exxos's Atari, Amiga & retro forum

Don't forget, according to the instruction page: https://exxosforum.co.uk/atari/last/DFB1X/

First thing to check is what voltage your PSU is providing. Measure on the motherboard to check, as there will be voltage drop. On the motherboard you need to be 5v, so the PSU will probably need to be something like 5.1v.

All checked and fine. Checked on the DFB1X board as well.

The system will run for hours as long as the DSP isn't accessed.

After the re-install today I had Frontbench running on the DFB1X for two hours without an issue.

I'm wondering if there's a timing issue with a 40MHz CPU rather than a 50MHz one, which was used for all the development testing.

Well, still got that delay option to play with & contact cleaning the expansion headers. Stuff to try is stuff to try :)

Before I blindly do things I would like input from the creators. They're the only ones who know what magic the CPLD is doing.

Maybe update the firmware to the latest. To my knowledge the models with 40 written on the CPU are the earliest versions.

The DSP was the last thing I got to work.

viewtopic.php?p=74755#p74755

It's a little unusual in that it latches data on the rising edge of the chip select line (driven by AS). The problem is that A0 is generated by the GALs when using the expansion port from UDS and LDS (which I also have to synthesize*). These normally deassert with AS so you get an undefined A0 value and trouble.

The solution was a delay to the deassertion of UDS and LDS as indicated above.

Perhaps something on your board is sufficiently near the tail of a particular distribution that this delay isn't enough. Or your GALs have different equations? Latter seems unlikely, though.

Is this the same Falcon that Frontbench crashed on when stock?

Now I should say this is all reference to the original DFB1 firmware. Exxos has released a few updates for DFB1X since then (the two firmwares are not compatible), but I do have a DFB1X at home with (I think) the latest firmware if you'd like me to try to replicate the problem.

Can you give me a small disc image with whatever exhibits the fault? I can't promise to fix it as 1) I only have one Falcon to test on and 2) it's not my firmware release but I think you've the same clock patch as me so I can at least double check if it's across the board.

BW

* The Falcon expansion port is a hybrid of the 68k bus and the 030 bus. It's a hot mess. I have to synthesize UDS and LDS based on DS and A0 from my 030. This then gets converted back to DS and A0 by the GALs for the motherboard to use and is the reason why we can't use the onboard FPU and why the DSP was such a pain to get to work.

Upgrading to the latest firmware didn't fix the problem, but did squeeze a few more frames out of Frontbench.

I don't see this as a huge problem. Anything where I'd need extra speed and TT-RAM I wouldn't want to use the DSP and for DSP based stuff I'd just disable the DFB1X using the switch.

@Badwolf useful information. It's the Joy demo which I'm using to test, I'll package it up for you on a floppy image later today.

Yes, it's the Falcon which didn't like Frontbench under TOS 4.04 (works fine on the DFB1X under EmuTOS though). It's very likely to be marginal timings and mine are just outside the margins some of the time.

stephen_usher wrote: 07 Jan 2025 11:10 @Badwolf useful information. It's the Joy demo which I'm using to test, I'll package it up for you on a floppy image later today.

:thumbup:

Yes, it's the Falcon which didn't like Frontbench under TOS 4.04 (works fine on the DFB1X under EmuTOS though). It's very likely to be marginal timings and mine are just outside the margins some of the time.

That was a really weird problem and definitely an indication of something awry. Be it in Frontbench or on the board itself.

Obviously not much overlap between the DSP and Frontbench, however.

...maybe that's a project for the future..? :lol:

BW

Badwolf wrote: 07 Jan 2025 11:14 That was a really weird problem and definitely an indication of something awry. Be it in Frontbench or on the board itself.

Given that Frontier itself runs flawlessly, as does everything else I throw at the machine in stock mode it would suggest something in Frontbench, but who knows? :-)

stephen_usher wrote: 07 Jan 2025 11:20 Given that Frontier itself runs flawlessly, as does everything else I throw at the machine in stock mode it would suggest something in Frontbench, but who knows? :-)

Aye. But obviously only something that triggers your machine! :shock:

BW