How the sampling frequency depends on the band width?

[roykyn]

Re: undersampling

hey see this......it is perfectly logical....and a picture some what against nyquist theorem.....i mean even at 100khz it cannot be reproduced to any thing that matches the original wave.....it is really confusing...but this is the answer for CMOS BABE

 

[echo47]

undersampling

Sampling a 50 kHz sinewave at 100 ksps violates Nyquist. You have to sample at *greater* than 100 ksps, such as 100.001 ksps. For example, at 100 ksps the samples could all land at zero volts, and you'd get zero output forever.

Beware, you cannot sample only five cycles of 50 kHz at 100.001 ksps and expect to reconstruct the original signal. That's because five cycles is not a sinewave, it's a sinewave burst that contains many high frequency components that would be corrupted by aliasing. If you sampled 1000 cycles at 100.001 ksps, then the burst's high frequency components would have much lower relative amplitude, the aliasing impact would be lower, and you could reproduce the original signal much better.

I sort-of learned all that stuff in school, but it never really sank into my head until years later when I began designing actual signal processing products. If anyone wishes to make a career of this stuff, I recommend getting MATLAB and maybe a DSP or FPGA development kit with analog I/O, and try experiments in your spare time.