Audio in v/Oct in pure data

[XponentOne]

Hi all,
So am working on a kind of random sequencer that takes pitch cv values from the adc~ via a snapshot~, am finding I need to offset and scale the input to get the output to reflect the input - is this correct? Am not at my machine right now to test stuff out, hoping someone can enlighten me if I maybe need to recalibrate or something? Or are my expectations wrong that sampling like this and passing the raw sampled value to dac~ should give same voltage?

[antisvin]

Yes, you would need to calibrate it to get voltage offset and scalar, then you can output correct voltage as V/Oct pitch CV. Your idea of calibration sounds almost correct, I think that current procedure for Lich relies on calibrating output with a multimeter, then using calibrated output as voltage generator to calibrate input.

Converting pitch CV to frequency is explained here: https://community.rebeltech.org/t/pure-data-conversion-v-oct-to-frequency/987/2?u=antisvin

To go the other way around and generate CV you'd need:

1. get number of volts = log2(FREQUENCY/440) , where 440 is your base frequency (can multiply or divide this by anything that is a power of 2 to get different base octave)
2. count the DAC value as out = volts / multiplier + offset

I suspect Martin has a stashed sample of PD code that does it, but this info should be added to our PD docs too.

[XponentOne]

Thanks for your help - yes I think I get it now, my confusion with the calibration was that I was assuming that the inputs would be presented to pd ready scaled so was surprised at then having to scale within pd to get it to match up. I've tried a little test cv-through patch and managed to get a fairly good mapping (as good as my cheap voltmeter can go anyway lol).

I assume there is no way for pd to get the saved scalars/offset directly from memory? Am just thinking that if I share my patch with a friend they'll need to update the values in the pd patch each time (or if I make public in the library).

[antisvin]

Reading scalar/offset values - is a reasonable feature request. For now only C++ and Faust support reading them from firmware. I'm not sure what could be a good way to expose them to user patches, maybe @mars has some plans for this.

[XponentOne]

Not the end of the world but would be handy I think!
Thanks for your help, got my sequencer working nicely now: