exxos's Atari, Amiga & retro forum

exxos wrote: 30 Nov 2022 21:39 Have you tried a slower oscillator say 40MHz ?

Not yet, but I'll have a look if I have any. By default the FPU runs at 50% of the oscillator speed right?

foft wrote: 30 Nov 2022 21:49 By default the FPU runs at 50% of the oscillator speed right?

Yes.

foft wrote: 30 Nov 2022 21:36 On my Falcon I have R18 + a recapped PSU and I read 5.00V.

On 3 DFBs with the release firmware I have the FPU working in one, sometimes, sometimes fails on a few ops in fputest. On the other two so far the FPU isn't working (bombs in fputest). I've not yet beeped them out or tried any mods.

I've built 4 DFB1's and every time I had a problem with the FPU it ended up being the CPLD that was faulty. (Chinese CPLDs from Ali :-( ) out of 6 CPLDs bought in Aliexpress, only 2 were good, the other 2 pieces worked ok but the FPU was unstable and the 5th did not start at all. I had to buy another 5 pieces to make 4 DFB1 work ok. :headbang:

markus0321 wrote: 03 Dec 2022 12:05 I've built 4 DFB1's and every time I had a problem with the FPU it ended up being the CPLD that was faulty. (Chinese CPLDs from Ali :-( ) out of 6 CPLDs bought in Aliexpress, only 2 were good, the other 2 pieces worked ok but the FPU was unstable and the 5th did not start at all. I had to buy another 5 pieces to make 4 DFB1 work ok. :headbang:

This is really interesting if it's not just co-incidence. I think a lot of people are using Aliexpress CPLDs as the shortage really bites. I've lost track of the provinence of many of my chips and it's entirely possible the board (that Exxos now has) where I couldn't get the FPU to work had a China chip on it.

I was very close to not putting the FPU on the board at all, then decided to throw it in as a bonus. I think 80%+ of all problems I've been contacted about are FPU-related. Now that Anders has provided a software FPU emulator, I'm tempted to just remove the FPU from the next revision.

There is a way to use the on-board FPU with one patch wire to a GAL. That's a third option.

BW

Badwolf wrote: 05 Dec 2022 11:20

markus0321 wrote: 03 Dec 2022 12:05 I've built 4 DFB1's and every time I had a problem with the FPU it ended up being the CPLD that was faulty. (Chinese CPLDs from Ali :-( ) out of 6 CPLDs bought in Aliexpress, only 2 were good, the other 2 pieces worked ok but the FPU was unstable and the 5th did not start at all. I had to buy another 5 pieces to make 4 DFB1 work ok. :headbang:

This is really interesting if it's not just co-incidence. I think a lot of people are using Aliexpress CPLDs as the shortage really bites. I've lost track of the provinence of many of my chips and it's entirely possible the board (that Exxos now has) where I couldn't get the FPU to work had a China chip on it.

Could this be down to tolerances though ? But like I said before, even the probe on X10 on one of the DSACK lines caused it to go nuts. is possible even the speed of the PLD could be a factor ? China ones most likely restamped speed wise for starters. As you know with the oscillator chaos really does not take much..

exxos wrote: 05 Dec 2022 11:50 Could this be down to tolerances though ? But like I said before, even the probe on X10 on one of the DSACK lines caused it to go nuts. is possible even the speed of the PLD could be a factor ? China ones most likely restamped speed wise for starters. As you know with the oscillator chaos really does not take much..

Well there ought not be any tolerance issues -- the timing margins are extremely lax. After all, the two chips can run at several multiples of each others' speeds.

So why the sensitivity at all?

BW

Badwolf wrote: 05 Dec 2022 12:00

exxos wrote: 05 Dec 2022 11:50 Could this be down to tolerances though ? But like I said before, even the probe on X10 on one of the DSACK lines caused it to go nuts. is possible even the speed of the PLD could be a factor ? China ones most likely restamped speed wise for starters. As you know with the oscillator chaos really does not take much..

Well there ought not be any tolerance issues -- the timing margins are extremely lax. After all, the two chips can run at several multiples of each others' speeds.

So why the sensitivity at all?

BW

There must be something strange about it because when I originally thought it was the FPU's fault, I had such a symptom that when I tried to synchronize with CLK with DFB1 (25MHz) and it worked, then I substituted an external 40MHZ oscillator and it caused a problem and even 33MHz did not fix it. But I changed the CPLD to another one and suddenly the FPU problems disappeared and even at 50MHz the FPU worked stably for me. I checked the third and fourth boards immediately during startup with an external generator for the FPU and if there was a problem, I changed the CPDL so many times until I found one that worked without a problem.

markus0321 wrote: 05 Dec 2022 18:00 But I changed the CPLD to another one and suddenly the FPU problems disappeared and even at 50MHz the FPU worked stably for me.

Yes, I don't doubt you -- such a failure mode would explain the 'demon' Exxos board too -- I just am not sure what the cause could be.

Other than FPUs are evil. Of this am I now convinced. ;)

BW

Badwolf wrote: 05 Dec 2022 18:55 Other than FPUs are evil. Of this am I now convinced. ;)

They are not evil, just misunderstood :lol: :hide:

Got my DFB1 Back from BW today. I noticed that as soon as it got to desktop and move the window the entire window contents would end up corrupted.

After a short fiddle about I re-flashed the release firmware just to make sure it was the correct one and it seems to have died totally :cry:

I have change the oscillator, removed the FPU, tried a different CPU. There is some activity on the LEDs But I do not get anything on the monitor whatsoever. It's dead Jim :cry:

No stock in Mouser for the PLD and they like £50 a pop at that :(

https://www.mouser.co.uk/c/semiconducto ... rt=pricing