The frequency bandpass for standard EMG amplifiers will be 10-20 Hz up to 450-1000 Hz. With these settings and a 256-Hz sampling rate, aliasing will occur. At the minimum, you should be sampling at twice the lowpass frequency setting (i.e., a sampling rate of ~2000 Hz). A more ideal sampling rate would be around 10,000 Hz. Read up on the Nyquist sampling theorem.

Thank you for your response. I know the sampling frequency should be at least two times the maximum frequency of the EMG but due to some limitations I have to use 256. The purpose only to find the movement onset and no further analysis required. Is this possible?

Hi,

If your purpose is "only" to find the onsets, you may use 256 Hz sampling rate but you would better have hardware filtering options which would allow for envelope detection. However, you may have few milliseconds range onset errors.

As you have stated, you intend to analyse surface EMG, for needle EMG that may not be possible.

Does your system for the surface EMG have hardware filtering options before sampling?

It you're willing to only detect the onset, 256 Hz is suitable (e.g. for rest/activity detection). The ideal would be at least 1 kHz, but I think this only make a difference if you are classifying patterns.

You can filter only the frequency spectrum above 10 Hz, as the Nyquist frequency limitation of f < fs/2 (128 Hz) will work as a low-pass filter. In my experience I observed that, for most cases, low frequencies are suitable for onset detection as you will probably rectify and smooth the signal for it.

Moreover, I recommend avoiding using power noise filters (50/60 Hz). If you power your equipment with batteries, these components would be almost irrelevant, so you don't need to reject useful information applying notch filters.

base on your level of detection and filtering used , you can go for it...Nyquist frequency matters here..