STE V1.5 32MHz BOOSTER - STATUS

[exxos]

So the board I have been using isn't booting at all now there just must be some bad connection on it still.. (This is the board I put double solder paste on the pins on the CPU). It just seemed to get progressively worse as the day go on until the point it stopped booting altogether.

Went back to my previous test board and that one has booted up.. So I will leave this one running for a while and see if this one locks up... This is the board I heated up with the heat gun to re-solder the CPU..

I think when my new PCBs arrive, I will use a hotter profile in my PCB oven.. Generally I always use the same one, because any hotter, PCB starts to burn, that is a problem I had with some of the earlier V1 STE boosters. Though I think it could be possible as the CPU is large, it is not actually getting up to hot enough temperature to melt the solder onto the pins correctly.

The thing is the original board I made was running for several hours several days without issue, so it's not like there's huge issue with the booster design itself. It is pretty much based on the tried and tested V1 STE booster anyway. So I think am just wasting my time trying to find faults which are probably not even there...

[PhilC]

Do you have a thermocouple you can place on the CPU when the board is in the oven to check the temps?

Could it also be that the pins you push into the socket are moving with heat? Am I correct in thinking that the previous set booster required removal of the CPU socket etc?

Of course I'm sure you've already checked all of this.

[exxos]

Do you have a thermocouple you can place on the CPU when the board is in the oven to check the temps?

Only a IR one, doesn't much help with the small oven window

Could it also be that the pins you push into the socket are moving with heat?

The pins on the PCB I solder by hand after the CPU (and other bits) have been soldered to the board in the oven. The plastic on the pins tends to melt in the oven.. some images in another thread here... viewtopic.php?f=18&t=334&p=2056#p2056

[exxos]

The board which died earlier.. I fluxed it to death and put it in the oven with a profile which stays hotter for longer.. I put the GAL in the socket as I knew the socket would melt and I would not get the chip in afterwards

(The previous board I had to take off test but it did run for almost an hour without problem, running out of time today but will try and test both boards out.)

Now this board boots again, before it was not even working with the ROM CS line connected.. Now it does.. Even more odd is that odd blurry image on my monitor also seems to have gone now

IMG_3065.JPG (136.26 KiB) Viewed 1670 times

[rubber_jonnie]

There will be much angular velocity involving a sledgehammer if it locks up again

Eeeek!

[exxos]

I left it running last night, got to around loop 50 and died.
Rebooted it ran until about loop 10.
Rebooted it ran until about loop 8.
Rebooted it ran until about loop 5.

This seems more like the CPU is overheating and crapping out. I left it overnight, turned it on this morning, ran up until about loop 9 (half hour run). This was running with 2 WS on ROM and 38MHZ.

I put my previous board in, thats running at 40MHz to see how long that runs for...

So I think I am leaning towards not all CPU's can run at 40MHz.. or so far, not even 38MHz. I will have to order a 32MHz OSC to try that.. I might get a 20MHz one.. its possible new CPU's may not even to be able to run at higher speeds. Which will be a pain if all 200 of them can't run faster than 20MHz!

So the overall plan would be I would have to build the boosters all at 40MHz.. leave them running all day, if they crash, swap it to a 32MHz OSC and try again and sell them as 32MHz ones. The ones which run at 40MHz will be sold as such. Problem being that's a hell of a lot of testing.. Also it would beg the question if it even be worth producing this booster if the CPU's can't run faster than 20MHz. Time will sell.

I think what's compounding the problem is the V1 booster ran at 8/32MHz, so the CPU wasn't being overclocked at all when accessing the bus. Obviously this overall runs a bit slower than running the CPU at constant speeds (40MHz etc). So the CPU heats up more. 8/32MHz switching might be equivalent "heat" to 20MHz constant speed for example. I'm not sure which one would ultimately be faster yet.. Though 20MHz constant run might be close to speeds of 8/32Mhz switching anyway.. That will be fun explaining to people why this 20Mhz booster is faster than the previous 32Mhz one!

Right so while typing this this other 40MHz one crashed at starting loop 8.... So time to start looking at slower OSC's...

Hacked on a 25MHz one.. see what happens...

IMG_3066.JPG (121.27 KiB) Viewed 1632 times

EDIT:

OK it crashed after 15mins Really don't understand it when the first one I made ran for like 100hours.. I need to re-try the 8/32 switching but I have a feeling that had some odd issues aswell, put that down to iffy soldering but maybe something else going on

I think this project will just have to go on hold for now until I got time and patience to keep messing about with it. Getting a backlog of machines to repair and upgrade so need to get cracking on those.

[rubber_jonnie]

So do they run fine at 8Mhz?

I know that isn't the plan, being a booster and all, but it might mean that the issue is either frequency or heat related if 8Mhz runs fine.

You could perhaps rule out heating by having a fan blow a constant flow of cool air over the CPU, plus a small stick on heat sink to rule out heating?

Unless of course it's a result of damage caused by heating in the oven. Any way you could hack on a socket for the CPU to try and rule that out?

Sorry if you've already thought of this, but it's where my brain is taking me!

[exxos]

So do they run fine at 8Mhz?

From previous tests no.. But that could have been bad soldering.. I will re-test later.

You could perhaps rule out heating by having a fan blow a constant flow of cool air over the CPU, plus a small stick on heat sink to rule out heating?

I don't think its a thermal problem now.. 15mins at 38MHz and 15mins at 25MHz.. doesn't add up. Running slower should give it more runtime if even 10mins. Unless these CPU's simply can't handle more than 20MHz. Though I think unlikely.

Unless of course it's a result of damage caused by heating in the oven. Any way you could hack on a socket for the CPU to try and rule that out?

I did wonder about hacking on a socket. Though be tricky to do and not sure it would prove much.

I will try 8MHz.. no code tricks, 8MHz in and out kinda thing, or might hardwire it bypassing the GAL.. but I did do that before and still had issues.. But again could have been bad soldering..

I did wonder about the length of the pins causes stability issues, but the V1 booster has the CPU a couple cm away anyway, and that works fine.

Though if 8MHz doesn't work, I think its will be game over for this project. I didn't want to do a plug in one because pushing 40MHz speeds via contacts isn't a good idea to start with. Its why I went with a solder-in type to avoid those issues. Though I am getting so many screwy issues randomly that I didn't have with the solder-in booster that I really think its just bad contacts from the booster to the CPU socket.

[rubber_jonnie]

From previous tests no.. But that could have been bad soldering.. I will re-test later.

I'd be interested in the results.

I don't think its a thermal problem now.. 15mins at 38MHz and 15mins at 25MHz.. doesn't add up. Running slower should give it more runtime if even 10mins. Unless these CPU's simply can't handle more than 20MHz. Though I think unlikely.

Fair enough. Is it possible to test the CPUs in an old style booster to see if they are stable at 32Mhz?

I did wonder about hacking on a socket. Though be tricky to do and not sure it would prove much.

I guess so, and not easy to do with that style of PCB either.

I will try 8MHz.. no code tricks, 8MHz in and out kinda thing, or might hardwire it bypassing the GAL.. but I did do that before and still had issues.. But again could have been bad soldering..

I did wonder about the length of the pins causes stability issues, but the V1 booster has the CPU a couple cm away anyway, and that works fine.

Though if 8MHz doesn't work, I think its will be game over for this project. I didn't want to do a plug in one because pushing 40MHz speeds via contacts isn't a good idea to start with. Its why I went with a solder-in type to avoid those issues. Though I am getting so many screwy issues randomly that I didn't have with the solder-in booster that I really think its just bad contacts from the booster to the CPU socket.

I did wonder if it could be some sort of pin contact issue, maybe resistance or capacitance that might be exacerbated by the clock speeds. Or even not contacting properly, something to do with the plastic block?

BTW, would it be possible to give the existing booster a faster clock speed using a separate oscillator? Just wondering is all.

[exxos]

I'd be interested in the results.

Its refusing to boot at all again It did once come up the Atari logo and "Bad ROM chips" error.. but the ROM is fine (works always with CPU direct in the MB socket).

Fair enough. Is it possible to test the CPUs in an old style booster to see if they are stable at 32Mhz?

Thats how I test all the CPUs before I use them or put them in my store. I don't think the CPU is to blame, I pretty much proved that earlier.

I did wonder if it could be some sort of pin contact issue, maybe resistance or capacitance that might be exacerbated by the clock speeds. Or even not contacting properly, something to do with the plastic block?

The plastic block pushes the pins together more, so it should help, but doesn't seem to matter either way currently.

BTW, would it be possible to give the existing booster a faster clock speed using a separate oscillator? Just wondering is all.

yep, I documented it a while ago, I had it running from a 38MHz OSC. I have my V1 running with a 32MHz osc in fact.

Its still possible it could be bad PCBs.. I just can't imagine my soldering on each one is *that* bad. Though the first 3 I sent off for assembly, 2 of them I couldn't get to work at all.. The CPU's and GAL's all tested before-hand.. buffer chips are new, and they all output 40MHz fine..

As my first one ran for 100 hours, its not like my design sux, and Some of these ran for several hours then just died for no reason. There could be some miro-fractures on the PCB tracks or vias which are not making good contact.. Then maybe heating it all up with the heatgun bends the board and connects them again for a short time...