Sie sind auf Seite 1von 29

Exam #4 Review

Challenge 25
Exam #4 Review
Topical Coverage: FIR/IIR

Exam #4 Review

Lesson 25 Review Example


A -3dB 2nd order analog lowpass Butterworth filter having 1000 r/s
passband is given by:

10 6
H s 2
s 2 103 s 106
If the sample frequency is fs=1000 Sa/s, what is the bilinear z-transform
version of H(s)?
Some aliasing should be expected when using the standard z-transform.
None, however, should be expected for the bilinear z-transform model.

Bilinear z-transform: s=(2/Ts)(z-1)/(z+1)

106
H z 2
s 2 103 s 106

106

4 10

z 1 2 2
z 1 2

2 106

2 z 1
Ts z 1

0.128 z 1
2
z .767 z 0.278
2

z 1 106
z 1

Digital filter poles are located at z= 0.384 j 0.362.

H(z) Impulse
Invariant
1000(-1/2j1 /2)
-1/2j1/2
0.375j0.32
Hp(s) Prototype
H(s)

Pole Comparisons
H(z) Bilinear
0.384j0.36

Lesson 25 Snapshot
Bi-linear Design Strategy
Step-by-step design procedure
Digital
Filter
Requirements

Step 1:
Prewarp digital frequencies

Hp(s)

Determine analog prototype


filter order N.
Compute the analog
prototype Hp(s)

Step-by-step design procedure


Step 2:

Hp(s)

Convert analog prototype Hp(s) into


analog filter H(s) using frequencyfrequency transformations.

H(s)

(critical analog frequencies have


been prewarped)

Step-by-step design procedure

Step 3:

H(s)

Convert analog H(s) into digital


filter H(z) using the bi-linear ztransform

H(z)

(prewarped analog frequencies are


automatically warped to their
original values)

Example
Design a Butterworth lowpass filter that meets the following
specifications:
Minimum passband attenuation = -3dB
Passband f[0,1]kHz
Minimum stopband attenuation = -10dB
Stopband f[2,5] kHz
Sample frequency fs=10kHz

Step 1: Given fp=1 kHz, fa=2 kHz and fs=10 kHz, determine the prewarped analog frequencies.

p 2 f p / f s 2 0.1 0.2
p

(normalized w.r.t. fs)

2
tan( p / 2) 20 103 tan(0.1 ) 6498r/s 1.0345kHz ~ f p
Ts

a 2 fa / fs 2 0.2 0.4 (normalized w.r.t. fs)


2
a
tan( a / 2) 20 10 3 tan(0.2 ) 14531r/s 2.312kHz fa
Ts
(General rule)

p p ; a a
9

Step 1 (continued): Design the analog prototype filter Hp(s).


From pre-warped data p, a, and magnitude frequency response
(Ap, Aa) it is determined that a 2nd order Butterworth solution is
required. A 2nd order Butterworth prototype filter has a transfer
function given by:
H p (s )

1
s 2 1.414.s 1

10

Step 2: Define analog filter model.


Perform a lowpass-to-lowpass frequency-frequency transform s=s/p,
relocating the passband cutoff frequency from p=1 r/s to the pre-warped
value of p=6.498103 r/s (1034Hz). The resulting analog filter
becomes:

H a ( s) H p ( s)

s s / 6498

4.3 10 7
H a ( s) 2
s 9.2 103 s 4.3 10 7

11

Step 3: Bilinear z-transform.


Apply the bi-linear z-transform to H(s) to obtain H(z):
0.0676( z + 1)2
H (z) 2
z 1.142z 0.412

This process has been reduced to software.


b=0.0676*[1,2,1]
a=[1, -1.142, 0.412]
[H,w]=freqz(b,a);
plot(w,abs(H))

Practice Exam #4

Exam #4 Review

13

Problem 1: FIR
A 7th order Type 1 FIR has the impulse response h[k]={-0.1, 0, 0.3, 0.5,
0.3, 0, -0.1} . The magnitude frequency response is shown for a
sample rate of fs=1k Sa/s.
Type 1: Odd (MATLAB even) order, even symmetry.
0. 6

0.5

0. 5

0. 4

0.3

0. 3

0. 2

0. 1

-0.1

-0. 1

0.3

-0.1

1.4

1.2

1.0
0.9

0.8

Nulls

0.6

0.4

0.2

0.0

0.5

1.5

2.5

Normalized positive baseband


Exam #4 Review

3.5

0.5
14

a)

What is the filters transfer function H(z)?


H(z)= -0.1+ 0z-1+ 0.3z-2+ 0.5z-3+ 0.3z-4+ 0z-5 -0.1z-6

b)

What is the filters DC gain?


H(z)|z=1= H(1)= -0.1+0+0.3+0.5+0.3+0-0.1=0.9

Exam #4 Review

15

c) What is FIRs worst case gain?


Gmax= |hi| = |-0.1|+|0|+|0.3|+|0.5|+|0.3|+|0|+|-0.1|=1.3 < 21.
Implies 1-interger bit of headroom be added to the filters
accumulator.
d) What is the FIRs group delay measured in seconds?
FIR is a Type I linear phase filter ()= -g, where for N=7
(multiplication by 0 still counts), g=(N-1)/2=3. At fs=1000 Sa/s, a
3 sample delay that corresponds to a 3 ms delay.

Exam #4 Review

16

e)

Assume that the input to the FIR is x[k]=cos(2(0.25)k). What is the


phase difference between the input and output signals?
The normalized input frequency is = /2 and the group delay is g
= (N-1)/2=3. The phase response of a Type I lowpass FIR is ()=
-g+0 or -3. Therefore the phase difference at = /2 is () =
-3/2.

Exam #4 Review

17

f)

Assume that a linear phase equiripple FIR is redesigned to have uniformly22 45


weighted errors of p = a=0.01 and a transition bandwidth of =0.1. What is
the approximate filter order?
The FIR order approximately is (Lesson 22, Slide 45)
N=(-10 log10 (ap)-15)/(14) or (-10(-4)-15)/(1.4)=17.8 ~ 18.

e)

A non-uniform error weighted filter (from 1.f) has p=0.1. What is the value of a?
uniform = p a or 0.01 = 0.1 a or a = 0.001

Exam #4 Review

18

Problem 2: IIR
A 1st order digital low pass filter, having a 3dB passband
ranging over f[0, 1/5] Hz, is sampled at a rate fs=1 Hz.
1.0
0.707
0

0.2

0.5

The analog filter H(s) is based on a 1st order Butterworth


analog prototype model:
Hp(s)=1/(s+1),
having its 3dB cut off frequency located at p =1 r/s.

Exam #4 Review

19

a. What is H(z)? (using the bilinear z-transform)


By the book. The digital filters passband cutoff frequency is f = 1/5 = 0.2
Hz. Therefore the normalized digital filters critical passband frequency is:
1/ 5
2 0.2 1.2566 r / s.
1

p 2

which defines the pre-warped analog filters critical passband cut off
frequency to be:

p
2
2 0.2
2 tan
p tan
1.45 r / s (0.23Hz )
Ts
2

2
p p

Exam #4 Review

20

Therefore the frequency-frequency transformed analog lowpass filter is:

Ha s H p s

s s / p

1
s
1
1.45

1.45
s 1.45

then apply the bilinear z-transform:


H z H a s s 2 z 1
Ts z 1

1.45
s 1.45 s 2 z 1

z 1

1.45
1.45 z 1
0.42 z 1

z 1 1.45 3.45z 0.55 z 0.16


2
z 1

Exam #4 Review

21

From this point on, assume that the digital filters transfer function is
H(z)=(1/2) (z+1)/(z-(1/2))
b. What is the DC gain of your digital filter?

H 1

0 .5 * 2
2
0 .5

c. What is the magnitude and phase response of your digital filter


measured at the Nyquist frequency?

H 1

0.5 z 1
z 0 .5

z 1

0 .5 * 0
0 0180
1 .5

Exam #4 Review

22

Problem 3: FIR
A 7th order lowpass FIR filter is designed to suppress signals above 6kHz
and pass signals out to 4kHz . The filter is clocked at 20k Sa/s and has an
assumed impulse response:
h[k] = {1/2, 1/4, 1/8, 1/16, 1/32, 1/64, 1/128}.

Exam #4 Review

23

a. Is the filter a linear phase FIR?


The FIR coefficients h[k] = {1/2, 1/4, 1/8, 1/16, 1/32, 1/64, 1/128} are not
symmetrically distributed. The FIR is therefore non-linear phase.
b. What is the FIRs transfer function (i.e., H(z))?
H(z)= 1/2 + 1/4 z-1+1/8 z-2+ 1/16 z-3+1/32 z-4+ 1/64 z-5+ 1/128 z-6
c. The FIRs input to your filter is assumed to be bounded by unity (i.e., |
x[k]|1.0). What is the maximum value of the output y[k] and what input
signal x[k] produces this maximal output?
6

G hi (
k 0

1 1 1
1
1
1
1

) 0.992 y k 1
2 4 8 16 32 64 128

or G 27 1 / 2 7 0.992 y k 1
and xworst_case_input[k]={1,1,1,1,1,1,1}

Exam #4 Review

24

Problem 4: IIR
A 1st order Butterworth digital lowpass IIR filter H(z) is to be designed
using the bilinear z-transform, and sampled at a 1 Sa/s rate. The IIR has
a -3dB passband defined over f[0, 0.25] Hz. The analog prototype is
given by:
Hp(s)=1/(s+1),
The analog and digital filters are illustrated below.
1
0.707

|H(ej)|
0

Exam #4 Review

25

a. What is the value of p (analog passband cutoff frequency) in


radians/second?
The normalized digital passand frequency is p = 2 f3db/fs = 2
(0.25)=1.5707 r/s.
Therefore the analog filters prewarped passband frequency is:

p
2
2 0.25

2 tan
p tan

2
tan

2 r/s
Ts
2

4
2
~ 0.318 Hz>0.25Hz

Exam #4 Review

26

b. Suppose the p from Part 4.a is p = 2 r/s. Derive the digital filters
transfer function using the bi-linear z transform.

For p 2; H s H p (s )

s
s
p

p
s p

2
; therefore
s2

2
2
2 z 1

s 2 z 2 z 1 2 z 1 2 2z 2 (2z 2)
Ts z 1
z 1
z 1
2 z 1 0.5 z 1

4z 0
z

H z H s z 2 z 1

Exam #4 Review

27

c. What is the filters DC gain?


H(z=1)=0.5(2)/1=1
d. What is the filters gain and phase response measured at f=0.25 Hz?
H z

0.5 z 1
z

z e j j

0.5 j 1
2

45
j
2

e. What is the output if x[k]=1.414 cos(2f/fs), where f=0.25 Hz?

H z

2
45 ( part 4.d )
2

Then y[k]= 1 cos((2f/fs) -45)


Exam #4 Review

28

Exam #4 Review

29

Das könnte Ihnen auch gefallen