You are on page 1of 45

Communication Systems, 5e

Chapter 6: Sampling and pulse


modulation

A. Bruce Carlson
Paul B. Crilly
(modified by J. H. Cho using Prof. W.J.
Song’s lecture note)
© 2010 The McGraw-Hill
Companies
Chapter 6: Sampling and pulse modulation

 Sampling theory and practice

 Pulse-amplitude modulation

 Pulse-time modulation

© 2010 The McGraw-Hill


Companies
Sampling theory and practice

© 2010 The McGraw-Hill


Companies
Chopper sampling

© 2010 The McGraw-Hill


Companies
Switching sampler (a) functional (b) waveforms (c) circuit

© 2010 The McGraw-Hill


Companies
Chopper sampler equations

Input signal: x(t )

chopper switch: s (t )

Chopper output  xs (t )  s (t )  x(t )


 xs (t ) is a sampled version of x(t )
© 2010 The McGraw-Hill
Companies
s (t ) can be expressed as a Fourier series


 s (t )  c0   2cn cos n 2f st
n 1

 xs (t )  s (t ) x(t )  c0 x(t )  2c1 x(t )cos 2f s t  2c2 x(t )cos 2 2 f s t  ...

© 2010 The McGraw-Hill


Companies
 

xs (t )  x(t ) s (t )  x(t )  c0   2cn cos n 2f s t 
 n 1 

Taking the Fourier transform of xs (t )  X s ( f )

X s ( f )  c0 X ( f )  c1  X ( f  f s )  X ( f  f s ) 
 c2  X ( f  2 f s )  X ( f  2 f s )   ...

© 2010 The McGraw-Hill


Companies
Spectra for switching sampling (a) message (b) sampled
message f s  2W (c) sampled message f s  2W

aliasing caused by
undersampling

© 2010 The McGraw-Hill


Companies
Accurate reconstruction

X ( f ) must be bandlimited  X ( f )  0 for f  0

and

1 1
fs   2W or f s   2W (pure sine waves)
Ts Ts

1
If f s  2W  under sampling  aliasing where spectra overlap
Ts
 aliasing prevents accurate reconstruction

© 2010 The McGraw-Hill


Companies
If a signal contains no frequency components for f  W ,
it is completely described by its sample values obtained by
sampling at a rate of f s  2W .

We call f s  W the Nyquist rate

© 2010 The McGraw-Hill


Companies
Why f s  2W for pure sine waves?

If f s  2W  it's possible to sample at zero crossing  xs ( nTs )  0

© 2010 The McGraw-Hill


Companies
Ideal sampling
Consider a periodic pulse train of


 t  kTs 
s (t ) 
k 
 
  


1
 s (t )lim s (t )     t  kTs 
 0 
k 


 x (t )  x(t ) s (t )  x(t )    t  kTs 
k 

 x (t )   x(kT )  t  kT 
k 
s s

 x(t ) t  kT  x(kTs )  the samples of x(t )


s

© 2010 The McGraw-Hill


Companies
In the frequency domain

X  ( f )  f s X ( f )  f s [ X ( f  f s )  X ( f  f s )]  ...

 fs  X ( f  nf )
n 
s

The spectrum is:

Note: with this and chopper sampling, the message spectrum is periodic
with period f s .
© 2010 The McGraw-Hill
Companies
Ideal reconstruction

To reconstruct the sampled signal, we use a LPF

H( f )

f
Y ( f )  H( f )X( f )

© 2010 The McGraw-Hill


Companies
Ideal LPF for reconstruction
Using the ideal LPF for reconstruction, and time delays td
we have

 f   j 2 ftd
H ( f )  K  e
 2B 

So that the reconstructed output is

Y ( f )  H ( f ) X  ( f )  Kf s X ( f )e  j 2 ftd
 original signal but delayed

© 2010 The McGraw-Hill


Companies
Reconstruction by interpolation
From the time/sequence domain perspective, we can reconstruct via
interpolation
Ideal interpolation  h(t )  1  H ( f )   2 BKsinc2 B(t  td )

 y (t )  h(t )  x( kTs )   x(kTs )h(t  kTs )


k

 2 BK  x(kT )sinc2 B(t  t
k 
s d  kTs )

If we sample at the minimum rate  B  f s / 2, with K  1/ f s


and td  0

 y (t )   k (kTs )sinc(f st  k )  ideal interpolation


k

© 2010 The McGraw-Hill


Companies
Ideal reconstruction via interpolation

© 2010 The McGraw-Hill


Companies
Practical sampling and
reconstruction
 Real samplers have finite durations pulses (e.g.
chopper sampling)
 Practical reconstruction filters are not ideal
 Sampled signals are time limited  not bandlimited
 cannot avoid some aliasing

© 2010 The McGraw-Hill


Companies
Non-ideal sampling
 As seen earlier, no loss of info with non-impulse
sampling
 Non-ideal LPF reconstruction can be overcome by
prefiltering the original message  equalization
 We can minimize aliasing by band limiting the input
signal  anti-alias LPF

© 2010 The McGraw-Hill


Companies
Practical reconstruction filter

© 2010 The McGraw-Hill


Companies
Reconstruction using a zero order hold (ZOH)
interpolation

 t  kTs 
y (t )   x (kTs )   
k  s T

 reconstruction filter  H ZOH ( f )  Tssinc(fTs )

© 2010 The McGraw-Hill


Companies
Reconstruction using a first order hold (FOH)
interpolation

 t  kTs 
y (t )   x(kTs )   
k  s T

 reconstruction filter  H FOH ( f )  Ts 1+(2fTs ) 2 sinc 2 (fTs )

© 2010 The McGraw-Hill


Companies
Signal reconstruction from sampled signal (a) ZOH (b)
FOH

© 2010 The McGraw-Hill


Companies
H FOH ( f ) and H ZOH ( f ) are low pass filter functions,
but they are not ideal LPFs.

The difference in frequency response between an ideal LPF and the


ZOH or FOH introduces aperture errors in the reconstruction
process.

These aperature errors can be compensated by prefiltering the data


to emphasize the higher frequencies.

Prefiltering  equalization

© 2010 The McGraw-Hill


Companies
RC anti-alias filter
1. Time limited signals have large bandwidth  to reduce
errors due to aliased components spilling over into the message
bandpass we incorporate an anti-alias filter.

2. Anti-aliasing filter often implemented via an RC LPF with


1
B .
2RC

3. Because RC LPFs do not have sharp cutoff  B W

 oversample the message signal

© 2010 The McGraw-Hill


Companies
Message spectrum (a) output of RC filter, (b) after sampling

(a) Original message spectrum, (b) spectrum after sampling


© 2010 The McGraw-Hill
Companies
Aliasing error:

 1/ 0.707 
Error %   
 1  ( f / B)2 
 a 

with

1
B , fa  fs  B
2RC

© 2010 The McGraw-Hill


Companies
If a signal has been over-sampled but with an
acceptable amount of aliasing, we then

 Feed it to our DSP for digital filtering to remove the


components above W.

 Down-sample the signal to some desired rate.

© 2010 The McGraw-Hill


Companies
Upsampling

 Some cases we can only sample a signal at the Nyquist rate;


but need more samples upsample the data.
 Upsampling  insert correct samples between the original
set  decrease effective sample interval  Ts  Ts / M
'

 Time domain: interpolation


 Frequency domain: zero padding

© 2010 The McGraw-Hill


Companies
Upsampling via linear interpolation

x '(k ')
x( k )
x (t ) x(t )

upsampling
 Ts'
Ts

t, k t, k '
k k 1 k 2 k 3 k' k ' 2 k ' 4 k ' 6
(b)
(a)

(a) Original signal sampled, (b) signal upsampled by factor of 2

© 2010 The McGraw-Hill


Companies
Why upsampling?

 Upsampling  greater time resolution

 Adaptive filtering methods may require more samples


than obtained by the Nyquist rate

 Other DSP algorithms that depend on time resolution


may require more than the minimum number of samples

© 2010 The McGraw-Hill


Companies
Upsampling does not add information

© 2010 The McGraw-Hill


Companies
6.2, 6.3 Pulse Modulation

© 2010 The McGraw-Hill


Companies
6.2 Pulse-Amplitude Modulation
(PAM)

© 2010 The McGraw-Hill


Companies
© 2010 The McGraw-Hill
Companies
Analog signal and corresponding PAM signal

© 2010 The McGraw-Hill


Companies
Flat-top sampling (a) sample & hold circuit (b) waveform

© 2010 The McGraw-Hill


Companies
PAM signals
x p (t )   x(kTs ) p(t  kTs ),
k

where x( kTs )  sampled message, p(t  kTs )  pulse that


carries the message and

1
  pulse duration and  Ts 
2W

PAM is the digital form of AM, and p(t  kTs ) is analogous


to AM's carrier (PAM also has a large DC component)

For adequate pulse resolution, the channel bandwidth, or


transmission bandwidth must be
1
BT  W
2
© 2010 The McGraw-Hill
Companies
PAM
 Rarely used for single channel communication
systems, but
 used in conjunction with instrumentation, data
telemetry, and instrumentation systems
 Time-division multiplexing (TDM) systems
 Basis for other digital modulation systems

© 2010 The McGraw-Hill


Companies
6.3 Pulse time modulation
 Pulse duration modulation (PDM). Also called pulse
width modulation (PWM)
 Pulse position modulation (PPM)
 Info in zero crossings  potential for wideband noise
reduction

© 2010 The McGraw-Hill


Companies
Types of pulse-time modulation

© 2010 The McGraw-Hill


Companies
(a) Generation of PDM and PM signals, (b) waveforms
© 2010 The McGraw-Hill
Companies
Conversion of PDM or PPM into PAM

© 2010 The McGraw-Hill


Companies
To preserve information content in the demodulation process
 received signal have fast rise time

 tr Ts

and
1
 BT   and extremely wideband system
2tr

© 2010 The McGraw-Hill


Companies

You might also like