Sampling Continuous-Time Signal

  • Thread starter Thread starter mickonk
  • Start date Start date
  • Tags Tags
    Sampling Signal
AI Thread Summary
The discussion revolves around the sampling of a continuous-time signal using an ideal low-pass filter with a cutoff frequency of 3π/4 and a sampling period of T = 1/125 seconds. Participants clarify that angular frequencies (wa1, wa2, wa3) should be expressed in rad/sec, though they can also be represented in Hz. The Fourier transform (FT) of the input signal is discussed, particularly how to handle time-shifted terms and the application of the translation property. There is confusion about the FT pairs and whether the provided definitions are correct, leading to a deeper examination of the properties of Fourier transforms. The conversation concludes with a concern about the filter's ability to process signals given the frequency constraints.
mickonk
Messages
7
Reaction score
0
This my homework:
Input signal to system is:

t1.png


where

t2.png


t3.png


H(exp(jw)) is transfer function of ideal low pass filter with cutoff frequency wg=3*pi/4 and zero phase characteristic. Sampling in A/D converter is done with period T=(1/125) seconds.
a) Calculate output signal ya(t)
b) Calculate sampling period T for ya(t)=xa(t)First thing: they said that wa1, wa2 and wa3 have dimension 1/s. Is that mistake? I think that it should be rad/sec.
I recently started studying digital signal processing and I'm not so good yet but here are my thoughts. I know that A/D sampling period tells us that every T seconds A/D converter will take value from input time signal.
Frequency of sampling would be (1/T) [Hz] and it must be at least two times bigger than biggest frequency in input signal. Ideal low pass filter will pass only signals with frequencies lower than cutoff frequency
I know that I should first find amplitude spectrum of input signal using Fourier transform but I don't know how to find x(n). Here is how I would find FT of input signal. We can write last term of xa(t) as $$\sin (wa3t+(\frac{\pi}{2}+\theta))$$ Fourier transform of sine wave $$A\sin (w_0t)$$ is $$Aj\pi[\delta(w+w_0)-\delta(w-w_0)]$$, So FT of first term of $$x_a(t)$$ will be $$1j\pi[\delta(w+w_{a1})-\delta(w-w_{a1})]$$, FT of second term $$(1/2)j\pi[\delta(w+w_{a2})-\delta(w-w_{a2})]$$. What would be FT for third term, since it is time shifted?
 
Last edited:
Physics news on Phys.org
This would be (hopefully) amplitude spectrum of first and second term of input signal:

t4.png
 
Last edited:
mickonk said:
First thing: they said that wa1, wa2 and wa3 have dimension 1/s. Is that mistake? I think that it should be rad/sec.
An angle in radians is a dimensionless quantity, so, strictly speaking, it's correct to use the unit hertz [1/s] for both frequency and angular frequency. We usually make it explicit, though, by using [rad/s] for angular frequency to avoid any confusion.

mickonk said:
What would be FT for third term, since it is time shifted?
Try going through your transform table again (Google one if you need to) - the translation/shifting property of the Fourier transform should be included in just about any of them.
 
Last edited:
Hi milesyoung.
That's FT pair from my book. Is it wrong?
I found time shifting property in my book: if function is $$f(t\pm t_0)$$, FT is $$e^{\pm jw_0t_0}F(jw)$$. So FT of time shifted term would be $$e^{\pm jw(\frac{\pi}{2}+\theta)} F(jw) $$, right?
 
Last edited:
mickonk said:
That's FT pair from my book. Is it wrong?
No, sorry, that was my mistake. I'm used to the unitary definition of the Fourier transform, so when I skimmed over it, I just noticed the lack of a scaling factor.

mickonk said:
FT is $$e^{\pm jw_0t_0}F(jw)$$
You probably mean ##F(\omega)##.

mickonk said:
So FT of time shifted term would be $$e^{\pm jw(\frac{\pi}{2}+\theta)} F(jw) $$, right?
Not quite for your function, but it's not something you have to do from scratch. See here:
http://en.wikibooks.org/wiki/Waves/Fourier_Transforms#Fourier_Transform_Pairs
 
If the filter cuts off at w = 3π/4 and the lowest signal frequency is w = 100π, and sampling does not generate frequencies below the signal frequency, then how can anything get past the ideal low-pass filter?
 

Similar threads

Back
Top