Scribd
Upload
5 views

Multirate_part1

The document discusses multirate signal processing, emphasizing its necessity in telecommunications where different sampling rates are required. It outlines methods for sampling rate conversion, particularly focusing on digital domain techniques that preserve signal quality without additional hardware. The document also explains downsampling, its effects on signal representation, and the importance of avoiding aliasing through appropriate filtering before downsampling.

Uploaded by

rakeshdhakharwal0507
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
5 views

Multirate_part1

The document discusses multirate signal processing, emphasizing its necessity in telecommunications where different sampling rates are required. It outlines methods for sampling rate conversion, particularly focusing on digital domain techniques that preserve signal quality without additional hardware. The document also explains downsampling, its effects on signal representation, and the importance of avoiding aliasing through appropriate filtering before downsampling.

Uploaded by

rakeshdhakharwal0507
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 13

Department of Electronics and Communication

Engineering

Digital Signal Processing

Subject Code – ECN-312

Multirate Signal Processing Part-1


Need for multi-rate signal processing
Whenever there is a requirement of changing the sampling
rate.
• For example, telecommunication system with transmit and
receive having different sampling speeds.
• The system has to deal with different signals having
different bandwidth requirements: different bandwidth
impose different sampling rate requirements as per Nyquist
Theorem.
• Reconstruction filters and antialiasing filters have some
restrictions on bandwidth of the signal to be processed. It
can be achieved with the Multirate signal processing.
• Filter banks and wavelet transforms depend on multirate
methods.
Methods for multi-rate signal processing
• The process of converting a signal from a given rate to a
different rate is called multi-rate signal processing/ sampling
rate conversion.
• Systems employing multiple sampling rates in the
processing of digital signals are called multi-rate digital
signal processing systems.
Method-1: Convert digital signal to analog using a D/A
converter, then resample the resulting analog signal at desired
rate using A/D converter.
The new sampling rate can be arbitrarily selected.
No need to have connection with older sampling rate.
Distortion introduced by conversion process lead to poor signal
quality
Require extra hardware components.
Method-2: Use Multirate signal processing
technique in digital domain.
• Carry out sampling rate conversion entirely in
the digital domain.
• Digital filtering can be used to preserve signal
quality by compensating for any distortion.
• No extra hardware is required.
• Some additional benefits such as configurable
according to signal bandwidth requirement or
filter limitations
This course concentrates on Multirate signal
processing technique in digital domain
Downsampling: Reducing sampling rate

• Decimation or Downsampling operation consists of taking


every Mth sample.

Time Domain Representation


x(n)
For M=3, if we apply above
operation, but observe the
signal with original
y(n) sampling rate Fs, only Mth
sample has useful
information.

For M=3, if we apply


y(m) above operation, and
observe output with lower
sampling rate (Fs/M).
Frequency Domain Representation
of Downsampling operation
Let us observe its behavior in Z-transform domain, for
simplicity, we take M=2: y (n)  x(2n) (1)
y (n)  {x(0), x(2), x(4),...,} (2)
Taking Z-transform of x(n) and y(n)
X ( z )  x(0)  x(1) z 1  x(2) z 2  ...  x(n) z  n  ... (3)
Y ( z )  y (0)  y (1) z 1  y (2) z 2  ...  y (n) z  n  ... (4)
By using relation between y(n) and x(n)
Y ( z )  x(0)  x(2) z 1  x(4) z 2  ...  y (2n) z  n  ... (5)

It is useful to introduce an intermediate variable v(n)


v(n)  {x(0), 0, x(3), 0, x(4),...,} (6)
V ( z )  x(0)  x(2) z 2  x(4) z 4  ...  x(2n) z 2 n  ... (7)
If we calculate X(-z) and add with X(z), we will achieve:
V ( z )  0.5{ X ( z )  X ( z )} (8)
From (5) and (7), we can say:
Y ( z )  V ( z1/2 ) (9)
From (8) and (9), we can say:
Y ( z )  0.5{ X ( z1/2 )  X ( z1/2 )} (10)
We can replace Z-transform representation by Fourier-
transform by using (z  e j )
Y (e j )  0.5  X (e j /2 )  X (e j /2 )  (11)

Y (e j )  0.5  X (e j /2 )  X (e j ( /2 ) ) 
or
(12)
Y ( )  0.5  X f ( / 2)  X f (   / 2)  (13)

We adopted a new nomenclature Y f ( )  Y (e j )


to simplify presentation X f ( )  X (e j )

Y f ( )  DTFT { 2  x(n)}


  2    
 0.5  X f ( / 2)  X f   (14)
  2 
Note:
 Y f ( ) repeats itself
 after
2 frequency
 span of 2π.
 X f ( / 2) and X f   repeats themselves
 2 
after frequency span of 4π.
So, how does LHS and RHS of (14) hold together?
X f ( )
Periodic with 2π

X f ( / 2)
Periodic with 4π

 2   
0.5 X f  
 2 
Periodic with 4π

Y f ( ) 
  2    
0.5  X f ( / 2)  X f  
  2 
repeats at 2π
Notice that:

The down-sampled signal y(n) does not look exactly like


 2   
original signal x(n). If we could remove term, X f  
with some gain, we could achieve original shape.  
2

The overlapping of second term with the first term is called


aliasing. Once Aliasing occurs, it is difficult to recover original
signal after downsampling.

We can surmise that If spectrum of X f ( ) would lie between


(0, π/2), no aliasing will occur.

We can extend these results to any M.


Downsampling by factor M

Y ( )  DTFT { M  x(n)}


f

1  M 1  2 k    
  Xf  
M  k 0  M 

• If plotted if older sampling frequency rate is considered,


the signal spectrum will expand M times.
• To avoid aliasing, the spectrum of X f ( ) should lie
between (0, π/M).
• In general information is lost in down-sampling
operation.
To avoid aliasing, a low-pass filter with bandwidth π/M
is applied before downsampling operation
v(n)

Let us see the example of processing a signal via down-


sampler of factor D:

Filter
Signal Specifications
spectrum
Downsampled
Filtered
Signal
signal

Example (Try yourself)


Find the length of FIR filter which is to be used for
downsampling operation by 2. Filter should have 0.1 dB ripple
in passband and down by 30dB in the stopband.

Hints:

You might also like