Dsp Processors III
-
Upload
rajendra-kumar-rayala -
Category
Documents
-
view
233 -
download
0
Transcript of Dsp Processors III
-
7/28/2019 Dsp Processors III
1/80
DSP PROCESSORS-III
-
7/28/2019 Dsp Processors III
2/80
Module 3
-
7/28/2019 Dsp Processors III
3/80
Syllabus
Digital Filters
IIR
FIR
Adaptive filters DIT and DIF algorithms
-
7/28/2019 Dsp Processors III
4/80
Discrete-time system
Discrete-t ime systemhas discrete-time input and output signals
-
7/28/2019 Dsp Processors III
5/80
-
7/28/2019 Dsp Processors III
6/80
-
7/28/2019 Dsp Processors III
7/80
A discrete-time filter is a discrete-time system that passescertain frequency components and stops others.
A Digital filter can be either an IIR or FIRfilter.
An IIRfilter usually requires less cost, i.e., less computationand memory.
However, an FIRfilter usually has better performance,
especially in the phase response( a generalized linear phase isrequired).
Discrete-Time Filters
-
7/28/2019 Dsp Processors III
8/80
IIR filter
IIR filter have infinite-duration impulseresponses, hence they can be matched to analog
filters, all of which generally have infinitely longimpulse responses.
-
7/28/2019 Dsp Processors III
9/80
-
7/28/2019 Dsp Processors III
10/80
Analysis of Ideal Filters
Consider an ideal low pass filter.
Over period (-, ), the frequency response of an ideal low passfilter is defined as
.
otherwise,0
||),jexp()(H
c
Let xi(n)=Aiexp (jin) be a frequency component of the input signal.
Then, the corresponding output signal of this filter is
yi(n)=Ai exp (jin) H (i).
If|i| c, thenyi(n)=Aiexp [ji (n-)].
ie, xi(n) is passed with a constant delay .
-
7/28/2019 Dsp Processors III
11/80
.
n,
n,)n(
)n(sin
)n(h
c
c
If|i|>c, thenyi(n)=0.
ie, xi(n) is stopped.
The impulse response of the ideal low pass filter is
Since h(n)0 for n
-
7/28/2019 Dsp Processors III
12/80
-
7/28/2019 Dsp Processors III
13/80
Prototype Analog Filters
IIR filter design techniques rely on existing analogfilters to obtain digital filters.
We designate these analog filters asprototype filters.
The prototypes widely used in practice
Butterworth lowpass Chebyshev lowpass (Type I and II)
-
7/28/2019 Dsp Processors III
14/80
Butterworth Lowpass Filter
This filter is characterized by the property that itsmagnitude response is flatin both passband and
stopband.
Moderate phase distortion
N
c
a jH 22
1
1|)(|
The magnitude-squared response of an N-order lowpassfilter
-
7/28/2019 Dsp Processors III
15/80
0 10 20 30 40 50 60 70 80 90 1000
0.2
0.4
0.6
0.8
1
1.2
N=1
N=2
N=200
N=100
Plot of the magnitude-squared response
-
7/28/2019 Dsp Processors III
16/80
Properties of Butterworth Filter
Magnitude Response
|Ha(0)|2 =1
|Ha(jc) |2 =0.5, for all N
|Ha(j)|2monotonically decrease for Approaches to ideal filter when N
-
7/28/2019 Dsp Processors III
17/80
-
7/28/2019 Dsp Processors III
18/80
Chebyshev Lowpass Filter
Chebyshev -I filtersHave equirippleresponse in thepassband
Chebyshev -II filtersHave equirippleresponse instopband
Butterworth filters
Have monotonic response in both bands
We note that by choosing a filter that has an equripple ratherthan a monotonic behavior, we can obtain a low-orderfilter.
(The maximum delay, in samples, used in creating eachoutput sample is called the orderof the filter )
Therefore Chebyshev filters provide lower orderthan
Buttworth filters for the same specifications.
-
7/28/2019 Dsp Processors III
19/80
Chebyshev -I filters
Chebyshev -II filters
-
7/28/2019 Dsp Processors III
20/80
The magnitude-squared response of Chebyshev -I filter
xx
xxNxT
T
jH
N
c
N
a
1)),(cosh(cosh
10)),(coscos()(
1
1|)(|
1
1
22
2
N is the order of the filter, Epsilon is the passband ripple factor
(a) For 0
-
7/28/2019 Dsp Processors III
21/80
-
7/28/2019 Dsp Processors III
22/80
Observations
At x=0 (or =0);
|Ha(j0)|2 = 1; for N odd;
= 1/(1+2); for N even
At x=1 (or = c);
|Ha(j1)|2 =1/(1+2);for all N.
For 0
-
7/28/2019 Dsp Processors III
23/80
-
7/28/2019 Dsp Processors III
24/80
-
7/28/2019 Dsp Processors III
25/80
-
7/28/2019 Dsp Processors III
26/80
Impulse Invariance Transformation
The digital filter impulse response look similar to that of a frequency-selective analog filter.
basic principle is sampling of impulse response of an analog filter
Sample ha(t) at some sampling interval T to obtain h(n)
h(n)=ha(nT)
Sincez = e jwon the unit circle ands = j on the imaginary axis,we have the following transformation from thes-plane to thez-plane:
z=e sT
also,
and
Tjjw eeorTw
k
ak
T
jsH
T
zH21
)((Frequency-domainaliasing
formula)
-
7/28/2019 Dsp Processors III
27/80
Complex-plane mapping in impulse invariance transformation
-
7/28/2019 Dsp Processors III
28/80
Properties:
Sigma < 0, maps into |z| 0, maps into |Z| >1 (outside of the Unit circle)
Causality and Stability are the same without changing Aliasing occur if filter not exactly band-limited
Let digital lowpass filter specifications be wp ws Rp and As
-
7/28/2019 Dsp Processors III
29/80
Let digital lowpass filterspecifications be wp,ws,Rp and As.
To determine H(z) , first design an equivalent analog filter and then
mapping it into the desired digital filter.
Design Procedure:
1. Choose Tand determine the analog frequencies:
p =wp/T, s = ws/T
2. Design an analogfilter Ha(s) using the specifications with one of theprototypes of the previous section.
3. Using partial fraction expansion, expand Ha(s) into
4. Now transform analog poles {pk} into digital poles {epkT} to obtain the
digital filter
Nk
k
ka
ps
RsH
1)(
N
k Tp
k
ze
RzH
k1 1
1
)(
Ad
-
7/28/2019 Dsp Processors III
30/80
Advantages
It is astable design and the frequencies and w are linearlyrelated.
Disadvantage
We should expect some aliasingof the analog frequencyresponse, and in some cases this aliasing is intolerable.
This design method is useful only when the analog filter isessentially band-limited to a lowpass or band pass filter in
which there are no oscillations in the stopband.
i i f i
-
7/28/2019 Dsp Processors III
31/80
Bilinear Transformation
transforms continuous-time system representationsto discrete-time and vice versa
This mapping is the besttransformation method.
It maps positions on the j axis, in the s-plane to theunit circle in the z-plane
For every feature in the frequency response of the analogfilter, there is a corresponding feature, with identical gain
and phase shift, in the frequency response of the digital
filter basic principle: application of the trapezoidal formula for
numerical integration of differential equation
2/1
2/1
1
121
1
sT
sT
zz
z
Ts
-
7/28/2019 Dsp Processors III
32/80
Complex-plane mapping in bilinear transformation
-
7/28/2019 Dsp Processors III
33/80
Observations
Sigma < 0|z| < 1,
Sigma = 0|z| = 1,
Sigma > 0|z| > 1
The entire left half-plane maps into the inside of theUnit circle. This is a stabletransformation.
The imaginary axis maps onto the Unit circle in aone-to-one fashion. Hence there is no aliasingin the
frequency domain.
Relation of to is nonlinear = 2tan-1(T/2) = 2tan(/2)/T;
Given the digital filter specifications wp ws Rp and As we want to
-
7/28/2019 Dsp Processors III
34/80
Given the digital filter specifications wp,ws,Rp and As, we want to
determine H(z).
The design steps:
1. Choose a value forT. this is arbitrary, and we may set T = 1.
2. Prewarp the cutoff freq. wp and ws;
ie, calculate p and s usingp = 2/T*tan(wp/2), s = 2/T*tan(ws/2)
3. Design an analog filter Ha(s) to meet the specifications.
4. Finally set
And simplify to obtain H(z) as a rational function in z-1
1
1
112)(
zz
THzH a
-
7/28/2019 Dsp Processors III
35/80
Advantage of the bilinear Transformation
It is a stable design There is no aliasing
There is no restriction on the type of filter that canbe transformed.
FIR FILTER
-
7/28/2019 Dsp Processors III
36/80
FIR FILTER
A finite impulse response filter is a type of a digital filter.
The impulse response is 'finite' because it settles to zero ina finite number of sample intervals. The impulse response of an Nth order FIR filter lasts for
N+1 samples, and then dies to zero.
Properties: Inherently stable - all the poles are located at the origin
and thus are located within the unit circle.
Require no feedback- any rounding errors are notcompounded by summed iterations. This also makesimplementation simpler.
can be designed to be linear phase - which means the phasechange is proportional to the frequency.
Disadvantage:
-
7/28/2019 Dsp Processors III
37/80
Disadvantage:
considerably more computation power is required comparedwith a similar IIR filter.
This is especially true when low frequencies are to be affected
by the filter.
Design methods:
Window design method Frequency Sampling method Weighted least squares design
Minimax design Equiripple design
-
7/28/2019 Dsp Processors III
38/80
Window design method
Simplest FIR filter design is window function technique Specific window should be chosen according to desiredconstraints
the impulse response of the causal FIR filter isobtained by windowing the ideal filter.
-
7/28/2019 Dsp Processors III
39/80
windowing hd(n), we can obtain the impulse response of a
causal FIR lowpass filter. ie,
h(n) = hd(n) w(n)
where w(n) is a window function and is equal to 0 for n
-
7/28/2019 Dsp Processors III
40/80
-
7/28/2019 Dsp Processors III
41/80
Window functions
The rectangular window is defined as
.otherwise,0
1Nn0,1)n(w
The Bartlett window is defined as
.
otherwise,0
1Nn1)/2(N),1N/(n22
2/)1N(n0),1N/(n2
)n(w
-
7/28/2019 Dsp Processors III
42/80
.otherwise,0
1Nn0,
1N
n2cos5.05.0
)n(w
The Hamming window is defined as
.
otherwise,0
1Nn0,1Nn2cos46.054.0
)n(w
The Blackman window is defined as
.
otherwise,0
1Nn0,1N
n4cos08.0
1N
n2cos5.042.0
)n(w
The Hann window is defined as
-
7/28/2019 Dsp Processors III
43/80
Bartlett WindowRectangular Window
Hann Window
-
7/28/2019 Dsp Processors III
44/80
Blackman Window
Hamming Window
-
7/28/2019 Dsp Processors III
45/80
It can be seen that when the length of the window is fixed, a narrow
transition band corresponds to large ripples
Type of Window
Width of
Transition Band
Peak Amplitude
of Ripples
Rectangular 4/N -21dB
Bartlett 8/(N-1) -25dB
Hann 8/(N-1) -44dBHamming 8/(N-1) -53dB
Blackman 12/(N-1) -74dB
Features of Commonly Used Windows.
-
7/28/2019 Dsp Processors III
46/80
The Kaiser window is defined as
.
otherwise0,
1Nn0,)(I
1
1N
n21I
)n(w
0
2
0
Here, I0() is the zero-order modified Bessel function of the first kindand is a shape parameter.
should be selected to obtain a good tradeoff between the width of
the transition band and the amplitudes of the ripples.
-
7/28/2019 Dsp Processors III
47/80
The windowing method is carried out in the following
steps:
1. Determine the type of the window according to the
specification for the amplitudes of the ripples.
2. Find the length of the window according to the
specification for the width of the transition band.
3. Find
4. Determine the impulse response of the causal FIR lowpass
filter
Ad ti filt
-
7/28/2019 Dsp Processors III
48/80
Adaptive filter An adaptive filter is a filter that self-adjustsits
transfer function according to an optimizingalgorithm.
Because of the complexity of the optimizingalgorithms, most adaptive filters are digital filters
that perform digital signal processing and adapt theirperformance based on the input signal.
In IIR and FIR filters process parameters are
known in advance or variation is assumed to beknown
Adaptive filters are best suited when signalconditions are slowly changing or there is large
uncertainty and filter has to compensate for that
Basic adaptive filter:
-
7/28/2019 Dsp Processors III
49/80
Here the adaptive filter output y is compared with adesired signal d to yield a error signal e which is fed
back to adaptive filterCoefficients of adaptive filter is adjusted or optimized
using a least mean square algorithm based on theerror signal
Adaptive filter
d
x
+
-
e
y
p
Output of adaptive filter:
-
7/28/2019 Dsp Processors III
50/80
p p
y(n)= wk(n) x(n-k)
where wk(n) represent N weights for a specific time n.
The performance measure is based on the error signal.
The error signal, e(n) = d(n)-y(n)
The weights of wk(n) are adjusted such that a mean squared
error function is minimized
Mean square error function is E[e2
(n)],where E represents theexpected value
K=0
N-1
Least mean square algorithm:
-
7/28/2019 Dsp Processors III
51/80
Least mean square algorithm:
The simple and effective means of updating the weightswithout averaging or differentiating
Wk(n+1) = wk(n)+2e(n). x(n-k) , k=0,1,,N-1Where x(n) is the input and is adaptive step size.
For each time n, weight wk(n) is updated or replaced bynew coefficients, unless the error function is zero.
After the output y(n), error signal e(n) and wk(n) areupdated for a specific time n, new sample is taken and theadaptation process is repeated
Applications of adaptive filters
-
7/28/2019 Dsp Processors III
52/80
Applications of adaptive filters
Noise cancellation
Signal prediction
Adaptive feedback cancellation
Echo cancellation
For noise cancellation:
-
7/28/2019 Dsp Processors III
53/80
For noise cancellation:
The basic adaptive structure is modified for a noisecancellation system
Adaptive filter
d+n
n
+
-
e
The desired signal d is corrupted by noise n.
The input to the adaptive filter is noise n that is correlated with noise n (noise
n could come from same source as n but modified by environment) The output y is adapted to noise n
When this happens the error signal approaches to desired signal d.
The overall output is e and not the adaptive filter output y.
y
For system identification:
-
7/28/2019 Dsp Processors III
54/80
y
The same input u is given to the unknown system (plant) and to theadaptive filter.
The error signal e is the difference between response d and the filter
output y. The error signal e is fed to the adaptive filter and is used to update thefilter coefficients, until y = d
When this happens the adaption process is finished and e approaches tozero.
Thus the unknown system is modeled.
Signal Prediction:
-
7/28/2019 Dsp Processors III
55/80
Signal Prediction: Used to provide a prediction of the present value of a random signal
This is like interference cancellation, but the adaptive filter uses adelayed version of the primary signal as the reference.
Parameters
u = input of adaptive filter = delayed version of random signal
y = output of adaptive filter
d = desired response = random signal
e = d - y =estimation error = system output
DFT
-
7/28/2019 Dsp Processors III
56/80
DFT
The discrete Fourier transform (DFT) is one of the
specific forms of Fourier analysis.
It transforms one function into another, which is calledthefrequency domain representation of the original
function (the time domain).
IDFT does the inverse operation
But the DFT requires an input function that is discrete andwhose non-zero values have a limited (finite) duration.
Such inputs are often created by sampling a continuousfunction, like a person's voice
-
7/28/2019 Dsp Processors III
57/80
-
7/28/2019 Dsp Processors III
58/80
The FFT is based on the divide-and-conquer paradigm
-
7/28/2019 Dsp Processors III
59/80
-
7/28/2019 Dsp Processors III
60/80
-
7/28/2019 Dsp Processors III
61/80
-
7/28/2019 Dsp Processors III
62/80
-
7/28/2019 Dsp Processors III
63/80
-
7/28/2019 Dsp Processors III
64/80
-
7/28/2019 Dsp Processors III
65/80
-
7/28/2019 Dsp Processors III
66/80
-
7/28/2019 Dsp Processors III
67/80
-
7/28/2019 Dsp Processors III
68/80
-
7/28/2019 Dsp Processors III
69/80
-
7/28/2019 Dsp Processors III
70/80
The DIF FFT is the transpose of the DIT FFT
-
7/28/2019 Dsp Processors III
71/80
The DIF FFT is the transposeof the DIT FFT
To obtain flowgraph transposes:
Reverse direction of flowgraph arrows Interchange input(s) and output(s)
DIT butterfly: DIF butterfly:
Comparing DIT and DIF structures:
-
7/28/2019 Dsp Processors III
72/80
Comparing DIT and DIF structures:
DIT FFT structure: DIF FFT structure:
Steps involved in DIF FFT algorithm:
-
7/28/2019 Dsp Processors III
73/80
Data shuffling not required N=2p & input is separated into two parts .
The first set consists of first N/2 input samples with n ranging from0 to N/2-1
Second set consists of the remaining N/2 input samples with n
ranging from N/2 to N-1
X(k) is decimated into even & odd numbered points (DIF)decomposing an N point DFT into 2 N/2Point DFTs
The procedure is repeated till we get a 2-point DFT.
The 2-point DFT is then converted to butterfly to get the completestructure.
-
7/28/2019 Dsp Processors III
74/80
-
7/28/2019 Dsp Processors III
75/80
-
7/28/2019 Dsp Processors III
76/80
DIF or DIT?
-
7/28/2019 Dsp Processors III
77/80
In terms of computational work load, both perform exactlythe same number of butterflies.
Each butterfly requires exactly one complex multiply andtwo complex adds.
The most significant difference between simple DIF
and DIT algorithms is that DIF starts with normalorder input and generates bit reversed order output.
In contrast, DIT starts with bit reversed orderinput and generates normal order output.
So use DIF for the forward transform and DIT forthe inverse transform
Advantages:
-
7/28/2019 Dsp Processors III
78/80
g
(i) Reasonably good saving in terms of computation, i.e. it
is efficient
(ii) Short Program required
(iii) Easy to understand
Disadvantages:
(i) A relatively large number of operations still required,
especially multiplications
(ii) Problem of the generation of WNfor n=0,1,2,N-1.
Applications of FFT
-
7/28/2019 Dsp Processors III
79/80
Applications of FFT
Linear convolution
(1) Append zeros to the two sequences of lengthsNand
M, to make them of lengths an integer power of two
that is larger than or equal toM+N-1.(2) Apply FFT to both zero appended sequences
(3) Multiply the two transformed domain sequences
(4) Apply inverse FFT to the new multiplied sequence
-
7/28/2019 Dsp Processors III
80/80
THE END