Professional Documents
Culture Documents
Digital Communication
System Concepts
Vijay K. G a r g 2.1 Digital Communication System .............................................................. 957
Department of Electrical and 2.2 Messages, Characters, and Symbols ......................................................... 957
Computer Engineering, 2.3 Sampling Process ................................................................................. 957
University of Illinois at Chicago,
Chicago, Illinois, USA 2.4 Miasing .............................................................................................. 959
2.5 Quantization ....................................................................................... 960
Yih-Chen Wang 2.6 Pulse Amplitude Modulation ................................................................. 960
Lucent Technologies, 2.7 Sources of Corruption .......................................................................... 961
Naperville, Illinois, USA 2.8 Voice Communication .......................................................................... 963
2.9 Encoding ............................................................................................ 964
2.9.1 EncodingSchemes
Digital
information II
Textual ~ II
informationl - [ Sampler Quantizer Coder WavefOrmencoder
~[Transmitter
I
I I
I I
I_ J
Channel
Binarydigits
Pulsewaveform
Digitalinformation
FIGURE 2.1 BlockDiagram of a Typical Digital Communication System
1 where g~(t) is the ideal sampled signal and where 8(t - nT~) is
the delta function positioned at time t = nTs.
A delta function is closely approximated by a rectangular
pulse of duration At and amplitude g ( n Ts) / A t; the smaller we
make At, the better will be the approximation:
-1 oo
FIGURE 2.2 Binary and Quaternary Systems gs(t) = f~ ~_, G(f - mr,), (2.2)
m = - ~x~
Ts(t)
g(t)
T
FIGURE 2.3 Sampling Process
sampling period T, = 1/2W, then the corresponding spec- width W Hz is called the Nyquist rate and 1/2 W sec is called
trum is given as: the Nyquist interval.
We discuss the sampling theorem by assuming that signal
g(t) is strictly band-limited. In practice, however, an infor-
C~(f) = g (~w)e j~- = fsG(f) + f~ G(f - mf~) mation-bearing signal is not strictly band-limited, with the
m--~, me0 result that some degree of under sampling is encountered.
(2.4) Consequently, some aliasing is produced by the sampling pro-
cess. Aliasing refers to the phenomenon of a high-frequency
Consider the following two conditions: component in the spectrum of the signal seemingly taking on
the identity of a lower frequency in the spectrum of its sampled
(1) G(f) = 0 for If[ > W.
version.
(2) fs = 2W.
We find from equation 2.4 by applying these conditions,
2.4 Aliasing
G(f)=2-~G~(f) - W < f < W. Figure 2.4 shows the part of the spectrum that is aliased due to
(2.5) under sampling. The aliased spectral components represent
.'. G(f) = ~ g e-(L~) - W < f < W. ambiguous data that can be retrieved only under special con-
n~--CXD ditions. In general, the ambiguity is not resolved and ambigu-
ous data appear in the frequency band between (fs -fro) and
Thus, if the sample value g(n/2W) of a signal g(t) is fro.
specified for all n, then the Fourier transform G(f) of the In Figure 2.5, we show a higher sampling rate Jj to eliminate
signal is uniquely determined by using the discrete-time Four- the aliasing by separating the spectral replicas.
ier transform of equation 2.5. Because g(t) is related to G(f) Figures 2.6 and 2.7 show two ways to eliminate aliasing
by the inverse Fourier transform, it follows that the signal g(t) using antialiasing filters. The analog signal is preflltered so
is itself uniquely determined by the sample values g(n/2 W) for that the new maximum frequency fm is less than or equal to
-cxD < n < c~. In other words, the sequence {g(n/2W)} has f J 2 . Thus, there are no aliasing components seen in Figure 2.6
all the information contained in g(t). because f~ > 2f'm. Eliminating aliasing terms prior to sampling
We state the sampling theorem for band-limited signals of is a good engineering practice. When the signal structure is
finite energy in two parts that apply to the transmitter and
receiver of a pulse modulation system, respectively.
lx~(ol
(1) A band-limited signal of finite energy with no fre-
quency components higher than W Hz is completely
described by specifying the values of signals at instants
of time separated by 1/2 W sec.
(2) A band-limited signal of finite energy with no fre-
quency components higher than W Hz may be com-
pletely recovered from a knowledge of its samples
taken at the rate of 2 W samples/sec.
This is also known as the uniform sampling theorem. The
sampling rate of 2 W samples per second for a signal band- FIGURE 2.4 SampledSignal Spectrum
960 Vijay K. Garg and Yih-Chen Wang
2.5 Quantization
,f
f'm fs-f'm t~
In Figure 2.8, each pulse is expressed as a level from a finite
number of predetermined levels; each such level can be repre-
FIGURE 2.6 Prefiltering to Eliminate Aliasing
sented by a symbol from a finite alphabet. The pulses in Figure
2.8 are called quantized samples. When the sample values are
quantized to a finite set, this format can interface with a digital
IX~(f)l
system. After quantization, the analog waveform can still be
recovered but not precisely; improved reconstruction fidelity
of the analog waveform can be achieved by increasing the
i
number of quantization levels (requiring increased system
1 bandwidth).
I,
fn,'
2.6 Pulse Amplitude Modulation
td2
There are two operations involved in the generation of the
FIGURE 2.7 Postfiltering to Eliminate Aliasing Portion of the
pulse amplitude modulation (PAM) signal:
Spectrum
(1) Instantaneous sampling of the message signal m(t)
well known, the aliased terms can be eliminated after sampling every Ts sec, where fs = 1/Ts is selected according to
with a linear pass filter (LPF) operating on the sampled data. the sampling theorem
In this case, the aliased components are removed by postfilter- (2) Lengthening the duration of each sample obtained to
ing after sampling. The filter cutoff frequency f'm removes the some constant value T
aliased components; i'm needs to be less than (f~-fm). It
should be noted that filtering techniques for eliminating the
aliased portion of the spectrum will result in a loss of some g(t)
signal information. For this reason, the sample rate, cutoff
bandwidth, and filter type selected for a particular signal
bandwidth are all interrelated.
Realizable filters require a nonzero bandwidth for the tran-
sition between the passband and the required out-of-band
attenuation. This is called the transition bandwidth. To min- t
imize the system sample rate, we desire that the antialiasing -*IT
filter has a small transition bandwidth. Filter complexity and
cost rise sharply with narrower transition bandwidth, so a FIGURE 2.8 Flattop Quantization
2 Digital Communication System Concepts 961
h(t)
1.0 f)l
ctrummagnitude
These two operations are jointly referred to as sample and (1) Assuming a sampling rate of 8 kHz, calculate the
hold. One important reason for intentionally lengthening the spacing between successive pulses of the multiplexed
duration of each sample is to avoid the use of an excessive signal.
channel bandwidth because bandwidth is inversely propor- (2) Repeat your calculations using Nyquist rate sampling.
tional to pulse duration.
The Fourier transform of the rectangular pulse h(t) is given 106
as (see Figure 2.9): T~ -- - - -- 125 Ixs.
8000
H(f) = Tsinc(fT)e -j2"~fr. (2.6) For 25 channels (24 voice channels +1 sync), time allocated for
each channel is 125/25 = 5 b~s. Since the pulse duration is
We observe that by using flattop samples to generate a PAM 1 izs, the time between pulses is (5 - 1) = 4 b~s.
signal, we introduce amplitude distortion as well as a delay of The Nyquist rate is 7.48 Hz (2.2 x 3.4).
T/2. This effect is similar to the variation in transmission In addition:
frequency that is caused by the finite size of the scanning
aperture in television. The distortion caused by the use of 106
Ts -- -- 134 p~s.
PAM to transmit an analog signal is called the aperture affect. 7480
This distortion may be corrected by using an equalizer
134
(see Figure 2.10). The equalizer has the effect of decreasing Tc -- -- 5.36 Izs.
25
the in-band loss of the filter as the frequency increases in such a
manner to compensate for the aperture effect. For T/Ts < O.1,
The time between pulses is 4.36 Vs.
the amplitude distortion is less than 0.5%, in which case the
need of equalization may be omitted altogether.
PAMsignal ]
s(t) IlL Filter(LPF)I~ [
EqualizerJ . Messagesignalm(t)
FIGURE 2.10 An Equalizer Application
962 Vijay iv(. Garg and Yih-Chen Wang
encoding the PAlVl waveform into a quantized wave- symbol duration and cause signal pulses to overlap. This
form involves discarding some of the original analog overlapping is called inter-symbol interference ISI), ISI
information. This distortion is called quantization causes system degradation (higher error rates); it is a
noise; the amount of such noise is inversely propor- particularly insidious form of interference because raising
tional to the number of levels used in the quantization the signal power to overcome interference will not im-
process. prove the error performance:
• Quantizer saturation: The quantizer allocates L levels to
the task of approximating the continuous range of inputs q/2 q/2
with a finite set of outputs (see Figure 2.11). The range of 0-2= Ie2p(e)de= Ie21de=q2
inputs for which the difference between the input and q 12
output is small is called the operating range of the
-q/2 -q/2
converter. If the input exceeds this range, the difference = average quantization noise power.
between the input and output becomes large, and we say
that the converter is operating in saturation. Saturation
errors are more objectionable than quantizing noise. Gen-
Vd z (V_~)2 (~)2_ L2q2~ (2.7)
erally, saturation is avoided by use of automatic gain
control (AGC), which effectively extends the operating
range of the converter. ~ q q2/12 -- 3L2. (2.8)
• Timing jitter: If there is a slight jitter in the position of
the sample, the sampling is no longer uniform. The effect In the limit as L ---, oc, the signal approaches the PAM format
of the jitter is equivalent to frequency modulation (FM) (with no quantization error) and signal-to-quantization noise
of the baseband signal. If the jitter is random, a low-level ratio is infinite. In other words, with an infinite number of
wideband spectral contribution is induced whose proper- quantization levels, there is zero quantization error.
ties are very close to those of the quantizing noise. Timing Typically L = 2R, R = Log2L, and q = G- 2vp = (2Vp)/2 R.
jitter can be controlled with very good power supply
isolation and stable clock reference.
• Channel noise: Thermal noise, interference from other 1"2Vp\ 2 ---=_1V2 2_2R
users, and interference from circuit switching transients
can cause errors in detecting the pulses carrying the
Let P denote the average power of the message signal re(t),
digitized samples. Channel-induced errors can degrade
and then:
the reconstructed signal quality quite quickly. The rapid
degradation of the output signal quality with channel-
induced errors is called a threshold effect. (SNR)o -- ~ -- 22R. (2.9)
° Intersymbol interference: The channel is always band- \ P/
limited. A band-limited channel spreads a pulse wave-
form passing through it. When the channel bandwidth is The output SNR of the quantizer increases exponentially
much greater than pulse bandwidth, the spreading of the with increasing number of bits per sample, R. An increase
pulse will be slight. When the channel bandwidth is close in R requires a proportionate increase in the channel band-
to the signal bandwidth, the spreading will exceed a width.
Example 5
We consider a full-load sinusodial modulating signal of ampli-
tude A that uses all representation levels provided. The average
signal power is (assuming a load of 1 fl):
L levels
1 The equations are written and solved as follows:
P ~--.
A2
2
(SNR) o
3
m
2
A
(I/3A22_2R)
2 _~
(22R) = 1.8 + 6R dB.
FIGURE 2.11 Uniform Quantization
2 Digital Communication System Concepts 963
L R [bits] SNR [decibels] In equation 2.10, IX is constant, x and y are the input
32 5 31.8 and output voltages, ~ = 0 represents uniform quantization,
64 6 37.8 and ix = 255 is the standard value used in North America.
128 7 43.8
• A-Law, used in Europe, is as follows:
256 8 49.8
y _ A(lx[/Xm~)sgnx, 0 < Ixl < 1 (2.11a)
yma× 1 + In A Xma~- A
2.8 V o i c e C o m m u n i c a t i o n 1 + ln(lx[/Xmax) 1 Ixl
< 1. (2.11b)
= 1 + in A sgnx, A < Xma~
For most voice c o m m u n i c a t i o n , very low speech volumes
predominate; about 50% of the time, the voltage characteriz- The A is the positive constant, and A = 87.6 is the standard
ing detected speech energy is less than 1/4 of the root-mean- value used in Europe.
square (rms) value. Large amplitude values are relatively rare;
only 15% of the time does the voltage exceed the rms value. Example 6
The quantization noise depends on the step size. When the The information in an analog waveform with maximum fre-
steps are uniform in size, the quantization is called the uniform quency fm= 3 kHz is transmitted over an M-ary PCM system,
q u a n t i z a t i o n . Such a system would be wasteful for speech where the number of pulse levels is M = 32. The quantization
signals; many of the quantizing steps would rarely be used. In distortion is specified not to exceed 4- 1% of the peak-to-peak
a system that uses equally spaced quantization levels, the analog signal.
quantization noise is same for all signal magnitudes. Thus, (1) What is minimum number of bits/sample or bits/
with uniform quantization, the signal-to-noise ratio (SNR)
PCM word that should be used?
is worse for low-level signals than for high-level signals. Non- (2) What is minimum sampling rate, and what is the
uniform q u a n t i z a t i o n can provide fine quantization of the
resulting transmission rate?
weak signals and coarse quantization of the strong signals.
(3) What is the PCM pulse or symbol transmission rate?
Thus, in the case of nonuniform quantization, quantization
noise can be made proportional to signal size. Improving the Solutions:
overall SNR by reducing the noise for predominant weak
signals, at the expense of an increase in noise, can be done for lel _<pyre,
rarely occurring signals. The nonuniform quantization can
be used to make the SNR a constant for all signals within the where p is fraction of the peak-to-peak analog voltage•
input range. For voice, the signal dynamic range is 40 dB.
N o n u n i f o r m q u a n t i z a t i o n is achieved by first distorting the • le axl- Vpp
original signal with logarithmic compression characteristics 2L '
and then using a uniform quantizer. For small magnitude
signals, the compression characteristics have a much steeper •
\2t
< pv,,).
slope than the slope for large magnitude signals. Thus, a given -
N" I
I
H I
I
+v
-V