I remember you could get a fairly high frequency (¿12kHz, maybe even 20kHz?) if you were willing to not sample uniformly. On a 6502, you would shift 8 1-bit sa

I remember you could get a fairly high frequency (¿12kHz, maybe even 20kHz?) if you were willing to not sample uniformly. On a 6502, you would shift 8 1-bit samples into a single byte at high frequency, spend some time writing the byte, rapidly sampled 8 more bits, etc. The “write a byte” step also wasn’t constant time, as increasing the address to write to took longer when you crossed 256-byte boundaries. So, recording missed samples at predictable times. The trick was to make the playback code use the exact same timings. I think there was an article in Compute! for doing that on Commodore machines, but cannot find it.