exxos's DFB1 trials

[exxos]

Could the voltage either before or after the oscillator be changed?

Probably will be better to run a 3.3V OSC. But not so simple currently..

I'm constructing a crude duty cycle adjuster to see if I can get that bodged into correct the problem.

[exxos]

The story so far... Chaos...

IMG_0276.JPG (270.75 KiB) Viewed 1125 times

IMG_0273.JPG (302.64 KiB) Viewed 1125 times

IMG_0274.JPG (227.8 KiB) Viewed 1125 times

IMG_0275.JPG (284.7 KiB) Viewed 1125 times

IMG_0277.JPG (199.7 KiB) Viewed 1125 times

[exxos]

Had a better idea just to go hunting for a different oscillators

So put in a 40MHz one which ironically gives a proper 50:50 duty cycle and yet I was failing to get past loading NVDI

So put back the previous 40MHz OSC and 100pF on the thing, now it passed all FPU tests and running the rendering test currently..

The capacitor on the FPU side is not in place this time though... Neither is the 33R to the PLD form the OSC.

EDIT

Saying that I am also running a screwy version of the firmware so I need to backtrack a little bit

EDIT2:

Passed FPU test on the original firmware..

EDIT3:

So the 50% 40MHz OSC with 100pf crashes while loading NVDI basically. so some quick mathematical foo, 40MHz 100pF = 40R. So slapped in something similar for a actual resistor and now it did actually boot to desktop but crashes on FPU first test.

With the resistor and original oscillator.. The duty cycle has actually dropped to 42% now... And this one crashes on the first FPU test.

EDIT4:

Slapped on a multi-turn 500R pot and went down to 3.00V right up to 6V on the OSC output with both OSC's and FPU test crashed on load. It's all rather quite odd.

[exxos]

In swapping between both of the 40MHz OSC's. I wonder if it is because the one which works is actually 39.68MHz and one which doesn't is 40.20MHz

The slightly faster oscillator not work even with the FPU removed.

[exxos]

@Badwolf

My semi-conclusion for tonight's this:

50MHz on the CPU doesn't work. Could be down to the slightly lower voltage or some other issues. So whatever they are, 40MHz is all I have tested with.

My 2 40MHz OSC's are slightly faster and slightly slower than 40MHz. slightly faster one just outright refuses to work with the FPU. slightly slower one does work with 100pF on the OSC.

I suspect the FPU *can* run at 40MHz *if* the duty cycle is exactly 50:50 which is is near-on impossible to do. The 100pF pulls the clocks closer to 50:50 and its *just* enough to get it to work.

So when the slightly faster OSC is used.. forget it basically..It tips it over the edge again.

So all things considered, I will speculate the top speed of the FPU is probably going to be more like 38MHz.

If everything was absolutely ideal and spot-on then I think the whole lot would probably run on the same 40MHz clock. But there is just too many tolerances and variables to really work through that spending probably months getting to the real fine art of electronics, board layout, tolerances etc etc. I do think things can be improved here and there as previously mentioned.

EDIT:
The only oscillator I can find is 25MHz. So that is in the FPU clock socket. now both the 40MHz OSC's work to drive the PLD.. But the 100pF is still needed. It *might* work better with a 3.3V oscs. Then the 100pf *might* not be needed then. Hard to say at this point.

50MHz still doesn't work on the PLD. I really need to try upping the 5V a bit somehow to see if that makes any odds. Indeed it is possible that could just be enough to get the FPU working at 40MHz again.

[Badwolf]

I'm a bit slow today -- might have overindulged last night. Take me with a pinch of salt.

@Badwolf

My semi-conclusion for tonight's this:

50MHz on the CPU doesn't work. Could be down to the slightly lower voltage or some other issues. So whatever they are, 40MHz is all I have tested with.

Mmm. That's interesting. It's a bit of a shame, but it's the first time I've had that reported.

I suspect the FPU *can* run at 40MHz *if* the duty cycle is exactly 50:50 which is is near-on impossible to do. The 100pF pulls the clocks closer to 50:50 and its *just* enough to get it to work.

I wonder if an 80MHz oscillator halved would perform better, in that case. Do you ahve any of your RCO boards laying around still?

If everything was absolutely ideal and spot-on then I think the whole lot would probably run on the same 40MHz clock. But there is just too many tolerances and variables to really work through that spending probably months getting to the real fine art of electronics, board layout, tolerances etc etc. I do think things can be improved here and there as previously mentioned.

EDIT:
The only oscillator I can find is 25MHz. So that is in the FPU clock socket. now both the 40MHz OSC's work to drive the PLD.. But the 100pF is still needed. It *might* work better with a 3.3V oscs. Then the 100pf *might* not be needed then. Hard to say at this point.

50MHz still doesn't work on the PLD. I really need to try upping the 5V a bit somehow to see if that makes any odds. Indeed it is possible that could just be enough to get the FPU working at 40MHz again.

So even when not driving the FPU, you need some capacitance on the oscillator for the rest of the board to work?

It's strange because back here you had it working at 50 and I think it was before any of this mods: viewtopic.php?p=93412#p93412

BW

[exxos]

Mmm. That's interesting. It's a bit of a shame, but it's the first time I've had that reported.

I don't think there is enough boards out there to get any sort of "feel" to what is likely to be an general issue or not.

I wonder if an 80MHz oscillator halved would perform better, in that case. Do you have any of your RCO boards laying around still?

I did have one hacked on a ST536 *somewhere* . I think I actually saw a 80MHz OSC in my box. It might be worth trying out but not sure I will have time today.

So even when not driving the FPU, you need some capacitance on the oscillator for the rest of the board to work?

Without the FPU and 40MHz it seems to work with or without the cap. Passed full test of GB6. But also consider now that the FPU is not driven from the master OSC either..

Without the cap on the master 40mhz OSC, with the FPU fitted (on its own 25MHz OSC) it crashes on GB6. So the cap *is* needed.

It's strange because back here you had it working at 50 and I think it was before any of this mods: viewtopic.php?p=93412#p93412

50MHz doesn't work without the FPU and with or without the cap. So unstable either way. I don't know why it worked before. Might have simply being because the whole thing was stone cold when I first tested it that it worked. I did prove several posts back the putting freezer spray on the CPU helped. So probably was just a fluke it worked originally.

In case you think it is the CPU acting up.. They were all tested on the ST536 and I even heated them up with the gas soldering iron to make sure they wasn't going to crap out when they got extremely hot.

So for whatever reason, the DFB1 simply is not stable at 50MHz. As previously noted, it could be because the voltage is slightly lower on the Falcon than on the ST. Or it could be some other issue.

I don't think it is realistically worth spending any more time trying to figure this all out. If anything all the current mods should be done on a re-spin of the board before trying things again.

EDIT:
Happy to send this board back to you in its current state if you want to perform any further analysis on it?

[Badwolf]

It's strange because back here you had it working at 50 and I think it was before any of this mods: viewtopic.php?p=93412#p93412

50MHz doesn't work without the FPU and with or without the cap. So unstable either way. I don't know why it worked before. Might have simply being because the whole thing was stone cold when I first tested it that it worked. I did prove several posts back the putting freezer spray on the CPU helped. So probably was just a fluke it worked originally.

Yes, possible. We went straight into FPU debugging before we had a chance to properly soak test it.

So for whatever reason, the DFB1 simply is not stable at 50MHz. As previously noted, it could be because the voltage is slightly lower on the Falcon than on the ST. Or it could be some other issue.

EDIT:
Happy to send this board back to you in its current state if you want to perform any further analysis on it?

I think we ought to compare 5V lines but then it might be worth swapping boards. I'll send you mine, you send me back that one. I know this one is is rock solid in my Falcon at 50/25. If it isn't in yours, then it's PSU or Falcon-Falcon variations.

The same if your board works flawlessly here at 50MHz.

In order to make sense we'd probably have to keep the whole ensemble. CPU, FPUs, oscillators.

Will probably have to finish my firmware work before we get into that, though.

BW

[exxos]

Yes, possible. We went straight into FPU debugging before we had a chance to properly soak test it.

Yep

I think we ought to compare 5V lines but then it might be worth swapping boards.

Would be useful to know what you measure on your 5V. I am testing on the bottom left of the booster board header pins.

My multimeter shows about 4.89V.

On the scope.. 10ms /timebase

IMG_0281.JPG (58.74 KiB) Viewed 1030 times

If you can document that firstly then that would be useful. If the voltage is higher on your machine and more stable than mine, can you run the Falcon on the bench PSU and drop the voltage to say 4.50V and see what happens ?

If yours fails at that voltage then It could be very good hint as to why 50MHz is unstable on my machine. Basically just the CPU being under-volted. Then of course, some may be more tolerant than others so it is not conclusive.

I'll send you mine, you send me back that one. I know this one is is rock solid in my Falcon at 50/25. If it isn't in yours, then it's PSU or Falcon-Falcon variations.

The same if your board works flawlessly here at 50MHz.

In order to make sense we'd probably have to keep the whole ensemble. CPU, FPUs, oscillators.

Will probably have to finish my firmware work before we get into that, though.

Would be logical for me to test your board in this machine. There is still the problem that because it is a completely different board with different tolerances, it may well just tolerate being in my machine anyway. OTOH if it out right fails and is unstable then we have proven something at least.

I can send you this board out on Monday if you want ?. I would send it complete and working with 40MHz OSC (25MHz on the FPU). It should work in your machine straight away.. Then if you put in a 50MHz OSC (try with and without the cap) and see what happens.

[Badwolf]

Would be useful to know what you measure on your 5V. I am testing on the bottom left of the booster board header pins.

My multimeter shows about 4.89V.

On the scope.. 10ms /timebase

If you can document that firstly then that would be useful.

Yep, can do that over the weekend. Not sure I'll have the time tonight, bit of a rush on today.

If the voltage is higher on your machine and more stable than mine, can you run the Falcon on the bench PSU and drop the voltage to say 4.50V and see what happens ?

If yours fails at that voltage then It could be very good hint as to why 50MHz is unstable on my machine. Basically just the CPU being under-volted. Then of course, some may be more tolerant than others so it is not conclusive.

I don't have a bench PSU capable of driving the Falcon, but I do have an ATX supply that runs a bit low. Can give that a crack.

Would be logical for me to test your board in this machine. There is still the problem that because it is a completely different board with different tolerances, it may well just tolerate being in my machine anyway. OTOH if it out right fails and is unstable then we have proven something at least.

I can send you this board out on Monday if you want ?. I would send it complete and working with 40MHz OSC (25MHz on the FPU). It should work in your machine straight away.. Then if you put in a 50MHz OSC (try with and without the cap) and see what happens.

Yeah, OK. Let's do that. Cheers.

What *did* happen to my fifth board? If I can find that I'll build another one too.

BW