Bug #67

Pokey two tone timing problem

Added by admin over 3 years ago. Updated 4 months ago.

Start date:
Due date:
% Done:



An old bug.. See

10 P=53760
20 POKE P+8,64
30 POKE P+1,10+32
31 POKE 53775,11
40 POKE P,20
50 GOTO 50

high pitched, low pitched on real Atari

Screenshot from 2018-06-20 23-02-19.png View (111 KB) foft, 06/20/2018 10:02 PM


#2 Updated by foft over 3 years ago

Been measuring cases on the scope...

Confirmed the extra cycle delays described in that forum post on my 600XL. Confirmed missing on the EclaireXL implementation.

#3 Updated by foft over 3 years ago

I was concerned that my clocks were off at first since I was running in VGA mode, but that changes vsync slightly (50Hz vs 4.86Hz) which of course also impacts audio frequency:-) Debugging in RGB mode gives a perfect match.

#4 Updated by foft over 3 years ago

A 2 cycle delay on the reset fixes both the 4 cycle and the 9 cycle case.

Now to check this specific example!

#5 Updated by foft over 3 years ago

Nope, seems to be some interaction with the noise filtering. I wonder if the delay also needs applying to the pulse from the timer, that then hits the noise filter.

#6 Updated by foft over 3 years ago

So, errr identical then! Going to have to think about this one...

B is Atari, A is EclaireXL.

#7 Updated by foft over 3 years ago

I found if I pokey 53775,11+128 on the Atari then it matches the Eclaire output. Unfortunately forcing the serious output on eclaire to 1 or 0 does not change anything. Still its something to go on!

#8 Updated by foft about 1 year ago

I have audio test pcbs now to allow synchronised writes to pokeymax vs pokey. Which should help track down this kind of thing.

#9 Updated by foft 4 months ago

This was fixed following investigation and a simple test case/detailed spreadsheet from Synthpopalooza.

I used these new audio test pcbs and a cycle exact fpga comparison program to check it against the real pokey.

Incidentally I found another few small intermittent cycle differences which I'll be able to track down with this too.

#10 Updated by foft 4 months ago

  • Status changed from New to Resolved

Also available in: Atom PDF