Reference: "Digital Communications" - John G. Proakis
Reference: "Digital Communications" - John G. Proakis
Reference: "Digital Communications" - John G. Proakis
Baseband digital signals - signals whose spectrum extend down to or near zero
frequency.
Noise
Transmitter Channel Receiver
The first term aih(0) represents the contribution of the i th transmitted pulse. The second
term represents the residual effect of all other transmitted pulses on the decoding of the i
th bit.
The residual effect due to the occurrence of pulses before and after sampling instant iT is
called intersymbol interference (ISI).
The presence of noise and ISI degrades the receiver performance and the objective is to
minimize these effects and hence deliver the data to its destination with the smalleset
error rate possible.
When the signal to noise ratio is high (telephone system PSTN) the operation of the
system is largely limited by ISI rather than noise.
Thus first we neglect the effect of noise and focus attention on ISI.
76
Nyquist’s Criterion for Distortionless Baseband Binary
Transmission
* Typically, the transfer function of the channel and the transmitted pulse shape are
specified.
Objective: To determine the transfer functions of the transmit and receive filters to
reconstruct the original data sequence.
1, i k
h(iT - kT) = 0, i k
(**)
1, n 0
h(nT) = 0, n 0
(***)
y(iT) = ai
Therefore condition given in (**) ensures “perfect reception in the absence of noise”.
If h(t) is sampled at a rate of 1/T, i.e., h(t) multiplied by a periodic impulse train with
period T,
i(t) = (t nT )
n
hs(t) = h(t) i(t) = h(t) (t nT ) = h(nT ) (t nT )
n n
Hs(f) = h(nT )e j2nT = 1 since h(0)=1, h (n0 T) = 0
n
or
77
hs(t) = h(0) (t) = (t)
Hs(f) = F{(t)} = 1
1
I(f) = ( f nf s )
T n
1
Hs(f) = H(f) * I(f) = H(f) * ( f nf s )
T n
1
= H( f nf s )
T n
1
H( f nf s ) = 1 or (fs = 1/T)
T n
H( f n / T ) = T
n
---> Nyquist’s Criterion for distortionless baseband transmission in the absence of noise.
* Important to remember that H(f) refers to the overall system, incorporating the transmit
filter, the channel, and the receive filter.
Theorem (Nyquist)
78
The necessary and sufficient condition for h(t) to satisfy
1, n 0
h(nT) = 0, n 0
H( f n / T ) = T
n
We have three cases. Let’s define B(f) = H( f n / T)
n
( Insight: Since obviously B(f) is periodic it can be expanded as a Fourier series, i.e.,
B(f) = bn e j2nfT and the condition says b0 = T and bn = 0 for n 0.
n
Also note that
Hs(f) = B(f)/T = 1/T bn e j 2nfT = h(nT )e j 2nT This implies bn = T h(-nT)
n n
79
1. When T < 1/2W or 1/T > 2W , B(f) consists of non-overlapping replicas of H(f)
separated by 1/T as shown below.
B(f)
There is no choice for H(f) to ensure B(f) T in this case and hence there is no way that
we can design a system with no ISI.
2. When T = 1/2W or 1/T=2W (the Nyquist rate), the replication of H(f) separated by 1/T
are as shown below.
B(f)
-1/T+W = -W 0 W = 1/T - W
It is clear that in this case there exists only one H(f) that results in B(f) = T, namely
T , (| f |W )
H(f) = 0, (otherwise)
H(f)
-W 0 W
80
which corresponds to the pulse
sin(t / T ) t
h(t) = sin c
t / T T
This means that the smallest value of T for which transmission with zero ISI is possible
is T=1/2W, and for this value h(t) has to be a sinc function.
-4 -3 -2 -1 0 1 2 3 4 t/T
t < 0, sinc[(t - t0)/T] 0. The sampling time should also be shifted to “nT + t0”.
* Another difficulty is that rate of convergence to zero is slow. The tails of h(t) decay as
1/t, thus a small timing “jitter” at the receiver or mistiming error in sampling could
produce an infinite series of ISI components.
Another thing is that the amplitude characteristic of H(f) should be flat from -W to W
and zero elsewhere. This is physically unrealizable because of the abrupt transitions at
the band edges W.
Therefore in practice trying to approach (sin x)/ x pulse at the receiver would usually be
counter productive.
81
* Example of ideal Nyquist signalling - Notice zero ISI at sampling points.
1 0 1 1 0 1
Case No 3: When T > 1/2W, B(f) consists of overlapping replications of H(f) separated
by 1/T as shown below. In this case there exist numerous choices for H(f) T.
B(f)
-W -1/T+W 0 1/T - W W
*The practical difficulties encountered with the ideal Nyquist channel can be overcome
by reducing the signaling rate from the maximum value 2W to an adjustable value
between W and 2W.
A particular pulse spectrum, for the T > 1/2W case, that has desirable spectral properties
and has been widely used in practice is the raised cosine spectrum.
82
The raised cosine frequency characteristic consists of a flat portion and a rolloff portion
that has a sinusoidal form as follows,
1
T 0 | f |
2T
H( f ) T T 1 1 1
1 cos | f | | f |
2 2T 2T 2T
1
0 | f |
2T
is called the rolloff factor and takes the values in the range 0 1.
The bandwidth occupied by the signal beyond the Nyquist frequency 1/2T is called the
excess bandwidth and is usually expressed as a percentage of the Nyquist frequency(
excess bw = 50% or 50% rolloff).
cos(t / T )
h(t )sin c(t / T )
1 4 2t 2 / T 2
The following figures illustrates the raised cosine spectral characteristics and the
corresponding pulses for = 0,0.5 and 1.0. The tails of h(t) decay as 1/t 3 for > 0. Thus
a mistiming error in sampling leads to a series of ISI terms that converges to a finite
value.
83
1
T=0.25 h(t)
0.8
0.6
0.4
=0 =1
0.2
T 2T 3T
0
-0.2
=0.5
-0.4
-0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8
0.25
T=0.25 H(f)
0.2 =0
=0.5
0.15
0.1
=1.0
0.05
0
-4 -3 -2 -1 0 1 2 3 4
84
Note that for the pulse reduces to h(t) = sinc(t/T) and the symbol rate 1/T = 2W.
When the symbol rate is reduced to 1/T = W.
(i.e. rate reduces by W for this value of or we can say if we keep the original signaling
rate (2W) to have no ISI we need a bandwidth of 2W which we considered to be fixed at
W in the foregoing analysis)
H(f) = GT(f)GR(f)
If the receiver filter GR(f) is matched to the transmitter filter GT(f) then
H(f) = | GT(f) |2
I.e., ideally
GT(f) = | H ( f )|e j 2ft0 and GR(f) = G*T(f) “square root raised cosine filter”
where t0 is some delay required to ensure the physical reliability of the filter. Note also
that an additional delay is necessary to ensure the physical realizability of the receiving
filter as well.
* Overall raised cosine spectral characteristic is split evenly between the transmitting
filter and the receiving filter.
85
Design of Bandlimited Signals with Controlled ISI - Partial-Response (PR)
Signaling or Correlative-level Coding
* To achieve zero ISI signaling rate has to be reduced from the Nyquist rate of 2W
symbols/s to realize practical transmitting and receiving filters.
* Here the objective is to maintain Nyquist rate of signaling allowing for a ‘controlled’
amount of ISI.
This is equivalent to
As we have discussed before it is impossible to satisfy this condition for T < 1/2W.
However for T=1/2W this is satisfied if
This pulse is called duobinary signal pulse where ‘duo’ implies doubling of the
transmission capacity of a binary system.
86
This can be represented in the following manner.
ak c Ideal Nyquist
k Channel H (f)
+
N
Delay
Filter
ck = ak + ak-1 ; ck { 0, 2}
This will change the two level uncorrelated ak sequence into a sequence of correlated
three level pulses ck. This correlation between the adjacent pulses may be viewed as
introducing ISI into the transmitted signal in an artificial manner.
However, the ISI introduced is under designer’s control, which is the basis of
“correlative coding”.
Hd(f) = 1 + e-j2fT
1
(1 e jf / W ) | f |W
2W
H(f) = T + T e-j2fT =
0 otherwise
same as before.
87
The spectrum is given below along with the pulse shape. Note that the spectrum decays
to zero smoothly which means that physically realizable filters can be designed that
approximate this spectrum very closely.
Another special case that leads to physically realizable transmitting and receiving filters
is specified by the samples
1 (n 1)
h(nT) = 1 (n 1)
0
(otherwise)
(t T ) (t T )
h(t) = sin c sin c T
T
1 f
(e jf /W e jf /W ) sin
j
| f | W
H(f) = 2W W W
0 | f | W
This pulse and its magnitude are illustrated in the figures below. It is called a “modified
duobinary signal pulse”. The spectrum has a zero or a “spectral null” at f=0 (d.c.),
making it suitable for transmission over a channel that does not pass d.c.
|H(f)|
W
f
-W
88
Generalized form of Correlative-Level Coding
(Partial-Response Signaling)
One can obtain other filter characteristics by selecting different values for the sample
h(nT) and more than two non zero samples. However as we select more samples getting
rid of controlled ISI becomes more complicated.
2
h(t) = h(nT ) sin c 2W t
,
T= 1/2W
n 2W
Corresponding spectra
1
h(nT )e jnf /W | f | W
H(f) = 2W n
0 | f | W
T T T
h0 h1 hN-1
Ideal filter
89