PAM with natural sampling

W(t) Ws(t)

t t

S(t) Analog bilateral switch

Ts
 Ws(t)
W(t)
=W(t)S(t)
t

Duty Cycle D=/Ts=1/3 S(t)

 Spectrum of PAM with natural sampling
|W(f)|
• Spectrum of input analog signal
• Spectrum of PAM
1
– D=1/3, fs=4B
f
– BT= 3fs = 12B -B B

s
i
nf


n
D
 
 f
Wf
( n
f
s)
|Ws(f)|

D=1/3 sin f
D
 f

-3fs -2fs -fs -B B fs 2fs 3fs

 PAM with flat-top sampling
W(t) Ws(t)

t t

S(t) Ts

Sample and Hold

t
Spectrum of PAM with flat-top sampling
• Spectrum of Input |W(f)|

• Spectrum of PAM
– /Ts=1/3, fs=4B 1
f
– BT= 3fs = 12B
-B B


1
|Ws(f)| H f
() W (f nfs)
T
s 
n

D=1/3  sin f
Ts  f

-3fs -2fs -fs -B B fs 2fs 3fs

 m(t)

Information signal
t

s(t)
Pulse signal 
t
Ts
Sampled signal (PAM)
ms(t)  ms(t)
   Ts 
t t
Ts Ts

Natural Sampling Flat-top Sampling

 Sampling

Let g (t ) denote the ideal sampled signal


g ( t )   g (nT )  (t  nT )
n  
s s (3.1)

where Ts : sampling period

f s  1 Ts : sampling rate
 
g( t )   ( t  nT
n  
s )

1 m
G( f ) 
Ts

m  
(f  )
Ts

 
m  
f s G ( f  mf s )

g (t )  f s  G( f
m  
 mf s ) (3.2)

or we may apply Four ier Transform on (3.1) to obtain


G ( f )   g ( nT
n  
s ) exp(  j 2 nf T s ) (3.3)

or G  ( f )  f s G ( f )  f s  G( f
m  
 mf s ) (3.5)
m 0

If G ( f )  0 for f  W and T s  1
2W

n j n f
G ( f )  
n  
g(
2W
) exp( 
W
) (3.4)
With
1.G ( f )  0 for f W
2. f s  2W
we find from Equation (3.5) that
1
G( f )  G ( f ) ,  W  f  W (3.6)
2W
Substituting (3.4) into (3.6) we may rewrite G ( f ) as
1 
n jnf
G( f ) 
2W

n  
g(
2W
) exp( 
W
) ,  W  f  W (3.7)

n
g (t ) is uniquely determined by g ( ) for    n  
2W
 n 
or  g ( )  contains all information of g (t )
 2W 
  n 
To reconstruct g (t ) from  g ( )  , we may have
 2W 

g (t )   G ( f ) exp( j 2ft )df


W 1 
n j n f

W 2W

n  
g(
2W
) exp( 
W
) exp( j 2 f t )df

n 1 W n 
  g( ) W exp  j 2 f (t  2W )df (3.8)
n   2W 2W

n sin(2 Wt  n )
  g( )
n   2W 2 Wt  n

n
  g( ) sin c( 2Wt  n ) , -   t   (3.9)
n   2W
(3.9) is an interpolation formula of g (t )
Sampling Theorem
Sampling Theoremfor strictly bandband-limitted
for strictly - limited signals
signals
1.a signal which is limited to  W  f  W , can be completely
 n 
described by  g ( ) .
 2W 
 n 
2.The signal can be completely recovered from  g ( )
 2W 
Nyquist rate  2W
Nyquist interval  1
2W
When the signal is not band - limited (under sampling)
aliasing occurs .To avoid aliasing, we may limit the
signal bandwidth or have higher sampling rate.
Figure 3.3 (a) Spectrum of a signal. (b) Spectrum of an undersampled version
of the signal exhibiting the aliasing phenomenon.

6
Figure 3.4 (a) Anti-alias filtered spectrum of an information-bearing signal. (b)
Spectrum of instantaneously sampled version of the signal, assuming the use of a
sampling rate greater than the Nyquist rate. (c) Magnitude response of
reconstruction filter.
Pulse Amplitude Modulation
Flat-top pulse, sample &
hold
 
s (t )   m(nT ) h(t  nT )
n  
s s (3.10)

 1, 0 t  T
1
h (t )   , t  0, t  T (3.11)
2
 0, otherwise

The instantaneously sampled version of m(t ) is

m (t )   m(nT ) (t  nT )
n  
s s (3.12)

m (t )  h(t )   m ( )h(t   )d


 


 m(nT ) (  nT )h(t   )d
n  
s s

 
  m(nT )
n  
s

 (  nTs )h(t   )d (3.13)

Using the sifting property , we have


m (t )  h(t )   m(nT )h(t  nT )
n  
s s (3.14)
 PAM
Flat-Top Sampling
 Recovering the original message signal m(t) from PAM signal

Where the filter bandwidth is W

The filter output is f s M ( f ) H ( f ) . Note that the
Fourier transform of h(t ) is given by
H ( f )  T sinc( f T ) exp( j f T ) (3.19)
amplitude distortion delay  T
2
 aparture effect
Let the equalizer response is
1 1 f
  (3.20)
H ( f ) T sinc( f T ) sin( f T )
Ideally the original signal m(t ) can be recovered completely.
 3.4 Other Forms of Pulse Modulation
a. Pulse-duration modulation (PDM)
b. Pulse-position modulation (PPM)
PPM has a similar noise performance as FM.
 PAM
• PAM is a general signalling technique
whereby pulse amplitude is used to convey
the message
• For example, the PAM pulses could be the
sampled amplitude values of an analogue
signal
• We are interested in digital PAM, where the
pulse amplitudes are constrained to chosen
from a specific alphabet at the transmitter
 PAM Scheme
Modulator
 
x
s(
t)a
kt
(

k
kT

) x
()
t 

k
a
kh
T

t
( kT
)

ak Pulse Transmit
generator filter
HT(w) hT(t)
Symbol
clock Demodulator HC(w)
Channel


Recovered y
(
t
) 
a
h
k(
t 
kT
)v(
t) hC(t)

k
symbols
Data Receive
+
slicer filter
Noise N(w)
Recovered HR(w), hR(t)
clock
 PAM
• In binary PAM, each symbol ak takes only
two values, say {A1 and A2}
• In a multilevel, i.e., M-ary system, symbols
may take M values {A1, A2 ,... AM}
• Signalling period, T
• Each transmitted pulse is given by
ak hT (t  kT )
Where hT(t) is the time domain pulse shape
 PAM
• To generate the PAM output signal, we may
choose to represent the input to the transmit
filter hT(t) as a train of weighted impulse
functions 
x
st)
( a
k

k
t
(

kT
)

• Consequently, the filter output x(t) is a train of

pulses, each with the required shape hT(t)

x
(t)a
k

k
h
Tt
(

kT
)
 PAM
 
x
st)
( a
k

k
t
(

kT
) x
(t)

k
a
k

h
Tt
( kT
)

xs (t) x(t)
Transmit
Filter
hT (t)
• Filtering of impulse train in transmit filter
 PAM
• Clearly not a practical technique so
– Use a practical input pulse shape, then filter to
realise the desired output pulse shape
– Store a sampled pulse shape in a ROM and read out
through a D/A converter
• The transmitted signal x(t) passes through the
channel HC(w) and the receive filter HR(w).
• The overall frequency response is
H(w) = HT(w) HC(w) HR(w)
 PAM
• Hence the signal at the receiver filter output is

y
(
t
) 
a
h
k(
t

k

kT
)

v(
t)

Where h(t) is the inverse Fourier transform of H(w)

and v(t) is the noise signal at the receive filter
output
• Data detection is now performed by the Data
Slicer
 PAM- Data Detection
• Sampling y(t), usually at the optimum instant
t=nT+td when the pulse magnitude is the
greatest yields

yn  y(nT  td )  ak h((n  k)T  td )  vn
k 
Where vn=v(nT+td) is the sampled noise and td is the
time delay required for optimum sampling
• yn is then compared with threshold(s) to
determine the recovered data symbols
 PAM- Data Detection
TX data ‘1’ ‘0’ ‘0’ ‘1’ ‘0’
TX symbol, ak +A -A -A +A -A
T
Signal at data
slicer input, y(t)
0
td
Sample clock Ideal sample instants
at t = nT+td
Sampled signal, 0 Data Slicer decision
yn= y(nT+td) threshold = 0V

Detected data ‘1’ ‘0’ ‘0’ ‘1’ ‘0’

Pulse Amplitude Modulation (PAM)
• Amplitude of periodic pulse train is varied
with a sampled message signal m
Digital PAM: coded pulses of the sampled and
quantized message signal are transmitted (lectures
12 and 13)
Analog PAM:
m(t) periodic pulse train with
s(t) =period
p(t) m(t) Ts is
the carrier (below)
p(t)

t
T Ts T+Ts 2Ts 12 - 28
 Analog PAM
• Pulse
Transmitted
amplitude
signal
varied with amplitude of 
s(t )   m(T n) h(t  T n) s s
sampled message sample hold
n  

Sample message every Ts

h(t)
Holdissample
a rectangular pulse (T < T )
for T seconds s
of duration T units  1 for 0  t  T
Bandwidth  1/T 
h(t )  1 / 2 for t  0, t  T
 0 otherwise
s(t) 
m(0) m(t)
h(t) m(Ts) As T  0,
1 1
h(t )   (t )
t t T
T T Ts T+Ts 2Ts
12 - 29
 Analog PAM
• Transmitted signal 
• Equalization of
s(t )   m(T
n  
s n) h(t  Ts n) sample and hold

  m(T s n)   (t  Ts n) * h(t )  distortion added in
n  
 

   m(Ts n)  (t  Ts n) * h(t )
transmitter
 n   
H(f) causes amplitude
msampled(t)
distortion and delay of
T/2
• Fourier
S( f )  M ( f ) H( f )
transform sampled 1

1

 f

Equalize amplitude
 fs M( f  f s k) H ( f ) H ( f ) T sinc( f T ) sin( f T )
k  
distortion by post- T
H ( f )  T sinc( f T ) e  j 2  f T /2
 0.1
filtering with T s
 T sinc( f T ) e  j  fT
magnitude response 12 - 30
 Sampling
• In many applications it is useful to represent a signal
in terms of sample values taken at appropriatelly
spaced intervals.
• The signal can be reconstructed from the sampled
waveform by passing it through an ideal lowpass
filter.
• In order to ensure a faithful reconstruction, the
original signal must be sampled at an appropriate
rate as described in the sampling theorem.
 Sampling
• sampling theorem
– A real-valued band-limited signal having no
spectral components above a frequency of B Hz is
determined uniquely by its values at uniform
intervals spaced no greater than seconds  2B1 
 
apart.
 Sampling
• Consider a band-limited signal f(t) having no
spectral component above B Hz.
• Let each rectangular sampling pulse have unit

amplitudes, seconds in width and occuring
at interval of T seconds.
 Sampling

f(t) A/D fs(t)

conversion

Sampling
 Sampling
• Denote the sampled signal by fs(t) and the periodic
gate function as PT(t), we have
fs(t)= f(t) PT(t)

• The periodic signal PT(t) can be represented by the

Fourier Series as 
PT (t )   Pe
n 
n
jno t

2
where o 
T
 Sampling
• The sampled signal can, therefore, be represented
as:

f s (t)= f(t)  Pn e jno t

n 
• By taking the Fourier transform of both sides, we
have:
 
jno t 
F  fs (t) = F f(t)  Pn e 
 n  

  P F f(t)e 
n 
n
jno t
 Sampling
• By using the frequency translation property of the
Fourier transform, the spectral density of fs(t) can
be written as:

Fs ( )=  P F(  n )
n 
n o


=Po F( ) +  P F(  n )
n 
n o

n0
 Sampling
• The spectral density of fs(t) is exactly like that of f(t).
It repeats itself periodically in frequency every .
o of the original spectral density are
The replicas
weighted by the amplitude of the Fourier series
coefficients of the sampling waveform
 Steps in sampling a band-limited signal.

0
1

2o o o 2o

2o o o 2o

0
1
 Effects of changing the sampling rate.
• If T decreases,  increases
o and all replicas of
F( )moves farther apart.
• If T increases,  decreases and all replicas of
o
F( )moves closer. Soon a point is reached beyond which
a reduction in the sampling rate will result in overlap
between spectral densities. This point is reached when

2
 4 B
T
1
T h e re fo re , T = .
2B
 Sampling
• To avoid spectral overlap:

1
T< .
2B

• Nyquist sampling rate.

 Sampling
• The following are the limitations on the use of
the full potentials of the sampling theorem:
– No ideal low-pass filter.
– No strictly band-limited signals.

• ALIASING
Aliasing in Frequency Domain
X   j 
1

s x 0  x s
X   j 
1

s x 0  x s
X   j 
1

s x 0  x s
 Impulse Sampling
• With an impulse sampler, the switching
function is a train of impulse functions:
• x(t) = n=- (t – nT)
Analog signal x(t) xs(t) Sampled signal

Switching function

T
 Impulse Sampling
• The impulse sampled waveform is
• xs(t) = x(t) x(t)
• = n=- x(t) (t – nT)
• = n=- x(nT) (t – nT)
• where x(nT) are the instantaneous sample
values selected by the impulse sampler at the
times nT.
 Impulse Sampling
Signal waveform Sampled waveform

0
0
1 201
1 201

Impulse sampler

0
1 201
 Impulse Sampling
with increasing sampling time T
Sampled waveform Sampled waveform

0 0
1 201 1 201

Sampled waveform Sampled waveform

0 0
1 201 1 201
• The Fourier transform of an impulse train in time
• x(t) = n=- (t – nT)
• is another impulse train in frequency
• X(f) = (1/T) n=- (f – n/T) = fs n=- (f – n fs)

• Fourier transform of the impulse sampled waveform is the

convolution
• Xs(f) = X (f) * X(f)
• = X (f) * fs n=- (f – n fs)
• = fs n=- X(f – n fs)
 Natural sampling
(Sampling with rectangular waveform)

• Consider a band-limited signal x(t) having no

spectral component above B Hz.
• Let each rectangular sampling pulse have
amplitude A, be  seconds in width and
occurring at interval of T seconds.
Analog signal x(t) xs(t) Sampled signal

Switching function

A

T
 Natural sampling
(Sampling with rectangular waveform)
Signal waveform Sampled waveform

0
0 1 201 401 601 801 1001 1201 1401 1601 1801 2001
1 201 401 601 801 1001 1201 1401 1601 1801 2001

Natural sampler

0
1 201 401 601 801 1001 1201 1401 1601 1801 2001
 Format analog signals
• To transform an analog waveform into a
form that is compatible with a digital
communication system, the following steps
are taken:
1. Sampling
2. Quantization and encoding
3. Baseband transmission

Lecture 2 51
 Sampling

Time domain Frequency domain

xs (t )  x (t )  x(t ) X s ( f )  X ( f )  X ( f )
x (t )
| X(f )|

x (t ) | X ( f ) |

xs (t )
| Xs( f )|

Lecture 2 52
Aliasing effect

LP filter

Nyquist rate

aliasing

Lecture 2 53
 Sampling theorem

Analog Sampling Pulse amplitude

signal process modulated (PAM) signal
• Sampling theorem: A bandlimited signal
with no spectral components beyond , can
be uniquely determined by values sampled
at uniform intervals of

– The sampling rate, is called

Nyquist rate.
Lecture 2 54
 Quantization
• Amplitude quantizing: Mapping samples of a continuous
amplitude waveform to a finite set of amplitudes.
Out

In
Average quantization noise power
Quantized

Signal peak power

values

Signal power to average

quantization noise power

Lecture 2 55
 Encoding (PCM)

• A uniform linear quantizer is called Pulse Code Modulation

(PCM).
• Pulse code modulation (PCM): Encoding the quantized signals
into a digital word (PCM word or codeword).
– Each quantized sample is digitally encoded into an l bits codeword
where L in the number of quantization levels and

Lecture 2 56
 Quantization example
amplitude
x(t)
111 3.1867

110 2.2762 Quant. levels

101 1.3657

100 0.4552

011 -0.4552 boundaries

010 -1.3657

001 -2.2762 x(nTs): sampled values

xq(nTs): quantized values
000 -3.1867
Ts: sampling time
PCM t
codeword 110 110 111 110 100 010 011 100 100 011 PCM sequence
Lecture 2 57
 Quantization error
• Quantizing error: The difference between the input and output of
a quantizer e(t )  xˆ (t )  x(t )

Process of quantizing noise

Qauntizer
Model of quantizing noise
y  q (x)
AGC x (t ) xˆ (t )
x(t ) xˆ (t )
x
e(t )

+
e(t) 
ˆ(t)x(t)
x

Lecture 2 58
 Quantization error …
• Quantizing error:
– Granular or linear errors happen for inputs within the dynamic range
of quantizer
– Saturation errors happen for inputs outside the dynamic range of
quantizer
• Saturation errors are larger than linear errors
• Saturation errors can be avoided by proper tuning of AGC
• Quantization noise variance:

 2
q 
E
{[
xq
(
x
)]
}
e(
x
)
p(
x
)
dx
22
Lin
Sat

2 2




12
qL/2
q2

2
2
Linp(
xl)
ql
l Uniform q. 
2
Lin 
12l
0 12

Lecture 2 59
 Uniform and non-uniform quant.
– Uniform (linear) quantizing:
– No assumption about amplitude statistics and correlation properties of
the input.
– Not using the user-related specifications
– Robust to small changes in input statistic by not finely tuned to a specific
set of input parameters
– Simple implementation
• Application of linear quantizer:
– Signal processing, graphic and display applications, process control
applications
– Non-uniform quantizing:
– Using the input statistics to tune quantizer parameters
– Larger SNR than uniform quantizing with same number of levels
– Non-uniform intervals in the dynamic range with same quantization
noise variance
• Application of non-uniform quantizer:
– Commonly used for speech

Lecture 2 60
 Non-uniform quantization
• It is achieved by uniformly quantizing the “compressed” signal.
• At the receiver, an inverse compression characteristic, called “expansion” is
employed to avoid signal distortion.

compression+expansion companding

y  C (x) x̂
x(t ) y (t ) yˆ (t ) xˆ (t )

x ŷ
Compress Qauntize Expand
Transmitter Channel Receiver
Lecture 2 61
 Statistics of speech amplitudes
• In speech, weak signals are more frequent than strong ones.

Probability density function 1.0

0.5

0.0
1.0 2.0 3.0
Normalized magnitude of speech signal
S
• Using equal step sizes (uniform quantizer) gives low  weak
for  signals
S  N q
and high for strong
  signals.
 N q
– Adjusting the step size of the quantizer by taking into account the speech statistics
improves the SNR for the input range.

Lecture 2 62
 Baseband transmission
• To transmit information through physical
channels, PCM sequences (codewords) are
transformed to pulses (waveforms).
– Each waveform carries a symbol from a set of size M.
k  log 2 M
– Each transmit symbol represents bits of
the PCM words.
– PCM waveforms (line codes) are used for binary
symbols (M=2).
– M-ary pulse modulation are used for non-binary
symbols (M>2).

Lecture 2 63
 PCM waveforms

• PCM waveforms category:

 Nonreturn-to-zero (NRZ)  Phase encoded
 Return-to-zero (RZ)  Multilevel binary
1 0 1 1 0 1 0 1 1 0
+V +V
NRZ-L -V Manchester -V

Unipolar-RZ +V Miller +V
0 -V
+V +V
Bipolar-RZ 0 Dicode NRZ 0
-V -V
0 T 2T 3T 4T 5T 0 T 2T 3T 4T 5T

Lecture 2 64
 PCM waveforms …
• Criteria for comparing and selecting PCM
waveforms:
– Spectral characteristics (power spectral density
and bandwidth efficiency)
– Bit synchronization capability
– Error detection capability
– Interference and noise immunity
– Implementation cost and complexity

Lecture 2 65
Impulse Sampling