alt Hacker NewsDo say more! What was the problem with the clock, more exactly? I believe you, I've had issues caused by clock skew and CAN bus for example, when you have a small error that is amplified on beach bit enough time, errors add up and you eventually get out of synch.
But in the case if sound, I would have expected the skew to be less of a problem. Also surprised how the orof instantly know. It took me a while to figure out. How did you fix it? Cool story!
One simplistic way is to successively add a small constant to a large integer, and generate the waveform from the most significant bits. A "cent," which is 1/100 of a semitone, is a factor of about 580 parts per million, so you can work out the precision needed for the constant. On a microcontroller, you can control the timing with a PWM, which runs independently of the processor and its timing foibles.
Proof is left as an exercise to the student. ;-)