# Effects of sampling rate approaching the Nyquist frequency? (not undersampling)

I understand that you need to sample at a frequency greater than or equal to the Nyquist frequency to reconstruct a signal/aliasing etc. I read though that a "more conservative" approach is to sample at 2x-5x the Nyquist frequency (Rampil '98, on EEG processing). Assuming signal bandwidth doesn't extend beyond $F_N/2$ (has been low-pass filtered), are there effects/advantages of sampling at higher frequencies than $F_N$? That is, if I have a signal with no components above 100Hz, is there an advantage to sample at 1kHz vs at 200Hz?

The context is that I'm resampling various signals (EEG, EMG, joint angles) to a common time basis and need to pick the target frequency. Processing time is fast enough to not be a huge concern, but I do want to decrease it somewhat.

Thanks!

What you're describing is called oversampling. You may want to take a look at the following -

http://en.wikipedia.org/wiki/Oversampling

• Ah, thanks. Before spending half an hour on google scholar with "effect of sampling above nyquist freq" and the like, I probably should have thought to wiki "oversampling." – dpbont Mar 18 '14 at 6:06

Sampling at 2*Fn is suitable only for signals that are infinitely long and perfectly bandlimited to below Fn, which conditions don't exist in actual signals being sampled or resampled (given non-perfect low-pass filtering), or in any non-infinite length signals.

A finite length rectangular window will itself introduce higher frequency content to the signal being sampled or recorded.

Therefore, there are major advantages to a sample rate higher than 2*Fn, in terms of not corrupting the signal, especially if any higher frequency (close to Fn) spectral content near the beginning or end of your recorded signal is important.

Added: Reconstructing a signal represented by data at a sample rate well above twice the highest frequency can greatly reduce the computational (or hardware) requirements of the reconstruction filter or interpolator.