidth than that of the corresponding analog signal.

3
PCM Transmission System

4
Sampling, Quantizing, and Encoding

5
 Gray Code

„ Encoding each quantized sample value into a digital word.

„ Gray code has only one bit change for each step change in the quantized
level. Consequently, single errors in the received PCM code word will
cause minimal errors in the recovered analog level, provided that the
sign bit is not in error.

6
 Bandwidth of PCM Signals

„ The bit rate is R = nf s

7
 Effect of Noise
„ The analog signal that is recovered at the PCM system
output is corrupted by noise. Two main effects produce
this noise or distortion:
„ Quantizing noise that is caused by the M-step quantizer at the
PCM transmitter.
„ Bit errors in the recovered PCM signal. The bit errors are caused
by channel noise as well as improper channel filtering, which
causes ISI.
„ The analog sample output of the PCM system for the kth
sampling time is
yk = xk + nk
„ The xk is the signal (same as the input sample) and nk is
noise.
8
PCM Communication System

9
 Effects of Noise
„ Noise due to the quantizing error
eq = Q( xk ) − xk
„ Noise due bit errors
eb = yk − Q( xk )
„ Thus

nk2 = eq2 + eb2

∞ δ /2 1 δ2 V 2
eq2 = ∫ eq2 f (eq )deq = ∫ eq2 deq = =
−∞ −δ / 2 δ 12 3M 2

10
 Effects of Noise
n n
e = [ yk − Q( xk )] = [V ∑ bkj 2 − V ∑ akj 2 − j ]2
2
b
2 −j

j =1 j =1

n n n
= V [∑ (bkj − akj )2 ] = V [∑ (bkj − akj )2 ][∑ (bkl − akl )2 −l ]
2 −j 2 2 −j

j =1 j =1 l =1

n n
= V [∑∑ (bkj bkl − akj bkl − bkj akl + akj akl )2 − j −l ]
2

j =1 l =1
n n
=V 2
∑∑ (b b
j =1 l =1
kj kl − akj bkl − bkj akl + akj akl )2 − j −l

n
=V 2
∑ kj kj kj kj
(b
j =1
2
− 2 a b + a 2
) 2 −2 j

bkj2 = (+1) 2 P (+1Rx) + (−1) 2 P(−1Rx) = 1

akj2 = (+1) 2 P (+1Tx ) + (−1) 2 P(−1Tx ) = 1

11
 Effects of Noise
akj bkj = (+1)(+1) P(+1Tx, + 1Rx) + (−1)(−1) P(−1Tx, − 1Rx)
+ (−1)(+1) P(−1Tx, + 1Rx) + (+1)(−1) P(+1Tx, − 1Rx)
= [ P(+1Tx, + 1Rx) + ( P(−1Tx, − 1Rx)] − [ P(−1Tx, + 1Rx) + ( P(+1Tx, − 1Rx)]
= [1 − Pe ] − [ Pe ] = 1 − 2 Pe
n
e =V
2
b
2
∑ [1 − 2(1 − 2 P ) + 1]2
j =1
e
−2 j

n n 1
− ( 14 ) n +1 1 − ( 1 n
4)
= 4V 2 Pe ∑ 2 − 2 j = 4V 2 Pe ∑ ( 14 ) j = 4V 2 Pe 4
= V 2
P
1 − 14
e 3
j =1 j =1 4

4 2 4 1 4 1 4 M 2
−1
= V Pe [1 − ( 14 ) ] = V Pe [1 − n 2 ] = V Pe [1 − 2 ] = V Pe
n 2 2 2

3 3 (2 ) 3 M 3 M2

12
 Effects of Noise
V2 4 2 M 2 −1
n =e +e =
2
k
2
q + V Pe
2
b
3M 2 3 M2
„ The peak signal to average noise ratio for the analog output is

⎛S⎞ ( xk ) 2max V2 3M 2
⎜ ⎟ = = 2 =
⎝ ⎠ pk out
N nk 2 V / 3M 2
+ 4V 2
Pe ( M 2
− 1) / 3 M 2
1 + 4( M 2 − 1) Pe

„ If the input analog signal has a uniform PDF over [-V, +V], the
average signal to average noise ratio is
V 1
∫
∞
∫ v2 dv
2
⎛S⎞ x 2 v f (v) dv −V ( 2V ) 2
/ 12
⎜ ⎟ = 2 = k −∞
= 2V =
⎝ N ⎠ out nk nk2 nk2 nk2
1V2 1⎛ S ⎞ M2
= = ⎜ ⎟ =
3 nk 3 ⎝ N ⎠ pk out 1 + 4( M 2 − 1) Pe
2

13
Effects of Noise

14
 Effects of Noise
„ In many practical systems Pe is negligible. If we assume that Pe
= 0, then, the peak SNR resulting from only quantizing errors is
⎛S⎞
⎜ ⎟ ( )
= 3M 2 = 3 ⋅ 2 n
2

⎝ N ⎠ pk out
„ And the average SNR due only to quantizing error is
⎛S⎞
⎜ ⎟ =M = 2
2 n
( ) 2

⎝ N ⎠ out
„ We may express SNR in decibels as

⎛S⎞
⎜ ⎟ = 10 log⎜ ⎟ = ⎨
[ 2
( )]
⎛ S ⎞ ⎧⎪10 log 3 ⋅ 2 n ≈ 6.02n + 4.77 ⎫⎪
⎬ = 6.02n + α
⎝ ⎠ dB
N ⎝ ⎠ ⎪⎩
N ( )
n 2
10 log 2 ≈ 6.02n ⎪⎭

„ This equation – called the 6dB rule – points out that an

additional 6-dB improvement in SNR is obtained for each bit
added to the PCM word.

15
 Example
„ EAssume that an analog voice-frequency (VF) signal, which
occupies a band from 300 to 3400 Hz, is to be transmitted over a
binary PCM system. The minimum sampling frequency would be
6.8 kHz. In practice the signal would be oversampled, and a
sampling frequency of 8 kHz is the standard used for VF signals in
telephone communication systems. Assume that each sample value
is represented by 8 bits, then the bit rate of the PCM signal is
R = nf s = (8 bits/sample) × (8k samples/s) = 64 kbits/s
„ If rectangular pulse shaping is used, the absolute bandwidth is
infinity and the first null bandwidth is
1
Bnull = R = = 64 kHz
Tb
„ The peak SNR is

⎛S⎞
⎜ ⎟ ( )
= 3M 2 = 3 2 n
2
( )
=32 8 2
= 196608 ≈ 52.94 dB
⎝ N ⎠ peak out
16
 Nonuniform Quantizing
• Voice analog signals are more likely to have amplitude values
near zero than at the extreme peak values allowed. For signals
with nonuniform amplitude distribution, the granular quantizing
noise will be a serious problem if the step size is not reduced for
amplitude values near zero and increased for extremely large
values. This is called nonuniform quantizing since a variable step
size is used.

17
 μ Law and A-Law
„ In the US, Canada, and Japan, a law type of compression
characteristic is used that is defined by

ln(1 + μ w1 (t ) )
w2 (t ) =
ln(1 + μ )
μtyp = 255

„ Another compression law, used mainly in Europe, is the A-law

characteristic, defined by
⎧ A w1 (t ) 1
⎪⎪ , 0 ≤ w (t ) ≤
+
1
w2 (t ) = ⎨ 1 ln A A
⎪1 + ln( A w1 (t ) ) , 1 ≤ w (t ) ≤ 1
⎪⎩ 1 + ln A A
1

Atyp = 87.6
18
 Piecewise Linear Chords
„ In practice, the smooth nonlinear characteristics are
approximated by piecewise linear chords as shown below

19
 Signal-to-Noise Ratio
„ When compression is used at the transmitter, expansion (i.e.,
decompression) must be used at the receiver output to restore
signal levels to their correct relative values. The expandor
characteristic is the inverse of the compression characteristic,
and the combination of a compressor and an expandor is called
a compandor.
„ The output SNR follows the 6-dB law.
⎛S⎞
⎜ ⎟ = 6.02n + α
⎝ N ⎠ dB
„ For uniform quantizing
2
⎛ 2 xrms ⎞
2
⎛S⎞ ⎛ Mxrms ⎞
2 2 2 n
x x x
⎜ ⎟ = 2 = 2 = 2 rms
= 3⎜ rms
⎟ = 3⎜⎜ ⎟⎟
⎝ N ⎠ out n ⎝ V ⎠
2
n V /(3M ) ⎝ V ⎠
⎛S⎞ ⎡ ⎛ 2 xrms ⎤
n
⎞
2

⎜ ⎟ = 10 log ⎢3⎜⎜ ⎟⎟ ⎥ = 20 log 2 n + 10 log 3 − 20 log(V / xrms )

⎝ N ⎠ dB ⎢⎣ ⎝ V ⎠ ⎥⎦
≈ 6.02n + 4.77 − 20 log(V / xrms )
20
 SNR
„ For μ-law companding
⎛S⎞
⎜ ⎟ = 6.02n + α , α ≈ 4.77 − 20 log[ln(1 + μ )]
⎝ N ⎠ dB

„ For A-law companding

⎛S⎞
⎜ ⎟ = 6.02n + α , α ≈ 4.77 − 20 log[1 + ln A)]
⎝ N ⎠ dB

„ Notice that the output SNR is a function of the input level for the
case of uniform quantizing, but is relatively insensitive to the
input level for companding. The ratio V/xrms is called the loading
factor.
(V / xrms )typ = 4

21
SNR Comparison

22
 Homework
„ LC 3-8, 3-11, 3-12, 3-16, 3-17

23