Multirate Digital Signal

Processing
 Multirate Digital Signal
Processing

What is multirate signal processing?

Processing of digital signal with
different sampling rates in the system.

Sampling Rate Conversion

 Multirate Digital Signal
Processing
Basic Sampling Rate Alteration Devices
• Up-sampler - Used to increase • Down-sampler - Used to decrease
the sampling rate by an integer the sampling rate by an integer
factor factor
 Multirate Digital Signal
Processing

Why sample rate conversion? (I)

• Compatibility: convert sample frequencies of
different stds.
• Efficiency: easier data processing
(computationally more efficient), less storage,
lower transmission speed, …
• All-digital: Change sample frequency in an
efficient manner
• Cost: Avoid need for expensive analogue anti-
aliasing filters
 Up-Sampler
Time-Domain Characterization
• An up-sampler with an up-sampling
factor L, where L is a positive integer,
develops an output sequence xu [n] with
a sampling rate that is L times larger
than that of the input sequence x[n]
• Block-diagram representation
x[n] L xu [n ]
 Up-Sampler
• Up-sampling operation is implemented by
inserting L  1 equidistant zero-valued
samples between two consecutive
samples of x[n]
• Input-output relation

 x[n / L], n  0,  L,  2 L,

xu [n]  
 0, otherwise
 Up-Sampler
• In practice, the zero-valued samples
inserted by the up-sampler are replaced
with appropriate nonzero values using
some type of filtering process
• Process is called interpolation and will be
discussed later
 Down-Sampler
Time-Domain Characterization
• An down-sampler with a down-sampling
factor M, where M is a positive integer,
develops an output sequence y[n] with a
sampling rate that is (1/M)-th of that of
the input sequence x[n]
• Block-diagram representation
x[n] M y[n]
 Down-Sampler
• Down-sampling operation is implemented
by keeping every M-th sample of x[n] and
removing M  1 in-between samples to
generate y[n]
• Input-output relation
y[n] = x[nM]
 Down-Sampler
• Figure below shows explicitly the time-
dimensions for the down-sampler

x[ n ]  xa ( nT ) M y[ n ]  xa ( nMT )

Input sampling frequency Output sampling frequency

1 ' FT 1
FT  FT  
T M T'
 Up-Sampler
• Figure below shows explicitly the time-
dimensions for the up-sampler

x[ n ]  xa ( nT ) L y[n]
 x ( nT / L ), n 0,  L, 2 L,
 a
 0 otherwise

Input sampling frequency Output sampling frequency

1 ' 1
FT  FT  LFT 
T T'
Basic Sampling Rate Alteration Devices
• The up-sampler and the down-sampler are
linear but time-varying discrete-time systems
• Consider a factor-of-M down-sampler defined
by y[n] = x[nM]
• Its output y1[n] for an input x1[n]  x[n  n0 ] is
then given by
y1[n]  x1[ Mn]  x[ Mn  n0 ]
• From the input-output relation of the down-
sampler we obtain
y[n  n0 ]  x[ M (n  n0 )]
 x[ Mn  Mn0 ]  y1[n]
 Up-Sampler

Frequency-Domain Characterization
• Consider first a factor-of-2 up-sampler
whose input-output relation in the time-
domain is given by
 x[n / 2], n  0,  2,  4,
x u [n ]  
 0, otherwise
 Up-Sampler

• In terms of the z-transform, the input-

output relation is then given by
 
n n
X u ( z)   u
x [ n ] z   x[ n / 2] z
n   n  
n even

 
m 
x[m] z 2 m  X ( z 2 )
 Up-Sampler

• In a similar manner, we can show that

for a factor-of-L up-sampler
L
X u ( z)  X ( z )
• On the unit circle, for z  e j , the input-
output relation is given by
j j L
X u (e )  X (e )
 Up-Sampler
• Figure below shows the relation between
j j
X (e ) and X u (e ) for L = 2 in the
case of a typical sequence x[n]
 Up-Sampler

• As can be seen, a factor-of-2 sampling

rate expansion leads to a compression
j
of X ( e ) by a factor of 2 and a 2-fold
repetition in the baseband [0, 2]
• This process is called imaging as we
get an additional “image” of the input
spectrum
 Up-Sampler
• Similarly in the case of a factor-of-L
sampling rate expansion, there will be L  1
additional images of the input spectrum in
the baseband
• Lowpass filtering of xu [n] removes the
images and in effect “fills in” the zero- L  1
valued samples in xu [n] with interpolated
sample values
 Down-Sampler
Frequency-Domain Characterization
• Applying the z-transform to the input-output
relation of a factor-of-M down-sampler

y[n]  x[Mn]
we get

n
Y ( z)   x[Mn] z
n  
• The expression on the right-hand side cannot
be directly expressed in terms of X(z)
 Down-Sampler

• To get around this problem, define a

new sequence xint [n] :
 x[n], n  0,  M ,  2 M ,
xint [n]  
 0, otherwise
• Then
 
n n
Y ( z)   x[Mn] z   xint [Mn] z
n   n  

k / M 1/ M
  xint [k ] z  X int ( z )
k  
 Down-Sampler
• Now, xint [n] can be formally related to x[n]
through
xint [n]  c[n]  x[n]
where periodic train c[n]
1, n  0,  M ,  2 M ,
c[n]  
0, otherwise
• A convenient representation of c[n] is given by
1 M 1 kn
c[n]  
M k 0
WM
where
 j 2 / M
WM  e
 Down-Sampler
• Taking the z-transform of xint [n]  c[n]  x[n]
and making use of
1 M 1
c[n]   WMkn
M k 0
we arrive at
   M 1
n 1 kn  n
X int ( z )   c[n]x[n] z    M
 W  x[ n ] z
n   M n    k 0 
1 M 1 
 
 1 M 1
kn  n k
  
  x
M k 0  n  
[ n ]WM z 
   X z W M
 M k 0
 Down-Sampler
• Consider a factor-of-2 down-sampler
with an input x[n] whose spectrum is as
shown below

• The DTFTs of the output and the input

sequences of this down-sampler are
then related as
1
Y (e )  { X (e j / 2 )  X (e j / 2 )}
j 
2
 Down-Sampler
• Now X (e j / 2 )  X (e j ( 2 ) / 2 ) implying
that the second term X (e j / 2 ) in the
previous equation is simply obtained by
shifting the first term X (e j / 2 ) to the right
by an amount 2 as shown below
 Down-Sampler
• The plots of the two terms have an overlap,
and hence, in general, the original “shape”
X ( e j )
of is lost when x[n] is down-sampled
as indicated below
 Down-Sampler
• This overlap causes the aliasing that takes
place due to under-sampling
• There is no overlap, i.e., no aliasing, only if
X ( e j )  0 for    / 2
• Note: Y (e j ) is indeed periodic with a
period 2, even though the stretched
version of X (e j ) is periodic with a period
4
 Down-Sampler
• For the general case, the relation between
the DTFTs of the output and the input of a
factor-of-M down-sampler is given by
1 M 1
Y ( e j )   X ( e j ( 2 k ) / M )
M k 0
• Y (e j ) is a sum of M uniformly
shifted and stretched versions of
X ( e j )
and scaled by a factor of 1/M
 Down-Sampler
• Aliasing is absent if and only if
X (e j )  0 for    / M
as shown below for M = 2
X (e j )  0 for    / 2
 Cascade Equivalences
• Two cascade equivalences are shown
below
Cascade equivalence #1
x[n ] M H (z ) y1 [ n ]

 x[n ] H (z M ) M y1 [ n ]
Cascade equivalence #2

x[n ] L H (z L ) y2 [ n]

 x[n ] H (z ) L y2 [ n]
 Filters in Sampling Rate
Alteration Systems
• The bandwidth of a critically sampled
signal must be reduced by lowpass
filtering before its sampling rate is reduced
by a down-sampler to avoid aliasing
• Likewise, the zero-valued samples
introduced by an up-sampler must be
interpolated by lowpass filtering to more
appropriate values for an effective
sampling rate increase
 Filter Specifications
• Since up-sampling causes periodic
repetition of the basic spectrum, the
unwanted images in the spectra of the up-
sampled signal xu [n] must be removed by
using a lowpass filter H(z), called the
interpolation filter,
filter as indicated below
xu [n]
x[n ] L H (z) y[n ]

• The above system is called an interpolator

 Filter Specifications
• On the other hand, prior to down-
sampling, the signal v[n] should be
bandlimited to   /M by means
of a lowpass filter, called the decimation
filter,
filter as indicated below to avoid aliasing
caused by down-sampling
x[n ] H (z) M y[n ]

• The above system is called a decimator

 Interpolation Filter
Specifications
• If we pass x[n] through a factor-of-L up-
sampler generating xu [n] , the relation
between the Fourier transforms of x[n] and
are givenxby u [n]
X u ( e j )  X ( e j L )
• It therefore follows that if xu [n] is passed
through an ideal lowpass filter H(z) with a
cutoff at /L and a gain of L, the output of
the filter will be precisely y[n]
 Interpolation Filter
Specifications
• If  c is the highest frequency that needs
to be preserved in x[n], then
 p  c / L
• Summarizing the specifications of the
lowpass interpolation filter are thus given
by
j  L,   c / L
H (e ) 
 0,  / L    
Decimation Filter Specifications
• In a similar manner, we can develop the
specifications for the lowpass decimation
filter that are given by

j 1,   c / M
H (e ) 
0,  / M    
• The design of the filter H(z) is a standard
IIR or FIR lowpass filter design problem
 Polyphase Decomposition
• The FIR filter is realized using direct form
• To avoid unnecessary calculations the decimator
is replaced with efficient transversal structure.
N 1
y (n)   h[m]xm [n]
m 0

• For the polyphase structure

M 1
y ( n )   p m [ n ] * xm [ n ]
m 0
 Polyphase Decomposition
Decomposition of H(z)=Σhm z -m in blocks of M:
H (z) = ... + h(−M )zM + h(−M + 1) z M −1 + ... + h(−1) z1
+ h(0)z0 + h(1) z−1 + ... + h(M − 1) z−(M −1)
+ h(M )z−M + h(M + 1) z−(M +1) + ... + h(2M − 1) z−(2M −1)
+ h(2M )z−2M + h(2M + 1) z−(2M +1) + ... + h(3M − 1) z−(3M −1) + ...

= z0[... + h(0) z0 + h(M ) z−M + ...] + z−1[... + h(1) + h(M + 1) z−M + ...]
+ z−2[... + h(2) + h(M + 2) z−M + ...] + ...
+ z−(M −1) [... + h(M − 1) + h(2M − 1) z−M + ...]
M −1 +∞

H (z) = ∑ z P (z ) ∑ z h(nM + i)
−i M −n
⇒ i where Pi (z) =
i=0 n=−∞
Polyphase Decomposition
 Implementation of Decimation
Using noble identity:

Operations performed at Operations at low rate ⇒

high rate more efficient
 Implementation of Decimation
• Using commutator:

one input per D pulses;

counter-clockwise rotation
THANK YOU