DFB1X bare bones build

[Badwolf]

I ran the FPU test at 50MHz and everything passes except the trig functions at the end, which IIRC is typical if the FPU is clocked too high.

That's been my experience, but don't junk the FPU. I've seen reports of FPUs failing at 50MHz on (the original) DFB1 but passing on other boards so it may be of use somewhere else. I don't know if I'm particularly hard on them somehow or if something amiss in the design, but I couldn't find an obvious smoking gun.

Obviously the FPU is no good at 50 so that was a lost cause but I'm still a bit concerned at the behaviour change? Mainly concerned that the behaviour before was consistent, and now it has changed, but is also now consistent. Has anyone seen this error before on one of these boards?

Not seen anything quite like that and the CRC error suggests something is being stepped on. Could the jiggling in the FPU removal and insertion be the sign of an intermittent connection on the board somewhere?

The difference in behaviour between FPU in and FPU out suggests it's probably FPU or FPU socket related, but difficult to see how it could be being selected incorrectly as its chip select ought to be pulled high.

That's worth a check I suppose. There should be some kind of pull up resistor on the back of the FPU socket, I'd have thought. Check its joints and that it's a short to the 5V rail on one side?

BW

[dml]

Not seen anything quite like that and the CRC error suggests something is being stepped on.

I thought the same but since everything works when switched to 40, it's a bit ..weird.

Could the jiggling in the FPU removal and insertion be the sign of an intermittent connection on the board somewhere?

The last 2 things I did before it started was messing with the FPU socket and jumpering JC1.

Maybe another IC has been disturbed, or some issue with the mainboard FPU socket... but whatever it is only happens at 50MHz and once it started happening, it kept happening.

I plan to give the mainboard a bath in flux cleaner as it's been gathering dust since 1992. Will be easier to inspect everything properly.

The difference in behaviour between FPU in and FPU out suggests it's probably FPU or FPU socket related, but difficult to see how it could be being selected incorrectly as its chip select ought to be pulled high.

It might be something disturbed near that socket or on the mainboard and only causing problems at higher MHz. The DFB1X board is very tidy though, couldn't find any defects. It also fits pretty smoothly on the expansion port without a ton of force so I don't think anything is cracked or lifted.

I'm not really keen on those surface mounted PLCC sockets though. The thru-hole ones are hard to remove and the surface mounted ones are hard to install and harder to tell if there's a lifted pad somewhere.

That's worth a check I suppose. There should be some kind of pull up resistor on the back of the FPU socket, I'd have thought. Check its joints and that it's a short to the 5V rail on one side?

I'll check that tomorrow, might find something on a second pass.

I did buzz out a few things around the oscillator pad and FPU socket but nothing suspicious. I bridged a poly film cap across the FPU socket 5v/gnd (just holding it in place) and ONE time I got it to boot at 50 without the CRC error - but only one time out of maybe 5. Might try a local oscillator at 50 to see if it behaves the same or not, out of interest.

Done with it for today. Swapped the NVRAM for a turned pin socket earlier. Tried 3 new Dallas chips, two were dead from the NOS tube. Third one worked.

Have two other machines with different and much bigger woes so I'm satisfied this one is mostly behaving itself - it is working fine at 40MHz or at 50 with no FPU.

[exxos]

I have only just got home and only just had a quick read through this thread..

But it sounds like as you have mentioned multiple CPUs and FPUs, that you have not purchased them from the store ?

20% of CPUs will not even work properly. Even at 40mhz. I have wrote tons of stuff about this already. The roundup is here https://exxosforum.co.uk/atari/last/MC6 ... /index.htm

60% of the FN40 FPUs will not work past 16mhz. Not many will work at 50mhz.

This is why I speed grade them from my store and test them properly before I sell them. Just because you have CPU's and FPUs stamped as RC50 or FN40 does not mean they can run at that. China restamp everything, its well known. Most FN40s are 16Mhz (or maybe 33MHz parts).

You can indeed use your own chips if you so desire, but you are just inherently going to have such problems with unverified chips. You will probably have to test 100 FPUs to find one which works reliable at 50mhz. It's what I have to do. It's why I spent such a long time building up test rigs...
https://exxosforum.co.uk/forum/viewtopi ... 88#p107888

I don't think have mentioned if your Falcon has a clock patch or not. If you don't have one then the 16MHz feed to the DFB1X will be FUBAR right from the start.

[Darklord]

Ran into this same issue with the Pak 68/3 accelerator board in my STacy.

I can run 50mhz with no FPU, or 40mhz with a 68882 FPU. I can run at
50mhz, but the 68882 always fails the test.

I've got 3-4 50mhz 68882's and the one successfully running at 40mhz
is the best of the lot. Never tried a 68881.

[dml]

But it sounds like as you have mentioned multiple CPUs and FPUs, that you have not purchased them from the store ?

I'm not complaining about anything from your store - If I was doing that I'd contact you directly not on here I did already have some chips to try but I had those a long time.

20% of CPUs will not even work properly. Even at 40mhz. I have wrote tons of stuff about this already. The roundup is here https://exxosforum.co.uk/atari/last/MC6 ... /index.htm

Yes they seem to be all over the place. With the relabelling thing going on it is hard to get good parts like this now.

60% of the FN40 FPUs will not work past 16mhz. Not many will work at 50mhz.

Which would be fakes/repainted parts, otherwise would not get past Motorola QA/binning.

This is why I speed grade them from my store and test them properly before I sell them. Just because you have CPU's and FPUs stamped as RC50 or FN40 does not mean they can run at that. China restamp everything, its well known. Most FN40s are 16Mhz (or maybe 33MHz parts).

The CPUs, FPUs and other ICs I have are not relabels. I know this because I've had these parts since the 90s when it was easy to get genuine ones. It's always possible some are faulty or got tarnished/corroded but at least genuine.

And I think it is confirmed by the fact they do work correctly at 40MHz Tried 3 of them and they all work exactly the same. All pass all of my own FPU tests (which are extremely exhaustive ).

I'm not even disappointed they don't work at 50MHz, I was just suspicious that the 50MHz ROM error started happening ONLY after I touched the FPU socket and JC1. That's what I was asking about, in case it had been seen before or indicates something familiar.

I don't think have mentioned if your Falcon has a clock patch or not. If you don't have one then the 16MHz feed to the DFB1X will be FUBAR right from the start.

The host machine has no buffer - vanilla unmodified - I use it just for testing stuff. The DFB1X is due to go into a different machine with the clock patch already done but it needs some other fixes first. I was aware from the start it might not work at all without a patch but it seems to be working to spec already which is a bonus.

[exxos]

I was just suspicious that the 50MHz ROM error started happening ONLY after I touched the FPU socket and JC1. That's what I was asking about, in case it had been seen before or indicates something familiar.

I'm not aware of any relating issues. I'd suggest you put your MB ROM back in, boot the DFB1X with the disabled jumper set and wipe and reflash the ROM with TOS404. Remove the MB ROM and remove the disabled jumper to boot with DFB1X enabled.

There really shouldn't be any issues with flash. I don't think it even runs faster until it gets to the desktop. I'd also be careful the header connectors are not even slightly bent as it can cause bad connections and all kinds of odd faults.

Also reseat the ROM select jumper a few times just in case there's a bad connection on it.

[dml]

Just a quick update to say I tested the ALT ram today at 50MHz and everything is fine. Ran some stress tests with FPU @40, all ok.

MAPROM doesn't like a couple of things I tried but I have an idea what might be going on there. I'll look into that separately.

Hardware wise it all seems happy.

[exxos]

[Badwolf]

Hardware wise it all seems happy.

Good news!

Did you figure out what caused the FPU shenangians?

BW

[dml]

Did you figure out what caused the FPU shenangians?

Not exactly - I have removed and put the board back a few times since, mainly due to tidying up other stuff and I noticed that behaviour stopped and it has gone back to failing the trig functions (but passing everything else and no ROM CRC errors on boot) when set to 50MHz. Which is what I was seeing to begin with. Behaviour at 40MHz didn't change throughout the whole sequence of events - worked fine.

So I think it was either a temporary anomaly due to running at 50MHz and an angry FPU - or an unidentified person-behind-the-steering-wheel issue which has also been fixed.

For now it seems ok, will be moving onto another machine now. Which might even be Mikro's nuked TT.