Sunteți pe pagina 1din 4

EE411 - Digital Signal Processing

(Spring 2014)

Assignment 4

Problem 1:
a) For the block diagram shown below, specify (and briefly justify) the conditions under which
() = () .

a) Referring to the block diagram given in part a, is there 1-1 correspondence between () and
[]? If so, write an equation that describes the relation between them. Otherwise, justify your
answer.
b) Referring to the block diagram given in part a, is there 1-1 correspondence between () and
[]? If so, write an equation that describes the relation between them. Otherwise, justify your
answer.
c) For the block diagram shown below, specify (and briefly justify) the conditions under which
() = () .

d) For the block diagram shown below, specify (and briefly justify) the conditions under which
[] = [].

e) A continuous-time signal is composed of the following tones: 250 Hz, 450 Hz, 1 kHz, 2.75 kHz,
4.05 kHz. The signals is sampled at Fs = 1500 samples/s and the resultant discrete-time sequence
is played through the speakers (assume ideal reconstruction). What tones do you expect to hear in
the resultant sound?

f) Referring to the block diagram given in part a, xc(t)=cos(1500t+/4), T = 1/850 seconds. Find
().
g) Assume that xc(t)=cos(2Ft), T =0.001s and xr(t)=cos(20t). What are two possible values of F
that could have resulted in the given pair of xc(t) and xr(t)? Write the expressions for x[n]
corresponding to those values of F.
Problem 2 (D/C conversion):
Suppose [] = [] + 2 [ 1] + 3[ 2] was obtained by sampling () at Fs = 1 sample/sec
while satisfying the Nyquists criterion.
a) Find numerical value of () for the following values of t: 0, 2, 1.5, 2.35. How many samples of
[] are needed to obtain () for a given value of t?
b) Sketch (). Is it limited in time? Explain.
c) Sketch the reconstructed signal () when the following reconstruction filters are used (you do
not need to use a calculator):
(zero-order hold)
a. _0 () = () ( )
b. _1 () =

1
0,

||
,

| |

(1st order hold)

How many samples of [] are needed to compute () using these reconstruction filters?

d) Compare the reconstruction filters given in part c to the ideal reconstruction filter in terms of
a. their ability to reconstruct/approximate () using its samples (justify your answers
using frequency domain analysis). Assume that Nyquists criterion is met.
b. possibility of real-time implementation
Problem 3 (Fractional delay):

Suppose [] = [] + 2 [ 1] + 3[ 2] was obtained by sampling () ([] = ()) at Fs


= 1 sample/sec while satisfying the Nyquists criterion.
a) Using [], compute the samples of a sequence 1 [] = ( 2) for n = 0, 1 and 2.
b) Using [], compute the samples of a sequence 2 [] = ( 0.2) for n = 0, 1 and 2.
Comment on why the process of obtaining 2 [] (in this case) is sometimes called sinc
interpolation. How many samples of [] are needed to compute one sample of 2 []?

Problem 4:

Consider the block diagram shown in figure 4.11. Assume that the xc(t) is band-limited (|Xc(j)| = 0 for
|| 40). The overall system (enclosed in the dotted rectangle) is a continuous-time system.
a) In general, what is the largest value of T that would ensure that () = (). () for
the band-limited input described above? (Assume that the discrete-time system is LTI).
b) Determine the largest possible value for T and the corresponding frequency response H(ej )for
the discrete-time system such that

(),
() =
0,

10 < || < 20

Hint: You may allow some aliasing, by looking at frequencies that are being eliminated.
c) Obtain impulse response for the discrete-time system.
Problem 5:
Use MATLAB to solve this problem. Submit the printouts of the code as well as properly labeled plots
along with your answers.
Load handel.m. Youll get two variables: y and Fs. Assume that () was sampled to obtain y while
satisfying the Nyquists criterion ([] = () = (/)).
a) Compute (using freqz) and plot | | for || < 3 and | ()| for || < 3/T. Listen to
the sound that can be obtained by D/C conversion of [] at Fs samples per second.

b) Consider a signal [] = . Plot | | for || < using the variables created in part

a (with appropriate scaling of the axes). You are not allowed to compute or approximate [] (in
time-domain) to solve this part.

( ), = 0, 3, 6,
c) Using [], obtain a signal [] = 3
0,

Write an expression for [] in terms of []. Specify values of n for which [] is equal
to [].
d) Listen the analog sound that can be obtained by D/C conversion of [] at T=1/(3T), or Fs =
3Fs. You should hear something that sounds like what you heard in part a in presence of some
noise. Comment on the frequency of the noise (compared with frequency of the original sound).
e) Compute (using freqz) and plot | | for || < . Compare it with | | (part a) and
explain the results using your theoretical understanding.
f) Using the variables created in part e (with appropriate scaling of the axes), plot the magnitude
spectrum of the sound that you listened to in part d for || < /. Compare it with | ()|
(part a) . Based on your answer, explain the origin and frequency-range of the noise you heard in
part d.
g) Filter [] using a filter with impulse response h = 3*fir1(500, .32). You do not need to
understand how that impulse response was obtained. Plot the magnitude response of the filter.
Listen to the filtered signal (at Fs samples per second), and plot its spectrum. Compare these
results with the results obtained in the previous parts and explain your findings based upon your
theoretical understanding.
h) Repeat part g using linear-interpolation (see equation 4.92, delay h[n] to make it causal).
Compare the quality of the resultant sound with that obtained in g, and explain your findings
using your theoretical knowledge.
Problem 6:
Suppose you have to increase the sampling rate of a signal by a factor 2/3. There may be two approaches
to achieve that:

i.
ii.

Decimate (prefilter and compress) the signal by a factor of 3, and interpolate (expand and filter)
the resulting signal by a factor of 2.
The approach shown in figure 4.28.

Which approach would result in less severe loss of spectral components due to changing the samping rate?
Explain your answer using a toy example.

S-ar putea să vă placă și