Description
Cybernoid1's tune is noisy when played back on real Mockingboard or Phasor (in MB mode) hardware, but less noisy on a Phasor (in Phasor mode) h/w. The difference is the AY8913 is clocked at 1MHz (MB mode) vs 2MHz (Phasor mode).
Here's a comparison in Audacity:
Top: Cybernoid1-Fix5-Phasor(MB-mode-1MHz).wav
Bottom: Cybernoid1-Fix5-Phasor(Phasor-mode-2MHz).wav
(Both "Split stereo to mono", then removed the 2nd channel)
This noisy section is ~960ms in duration or 48 x 20ms (48 x 50Hz interrupts).
(In the above Audacity screenshot, I included a non-noisy block at the start & end - ie. where it's the same top & bottom.)
IE, the 48 x AY registers in this code: https://github.com/tomcw/Cybernoid/blob/master/CybernoidTest2/CybernoidTest2.a
Only AY channel-A is used, and the frequency (A-period) & amplitude (A-vol) is changed each 20ms (50Hz) interval.
There are 8 blocks of 6 AY-regs, where A-Vol changes like this each block: {$D,$C,$B,$A,$9,0}
You can see the 8 sections on the bottom (2MHz) trace; and less clearly in the top trace, see the 8 sections (ie. pick out the spike from the starting $D amplitude).
For some reason at the 1MHz clock, the same AY chip is unable to respond to the new frequency. Only a few times in the top (noisy section) does it oscillate, the remainder of the time it decays to zero, then there's a new spike when the amplitude changes.
NB. there are still a few parts in the 2MHz trace where it doesn't oscillate... in fact it's typically alternate parts.