Super frustrated with A1200 resetting in some weird cases

[xnopasaranx]

Hi there, I just registered after having searched this and other Amiga related forums for months now.

I have aquired an A1200 about a year ago and fitted a gotek and ACA1233n-40 after recapping it. I also installed a 2.5" IDE HDD and had to repartition it several times because of the known MAXTRANSFER limitation... well that's how you learn. It should be corruption free now.

The A1200 has never really worked in a satisfactory manner.

When the ACA1233n is not used, the A1200 seems to run stable for hours. As a test we had Elite 2 run for several hours and some other games, loaded from floppy and also from the gotek.

The four caps mentioned on the icomp wiki had been removed by the previous owner already. Board is a Rev. 1D

But when the ACA1233n is in the slot the system will reset at times. It seems to me, that this is when the HDD is being accessed. For example: I will play a few levels of lemmings and have the system powered on doing some fiddling around in workbench. Then I'll play eye of the beholder and try to save and the amiga resets.

The only thing left to try is that the PSUs I own could be going to sh** and so I have ordered a new PSU from amigastore.eu which is supposed to be beefier than the original commodore PSUs. Well... and it most certainly has no leaky caps as everything is brand new.

The power connection is clean, I have cleaned it thoroughly. Wiggling the connector shows no issues. Hitting the table also does not produce any crash, so I guess the ACA1233n is seated firmly.

I have measured both PSUs with a voltmeter, but not under load as I was afraid to short something out and kill the machine. If this was cheaper hardware I would probably get over myself and take some actual measurements, I also have an oscilloscope to have a look at the curves. But I don't have a lot of experience with it yet and would rather not blow up my 500+ euro investment.

Anyway, I am close to selling everything off again, because I just can't enjoy it at all if the systems crashes all the time after some use.

I am at my wits end. What else could it be? Is there some sort of stability test I can run to get some sort of insight or at least reproduce the resets somehow to narrow this down?

[SpocksBeer]

Standard advice, at least as far as I can see recently, is that the stability (low ripple) of the 5V power rail is critical to Amigas running well. Doubly so with any accelerators present. You won't be able to detect an unstable 5V rail with a multimeter in most cases, only your oscilloscope, and only when the PSU is under load.

Given when you reduce the overall power requirements of the A1200 by removing the accelerator things improve, I'd say you're on the right track with your replacement PSU.

I feel your pain, and good luck.

[xnopasaranx]

Thanks for the encouraging feedback! I hope the new PSU arrives before the weekend, so I can test a little.

Do you have any advice on where I can safely measure with my scope? It doesn't have a memory function, so maybe I can just have a check for ripple/decay when HDD is accessed, but I can't reproduce the crash reliably. Anyway that would be helpful (and also get me some more practice with the scope)

[indigolemon]

Originally Posted by xnopasaranx

Do you have any advice on where I can safely measure with my scope?

I believe most people check at the floppy drive power connector.

[xnopasaranx]

@indigolemon: awesome, thanks for the tip.

I received an RGB scart cable today and noticed a lot of distortion effect on the signal. horizontal lines and wobbles. So I switched to the other PSU I still have have. Was using a light A500 PSU with 4.5A and am now on an A1200 PSU with 3A. The lines instantly disappeared and the picture is super crisp now. So the older PSU definitely is at the end of it's life without a recap now. Not sure if I want to do that, I am still awaiting the brand new PSU.

I guess the old PSU was having weird effects and the slightly newer one isn't quite safe anymore but still in slightly better shape. Once the new PSU is here I'll do a burn in test and see if the machine can stay up for a couple of hours. I am very relieved right now.

I can't believe how crappy that composite signal is, I never noticed the distortions so much and attributes them to the overall quality of composite.

[xnopasaranx]

okay the new PSU has arrived. The amiga does not reset anymore, but it is still unstable after a few hours. I was not able to let it run over night as I planned to do originally, because after about 3 hours of fiddling around and playing lemmings with my girlfriend, whdload games started to randomly crash. I played pinball fantasies AGA and it behaved weird, crashed and when started again had horrible graphics corruption (colours wrong/inverted, artifacts flickering on sprites). Then the game would crash again and I restarted it, the graphics were back to normal but as soon as I lost a ball the game would crash again, back to workbench.

I am thinking this might be an overheating problem. Can't really see why that would happen after some time otherwise. RAM issues would set in more randomly and eventually earlier on, right? I need to run a memory checker to make sure the RAM is okay onboard and on the ACA...

[d3cka2d3pwood]

It might be a faulty accelerator.
Where did you get it from? How old is it?
I had same problems with my ACA 1221ec and after an awful lot of discussion and fault finding with the supplier it came down to a faulty card.
I upgraded to the ACA 1233n and runs sweet as a nut now.

[xnopasaranx]

I have the ACA 1233n and ordered it from vesalia.de

Dang... if that is true and my accelerator has a fault, I hope I can still exchange it with them. I ordered the card in feb 2017. Initially I said I got it about a year ago, but that was my mind playing tricks on me apparently. Just looked at the order confirmation.

I need to test with and without the accelerator installed to narrow it down. Hopefully the reseller will understand that it was very hard to troubleshoot and exchange the card.

EDIT: Well, I guess the warranty should be for two years, as that is mandatory in Europe I believe. So if the shop can't help, I'll have to get in contact with individual computers themselves. I wrote vesalia an email describing my dilemma, anyway. Now I need to run some more tests.

UPDATE: okay it is definitely the ACA1233n... I ran systest from adf with and without the card inserted. with the accelerator I get a freeze once certain memory areas are being written to. I will need to contact the reseller and see if I can get it fixed through them

[d3cka2d3pwood]

Hope you get it sorted mate.
I had a big head ache with getting mine sorted and it cost me a lot of time and expensive of returning back to the supplier at least 5 times which was never reimbursed.
I heard vesalia are good though so you might be in luck.

[xnopasaranx]

Argh! I had an email exchange with Vesalia and attached screenshots of the error messages I took when running systest...

they advised that the address range is infact not on the ACA card but the chipmem itself. So yeah... not as "easy" as I thought. Looks like I have to replace the onboard ChipMem myself.

I have ordered a few V53C16256HK50 EDO Ram chips as suggested in another thread here on EAB. They will ship from mexico, so this will take a bit until I can finally get my Amiga fixed.

Should I put in sockets for the chipmem while I am at it, or is it better to directly solder these? I have a desoldering vacuum gun, so hopefully it should be an easy repair...

Oh yeah, I ordered 3x 4 RAM chips, to be sure I don't have a non-working one and have to order again. I am sure the left overs will find another home in the community...

Attached are the screenshots of systest for reference...

[d3cka2d3pwood]

Hmmm why would you Amiga work fine without the card in then?
This is beyond me sorry but there are some very knowledgeable people in here though so fingers crossed they they see your thread.

[xnopasaranx]

Originally Posted by d3cka2d3pwood

Hmmm why would you Amiga work fine without the card in then?
This is beyond me sorry but there are some very knowledgeable people in here though so fingers crossed they they see your thread.

Yeah I was wondering about that, too. But if the RAM is wonky, I guess the problems will surface when an accelerator card puts additional stress on them?

Maybe someone here can clear this up.

I'll have to run systest without the accelerator for an extended time to see if it freezes then, too.

[d3cka2d3pwood]

@Daedalus can you help xnopasaranx.

[xnopasaranx]

thanks, I also pm'd mr. schoenfeld himself. maybe he can give some tipps to narrow down the problem as well.

[Daedalus]

Hmmm... I don't know if I can add too much here, the PSU is the first thing to look at but that's already sorted. Still, you might have a look at the 5V rail after the machine's been on long enough for the problem to occur, with the accelerator and without, just in case it's dropping with the extra load. Usually the A1200 resets before the rail drops too low for the chips, but there could always be something else going on too.

I can't say for sure why an error might happen in chip RAM only when the accelerator is attached. The chip RAM bus is isolated from the accelerator by Budgie and Alice, so it's more likely down to some issue with timing, crossing the clock threshold between the CPU and the AGA chipset. With the timing fix capacitors already removed, I don't really know what else can be done there however, other than the usual - cleaning contacts on the motherboard connector, slowing down the ACA clock speed to see if it goes away, etc.

[xnopasaranx]

thanks for the tipps. seems like I am on the right track now, but still I am not quite sure how to proceed. To clock down the ACA, I believe I would have to swap the divider resistor? and probably the quartz as well? or is another divider enough?

On the weekend, I finally have some free time, so I can measure with my scope what is going on on the 5V rail. Pretty sure though, that my issues are not power related anymore.

I'll also test the chipmem without the ACA for an extended period of time.

Some additional results show, that the crash appears quicker than I thought when hammering the RAM with systest. One of the screenshots shows that the amiga already froze on the first cycle (Round 1.5). So that must have happened after 30mins or so already. Guessing that the memory access with games is usually not as thorough and that is why the problem appears a lot later in normal use.

@Daedalus do you think that swapping out the chipmem will yield better stability? and would it be a goog idea to put in sockets for the RAM chips? or is that a bad idea? I think I'd go for precision sockets anyway (the ones with a locking mechanism and a little lever), rather then the normal ones.

EDIT: Oh and I have of course tried to clean the contacts before installing the accelerator. I used a rubber eraser and isoprop alcohol after initial cleaning. But I will try to pull the card out a bit to see if the connection gets better and rule that out.

[Daedalus]

Sorry, scratch the slowing down suggestion - don't go changing crystals just for troubleshooting! I thought there was a software way of adjusting the clock speed, but it's only on some ACA boards.

Yep, games will have a lot of their code in fast RAM as well when that's present, so you're using less chip RAM, and using it less frequently too, so it probably takes more time to hit a bad patch.

Swapping out the RAM might improve things if the RAM is marginal, but it's hard to know really. I'm undecided about sockets - on the one hand, it makes things simpler if it needs more shuffling of RAM chips, but on the other hand, sockets can introduce other variables and problems into the mix.

To me this feels like the machine is tripping over on the fine timings involved in CPU transfers to chip RAM, but it's impossible to dell without a more detailed look.

[Lord Aga]

Originally Posted by Daedalus

I thought there was a software way of adjusting the clock speed, but it's only on some ACA boards.

Well, yes. With the ACA 1233n 40MHz version you can change clocks to 16, 21 and 40 steps from Workbench using ACA config tools.

[Daedalus]

Ah, thanks. That's what I was suggesting to try!

[Lord Aga]

Frankly, I am surprised this wasn't the first troubleshooting step.
Maybe the card just isn't stable at 40MHz, but will work properly at lower speeds.