exxos's DFB1 trials

[exxos]

@dml

IMG_0254.JPG (342.76 KiB) Viewed 965 times

[dml]

success!

[exxos]

@Badwolf Going back to what was said yesterday about the thing getting more complicated. Maybe wouldn't be a bad idea just to keep basically the design as you have it now for a "minimal" type of design, and just have the FPU downclocked to 25MHz as standard anyway. Then just not bother.I doubt many people out there would be using the FPU to start with. There may not even bother buying a FPU for it. I think other than the capacitor on the master oscillator it is pretty much production ready I think. At least on my sample set of one

There the people want a more "advanced board" then have the "exxos edition" which is harder to build but will allow people to experiment more with clock speeds and whatnot. Though I would probably get some birdseed boards done at JLC like I did with the TF536 anyway.

[exxos]

50MHz OSC in place. it passed all the FPU tests..

BUT..

IMG_0257.JPG (195.47 KiB) Viewed 954 times

[dml]

50MHz OSC in place. it passed all the FPU tests..
BUT..

I noticed from an earlier post that the mandelbox started working after you changed it to load from FastRAM.

Is there something about STRam that is affecting the FPU stability? What happens if you run this same mandelbrot test from FastRAM vs STRam?

[Badwolf]

@Badwolf Going back to what was said yesterday about the thing getting more complicated. Maybe wouldn't be a bad idea just to keep basically the design as you have it now for a "minimal" type of design, and just have the FPU downclocked to 25MHz as standard anyway. Then just not bother.I doubt many people out there would be using the FPU to start with. There may not even bother buying a FPU for it. I think other than the capacitor on the master oscillator it is pretty much production ready I think. At least on my sample set of one

Whilst I agree with all that there's only one minor problem:-

That's exactly how your board was set up up here (bar the master osc cap) and your board did not work when it left me.

So unless that master OSC cap is all the difference (and I wasn't using the same oscillator as you -- I was using an osc and an FPU that worked in this board here), we still haven't found out why your board didn't work for me either.

So I still feel like we're *just* missing something.

BW

[exxos]

I noticed from an earlier post that the mandelbox started working after you changed it to load from FastRAM.

Is there something about STRam that is affecting the FPU stability? What happens if you run this same mandelbrot test from FastRAM vs STRam?

ah he started working when I added the 33pF onto the clock.. I'm not sure where was ran from had any difference.

I just removed the 33pF and now it is failing on those few tests again as it was before...

IMG_0258.JPG (66.07 KiB) Viewed 949 times

So I think it seems like the FPU is VERY sensitive to clock issues.

[Badwolf]

Is there something about STRam that is affecting the FPU stability? What happens if you run this same mandelbrot test from FastRAM vs STRam?

Shades of BadMood ST-RAM/TT-RAM crossover? That reminds me, I must answer your last question on that thread...

@exxos, there's a test for your new found working board! viewtopic.php?p=93269#p93269

BW

[exxos]

So unless that master OSC cap is all the difference (and I wasn't using the same oscillator as you -- I was using an osc and an FPU that worked in this board here), we still haven't found out why your board didn't work for me either.

So I still feel like we're *just* missing something.

Maybe, but when things are borderline like these such things then it is not really a surprise. Your OSC may well work in some boards but not in others. Those boards themselves are not identical so you cannot make the assumption that the OSC is "ok".

[exxos]

I messed around a bit with the 50MHz OSC. But not having much luck. Though with there being noise on the databus and DSACK lines etc, it may be aggravating it enough to not work. Also it is simply possible the FPU may not work at those speeds anyway. Might be worth revisiting 50MHz on a new board re-spin but I don't think it is currently worth spending the time getting those extra few MHz running at this point.