On Tue, Feb 10, 2015 at 1:00 PM, F. Silvain <silvain(a)freeshell.de> wrote: Hey! Cool - I've done this in software, but JACK makes that ~ the same thing. I used a rather more crude approach though - detailed below. Yes - I would advise oversampling at the capture stage - more frequency headroom and better accuracy in general. In order to capture all the harmonics of the sound, you'll probably want to play a very low frequency on the hardware synth - but not too low that its DC offset highpass-filter removes the lowest sound. I was working with a bassline - just playing the notes on a MIDI keyboard and listening carefully allowed me to select the "ideal" frequency as reproduced by the synth. As Clemens noted - don't try to sample one cycle of exactly 65536 samples. Instead, record the "ideal tone", then resample one cycle of that waveform to the desired 65536 lenght. How I resampled to that lenght? Audacity - a little zooming + scrolling. This is the crude part of this solution. This could be done by correlating the signal to itself - a brute force approach to finding the closes match to itself would reveal the exact lenght of one osciallation[1]. On cutting away all extra parts of the recording, resample the single-cycle waveform to desired lenght and use. The wavetables produced are used in OpenAV's Sorcer synth[2]. Note that the playback method of the wavetable in Sorcer is far from ideal - don't copy it for your own sake :) Cheers, -Harry [1] assuming the sound is static - if it isn't theres little point in sampling one oscillation, as you're only capturing the one cycle of a wave it has by definition a static frequency spectrum if repeated to infinity. [2] http://openavproductions.com/sorcer and https://github.com/harryhaaren/openAV-Sorcer/tree/master/wavetables -- http://www.openavproductions.com