CT63 fault diagnosis

[dml]

I'm writing this up - but keeping it brief - in case it helps someone else debug a similar looking problem in future.

I obtained a CT63 some time ago with the intention of using it to test stuff I was working on. Remote-debugging compatibility & timing problems over forum threads doesn't always work :) so it made sense to have one ready to test.

However due to cirumstances, I put it away - untested - for a future time. The untested part was my first mistake. When I did finally bring it out and set it up - I found it didn't work. :/


The symptoms looked like this:

- machine boots and runs perfectly in 030 mode, no issues.
- machine shows the 060 splash screen in 060 mode, then halts on the TOS boot screen reporting no SDRAM, no CTCM (a CTCM is present!)
- flashing the CT60TOS.BIN from 030 mode would succeed. the flashed version reported correctly in the flashing tool before and after.
- 60MHz clock measured ok on CTCM pin 2 (and the 060 is clearly running, to show the SDRAM/CTCM errors) so it's not like the CTCM is dead.

error.jpg

Cutting a longer story short, I made several attempts over the last year or so at getting the board to work in different machines, with different CPUs and two SDRAM sticks, all of which had been checked in another machine. Always the same result.

Since the top CPLD is responsible for SDRAM and also tied to the CTCM I suspected that being faulty, a solder bridge, a dry pad somewhere or just needing re-flashed.

On two separate occasions I took a pass over the board with a microscope, nudging the CPLD pins with a needle and looking for solder bridges. I could see several obvious bridges on the CPLD nearest the RAM but they looked very intentional and correlated with the schematic. Couldn't find anything wrong - soldering all looked good, pins were firm and joints were shiny.

At that point I sort of gave up on finding anything visible. It was JTAG time, or hot air time (wasn't super excited about either option).


Fast-forward to this week.

After a bit of to-and-fro with Mikro over CT6x known issues, hints, flashing procedures and stuff to check, I went back to the schematic and the microscope and AGAIN started checking out pins and traces between the CPLDs and areas of interest (CTCM, SDRAM, CPU in that order).

...and then at some point, with the lighting at just the right angle, I saw this...

bridge.jpg

...what looks like a very, very tiny bridge - and not on the pcb itself but just weirdly stretching from one pin to another (60-61). I checked them for continuity and the beeps confirmed it.

(This was not really visible when viewing straight down unless the lightsource is in certain magic positions)


Looking at the schematic, those are address pins and no way they should be joined. So I dabbed some flux and just touched the pin with the iron, cleared the bridge, put the board back in the Falcon and fired it up.

On the first try, the machine boots to the desktop and appears, so far, to be fully working.

board.jpg

So after all that, I think the USEFUL TAKEAWAY here is that an addressing fault in the lower CPLD can cause the machine to just not see devices which appear to be managed by the upper CPLD at the other end of the board. I probably didn't give the lower one enough attention on the last 2 passes because it did not seem to be involved...

[exxos]

Centurion strikes again :roll: There was a thread a while back that he flipped the power connector and ended up with thin tracks causing a power drop and all sorts of issues. There thread is somewhere on this forum.. I think the guy bodged a thick wire in the end to solve the voltage drops.. Maybe someone else remembers the threads..

EDIT:

Think it was this one..
viewtopic.php?p=91433#p91433

[Steve]

How did he manage to sell this to someone untested? Did you get it from him direct dml. There is no way this would have worked from the get-go. It's worrying to think he's selling expensive stuff like this without even bothering to test functionality.

[dml]

I really just tried to document this so others can recognise it if a similar problem happens on a board (manufacturing fault, soldering accident, loose offcuts/solderballs, lifted pin etc. etc.). There are a lot of address/data pins on these things so if any pins do get bridged, chances are it will be a pair of address or data pins.

The CT6x boards (all revs) and Falcon machines can have so many weird problems and it's difficult to search for cases. There isn't really a 'log' of faults vs solutions its a case of digging through forums and hoping you hit something relevant :)


In this case, I got the board direct and it does seem as if it was like this from the start. However I didn't set up to check it on arrival and put it aside, which was my fault. I can't be 100% sure this bridge wasn't a nearly-bridge and became a real one after getting mashed/pressed or whatever after shipping or when I first tested it so I'm not going to complain...

However... do make an effort check any custom-made boards at the time they arrive and not a year or more later :P

[officer960]

@dml what power supply did you use to get your CT63 to boot? I've tried five different power supplies and basically gave up and re-installed my DFB1X.

My CT63 same situation - got it and took a while to source an 060. It wouldn't boot even in 030 mode. Examined it closely and saw tons of solder related issues everywhere. I sent it back to him on my dime and he said he fixed it:
I have checked your CT63. Not only I found cracked capacitor and hairline short on memory CPLD, but also an entire set of all three bus drivers wer completely solder free on one side. A complete, utter f-up of me. I do apologize for all the inconvenience. I will clean the board, assemble with your RAM stick and CPU and ship it back to you ASAP:
I want to compensate you for your trouble. Is there anything you would like from my eshop?
Please let me know.

It still only boots one out of ten tries. I'd pretty much given up on ever getting it to work, but your post got me motivated to give this $700 paperweight another try.

[exxos]

Man that's bad.. He can't be testing them.

Bit of a thread hijack but I'm half tempted to make them myself due to all the screwups which keep happening. But with me not being able to ship easily to EU anymore.. But I'd rather produce the one Willy made but someone else would have to track him on AF and ask him if a deal could be done.

[dml]

@dml what power supply did you use to get your CT63 to boot? I've tried five different power supplies and basically gave up and re-installed my DFB1X.

I did have some weird problems with it not booting, when the same supply would boot a CT60. I didn't really get to the bottom of it but did notice there is some significant difference in the way the momentary power button (power/reset/LED headers on the back top/back of the board) debounce is being handled on the board so it is really sensitive to the amount of time the button is 'on' when pressed.

So I would e.g. poke at it with something to short the power pins and the fan would twitch but not spin up. I could do this maybe 20 times and get it to start once. It would not start successfully with the momentary button though.

With the CT60 it would start 1 in 2 times poking it with a short - and would nearly always start with the momentary button attached.

This was some older micro-ATX supply. It was more successful (but not 'fixed') with a second one but no clear difference between the two in power handling etc.

I figure its a timing thing with both the PSU power-up ramp vs power-good signal, and the momentary button timing. Both combined make it a PITA to start up for reasons not very obvious.

I think it was working better with an old PC brick ATX but I would need to check again what the situation is with it. I was able to get it booting though with a momentary button.

This particular problem involved powering up at all - i.e. the fan would not spin up. There's a second problem with booting a black screen which can also involve power - see below.

Examined it closely and saw tons of solder related issues everywhere. I sent it back to him on my dime and he said he fixed it:

Well good that it was fixed but clearly that could not have been tested before it left :/

It is really too complicated and the soldering is too fine to have a 100% pass rate after assembly, really needs to be in a machine and through some software tests.

It still only boots one out of ten tries. I'd pretty much given up on ever getting it to work, but your post got me motivated to give this $700 paperweight another try.

Does the fan always spin up but shows a black screen? Or does it just not power up most of the time?

In the case where the fan spins up but 90% of the time you just get a black screen - that can be a reset timing issue. This again involves the power supply ramp-up and the Falcon's own reset timing capacitor. I have the same problem on mine and it seems common - but the success rate for successful startup is different on every machine.

One hack is to change the Falcon's reset timing cap to something like 10x the normal value. Another is to try an old ATX brick (i.e. non-micro-atx) to see if the behaviour changes. And hitting the reset button on the back of the machine multiple times after powering on.

Also clean the pins on the expansion port with contact cleaner. I've had some problems with this even using the DFB1X. The pins are supposed to be gold plated but its almost non-existent, there can still oxidation and it can behave differently with different boards if there is any resistance at all.

It could still be some other problem but it sounds like the reset timing thing, if it can start *some* of the time. Careful checking with a microscope, pushing every pin with a needle, also seems like a good idea just in case.

[mikro]

I think it was working better with an old PC brick ATX

Exactly the same happened to me a few days ago, when trying to bypass the faulty 030. I had the Centuriontech's CT60 here, it wouldn't start for a world. And I consider myself a "screwdriver veteran" when it comes to short-circuiting the power pins. In the end I had to move the CPU+SDRAM to my old CT60 where it started on the first go.

Before all of that, more than a year ago, I could confirm your findings: https://www.atari-forum.com/viewtopic.p ... 72#p460972. So the latest CT63 is really extremely sensitive to the press-and-release interval. Sure, if you use Centuriontech's CT60PSU, it's all fine but that uses a different mechanism.

[coonsgm]

I myself purchased my first CT63 which was a basket case which would recognize SDRam in the bootup sequence but not once I got to the desktop. I tried 3 different 68060's at least 15 different RAM sims without any better results.

I reflowed and checked every pin on the Xilinx chip closest to the RAM socket as a last resort and finally got it working after about 6 months. After a while it began to have issues again so using a microscope I discovered one of the legs on the Xilinx was severely bent and was just barely connected. Using low temp solder I removed the chip with the intent to fix the leg and and resolder back to the board.

I'm not sure if the leg had been damaged previously or if it was my fault, but literally just touching the leg caused it to break off.

I ended up ordering a new CT63 (rolling the dice) and this time got one that has worked perfectly the last year and a half......I would love to fix the previous one....but I'm not sure how to program a new Xilinx chip.

Any ideas?

[exxos]

I myself purchased my first CT63 which was a basket case which would recognize SDRam in the bootup sequence but not once I got to the desktop. I tried 3 different 68060's at least 15 different RAM sims without any better results.

Damn thats another bad one then :(

Any ideas?

I wrote this a longgggg time ago..

https://exxosforum.co.uk/atari/last/CT60/flash.htm