posted 20 May 2013 10:23            

But in kyma I can only check "smooth delay changes" which makes the changes too smooth, too slow. Or I can leave it unchecked but have some very brutal artefacts in the output, due to (i guess) to low sampling rate of the modulation signal.

posted 20 May 2013 13:34            

default * (1 - (1 repeatingFullRamp))

posted 20 May 2013 14:24            

You could also try a waveshaper (along with a matching memory writer) to play back at different rates and directions. The value you put in the index input is where along the memory you play, but this is controlling it at sample rate, not control rate. It means that you need to supply a ramp in the index input just to hear the sound and control it with modules knitted together, as capytalk will put it back to control rate again.

You can use lo pass filters to smooth the index and get a cleaner sound, but you will probably have problems getting a clean join when the ramp jumps from +1 back to-1. I don't have a solution for that unless you are using a Capybara.

posted 17 June 2013 20:22            

@ pete, thanks! So going with a ramp or other modulation rate signal through a waveshaper with a linear lookup table would perform some kind of upsampling? I think I don't quite get it and hav no time to try it out right now, and would really like to understand what you mean.

posted 20 June 2013 15:49            

You have to twist the waveshaper around in your head and look at it in a different way.

What the wave shaper does is take an input ranging from -1 to +1 and outputs the value that it finds in the wave table that matches the position. Input -1 is the start of the table and +1 is the end. Now think of the table as a big sample ( good few seconds ) and think of the input as an index to that points to the individual samples inside that sample. The index and the look up both work at sample rate, so you can pick out any individual sample in the table (the long sample) instantly and can jump around the table freely at sample rate. Putting in a random index input will give us a random output, but if we put in a ramp that takes the total duration of the sample to get from the beginning to the end (-1 to +1) then the output will be the original sample un-effected. It will play through the sample sound, one sample at a time.

If the ramp moved 2 times as fast, then the output would miss out every other sample in the table and the signal will be twice as fast and twice the pitch (like verispeed). If the ramp took 2 times as long (to go from -1 to +1) we would only move along one sample in the table for every two samples in time but the holes between the samples will be filled in with the average of the two samples either side. This is because the wave shaper will interpolate for us (as long as you have the interpolate box ticked).

If the input ramp stops the sample stops, and if the ramp then goes downwards the sample will play backwards at a rate determined by the angle of the ramp . This is all done at sample rate, so you can change rate, direction and even jump around the sample on the exact sample in time you want without the restrictions of capytalk control rate. That said you have to control the ramp at sample rate as well to achieve this, which can be tricky.

If you have a Capybara you can use Pete's DSP modules to play with ramps at sample rate, but they don't work on the paca so you have to be a bit more creative.

I hope this makes sense

posted 21 June 2013 18:32            

SampleRateDelay.kym

Attached is a live input delay line with sample rate control of the delay, but it uses a completely different approach.

It relies on the oscillator modules FM input (that is really Phase modulation) being at sample rate. This deals with all the looping and wrap around ramps and interpolation at the borders internally, but it is limited to only 4096 samples. Here I've cascaded 4 of them giving a total max delay of 1/3 sec at 44.1Khz sample rate.

If you need longer delays you could cascade more.

Having 4 cascaded actually makes the interpolation at fast changes better but it does mean that the change happens at four different times depending on the delay. I've added a sample rate smoother in there (must keep the value just below +1) and it is controlled by a capy talk delay control, but you can replace these with sample rate delay control, and must be in the range of 0 to+1.

Hope this helps