(24 intermediate revisions by 2 users not shown)
Line 6: Line 6:
  
 
=[[HW4ECE38F15|Homework 4]] Solution, [[ECE438]], [[2015_Fall_ECE_438_Boutin|Fall 2015]], [[user:mboutin|Prof. Boutin]]=
 
=[[HW4ECE38F15|Homework 4]] Solution, [[ECE438]], [[2015_Fall_ECE_438_Boutin|Fall 2015]], [[user:mboutin|Prof. Boutin]]=
 +
 +
!!!!!!!!!!!!!!!!
 +
INSTRUCTOR'S NOTE: THERE IS A MISTAKE BELOW. THE AMPLITUDE OF THE DT FILTER SHOULD NOT BE MULTIPLIED BY 1/TS. -> Corrected!
 +
  
 
==Question ==
 
==Question ==
 
A continuous-time signal x(t) is such that its CTFT X(f) is zero when when |f|>1,400 Hz. You would like to low-pass-filter the signal x(t) with a cut off frequency of 800Hz and a gain of 7. Let's call this desired filtered signal y(t).  
 
A continuous-time signal x(t) is such that its CTFT X(f) is zero when when |f|>1,400 Hz. You would like to low-pass-filter the signal x(t) with a cut off frequency of 800Hz and a gain of 7. Let's call this desired filtered signal y(t).  
 +
 +
:<math> x(t) \rightarrow 
 +
\left[ \begin{array}{ccc} & & \\
 +
& \text{CT filter with frequency response } H(f) & \\
 +
& & \end{array}\right]
 +
\rightarrow y(t)
 +
</math>
 +
 +
[[Image:CT_system_edited.jpg|1000px|thumb|left]]
  
 
a) Assume that you are only given a sampling of x(t), specifically a sampling obtained by taking 6000 samples per second (samples equally spaced in time). Can one process this sampling in such a way that a band-limited interpolation of the processed (output) DT signal would be the same as y(t)? Answer yes/no. If you answered yes, explain how. If you answered no, explain why not.  
 
a) Assume that you are only given a sampling of x(t), specifically a sampling obtained by taking 6000 samples per second (samples equally spaced in time). Can one process this sampling in such a way that a band-limited interpolation of the processed (output) DT signal would be the same as y(t)? Answer yes/no. If you answered yes, explain how. If you answered no, explain why not.  
  
:Answer: No. Since the sampling frequency is 6000 Hz, there will be no aliasing when we sample the signal which has the maximum frequency of 1400 Hz (fs > 2*fm). However, when we apply LPF with 800 Hz cutoff, we loose signal in the range 800 Hz < |f| < 1400 Hz. So the low-passed signal cannot be reconstructed perfectly.
+
:<math> x(t) \rightarrow 
 +
\left[ \begin{array}{ccc} & & \\
 +
& \text{C/D Converter} & \\
 +
& & \end{array}\right]
 +
\rightarrow
 +
\left[ \begin{array}{ccc} & & \\
 +
& \text{DT filter with frequency response } H_d[\omega] & \\
 +
& & \end{array}\right]
 +
\rightarrow
 +
\left[ \begin{array}{ccc} & & \\
 +
& \text{D/C Converter} & \\
 +
& & \end{array}\right]
 +
\rightarrow y(t)
 +
</math>
 +
 
 +
[[Image:DT_system1_edited1.jpg|1000px|thumb|left]]
 +
 
 +
 
 +
:Answer: Yes. Since the sampling frequency is 6000 Hz, there will be no aliasing when we sample a signal which has a maximum frequency of 1400 Hz (fs > 2*fm). The sampled signal Xd(w) has a cutoff frequency <math>7\pi/15 (= 1400*2\pi/6000)</math>. If we have a discrete-time LPF with a cutoff frequency o<math>4\pi/15 (= 800*2\pi/6000)</math>, we can have a signal reconstructed same as y(t).
  
  
 
b) Now assume that the sampling from Part a) is downsampled by a factor 2. Can one process this downsampled signal in such a way a band-limited interpolation of the processed (output) DT signal would be the same as y(t)? Answer yes/no. If you answered yes, explain how. If you answered no, explain why not.  
 
b) Now assume that the sampling from Part a) is downsampled by a factor 2. Can one process this downsampled signal in such a way a band-limited interpolation of the processed (output) DT signal would be the same as y(t)? Answer yes/no. If you answered yes, explain how. If you answered no, explain why not.  
  
:Answer: No. The same reason as part (a).
+
[[Image:DT_system2_edited2.jpg|1000px|thumb|left]]
 +
 
 +
:Answer: Yes. Since the sampling frequency is 3000 Hz, there will be no aliasing when we sample a signal which has a maximum frequency of 1400 Hz (fs > 2*fm). The sampled signal Xd(w) has a cutoff frequency <math>14\pi/15 (= 1400*2\pi/3000)</math>. If we have a discrete-time LPF with a cutoff frequency <math>8\pi/15 (= 800*2\pi/3000)</math>, we can have a signal reconstructed same as y(t).
  
 
----
 
----
Line 27: Line 60:
 
----
 
----
  
[[HW2ECE38F15|Back to Homework2]]
+
[[HW4ECE38F15|Back to Homework4]]
  
[[2015_Fall_ECE_438_Boutin|Back to ECE438, Fall 2014, Prof. Boutin]]
+
[[2015_Fall_ECE_438_Boutin|Back to ECE438, Fall 2015, Prof. Boutin]]

Latest revision as of 15:23, 20 October 2015


Homework 4 Solution, ECE438, Fall 2015, Prof. Boutin

!!!!!!!!!!!!!!!! INSTRUCTOR'S NOTE: THERE IS A MISTAKE BELOW. THE AMPLITUDE OF THE DT FILTER SHOULD NOT BE MULTIPLIED BY 1/TS. -> Corrected!


Question

A continuous-time signal x(t) is such that its CTFT X(f) is zero when when |f|>1,400 Hz. You would like to low-pass-filter the signal x(t) with a cut off frequency of 800Hz and a gain of 7. Let's call this desired filtered signal y(t).

$ x(t) \rightarrow \left[ \begin{array}{ccc} & & \\ & \text{CT filter with frequency response } H(f) & \\ & & \end{array}\right] \rightarrow y(t) $
CT system edited.jpg

a) Assume that you are only given a sampling of x(t), specifically a sampling obtained by taking 6000 samples per second (samples equally spaced in time). Can one process this sampling in such a way that a band-limited interpolation of the processed (output) DT signal would be the same as y(t)? Answer yes/no. If you answered yes, explain how. If you answered no, explain why not.

$ x(t) \rightarrow \left[ \begin{array}{ccc} & & \\ & \text{C/D Converter} & \\ & & \end{array}\right] \rightarrow \left[ \begin{array}{ccc} & & \\ & \text{DT filter with frequency response } H_d[\omega] & \\ & & \end{array}\right] \rightarrow \left[ \begin{array}{ccc} & & \\ & \text{D/C Converter} & \\ & & \end{array}\right] \rightarrow y(t) $
DT system1 edited1.jpg


Answer: Yes. Since the sampling frequency is 6000 Hz, there will be no aliasing when we sample a signal which has a maximum frequency of 1400 Hz (fs > 2*fm). The sampled signal Xd(w) has a cutoff frequency $ 7\pi/15 (= 1400*2\pi/6000) $. If we have a discrete-time LPF with a cutoff frequency o$ 4\pi/15 (= 800*2\pi/6000) $, we can have a signal reconstructed same as y(t).


b) Now assume that the sampling from Part a) is downsampled by a factor 2. Can one process this downsampled signal in such a way a band-limited interpolation of the processed (output) DT signal would be the same as y(t)? Answer yes/no. If you answered yes, explain how. If you answered no, explain why not.

DT system2 edited2.jpg
Answer: Yes. Since the sampling frequency is 3000 Hz, there will be no aliasing when we sample a signal which has a maximum frequency of 1400 Hz (fs > 2*fm). The sampled signal Xd(w) has a cutoff frequency $ 14\pi/15 (= 1400*2\pi/3000) $. If we have a discrete-time LPF with a cutoff frequency $ 8\pi/15 (= 800*2\pi/3000) $, we can have a signal reconstructed same as y(t).

Discussion

You may discuss the homework below.

  • write comment/question here
    • answer will go here

Back to Homework4

Back to ECE438, Fall 2015, Prof. Boutin

Alumni Liaison

Ph.D. 2007, working on developing cool imaging technologies for digital cameras, camera phones, and video surveillance cameras.

Buyue Zhang