Re: NTSC VIC-II timing

On Wed, Jun 21, 2017 at 04:40:03PM +0200, silverdr@wfmh.org.pl wrote:
>According to the famous document by Chris Bauer, there are three timing 
>variants of the VIC-II. While PAL is always 63 cycles per  line, with 
>NTSC there can be either 64 or 65 cycles per line and I am in fact able 
>to reproduce the difference with -model ntsc and -model oldntsc in the 
>current VICE.

You should have the 6567R56A and 6567R8 that I gave you, right?

>The question (before I spend weekend on trial'n error counting cycles 
>and possibly reinventing the hammer ;-) is: do we have ane established, 
>reliable software method for detecting which NTSC VIC-II is installed 
>in the machine? I guess it must have been done multiple times by now 
>and used in some NTSC games/demos..

The simple way is to disable interrupts and write a loop that samples 
$d012, counting the cycles that go between changes. I do not know if 
exactly such implementation is available anywhere.

Andreas Boose introduced me to an algorithm that first does a "coarse" 
synchronization, like this:

	sei
	lda $d012
	bit $d012
	bne .-2

This creates a jitter of 4+3=7 cycles, assuming that .-2 is on the same 
page than .+2 (the bne takes 3 cycles until the loop is exited).

Then, for 7 lines, spend N_CYCLES_PER_LINE-1, sampling $d012 at the end 
of each iteration. Do a conditional branch to .+2 (relative offset $00) 
if you were ahead of the schedule. After these 7 lines, you should be in 
perfect sync (a deterministic number of cycles have elapsed since $d012 
last changed).

You should revise this algorithm so that it can somehow detect the 
N_CYCLES_PER_LINE, which can be 63 (PAL), 64 (6567R56A), or 65 (6569).

One possibility could be to start a CIA timer after the coarse 
synchronization, but bear in mind that there are some differences in the 
CIA timer operation. I think that it is doable with the processor only.

	Marko

       Message was sent through the cbm-hackers mailing list