exxos's Atari, Amiga & retro forum

Badwolf wrote: 08 Jan 2025 22:52 OK, I had a quick go with AceTracker with both DFB1 and DFB1X.

The screen had problems under acceleration on both of them with it failing to initialise probably three quarters of the time both under VGA and RGB interlaced modes.

Yes, that's one of the symptoms, and the Joy demo sometimes has the same problem some of the time.

Definitely a timing problem as it's not consistent and on the edge of working. Maybe it happens to be the phase between the motherboard clock and the accelerator clock at the time the CPU tries to interact with the DSP. The timings will change with the phase discrepancy between the motherboard clock and the accelerator one, which would be 2.5Hz for the 40MHz clock and 3 1/8th Hz for the 50Hz clock.

From observation it seems as if both Joy and AceTracker initialise the DSP just before setting the resolution and however they're doing this may be causing a problem on the bus which is preventing the video set-up from working.

@stephen_usher I know you're probably sick of me trying to help, but please let me persist since it seems to be a timing issue. You might not be aware but the clockpatch you installed has SDMA delay turned 'ON' by default. This could have implications as the delay is designed for bus acceleration. I think it is worthwhile trying it 'OFF', even if it doesn't improve the situation at least it's been tried. FYI the clock patch Exxos tested the joy demo with (V4) does not have delay, as far as I am aware.

Steve wrote: 09 Jan 2025 08:23 Out of interest BW, what clock patch are you using on your system?

I have the V2 without SDMA mod:

viewtopic.php?p=49426#p49426

stephen_usher wrote: 09 Jan 2025 08:56

Badwolf wrote: 08 Jan 2025 22:52 The screen had problems under acceleration on both of them

Yes, that's one of the symptoms, and the Joy demo sometimes has the same problem some of the time.

OK, I'll see what Joy does when I can.

From observation it seems as if both Joy and AceTracker initialise the DSP just before setting the resolution and however they're doing this may be causing a problem on the bus which is preventing the video set-up from working.

Interesting theory. I had assumed the screen problem was a purely videl interaction, but if there's a chat to the DSP immediately (as in one cycle) before the videl access, perhaps that half-cycle deassertion delay has some kind of ill effect?

In that case disabling the instruction cache and running from ST-RAM (to insert a motherboard read access between any register accesses) would be an obvious test.

BW

Given that the issue was identical before the clock patch was installed, it's unlikely to be the delay that's the issue. :-)

Anyway, I can't really work on the Falcon until at least the middle of next week now.

@Badwolf Please see my post to Stephen above, as mentioned, you probably have sdma delay 'on' too. It's worth turning it on/off to see if the behavior changes. And I appreciate @stephen_usher it might be unlikely, but since testing the theory only takes moving a solder blob, it's worth testing.

Steve wrote: 09 Jan 2025 09:48 You might not be aware but the clockpatch you installed has SDMA delay turned 'ON' by default.

What's could be influence of the SDMA clock here, you think? Purely on the audio side?

I was thinking about the DSP clock, but the DSP is clocked faster than the main bus IIRC, so presumably has an independent (or partially dependent) clock source? [sorry, not time to look it up right now]

Steve wrote: 09 Jan 2025 09:51 @Badwolf Please see my post to Stephen above, as mentioned, you probably have sdma delay 'on' too. It's worth turning it on/off to see if the behavior changes. And I appreciate @stephen_usher it might be unlikely, but since testing the theory only takes moving a solder blob, it's worth testing.

Hmm. Interesting I hadn't realised.

BW

Badwolf wrote: 09 Jan 2025 09:53 I was thinking about the DSP clock, but the DSP is clocked faster than the main bus IIRC, so presumably has an independent (or partially dependent) clock source? [sorry, not time to look it up right now]

Yes, the schematic shows it having a separate crystal.

Badwolf wrote: 09 Jan 2025 09:53 What's could be influence of the SDMA clock here, you think? Purely on the audio side?

I was thinking about the DSP clock, but the DSP is clocked faster than the main bus IIRC, so presumably has an independent (or partially dependent) clock source? [sorry, not time to look it up right now]

The DMA and the DSP seem to be intrinsically linked. These strange DSP synchronisation issues used to be caused, for years and years, by SDMA clock issues. I think that this feature could certainly have an effect on the situation.

smoore100 wrote: 08 Jan 2025 23:44 Ace tracker works fine on my DFB1X

Thanks for this data point!

Falcons are weird. :cry:

:D

BW

Falcons are only on the edge of actually working!