Looking in the Erlang scheduler at schedule_internal_call, I noticed some time conversions (nanoseconds → seconds → miliseconds) that make it look like Sonic Pi has milisecond accuracy. Is this correct?
The accuracy of Sonic Pi’s audio is determined by the accuracy of SuperCollider’s scheduler. The granularity of this is (I believe) related to the sample rate and audio buffer size of the audio hardware (in addition to SuperCollider’s own buffer size). So it should be in the very low millisecond range on typical hardware.
Note that this is all unrelated to the Erlang scheduler which only manages the timing of outgoing MIDI and OSC messages. The accuracy of this isn’t related to the conversions themselves but the resolution of Erlang’s internal scheduler. From my experience this is different for each OS with Windows improving dramatically a few years ago. It should be < 10ms in terms of accuracy.
Okay, thanks for the information!