Chapter 9: Filter Design, Multirate, and Correlation
Test Score
Question
Score
9.1
0 / 20
9.2
0 / 20
9.3
0 / 20
9.4
0 / 20
9.5
0 / 20
9.6
0 / 20
9.7
0 / 20
9.8
0 / 20
9.9
0 / 20
9.10
0 / 20
9.11
0 / 20
9.12
0 / 20
9.13
0 / 20
9.14
0 / 20
9.15
0 / 20
9.16
0 / 20
9.17
0 / 20
9.18
0 / 20
9.19
0 / 20
9.20
0 / 20
9.21
0 / 20
9.22
0 / 20
9.23
0 / 20
9.24
0 / 20
9.25
0 / 20
9.26
0 / 20
9.27
0 / 20
9.28
0 / 20
9.29
0 / 20
Total
0 / 100
Test 9.1
The result of applying a Bartlett (triangular) window to
$\{7,2,6,4,5\}$ is:
Solution: Multiply $\{7,2,6,4,5\}$
point-by-point by $\{0,\frac1{2},1,\frac1{2},0\}$.
Test 9.2
We observe $L$ samples of the sum of two sinusoids. We compute
the spectrum using an $N$-point DFT. To reduce sidelobes in the
spectrum, we should:
Solution: Increasing $N$ or $L$ does not
reduce sidelobes.
Test 9.3
We observe $L$ samples of the sum of two sinusoids. We compute
the spectrum using an $N$-point DFT. To resolve (split the peaks)
the two sinusoids, we should:
Solution: Increasing $N$ does not help
resolve peaks. A window makes it harder.
Test 9.4
We observe $L$ samples of the sum of two sinusoids. We compute
the spectrum using an $N$-point DFT. To smooth the spectrum, we
should:
Solution: Increasing $N$ or using a window
smoothes the spectrum.
Test 9.5
A filter with impulse response
$h[n]=\{a,b,\underline{0},-b,-a\}$ is in general:
A continuous-time lowpass filter has ${\bf H} ({\bf s})=\frac1{{\bf s}+1}$ and
$h(t)=e^{-t}\;u(t)$.
Design a discrete-time lowpass filter using impulse invariance with
$T=2$.
Solution: Using Eq. (9.74),
$h[n]=Th(nT)=2e^{-2n}\;u[n]$.
Test 9.10
A continuous-time lowpass filter has ${\bf H} ({\bf s})=\frac1{{\bf s}+1}$ and
$h(t)=e^{-t}\;u(t)$.
Design a discrete-time lowpass filter using bilinear transformation
with $T=2$.
To design a discrete-time lowpass filter with cutoff
$\Omega_0=\frac{\pi}{2}$ from a continuous-time lowpass filter with cutoff
$\omega_0$ using bilinear transformation with $T=0.001$, $\omega_0$ should
be:
Solution: The prewarping formula Eq. (9.98)
is
$$
\omega_0=\frac{2}{T}\tan\frac{\Omega}{2} =\frac{2}{0.001}\tan\frac{\pi/2}{2}=2000.
$$
Test 9.13
FIR filter $H(e^{j\Omega})$ minimizing
$\displaystyle
\int_{-\pi}^{\pi}\left|H(e^{j\Omega})-H_{\rm ideal}(e^{j\Omega})\right|^2
\;d\Omega
$ is designed by:
Solution: By Parseval's theorem, this is
$$
2\pi\sum_{n=-\infty}^{\infty}\left|h[n]-h_{\rm ideal}[n]\right|^2.
$$
So set $h[n]=h_{\rm ideal}[n]$ where possible, which amounts to using
a rectangular window.
Test 9.14
Design a discrete-time bandpass filter with cutoffs $\frac{\pi}{4}$
and $\frac{3\pi}{4}$ using using a 5-point rectangular window:
Design a discrete-time bandpass filter with cutoffs $\frac{\pi}{4}$ and $\frac{3\pi}{4}$ using
frequency sampling with ${\bf H} (e^{j0})={\bf H} (e^{j\pi})=0$ and $|{\bf H} (e^{j\pi/2})|=1$:
Solution: Use $h[n]=\{a,\underline{0},-a\}$ for some constant $a$.
Then ${\bf H} (e^{j0})={\bf H} (e^{j\pi})=0$
and ${\bf H}
(e^{j\Omega})=ae^{j\Omega}-ae^{-j\Omega}=j2a\sin(\Omega)$.
Then
$|{\bf H} (e^{j\pi/2})|=2a\sin(\pi/2)=1$ makes $a=\frac1{2}$.
Test 9.16
The filter type designed by the MATLAB command
firpm(10,[0,0.2,0.3,0.7,0.8,1],[0,0,1,1,0,0]) is:
Solution: Rejects $0<\Omega<0.2\pi$ and $0.8\pi<\Omega<\pi$; passes $0.3\pi<\Omega<0.7\pi$.
Test 9.17
Which of the following is an application of multirate signal
processing?
Solution: See Sections 9-10 and 9-11.
Test 9.18
The easiest way to raise the pitch of a musical signal by one
octave (a factor of 2 in frequency) is to:
Solution: Deleting every other sample of a
sampled sinusoid doubles its frequency.
Test 9.19
We reconstruct from its samples a signal Nyquist-sampled at 2000
samples/s. Our reconstruction filter is poor: it passes 1 kHz and
rejects only above 1 MHz. We can use this poor filter by upsampling
and interpolating the signal by what factor?
Solution: Interpolation by a factor of 1000
removes all copies of spectrum of the sampled signal between 1 kHz and
1 MHz. This suffices: bandwidth 1 kHz $\ll1$ MHz.
Test 9.20
We reconstruct from its samples a signal Nyquist-sampled at
20000 samples/s. Our reconstruction filter is poor: it passes 10
kHz and rejects only above 90 kHz. We can use this poor filter by
upsampling and interpolating the signal by what factor?
Solution: We need to separate the centers of
the copies of the spectrum induced by sampling by 100 kHz, since
baseband occupies $-10<f<10$ kHz and the first image occupies
$S-10<f<S+10$ kHz, where $S$ is the sampling rate after upsampling. To
reject all of the spectrum of the first image, we need $S-10=90\to
S=100$ kHz, by upsampling by a factor of $\frac{100}{20}=5$.
Test 9.21
A sinusoid is sampled at 1000 samples/s, input into the DSP
system shown, and then reconstructed.
300 Hz$\to$
$\bf\uparrow 3$
$\to$
$\bf\downarrow 3$
$\to y(t)$
What is the frequency in Hz of the output sinusoid?
A sinusoid is sampled at 1000 samples/s, input into the DSP
system shown, and then reconstructed.
300 Hz$\to$
$\bf\downarrow 3$
$\to y(t)$
What is the frequency in Hz of the output sinusoid?
Solution: 300 Hz becomes $3(300)=900$Hz,
which aliases to $(1000-900)=100$Hz.
Test 9.23
A sinusoid is sampled at 1000 samples/s, input into the DSP
system shown, and then reconstructed.
200 Hz$\to$
$\bf\uparrow 4$
$\to y(t)$
What is the frequency in Hz of the output sinusoid?
Solution: 200 Hz becomes
$\frac1{4}(200)=50$ Hz, with copies at $\frac1{4}(1000-200)=200$ Hz,
$\frac1{4}(1000+200)=300$ Hz, $\frac1{4}(2000-200)=450$ Hz, and
$\frac1{4}(2000-200)=550$ Hz, which along with higher frequencies is
filtered out by the reconstruction lowpass filter.
Test 9.24
A sinusoid is sampled at 1000 samples/s and input into the DSP
system shown.
500 Hz$\to$
$\bf\downarrow 2$
$\to$
$\bf\uparrow 2$
$\to y[n]$
What is the output?
Solution: $$
x[n]=\cos\left(2\right)\pi\frac{500}{1000}n=\cos(\pi n)=(-1)^n.
$$ Downsampling by 2 removes all of the $-1$'s; upsampling by 2
replaces them with 0's.
Test 9.25
The main application of autocorrelation is to:
Solution: See Section 9-12.2.
Test 9.26
The main application of cross-correlation is to:
Solution: See Section 9-12.4.
Test 9.27
The main application of correlation is to:
Solution: See Section 9-12.6.
Test 9.28
The autocorrelation of $\{\underline{1},2,3\}$ is:
Solution: From Eq. (9.169) and Example 9-20,
$$
\{\underline{1},2,3\}*\{3,2,\underline{1}\}
=\{3,8,\underline{14},8,3\}.
$$
Test 9.29
The cross-correlation of $\{\underline{1},2,3\}$ and
$\{\underline{4},5,6\}$ is:
Solution: From Eq. (9.177) and Example 9-21,
$$
\{\underline{1},2,3\}*\{6,5,\underline{4}\}=\{6,17,\underline{32},23,12\}.
$$