FIR Filter Design

Rashmi Gupta
• Finite Impulse Response (FIR) Filters:

• In FIR system the impulse response sequence is of finite duration i.e.

it has a finite number of nonzero terms.

• Infinite Impulse Response (IIR) Filters:

• In IIR system the impulse response sequence is infinite duration i.e.

The IIR system has infinite number of nonzero terms.

Rashmi Gupta
 Infinite Impulse Response Filters (IIR Filters)
 Impulse response is of infinite duration i.e. infinite number of non zero terms.
 Response is dependent on past and present values of the input and past values
of the ouput.

 Used when linear phase is not required in pass band and preferred if some
phase distortion is tolerable.

 lesser number of side lobes in the stop band with same number of parameters.

 Implementation involve fewer parameters (less memory requirements and

fewer computational complexity). Thus, if some phase distortion is toper able
or unimportant than IIR filter is preferred.

 Casual and stable IIR filters cannot have linear phase. In its design, only the
desired magnitude response is mentioned and obtained phase response from
design methodology is accepted.

 Physically stable IIR filters cannot have a linear phase because according to
condition of linear phase h(n)=h(M-1-n), the filter would have mirror image,
pole outside the circle and pole inside
Rashmiit thus results in unstable filter.
Gupta
Finite Impulse Response (FIR) Filters

 The impulse response is of finite duration

 No feedback (non-recursive)

 Its response depends only on past and present inputs

 Linear-phase characteristics

 Always stable

 In speech processing and data transmission, it is always desirable, that output

should only be delayed by some amount. This is possible only if the filter is
of linear phase & hence FIR filter is must in such applications.

Rashmi Gupta
Advantages of FIR filter over IIR Filter:
 They can have an exact linear phase
 They are always stable because all poles always lies on origin
 The design methods are generally linear
 They can be realized efficiently in the hardware
Disadvantages of FIR filter over IIR Filter:
 Require higher order of FIR filter. So have higher computationally
complexity and large memory requirements
This is main drawback
 Delay increases with increase in order of filters

Rashmi Gupta
 Difference between IIR and FIR Filters
IIR FIR

1. IIR digital filters are FIR digital filters are

General characterized by rational characterized by system
Equation system function as function which is not rational
as

Where and are filter

coefficients Where are filter
An IIR filter describe by the coefficients
difference equation An FIR filter describe by the
difference equation

Rashmi Gupta
2. Impulse response of IIR filters are Impulse response of FIR
Impulse computed for infinite number of digital filters have finite
Response samples i.e. number of samples i.e.

h(n)≠0 0≤ n ≤∞ h(n)≠0, 0 ≤ n ≤ M-1

=0, else
where h(n) is the impulse response
of filter.

Rashmi Gupta
3. IIR filters do not have linear phase FIR filters have linear
Linear and they are used where some phase characteristics.
Phase phase distortion is tolerable These filters are used in
speech processing to
eliminate the adverse effects
of frequency dispersion due
to non linearity of phase.
4. IIR filters are theoretical stable. FIR filters are always
Stability After truncation their coefficients stable. Has only zeroes
these become unstable. Has both (poles lie on origin)
zeroes & poles
5. IIR filters are less flexible for FIR filters have greater
Flexibility obtaining non standard frequency flexibility to change the
responses or for those filters for shape of their magnitude
which analog filter design response and realization
techniques are not available. efficiency.
Rashmi Gupta
6. IIR filters are usually FIR filters are generally
Realization realized by recursive realized by non recursive
(feedback) method. The method. There is no feedback
present output of these filters
system. The present output of
depends on previous outputs
as well present and past these filters does not depend
outputs. on previous outputs.

7. IIR filters are more FIR filters have less severe

Noise susceptible to round off effects.
noise associated with finite
precision arithmetic
quantization error and
coefficient in accuracies.

Rashmi Gupta
8. Time In IIR filters time delay is less Delay increases with increase
Delay in order of filters.
9. Require lesser number of For sharp amplitude
Complexity arithmetic operations and these responses, we require higher
have lower computationally order of FIR filter. This is
complexity and smaller main drawback.
memory requirements.
10. IIR filters have resemblance FIR filters are unique to the
Resemblance with analog filter. The discrete time domain. These
to analog common method for IIR digital cannot be derived from
filters filter design is to design an IIR analog filters.
analog filter followed by A-D
converter.( IIT, BLT)

Rashmi Gupta
Linear Phase in FIR Filter:

Following are the two conditions to achieve Linear Phase In FIR filter

Proof: A filter with linear phase is one which has constant phase delay
& constant group delay.

Rashmi Gupta
Derivation:

Now since the filter has to be casual & finite length for the practically realizable filter

Rashmi Gupta
Now for linear phase filter

Rashmi Gupta
Now, can be taken inside the summation, as they are
independent of n,

A Generalized solution to this equation according to Fourier series

expansion is:- Rashmi Gupta
Thus, h (n) has to be symmetric about for a linear phase filter.

Also, when constant group delay alone is required, the impulse

Response h (n) has to be anti symmetric :-

Rashmi Gupta
Gibb’s Phoenomenon

Truncated F.S. approximation of discontinuous signal will

exhibit high frequency ripples and overshoot near the
discontinuities of signal. This phenomenon is known as Gibb’s
Phenomenon.
As the N increases, ripples also increases but their amplitude
decreases.
 The casual FIR filter obtained by simply truncating the impulse
response coefficients of ideal filter exhibits an oscillatory behavior
in the magnitude response.
“The oscillatory behavior near the band edge of the filter is called
the Gibb’s Phenomenon”. Slow convergence in the Fourier series
results in pass-band & stop-band oscillations. These oscillations are
caused by the discontinuity at the pass-band edge.
Rashmi Gupta
Rashmi Gupta
Ways to reduce Gibb’s Oscillattions:-

 The discontinuity between pass-band & stop-band in the frequency

response is avoided by introducing transition band between pass-band
& stop-bands.

 Another technique used for the reduction of Gibb’s oscillation is to

precondition the impulse response. This precondition is imposed by
using window functions. Window functions are a class of time-domain
functions & by using different kinds of windows, we can improve
abrupt truncation of the impulse response & hence reduce oscillations.

Rashmi Gupta
FIR Filter Design Methods

 Window Technique
 Frequency Sampling Method
 Digital Differentiator
 Hilbert Transformer

Rashmi Gupta
Design of Linear-Phase FIR Filters using Windows

In this method we begin with the desired frequency response

specification & determine the corresponding unit sample response

Rashmi Gupta
Window Functions

Rashmi Gupta
By tapering the windows smoothly to zero at each end, the height of side
lobes can be diminished at expense of wider main lobe & thus wider
transition at the discontinuity.

Rashmi Gupta
Frequency Response of FIR Filters

 Type I: Symmetric Impulse Response with Odd Length (M Odd)

 Type 2: Symmetric Impulse Response with Even Length (M Even)
 Type 3: Antisymmetric Impulse Response with Odd Length (M Odd)
 Type 4: Antisymmetric Impulse Response with Even Length (M Odd)

Rashmi Gupta
Q. Design a LPF with given using Hanning windows with M=5

Sol.

Rashmi Gupta
For n = 2, from (1)

For M = 5

Rashmi Gupta
Hanning window functions

Rashmi Gupta
Or

Rashmi Gupta
Ques. Design a high pass linear phase FIR filter with desired frequency
response given by

For M=7 and

Use a) Rectangular window function

b) Hamming window function
c) Hanning window Function

Sol. Since

Rashmi Gupta
The impulse response

Rashmi Gupta
 From (1)

From (2)

Rashmi Gupta
Hanning window function

Rashmi Gupta
The frequency Response

Rashmi Gupta
Design of FIR Filter By Frequency Sampling Method

Rashmi Gupta
In complex conjugate pair, magnitudes are equal & 180° phase difference is there e.g.

real term, the matching terms are

thus,

then only they will form complex conjugate hair.

For N odd

For N even

for N odd

for N even

Rashmi Gupta
 Ques. A low pass filter desired freq. response is given below

Determine the filter coefficient for N = 7, using type 1 Freq. Sampling

technique

Sol. The samples are uniformly sampled at

(2)

Rashmi Gupta
Rashmi Gupta
DESIGN OF FIR DIGITAL DIFFERENTIATOR (DD) [JOHNY R JOHNSON]

The freq. response of ideal DD is linearly proportion to frequency. It is

given by

The phase response is frequencies.

The impulse response of ideal DD is

Rashmi Gupta
If linear phase is desired, then frequency response is

Rashmi Gupta
If N is odd, the τ is an integer
If N is even, then for any n

For N even

Both have property Rashmi Gupta

Design of FIR Hilbert Transform
The frequency response of an ideal Hilbert transform is given by

The impulse response of an ideal Hilbert transformer is

Rashmi Gupta
is infinite in duration and non-casual.

Rashmi Gupta
The Freq. Response of a linear phase Hilbert transformer is given by

The requirement of linear phase introduces a delay of τ samples. The

impulse response of a linear phase Hilbert transform is:

Rashmi Gupta
Ques. Design an Hilbert transformer having frequency response

Using (a) rectangular window (b) Blackman window

For N = 11, plot freq. response in both the cases.

Sol. The ideal freq. response is shown below:

Rashmi Gupta
The filter coefficients are anti-symmetric about n = 0
satisfying

Rashmi Gupta
 Rectangular window

The realizable transfer function

Rashmi Gupta
Blackman window

Rashmi Gupta