How to efficiently Sample a wideband signal

[sedarash]

Dear All,

I intend to sample and store some waveforms. I get the FFT of the waveform and it looks like the attached graph. My problem is that I can not determine the Nyquist frequency on a graph like this, since its a wide band signal.

​

Kindly please bear in mind that I have to find a balance between accuracy and storage cost. I would like to know how people normally sample such signals and how much accuracy cost is normally affordable. Any clue to follow up or any guidance backed up by theory or a cite-able practice would be most appreciated.

Thanks
Arash

 

[barry]

I think you're doing this backwards. YOU pick the Nyquist frequency by virtue of your sample rate, the signal doesn't pick it. What is the upper frequency of your 'wideband' signal? Is it infinite? You need to determine the upper limit of your frequency of interest, and then sample at least twice as fast as that. And you'll also probably need to filter your input signal if you can't guarantee that its bandwidth will be less than 1/2 the sampling frequency.

Barry

 

[FvM]

I guess, you should start to think about the nature of the signal and which information shall be preserved in signal acquisition. As simple as trivial, if you see spectral components up to fs/2, the signal bandwidth (is it actually a signal?) obviously exceeds the Nyquist frequency. You should know if it matters.