Digital Communication

Lecture 5
Sampling Theorem and PAM

Electronics and Communication Engineering

 Pulse modulation
Once the information source has been sampled there are 4 methods commonly
used for transmitting the amplitude of the sampled value. All these methods
utilizes modulating the carrier with a pulse whose parameters are made to vary.

1. Pulse Amplitude Modulation (PAM), whereby the height of a pulse is made

proportional to the sampled value.
Pulse Time Modulation (PTM)
2. Pulse Duration / Width Modulation (PDM/PWM), whereby the width of a pulse
is varied as a function of the sampled value.
3. Pulse Position Modulation (PPM), whereby the position of a pulse is changed as
a function of the sampled value with reference to a standard position/
clocking signal.
4. Pulse Code Modulation (PCM), whereby the sampled value is first coded into a
digital code and the code group then transmitted.

Another type – DM, ADM, etc.

In case of PAM, PWM and PPM, as the signal is not digitized/ digital ,
they are called analog pulse modulation.

ECE 2
 PAM

Either instantaneous sampling (Flat-top) or

natural sampling can be used.

• The flat-top is more useful to conversion to PCM.

• The naturally sampled type is easier to generate and is
used in other application.

ECE 3
 Sampling theorem
Consider an analog signal g(t)
that is continuous in both time
and amplitude and having
infinite duration and finite
energy. It is as in Fig 5.1(a).

Sample values at times t = 0,

Ts, 2Ts,…is denoted by the
series
 g  nT  , n  0,  1,  2,.......
s

Ts = Sampling period
fs = Sampling rate
= 1/ Ts

Fig 5.1: Illustration of the ideal Sampling process.

(a) Analog signal (b) Discrete time signal
ECE 4
 Sampling theorem (2)
The discrete time signal g(t) , that results from sampling process as


g  t    g  n T s    t  n T s  (5.1)
n   See Fig. 5.1 (b)
Where   t  n T s  = Dirac delta function at time t= nTs

From the definition of a delta function

g  t  δ  t  n T s   g  t  δ  t  n T s 


From (5.1) g t  gt    t  nT s
n  
 g  t  t (5.2)
Ts

Where,  is t Dirac

 comb or ideal sampling function .
Ts
ECE 5
Sampling theorem (3)

According to (5.2), g(t) is the Modulating Impulse

Instantaneously
output of an impulse modulator, wave g(t) modulator
sampled wave
which operates with g(t) as the
modulating wave and Ts(t) as g(t)

the carrier wave .

This circuit theoretic

Carrier
interpretation of g(t)
wave Ts(t)
is depicted in Fig 5.2.

Fig 5.2: circuit theoretic interpretation of the

ideal sampling process as impulse modulation.

ECE 6
Sampling theorem (4)
Let G(f) and G(f) denote the Fourier transforms of g(t) and g (t). Fourier transforms
of Ts (t) -
 
F  T t  f s
s


  f  mf s  (5.3)
m
Where, F[.] = Fourier transforms operation

Transforming (5.2) into frequency domain,

(5.4)

G  f   G ( f )   f s


  f  mf s  

 m    
Where, * = Convolution

Interchanging the order

G  f   f s

 G ( f )   f  mf 
s
 (5.5)
m

 fs

 
G f  mf
s
 (5.6)
m
ECE 7
Sampling theorem (5)

From (5.6), G(f) is a spectrum that is periodic in frequency f with period fs,
but not necessarily continuous. In other words, the process of uniformly
sampling a signal in the time domain results in a periodic spectrum in
frequency domain with a period equal to sampling rate. Thus G(f) is a
periodic extension of the original spectrum G(f).

If
 
F   t  n T s   exp  j 2 nf T  s

By taking the Fourier transform of both sides of (5.1),


G  f    g  n T  exp 
n 
s
j 2 nf T  s
5.7

ECE 8
Sampling theorem (6)
Suppose, g(t) is strictly band limited, with no frequency components higher than W
hertz. So, G(f) = 0 for f  W as illustrated in Fig 5.3a.

H(f) -fs
(a) (b)
1/2W

-W W

(c)
Fig 5.3: (a )Spectrum signal of g(t). (b)Spectrum of sampled signal g(t) for a
sampling rate fs = 2W. (c) Ideal amplitude response of reconstruction filter.

ECE 9
Sampling theorem (7)
Putting Ts = 1/2W in (5.7) ,

 n   j nf 
G  f    g   exp   (5.8)
n   2W   W 
Putting fs = 2W in (5.6) ,
1
G f   G  f   W  f W (5.9)
2W
From (5.8) ,
1 
 n   jnf 
G f    g   exp   W  f W (5.10)
2W n   2W   W 

If the sampled values g(n/2W) are specified, then G(f) is uniquely

determined by using (5.10). As g(t ) is related to G(f) by inverse FT, g(t )
is uniquely determined by {g(n/2W)} for -.n . So, {g(n/2W)}
contains all the information of g(t ) .

ECE 10
Sampling theorem (8)
Reconstructing g(t):

Substituting (5.10) in formula of IFT


g  t    G  f  exp j 2ft  df


1 
 n   jnf 
  exp j 2ft  df
W
 g   exp
W 2W   2W   W 

 n  1 W  n 
 

g 
 2W
 W 
 2W
exp j 2f 

t 
2W
 df

(5.11)

 n  Sin  2Wt  n  (5.12)
  g 
  2W   2Wt  n 

ECE 11
 Sampling theorem (9)
sinc function can be defined as
Sin  x 
sin c x   (5.13)
x
The sinc function exhibits interpolatory property as
for x  0
sin c x   
1 (5.14)
0 for x   1,  2.......

Where, x is independent variable.

Rewriting (5.12),

 n  (5.15)
g t    g  sin c 2Wt  n 
  2W 

(5.15) provides an interpolation formula for reconstructing the original

g(t) with sinc(2Wt) playing the role of an interpolation function.

ECE 12
 Natural Sampling (Gating)

DEFINTION: If w(t) is an analog waveform bandlimited to B hertz,

the PAM signal that uses natural sampling (gating) is

ws(t) =w(t)s(t) Where

S(t) is a rectangular wave switching waveform and

fs = 1/Ts ≥ 2B.
(.) is single rectangular pulse.

S(t) may be represent by Fourier series

 jn  s t, c  the Fourier series coefficien ts of the switching waveform.
St   c e n
n
 sin  nd 
d
nd

d  Duty cycle 
Ts
ECE 13
 Natural Sampling (Gating) (2)

Something 
in Haykin c  t   f sTA  sin c nf sT  exp j 2nf s t 


Lathi S  t   co   2cn cos n s t 
m 1


Where, co  , cn  f s sin c nf s 
Ts
 s  2f s

So, W  t   W  t  S  t 
s 
sin( nd )

Ws ( f )  F[ws (t )]   cnW ( f  nf s )  d  W ( f  nf s )
n  n   nd

S  f   f sTA  sin c mf sT  G  f  mf s  (Haykin)
m
ECE 14
Natural Sampling (Gating) (3)

w(t)

s(t)

ws(t) =w(t)s(t)

ECE 15
Generating Natural Sampling
The PAM wave form with natural sampling can be generated
using a CMOS circuit consisting of a clock and analog switch
as shown.

ECE 16
Spectrum of Natural Sampling
 
sin( nd )
Ws ( f )  F[ws (t )]  
n 
cnW ( f  nf s )  d 
n   nd
W ( f  nf s )

• The duty cycle of the switching

waveform is d = τ/Ts = 1/3.
• The sampling rate is fs = 4B.


sin( nd )
Ws ( f )  d 
n   nd
W ( f  nf s )

sin( nd )
d
 nd

17
ECE
Recovering Naturally Sampled PAM

 At the receiver, the original analog waveform, w(t), can be recovered

from the PAM signal, ws(t), by passing the PAM signal through a low-
pass filter where the cutoff frequency is: B <fcutoff < fs -B

 If the analog signal is under sampled fs < 2B, the effect of spectral
overlapping is called Aliasing. This results in a recovered analog
signal that is distorted compared to the original waveform.

LPF Filter
B <fcutoff < fs -B

ECE 18
Demodulation of PAM Signal
 The analog waveform may be recovered from the
PAM signal by using product detection,

ECE 19
Instantaneous Sampling (Flat-Top PAM)(2)
 DEFINITION: If w(t) is an analog waveform bandlimited to B Hertz, the instantaneous
sampled PAM signal is given by
 
 

ws (t )   w(kTs )h(t  kTs )  h(t )   w(kTs ) (t  kTs )  h(t )   w(t )   (t  kTs ) 
k  k   k  

Where h(t) denotes the sampling-pulse shape and, for flat-top sampling, the pulse shape is,

THEOREM: The spectrum for a flat-top PAM signal is:


1
Ws ( f )  H ( f )  W ( f  nf s )
Ts k 

 sin  f 
H ( f )    h(t )     
  f 

ECE 20
Instantaneous Sampling (Flat-Top PAM)(3)


S f   fs  G  f  mf s  H  f 
m
Haykin where
H  f   T sin c ft  exp  jft 

ECE 21
 The spectrum of the flat-top PAM
Analog signal maybe recovered from the flat-top PAM signal by the use of a LPF.
LPF Response
Note that the recovered signal
has some distortions due to the
curvature of the H(f).
Distortions can be removed by
using a LPF having a response
1/H(f).

ECE 22
 Self study:

Sampling and their frequency spectrum

L.W. Couch 130-131 p

134-135p

ECE 23
 Recovery of PAM: Simple low pass filter at the receiving end will
bypass the pulse rate frequency and fill in the areas between the
pulses sufficiently to restore the fidelity of the message signal.

Bandwidth requirements: The transmission of either naturally of

instantaneously sample PAM over a channel requires a very wide
frequency response because of narrow pulse width. The bandwidth
requirement is very larger than that of original analog signal.

P – 132/136(fig) OHP sheet

106

For each sample the frequency spectrum shows that the magnitude
exits in higher frequency due to harmonics which is required to
reconstruct the signal.
 is not very good for long distance transmission
 provides a means to other modulation.
 provides multiplexing (TDM)

ECE 24
 Pulse Width Modulation

ECE 25
Pulse Width Modulation(2)
 A square wave is generated and
applied to an integrator to
produce a triangle wave at a in
both figure.

 The message signal (b) is mixed

with triangle wave in a linear
mixer (adder).

 The triangle wave should have

a minimum frequency that is an
odd multiple of the highest
message frequency .

 The summed wave are applied

to a comparator.

Fig: 2.21
ECE 26
 Pulse Width Modulation(3)
Either instantaneous or natural
sampling can be used.

ECE 27
 Pulse Width Modulation(4)

ECE 28
Pulse Width Modulation(5)
Expression for PWM wave: Assume information source have

Maximum frequency=fm.
Maximum permissible pulse width = 1/2fm = 
m information source multiplexed, then 1/2fm = 
any given information source , em(t),
T= Sampling period for 1 channel = 1/2fm
k= Arbitrary constant.

pulse width will vary as {1+em(t)}.

t 
T
2m

1  ke  t 
m


If +1 em(t)  -1, T

Minimum pulse width = 1  k 
2m
T
Maximum pulse width= 1  k 
2m
ECE 29
Pulse Width Modulation(6)

Pulse wave may be s  t   c  cos n s t 
o 2c
n 1
n

where,

co 
Ts

cn  sin c nf s 
Ts
 s  2f s

T 
T
s t    2 sin c nf s T  cos n s t 
Ts n 1 Ts

ECE 30
Pulse Width Modulation(7)
Substituting the value of t
1 k 
T
s t    em  t   2 sin c nf s T  cos n s t 
2 m 2m n 1 T

 The 1st term is a DC or average component.

 The 2nd term is the information having frequency spectrum equal to that
of information source em(t).
 The 3rd term of Bessel's function to yield frequency component at higher
frequencies.

Thus the information may be recovered by passing the PDM wave through a
low pass filter having a bandwidth equal to that of em(t).

ECE 31