I'm fairly new to DSP and am trying to see the effect of undersampling on bandwidth limited white noise in Simulink. I took the bandwidth limited white noise source (set the BW to 1kHz), put it through a zero order hold with Sampling rate set to initally 5kHz and the output of this is connected to the FFT spectrum scope block (I buffer 2048 points and take a FFT of the same length). When I change the sampling rate of the zero order hold block to 1kHz and lower, I expected the aliased signal to have a higher magnitude than the unaliased spectra. But the undersampled signal shows a flat spectrum with a magnitude that doesn't change regardless of how low the sampling frequency gets (1kHz, 500Hz, 250Hz). Can anyone tell me what I'm doing wrong? I expected the spectra to add up as the root sum square Thanks!
expected spectrum of aliased BW limited white noise
Started by ●November 12, 2007
Reply by ●November 12, 20072007-11-12
nombwa wrote:> I'm fairly new to DSP and am trying to see the effect of undersampling on > bandwidth limited white noise in Simulink. > > I took the bandwidth limited white noise source (set the BW to 1kHz), put > it through a zero order hold with Sampling rate set to initally 5kHz and > the output of this is connected to the FFT spectrum scope block (I buffer > 2048 points and take a FFT of the same length). > > When I change the sampling rate of the zero order hold block to 1kHz and > lower, I expected the aliased signal to have a higher magnitude than the > unaliased spectra. But the undersampled signal shows a flat spectrum with > a magnitude that doesn't change regardless of how low the sampling > frequency gets (1kHz, 500Hz, 250Hz). > > Can anyone tell me what I'm doing wrong? I expected the spectra to add up > as the root sum squareNothing. Aliases are parts of the original signal shifted and possibly inverted in frequency. A flat spectrum part makes a flat alias. Why do two (or more) flat noise spectra remain flat when added, while two (or more) noise signals (usually) approach a Gaussian distribution when added? That's left as an exercise. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������