Chapter 14

Chapter 14
Finite Impulse Response (FIR) Filters

Dr. Naim Dahnoun, Bristol University, (c) Texas Instruments 2004 Chapter 14, Slide 2
Learning Objectives
Introduction to the theory behind FIR
filters:
Properties (including aliasing).
Coefficient calculation.
Structure selection.
Implementation in Matlab, C, assembly
and linear assembly.
Introduction
Amongst all the obvious advantages that
digital filters offer, the FIR filter can
guarantee linear phase characteristics.
Neither analogue or IIR filters can achieve
this.
There are many commercially available
software packages for filter design.
However, without basic theoretical
knowledge of the FIR filter, it will be
difficult to use them.
Properties of an FIR Filter
Filter coefficients:
| | | |
=
=
1
0
N
k
k
k n x b n y
x[n] represents the filter input,
b
k
represents the filter coefficients,
y[n] represents the filter output,
N is the number of filter coefficients
(order of the filter).
| | | |
=
=
1
0
N
k
k
k n x b n y
z
-1
+
z
-1
z
-1
+ +
z
-1
y(n)
x(n)
x x x x
b
0
b
1
b
2
b
N-1
FIR equation
Filter structure
| | | |
=
=
1
0
N
k
k
k n x b n y
If the signal x[n] is replaced by an impulse
o[n] then:
| | | |
=
=
1
0
N
k
k
k n b n y o
| | | | | | | | N b b b y
k
+ + + = o o o 1 0 0
1 0
| | | |
=
=
1
0
N
k
k
k n x b n y
o[n] then:
| | | |
=
=
1
0
N
k
k
k n b n y o
| | | | | | | | N n b n b n b n y
k
+ + + = o o o 1
1 0
| | | |
=
=
1
0
N
k
k
k n x b n y
o[n] then:
| | | |
=
=
1
0
N
k
k
k n b n y o
| |
=
=
=
k n for 0
k n for 1
k n o
| | | |
=
=
1
0
N
k
k
k n x b n y
Finally:
| |
| |
| | k h b
h b
h b
k
=
=
=
1
0
1
0
| | | |
=
=
1
0
N
k
k
k n x b n y
With:
| | k h b
k
=
The coefficients of a filter are the same as
the impulse response samples of the filter.
Frequency Response of an FIR Filter
By taking the z-transform of h[n], H(z):
Replacing z by e
je
in order to find the
frequency response leads to:
( ) | |
=
1
0
N
n
n
z n h z H
( ) ( ) | |
=
= =
1
0
N
n
jn j
e z
e n h e H z H
j
e e
e
Since e
-j2tk
= 1 then:
Therefore:
( ) | |
( )
| |

=
=
+
=
= =
+
1
0
1
0
2
2
N
n
jn
N
n
jn
e z
e n h e n h z H
e t e
t e
( ) ( )
e t e j k j
e H e H =
+2
FIR filters have a periodic frequency
response and the period is 2t.
Frequency
response:
FIR y[n] x[n]
F
s
/2 F
s
/2
Freq
Freq
x
[
n
]

y
[
n
]

( ) ( )
e t e j k j
e H e H =
+2
Solution: Use an anti-aliasing filter.
FIR y[n]
x[n]
ADC
Analogue
Anti-Aliasing
x(t)
F
s
/2 F
s
/2
Freq
Freq
x
(
t
)

y
[
n
]

Phase Linearity of an FIR Filter
A causal FIR filter whose impulse
response is symmetrical is guaranteed to
have a linear phase response.
0
n
h(n)
1 n n+1 2n+1 2n
N = 2n + 2
0
n
h(n)
1 n n+1 2n 2n-1 n-1
N = 2n + 1
Even symmetry Odd symmetry
A causal FIR filter whose impulse
response is symmetrical (ie h[n] = h[N-1-n]
for n = 0, 1, , N-1) is guaranteed to have
a linear phase response.
Condition Phase |
.
|
\
|
=
2
1 N
k Phase Property Filter Type
| | | | 1 = n N h n h
Positive Symmetry
e k
Linear phase
Odd Symmetry Type 1
Even Symmetry Type 2
Application of 90 linear phase shift:
Signal
separation
90
o
delay

90
o
delay

+
+
+
-
I
Q
Reverse
Forward
IH
QH
t B t A I
r f
e e sin cos + =
t B t A Q
r f
e e cos sin + =
t B t A
t B t A IH
r f
r f
e e
t
e
t
e
cos sin
2
sin
2
cos
+ =
|
.
|
\
|
+ + |
.
|
\
|
+ =
t B Q IH
r
e cos 2 = +
t B I QH
f
e sin 2 =
Design Procedure
To fully design and implement a filter five
steps are required:
(1) Filter specification.
(2) Coefficient calculation.
(3) Structure selection.
(4) Simulation (optional).
(5) Implementation.
Filter Specification - Step 1
(a)
1
f (norm)
fc : cut-of f f requency
pass-band stop-band
pass-band stop-band transition band
1
s
o
pass-band
ripple
stop-band
ripple
fpb : pass-band f requency
fsb : stop-band f requency
f (norm)
(b)
p
1 o +
s
A
p
A
0
-3
p
1 o
fs/2
fc : cut-of f f requency
fs/2
|H(f )|
(dB)
|H(f )|
(linear)
|H(f )|
Coefficient Calculation - Step 2
There are several different methods
available, the most popular are:
Window method.
Frequency sampling.
Parks-McClellan.
We will just consider the window method.
Window Method
First stage of this method is to calculate
the coefficients of the ideal filter.
This is calculated as follows:
( ) ( )
( )
=
=
=
=
=
}
}
0 n for
0 n for
2
sin 2
1
2
1
2
1
c
c
c c
n j
n j
d
f
n
n f
d e
d e H n h
c
c
e
e
e
t
e e
t
e
e
e
e
t
t
Window Method
Second stage of this method is to select a window
function based on the passband or attenuation
specifications, then determine the filter length based on
the required width of the transition band.
Window Type
Normalised Transition
Width (Af(Hz))
Passband Ripple(dB)
Stopband Attenuation
(dB)
Rectangular
N
9 . 0
0.7416 21
Hanning
N
1 . 3
0.0546 44
Hamming
N
3 . 3
0.0194 53
Blackman
N
5 . 5
0.0017 74
Kaiser
54 . 4
93 . 2
= |
N
96 . 8
71 . 5
= |
N
0.0274
0.000275
50
90
( )
132 8
4 . 1 2 . 1
3 . 3 3 . 3
=
=
A
= kHz
kHz f
N
Using the Hamming
Window:
Window Method
The third stage is to calculate the set of
truncated or windowed impulse response
coefficients, h[n]:
( ) ( ) ( ) n W n h n h
d
=
even N for
odd N for
2 2
2
1
2
1
=
=
s s
s s
N
n
N
N
n
N
( )

133
2
cos 46 . 0 54 . 0

2
cos 46 . 0 54 . 0
|
.
|
\
|
+ =
|
.
|
\
|
+ =
n
N
n
n W
t
t
for
Where:
66 66 s s n
for
Window Method
Matlab code for calculating coefficients:
close all;
clear all;

fc = 8000/44100; % cut-off frequency
N = 133; % number of taps
n = -((N-1)/2):((N-1)/2);
n = n+(n==0)*eps; % avoiding division by zero

[h] = sin(n*2*pi*fc)./(n*pi); % generate sequence of ideal coefficients
[w] = 0.54 + 0.46*cos(2*pi*n/N); % generate window function
d = h.*w; % window the ideal coefficients

[g,f] = freqz(d,1,512,44100); % transform into frequency domain for plotting

figure(1)
plot(f,20*log10(abs(g))); % plot transfer function
axis([0 2*10^4 -70 10]);

figure(2);
stem(d); % plot coefficient values
xlabel('Coefficient number');
ylabel ('Value');
title('Truncated Impulse Response');

figure(3)
freqz(d,1,512,44100); % use freqz to plot magnitude and phase response
axis([0 2*10^4 -70 10]);
Window Method
0 0.5 1 1.5 2
x 10
4
-6000
-4000
-2000
0
Frequency (Hz)
P
h
a
s
e

(
d
e
g
r
e
e
s
)
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
x 10
4
-60
-40
-20
0
Frequency (Hz)
M
a
g
n
i
t
u
d
e

(
d
B
)
0 20 40 60 80 100 120 140
-0.1
0
0.1
0.2
0.3
0.4
Coefficient number, n
h
(
n
)

Truncated Impulse Response
Realisation Structure Selection - Step 3
( )

=
1
0
N
k
k
k
z b z H
( ) ( ) ( ) z X z H z Y =
( ) ( ) ( ) ( ) 1 .... 1
1 1 0
+ + + + =

N n x b n x b n x b n y
N
z
-1
z
-1
z
-1
+ + +
b
0
b
2
b
N-1
y(n)
x(n)
b
1
Direct form structure for an FIR filter:
( )

=
1
0
N
k
k
k
z b z H
Linear phase structures:

N even:

N Odd:
( ) ( )
=

+ =
1
2
0
1
N
k
k N k
k
z z b z H
( ) ( )

+ + =
2
1
0
2
1
2
1
1
N
k
N
N
k N k
k
z b z z b z H
(a) N even.
(b) N odd.
+
b
0
+
+
+
+
b
1
+
b
2
+
b
N/2-1
y (n)
(a)
z
-1
z
-1
z
-1
z
-1
z
-1
+
b
0
+
+
+
+
b
1
+
b
2
+
b
(N-3)/2
y (n)
x(n)
b
(N-1)/2
+
(b)
z
-1
z
-1
z
-1
z
-1
z
-1
z
-1
( )

=
1
0
N
k
k
k
z b z H
Cascade structures:
( )
( )
( )
( )
[
+ + =
(
+ + + + =
+ + + + = =
M
k
k k
N N
N
N
N
k
k
k
z b z b b
z
b
b
z
b
b
z
b
b
b
z b z b z b b z b z H
1
2
2 ,
1
1 , 0
1
0
1 2
0
2 1
0
1
0
1
1
2
2
1
1 0
1
0
1
... 1
...
( )

=
1
0
N
k
k
k
z b z H
Cascade structures:
z
-1
+
b
1,1
x(n)
z
-1
+
b
1,2
z
-1
+
b
2,1
z
-1
+
b
2,2
z
-1
+
b
M,1
z
-1
+
b
M,2
y (n)
b
0
Implementation - Step 5
Implementation procedure in C with
fixed-point:
Set up the codec (\Links\CodecSetup.pdf).

Transform: to C code.

(\Links\FIRFixed.pdf)
Configure timer 1 to generate an interrupt at
8000Hz (\Links\TimerSetup.pdf).
Set the interrupt generator to generate an
interrupt to invoke the Interrupt Service
Routine (ISR) (\Links\InterruptSetup.pdf).
| | | |
=
=
1
0
N
k
k
k n x b n y
Implementation procedure in C with
floating-point:
Same set up as fixed-point plus:
Convert the input signal to floating-point
format.
Convert the coefficients to floating-point
format.
With floating-point multiplications there is
no need for the shift required when using
Q15 format.
See \Links\FIRFloat.pdf
Implementation procedure in assembly:
Same set up as fixed-point, however:
is written in assembly.

(\Links\FIRFixedAsm.pdf)

The ISR is now declared as external.

| | | |
=
=
1
0
N
k
k
k n x b n y
Implementation procedure in assembly:
The filter implementation in assembly is
now using circular addressing and
therefore:
The circular pointers and block size register
are selected and initialised by setting the
appropriate values of the AMR bit fields.
The data is now aligned using:

Set the initial value of the circular pointers,
see \Links\FIRFixedAsm.pdf.
#pragma DATA_ALIGN (symbol, constant (bytes))
y0 = b0*x0 + b1*x1 + b2*x2
+ b3*x3
Circular addressing link slide.
time
y
[
n
]

0 1 2
b
0
b
1
b
2
b
3
x
0
x
1
x
2
x
3
x
4
x
1
x
2
x
3
y0 = b0*x0 + b1*x1 + b2*x2
+ b3*x3
y1 = b0*x4 + b1*x1
+ b3*x3
+ b2*x2
time
y
[
n
]

0 1 2
b
0
b
1
b
2
b
3
x
4
x
5
x
2
x
3
y0 = b0*x0 + b1*x1 + b2*x2
+ b3*x3
y1 = b0*x4 + b1*x1
+ b3*x3
+ b2*x2
y2 = b0*x4 + b1*x5
+ b3*x3
+ b2*x2
time
y
[
n
]

0 1 2
b
0
b
1
b
2
b
3
FIR Code
Code location:
Code\Chapter 14 - Finite Impulse Response Filters
Projects:
Fixed Point in C: \FIR_C_Fixed\
Floating Point in C: \FIR_C_Float\
Fixed Point in Assembly: \FIR_Asm_Fixed\
Floating Point in Assembly: \FIR_Asm_Float\
Chapter 14
Finite Impulse Response (FIR) Filters
- End -

Chapter 14 - FIR Filters

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Chapter 14 - FIR Filters

Uploaded by

Copyright:

Available Formats

Finite Impulse Response (FIR) Filters

You might also like