Check out above where I describe the patch for two waveforms with Brian but this is how it works.
Sample file with a number of single waveforms (N) concatenated that you decide when making the file (when building the patch I started with samples with 2 then 10 & then finally 64 waveforms to help me figure out how to get it working).
So if you want to scan a specific portion of the sample and in our case only one of our single waveforms then we want to scan only 1/N of the sample - to be able to do this we then need to condition the the incoming signal so that it’s peak-to-peak amplitude is also 1/N in size. Using an aliasing saw of amplitude of 1 results in a wave that goes from -1 to +1 so this is were the division by 64 comes in. After using the two rational VCAs the aliasing saw now ranges from -1/64 to +1/64 … So almost there but the range is still too big I.e. x 2 too big.
The final conditioning step is to add the limiter with an output of -7dB - this in effect divides the amplitude by 2 as needed (strictly speaking it should be 6.02dB but I reduced it slightly more by viewing the scope as I was getting occasional scanning of the next waveform) and stops any amplitudes larger passing.
Now the input audio has a peak-to-peak amplitude of 1/N. The next part is to then set the phase parameter and 64 stepped increments using the grid quantiser.
With the center parameter = 0 then the phase parameter to address the first waveform is -0.492 (derived as follows: 1/64 = 0.015625, so each waveform is 0.015625 in width and the centre of the waveform is 0.015625/2 = 0.0078125, and with the centre of the sample set to 0 then the sample with respect to the phase spans -0.5 to 0.5 so the centre of the first waveform will be -0.5 + 0.0078125 = −0.4921875 which rounds to -0.492 which you see is the bias value for the phase parameter).
Hope that explains how it works.