UNIT 5

Multirate Digital Signal

Processing
WHAT IS MULTIRATE SIGNAL PROCESSING

• To process data at more than one sampling rate is called

multirate signal processing.
• The two primary operations are 1]decimation 2]Interpolation
• Decimation reduces sampling rate (i.e. sampling frequency) by
effectively compressing data and retaining only the desired
information.
• Interpolation increases the sampling rate so that it is easier to
process or to achieve compatibility with another system.

2
NEED OF MULTIRATE DIGITAL SIGNAL PROCESSING

System
x(n) y(n)

3
SAMPLING RATE REDUCTION: DECIMATION BY INTEGER FACTOR(M/D)
• Figure shows the process of
decimating a signal x(n) by an
integer factor M.
• It consists of digital anti aliasing
filter h(k).
• Sampling rate compressor
,symbolized by down arrow and
decimation factor M.
• The rate compressor reduces the
To prevent aliasing at lower rate the digital filter sampling rate from Fs to Fs/M.
used to bandlimit the input signal to less than Fs/2M
beforehand. Thus signal x(n) is first band limited
4
The sampling rate reduction is achieved by discarding M-1 samples for every M samples
of the filtered signal, w(n).
For M=3 ,Two samples out of
every three samples of x(n) are
discarded.

5
SAMPLING RATE INCREASE: INTERPOLATION BY INTEGER FACTORS
Interpolation is the digital equivalent of D-A Converter
X(n) is signal at sampling frequency Fs ,the interpolation process increases the sampling
rate by L to LFs.

6
For each sample of x(n) ,the
expanlder inserts L-1 zero
valued samples to form the
new signal w(m) at a rate of
LFs.
This signal is then applied to
LPF to remove images created
by the rate increase to yield
y(m).
If L=3 ,for each sample of
x(n,)three output samples are
produced.
This is due to two zero valued
samples inserted by expander.
7
SAMPLING RATE CONVERSION BY NON INTEGER FACTORS
OR L/M OR I/D CONVERSION
• In many applications we need to change the sampling rate by non integer factor.
• In practice such non-integer factor is represented by rational number that is the ratio of
two integers i.e L/M or I/D.
• L and M are integer such that Ratio of L/M or I/D is as close to the desired factor as
possible.
• The sampling frequency change is achieved by first interpolating the data by L and then
decimating the data by M.

8
The two LPF h1(k) and h2(k) can be combined into a single filter since they are in
cascade and have common sampling frequency.
If M>L then resulting operation is decimation process by a non integer and when M<L
then it is interpolation.

9
• In figure interpolation by
factor 3/2.
• The sample rate is first
increased by 3 by inserting
two non zero valued samples
for each sample of x(n).
• Then LPF is applied.
• The filtered data is then
reduced by factor of 2

10
NUMERICAL: REFEREE SOLVED PROBLEMS IN IFEACHOR BOOK(PROBLEM 9.1 TO 9.4)

11
MULTISTAGE APPROACH TO SAMPLING RATE CONVERSION
• When large change in the sampling rate is required it is more efficient to change
the two or more stages than in one single stage.
• In fact more practical multirate systems employ the multistage approach because it
allows a gradual reduction or increase in the sampling rate, leading to a significant
relaxation in the requirements of anti-aliasing or anti-imaging filter at each stage.

12
Multistage Decimation Process
Figure shows I-stage decimation process. Each stage is independent decimator.
The overall decimation factor M is expressed as:
M=M1M2M3…….Mi
If M>>1 ,the multistage approach leads to much reduced computational and
storage requirments.

13
FILTER REQUIRMENTS FOR INDIVIDUAL STAGES:
For multistage decimator the filter requirements for each stage to ensure that the
overall filter requirments are met are:
Fi=output sampling frequency
Fi=𝐹𝑖−1 /𝑀𝑖 i=1,2,3….I

14
b)Multistage structure
c)Filter Specifications for stage i

15
DETERMINING THE NUMBER OF STAGES AND DECIMATION FACTORS

Multiplication per second=MPS=σ𝐼𝑖=1 𝑁𝑖𝐹𝑖

Total storage requirement=TSR= σ𝐼𝑖=1 𝑁𝑖

IN general for optimum MPS and TSR the

decimation facor satisfy the following relation
M1> M2 > ………MI

16
POLYPHASE FILTER STRUCTURE
Polyphase structure for FIR filter were developed for the efficient implementation of
sampling rate converters.
M-component polyphase decomposition and Pi(z)
is polyphase components of H(z).

Each subsequence is obtained by down sampling

a delayed version of the original impulse
response

17
Blk diagm of polyphase filter structure for M=3

18
APPLICATIONS OF MULTIRATE DSP:
Subband Coding of Speech Signal:
• A variety of technique have been developed to efficiently represent
speech signals in digital form for either transmission or storage.
• Most of the speech energy is contained in lower frequencies, we would
like tot encode the lower frequency band with more bits than high
frequency band.
• Subband coding is a method where speech signal is subdivided into
several frequency bands and each band is digitally encoded
separately.

19
• Speech signal is sampled at a rate Fs samples per second.
• The first frequency subdivision splits signal spectrum into two equal width segments
• Low pass signal (0≤F≤Fs/4) and High pass signal (Fs/4≤F≤Fs/2).
• The second frequency subdivision splits low pass signal from first stage into two equal bands
low pass signal (0≤F≤Fs/8) and high pass signal (Fs/8≤F≤Fs/4).
• Similarly third subdivision takes place.

20
• Decimation by a factor 2 is performed after frequency subdivision.
• By allocating a different number of bits per samples to the signal in the four subbands, we can
achieve a reduction in the bit rate of the quantized speech signal.
• To avoid aliasing use quadrature mirror filter

21
Subband speech coder
• The synthesis method for subband
encoded speech signal is basically
the reverse of the encoding process.
• The signals in adjacent lowpass and
highpass frequency bands are
interpolated, filtered and combined
as shown in fig.

• Application of subband coding:

• It is efficient method to achieve
data compression in image signal
processing.

22
Synthesis of subband encoded signals
TRANSMULTIPLEXERS
• Transmultiplexers are devices for converting between time-division-multiplexed (TDM)
signals and frequency–division –multiplexed(FDM) signals.
• Consider FDM to TDM Conversion:

23 FDM TO TDM transmultiplexer

TDM TO FDM TRANSMULTIPLEXER

24
REFERENCES:

2 .Emmanuel Ifeachor, “Digital Signal Processing”, 2nd edition.

25