Beruflich Dokumente
Kultur Dokumente
CHRISTIAN-ALBRECHTS-UNIVERSITY
OF KIEL
DIGITAL
SIGNAL PROCESSING AND
SYSTEM THEORY
Initial Remarks
Sebastian Rohde
Digital Signal Processing and System Theory (DSS)
Office: Audio Lab
Phone: 0431 880-6141
e-mail: ser@tf.uni-kiel.de,
Office hours: Vary. Please make an appointment by email!
Digital Signal Processing and System Theory, Prof. Dr.-Ing. Gerhard Schmidt, www.dss.tf.uni-kiel.de
Advanced Digital Signal Processing, Exercises WS 2014/2015
TECHNICAL FACULTY,
CHRISTIAN-ALBRECHTS-UNIVERSITY
OF KIEL
DIGITAL
SIGNAL PROCESSING AND
SYSTEM THEORY
Notation
Symbol
Meaning/Usage
0 (t)
0 (n)
1 (n)
Analog frequency in Hz
fs
Sampling frequency in Hz
v(t)
s V (j)
v(n)
s V (ej )
v(n)
s V ()
Digital Signal Processing and System Theory, Prof. Dr.-Ing. Gerhard Schmidt, www.dss.tf.uni-kiel.de
Advanced Digital Signal Processing, Exercises WS 2014/2015
DIGITAL
SIGNAL PROCESSING AND
SYSTEM THEORY
TECHNICAL FACULTY,
CHRISTIAN-ALBRECHTS-UNIVERSITY
OF KIEL
Problem 1
A complex-valued continuous-time signal va (t) has the Fourier transform shown in figure
1. This signal is sampled to produce the sequence v(n) = va (nT ).
Va (j)
2 .
(b) What is the lowest sampling frequency that can be used without incurring any
aliasing distortion, i.e. so that va (t) can be recovered from v(n)?
Digital Signal Processing and System Theory, Prof. Dr.-Ing. Gerhard Schmidt, www.dss.tf.uni-kiel.de
Advanced Digital Signal Processing, Exercises WS 2014/2015
DIGITAL
SIGNAL PROCESSING AND
SYSTEM THEORY
TECHNICAL FACULTY,
CHRISTIAN-ALBRECHTS-UNIVERSITY
OF KIEL
Problem 2
domain)
Figure 2 shows an overall system for filtering a continuous-time signal using a discretetime filter. The frequency response of the ideal reconstruction filter Hr (j) and the
discrete-time filter are shown below.
p(t) =
va (t)
n= 0 (t
vi (t)
nT )
Convert from
impulse train
v(n)
y(n)
H(ej )
to discrete-time
sequence
Convert to
impulse train
Hr (ej )
5 105
yi (t)
Hr (ej )
yr (t)
H(ej )
2104
/4
/4
2104
Heff (j)
2104
Digital Signal Processing and System Theory, Prof. Dr.-Ing. Gerhard Schmidt, www.dss.tf.uni-kiel.de
Advanced Digital Signal Processing, Exercises WS 2014/2015
DIGITAL
SIGNAL PROCESSING AND
SYSTEM THEORY
TECHNICAL FACULTY,
CHRISTIAN-ALBRECHTS-UNIVERSITY
OF KIEL
Problem 3
(quantization)
Problem 4
(c) Let z(n) be the result of the linear convolution of h(n) and v(n): z(n) = h(n)v(n).
Is z(n) = y(n)?
Digital Signal Processing and System Theory, Prof. Dr.-Ing. Gerhard Schmidt, www.dss.tf.uni-kiel.de
Advanced Digital Signal Processing, Exercises WS 2014/2015
DIGITAL
SIGNAL PROCESSING AND
SYSTEM THEORY
TECHNICAL FACULTY,
CHRISTIAN-ALBRECHTS-UNIVERSITY
OF KIEL
Problem 5
(DFT)
Problem 6
Digital Signal Processing and System Theory, Prof. Dr.-Ing. Gerhard Schmidt, www.dss.tf.uni-kiel.de
Advanced Digital Signal Processing, Exercises WS 2014/2015
DIGITAL
SIGNAL PROCESSING AND
SYSTEM THEORY
TECHNICAL FACULTY,
CHRISTIAN-ALBRECHTS-UNIVERSITY
OF KIEL
(h) Sketch va (t), v(n), the Fourier transform V (ej ) and the DFT VM () for L = 15
and M = 30 (zero padding).
Problem 7
(FFT)
Problem 8
(FFT)
2
The M -point DFT of the M -point sequence x(n) = ej(/M )n , for M even, is
2
X() = M ej/4 ej(/M ) .
2
Suppose that an FFT program is available that computes the DFT of a complex sequence.
If we wish to compute the DFT of a real sequence, we may simply specify the imaginary
part to be zero and use the program directly. However, the symmetry of the DFT of a
real sequence can be used to reduce the amount of computation.
(a) Let x(n) be a real-valued sequence of length M , and let X() be its DFT with
real and imaginary parts denoted XR () and XI (), respectively; i.e.,
X() = XR () + j XI ().
Show that if x(n) is real, then XR () = XR (M ) and XI () = XI (M )
for = 1, ..., M 1.
(b) Now consider two real-valued sequences x1 (n) and x2 (n) with DFTs X1 () and
X2 (), respectively. Let g(n) be the complex sequence g(n) = x1 (n) + j x2 (n),
Digital Signal Processing and System Theory, Prof. Dr.-Ing. Gerhard Schmidt, www.dss.tf.uni-kiel.de
Advanced Digital Signal Processing, Exercises WS 2014/2015
DIGITAL
SIGNAL PROCESSING AND
SYSTEM THEORY
TECHNICAL FACULTY,
CHRISTIAN-ALBRECHTS-UNIVERSITY
OF KIEL
with corresponding DFT G() = GR () + j GI (). Also, let GOR (), GER (),
GOI () and GEI () denote, respectively, the odd part of the real part, the even
part of the real part, the odd part of the imaginary part, and the even part of the
imaginary part of G(). Specifically, for 1 M 1,
GOR () = 1/2{GR () GR (M )},
GER () = 1/2{GR () + GR (M )},
GOI () = 1/2{GI () GI (M )},
GEI () = 1/2{GI () + GI (M )},
and GOR (0) = GOI (0) = 0, GER (0) = GR (0), GEI (0) = GI (0). Determine expressions for X1 () and X2 () in terms of GOR (), GER (), GOI () and GEI ().
Problem 10
The signal flow graph in figure 4 describes the input-output relationship of v(k) and
y(k).
3
v (k )
1/2
y (k )
Y (z)
V (z)
Digital Signal Processing and System Theory, Prof. Dr.-Ing. Gerhard Schmidt, www.dss.tf.uni-kiel.de
Advanced Digital Signal Processing, Exercises WS 2014/2015
DIGITAL
SIGNAL PROCESSING AND
SYSTEM THEORY
TECHNICAL FACULTY,
CHRISTIAN-ALBRECHTS-UNIVERSITY
OF KIEL
Problem 11
()
v k
()
y k
z 1
(
0)
(
0)
r os
r os
r 2 sin2
z 1
(
0 )
()
()
v k
y k
z 1
2r
os(
0)
z 1
z 1
r2
Digital Signal Processing and System Theory, Prof. Dr.-Ing. Gerhard Schmidt, www.dss.tf.uni-kiel.de
Advanced Digital Signal Processing, Exercises WS 2014/2015
DIGITAL
SIGNAL PROCESSING AND
SYSTEM THEORY
TECHNICAL FACULTY,
CHRISTIAN-ALBRECHTS-UNIVERSITY
OF KIEL
Problem 12
y (k)
1/4
.5 for n 0
0 for n < 0
.5(1)n for n 0
0
for n < 0
y (k)
1/4
Digital Signal Processing and System Theory, Prof. Dr.-Ing. Gerhard Schmidt, www.dss.tf.uni-kiel.de
Advanced Digital Signal Processing, Exercises WS 2014/2015
DIGITAL
SIGNAL PROCESSING AND
SYSTEM THEORY
TECHNICAL FACULTY,
CHRISTIAN-ALBRECHTS-UNIVERSITY
OF KIEL
Problem 13
Determine the variance of the round-off noise at the output of the two cascade realizations
of the filter with system function
(1)
(2)
y(n)
v(n)
z 1
z 1
1/2
1/4
e1 (n)
e2 (n)
v(n)
y(n)
z 1
z 1
1/2
1/4
e1 (n)
e2 (n)
Digital Signal Processing and System Theory, Prof. Dr.-Ing. Gerhard Schmidt, www.dss.tf.uni-kiel.de
Advanced Digital Signal Processing, Exercises WS 2014/2015
DIGITAL
SIGNAL PROCESSING AND
SYSTEM THEORY
TECHNICAL FACULTY,
CHRISTIAN-ALBRECHTS-UNIVERSITY
OF KIEL
Problem 14
(filter design)
Determine the unit sample response hi of a linear-phase FIR filter of length L = 4 for
which the amplitude frequency response H0 () at = 0 and = /2 is specified as
H0 (0) = 1,
Problem 15
H0 (/2) = 1/2.
(filter design)
10
DIGITAL
SIGNAL PROCESSING AND
SYSTEM THEORY
TECHNICAL FACULTY,
CHRISTIAN-ALBRECHTS-UNIVERSITY
OF KIEL
(d) What is the delay of the system if L = 21? Sketch (use matlab) the magnitude
of the frequency response of the FIR approximation for this case, assuming a
rectangular window.
(e) What is the delay of the system if L = 20? Sketch (use matlab) the magnitude
of the frequency response of the FIR approximation for this case, assuming a
rectangular window.
Problem 16
(filter design)
Consider a type III linear-phase FIR filter with an amplitude response given by
H03 () = 2
S1
X
i=0
hi sin((S i)).
S
X
c(i) sin(i).
i=1
Show that if the amplitude response is symmetric, i.e., H03 () = H03 ( ), then the
even-indexed impulse response samples hi are zero, if S is even.
Problem 17
(filter design)
Digital filter specifications are often given in terms of the loss function,
Hl () = 20log10 (|H(ej )|), in dB. In this problem the peak passband ripple p and
the minimum stopband attenuation s are given in dB, i.e., the loss specifications of the
digital filter are given by
p = 20log10 (1 1 )dB,
d = 20log10 (2 )dB.
(a) Estimate the order of an optimal equiripple linear-phase lowpass FIR filter with the
following specifications: passband edge Fp = 1.8kHz, stopband edge Fs = 2kHz,
p = 0.1dB, s = 35dB, and sampling frequency FT = 12kHz.
The estimation formula can also be used to estimate the length of highpass, bandpass,
and bandstop optimal equiripple FIR filters. Then the width of the smallest transition
band is used to estimate the filter order.
(b) Estimate the order of an optimal equiripple linear-phase bandpass FIR filter with
Digital Signal Processing and System Theory, Prof. Dr.-Ing. Gerhard Schmidt, www.dss.tf.uni-kiel.de
Advanced Digital Signal Processing, Exercises WS 2014/2015
11
DIGITAL
SIGNAL PROCESSING AND
SYSTEM THEORY
TECHNICAL FACULTY,
CHRISTIAN-ALBRECHTS-UNIVERSITY
OF KIEL
the following specifications: passband edges Fp1 = 0.35kHz and Fp2 = 1kHz,
stopband edges Fs1 = 0.3kHz and Fs2 = 1.1kHz, passband ripple 1 = 0.002,
stopband ripple 2 = 0.001, and sampling frequency FT = 10kHz.
Problem 18
2
1
e0.2 z 1
1
1
e0.4 z 1
(a) Assume that this discrete-time filter was designed by the impulse invariance method
with T = 2, i.e. hi = ha (iT ), where ha (t) is real. Find the system function Ha (s)
of a continuous-time filter that could have been the basis for the design. Is your
answer unique? If not, find another system function Ha (s).
(b) Assume that H(z) was obtained by the bilinear transform with T = 2. Find the
system function Ha (s) that could have been the basis for the design. Is your answer
unique? If not, find another Ha (s).
Problem 19
1
2N
1 + ( cut
)
|H(e )| 0.17783,
0 || 0.2,
0.3 || .
Assume that aliasing will not be a problem, i.e., design the continuous-time Butterworth
filter to meet passband and stopband specifications as determined by the discrete-time
filter.
(a) Sketch the tolerance bounds on the magnitude of the frequency response, |H(j)|,
of the continuous-time Butterworth filter such that after application of the impulse
invariance method, the resulting discrete-time filter will satisfy the given design
specifications. Do not assume that T = 1.
(b) Determine the integer order N and the quantity cut T such that the continuous-
Digital Signal Processing and System Theory, Prof. Dr.-Ing. Gerhard Schmidt, www.dss.tf.uni-kiel.de
Advanced Digital Signal Processing, Exercises WS 2014/2015
12
DIGITAL
SIGNAL PROCESSING AND
SYSTEM THEORY
TECHNICAL FACULTY,
CHRISTIAN-ALBRECHTS-UNIVERSITY
OF KIEL
time Butterworth filter exactly meets the specifications determined in part (a) at
the passband edge.
Problem 20
with = 1/22 1. The figure shows the characteristical parameters for the given
specifications.
Problem 22
Consider the system shown in the figure. For each of the following input signals x(n),
indicate whether the output y(n) = x(n).
(a) x(n) = cos(n/4)
(b) x(n) = cos(n/2)
(c) x(n) = ( sin(n/8)
)2
n
Digital Signal Processing and System Theory, Prof. Dr.-Ing. Gerhard Schmidt, www.dss.tf.uni-kiel.de
Advanced Digital Signal Processing, Exercises WS 2014/2015
13
DIGITAL
SIGNAL PROCESSING AND
SYSTEM THEORY
TECHNICAL FACULTY,
CHRISTIAN-ALBRECHTS-UNIVERSITY
OF KIEL
| H(ej ) |2
1
1
1+2
22
p
x(n)
s
y(n)
H(ej )
H(ej )
1
3
Problem 23
Consider the systems shown in the figure. Suppose that H1 (ej ) is fixed and known.
Find H2 (ej ), the frequency response of an LTI system, such that y2 (n) = y1 (n), if the
inputs to the systems are the same.
x(n)
y1 (n)
2
x(n)
H1
(ej )
H2
y2 (n)
(ej )
Digital Signal Processing and System Theory, Prof. Dr.-Ing. Gerhard Schmidt, www.dss.tf.uni-kiel.de
Advanced Digital Signal Processing, Exercises WS 2014/2015
14
DIGITAL
SIGNAL PROCESSING AND
SYSTEM THEORY
TECHNICAL FACULTY,
CHRISTIAN-ALBRECHTS-UNIVERSITY
OF KIEL
Problem 24
The system shown in the figure approximately interpolates the sequence x(n) by a factor
L. Suppose that the linear filter has impulse response h(n) such that h(n) = h(n) and
h(n) = 0 for |k| > RL 1, where R and L are integers; i.e., the impulse response is
symmetric and of length 2RL 1 samples.
x(n)
L
v(n)
y(n)
H(ej )
(a) In answering the following, do not be concerned about causality of the system; it
can be made causal by including some delay. Specifically, how much delay must
be inserted to make the system causal?
(b) What conditions must be satisfied by h(n) in order that y(n) = x(n/L) for
n = 0, L, 2L, 3L, . . . ?
(c) By exploiting the symmetry of the impulse response, show that each sample of
y(n) can be computed with no more than RL multiplications.
(d) By taking advantage of the fact that multiplications by zero need not to be done,
show that only 2R multiplications per output sample are required.
Problem 25
Consider the noninteger sampling rate conversion in the figure. Develope step by step
an efficient structure for the sampling rate conversion, where most calculations are done
in the lowest possible sampling rate.
Y (z)
X(z)
G(z)
Digital Signal Processing and System Theory, Prof. Dr.-Ing. Gerhard Schmidt, www.dss.tf.uni-kiel.de
Advanced Digital Signal Processing, Exercises WS 2014/2015
15