A1200 motherboard stopped to work suddenly

[voyager]

Cosmos sells capacitors packs on Ebay, or send him a PM.

[rkauer]

Cosmos sells capacitors on Amibay, too. And Amibay have no site fees.

[whitebird]

Hello,

After a few month of standby I decided to try again to solve the problem with the 1200 mobo.

The phenomena has changed and now is completely deterministic.

After turning on the PSU, I get a black screen during 0,5 seconds, then Yellow screen (Power LED flashes 10 times) and again black and yellow screen and so on.

Now I know the problem come from an exception occuring before the guru was installed.

Note: I changed some of the capacitors but same result.

So with the help of 4 AM29F010 flash memory and after splitting the kickstart into 4 128K files, I made a circuit that gives me possibility to catch what interrupt is occuring (by modifying the kickstart code).

And I found 2 exceptions:

-Address Error
-Illegal Instruction.

Strange as I first thank it should be a hardware exception (interrupt).

But what I don't know is the instruction that makes such an exception (reading a word at an odd address)...

Is there a way in the exception routine to know what instruction (referenced by its address) triggers the exception?

Thanks

[whitebird]

I went deeper in the analysis: with two 32K 8 bits static RAMS, I was able to catch what was going on the address bus before the exception.
As a second step a PIC MCU reads the RAMs and sends the address to a PC so I have a text file showing them.

So I know what triggers the exception:

It's the routine called from f80e00 and located at address f818da.

it's the instruction
000018fc 2089 MOVE.L A1,(A0)

Of course there is no reason to have an error (word access to an odd address).
Does someone know what this routine is doing?? This will help me find if the problem may be due to bad address decoding from custom chip or if the fault is coming from the CPU.

[whitebird]

After some months of standby, I finally managed to find what was preventing the board from running correctly.

My first supposition was the good one: the RAM chips are faulty. I ran the logica diagnostic software which showed faults in the first 1MB segment, but everything good for the second segment.

To be sure the diagnostic was telling the truth, I remapped the second 1MB segment to address 0 by connecting RAS_0 (from budgie) to the /RAS pins of U18 U19 and be living /RAS of U16, U17 open (undriven).

Now I have a working board (I managed to run some games from a modified PC floppy drive) but with only 1MB RAM.

Because the original HM514260 chips are hard to find, I will replace them with TMS45160 which have the pinout. Need to check if timings are compatible.

[whitebird]

So, I bought 1 TMS45160 RAM chip (only 6€) as a replacement for faulty U17. Everything is working fine, the motherboard has again 2MB chip RAM.

Good to know the TMS chip is equivalency for the original A1200 DRAM.

[kipper2k]

Originally Posted by whitebird

So, I bought 1 TMS45160 RAM chip (only 6€) as a replacement for faulty U17. Everything is working fine, the motherboard has again 2MB chip RAM.

Good to know the TMS chip is equivalency for the original A1200 DRAM.

Good to hear, where did you get the ram from ?

[Retrofan]

A newbie question: What happens is you piggy back another ram?

[whitebird]

Got it from from the french reseller "électronique diffusion":

http://www.electronique-diffusion.fr/

They sell through the web but also have regular shops.

Retrofan, I'm not sure I understand your question well:

By "piggy backing" another RAM you mean put the new RAM chip over the one that is faulty without disoldering?

[Retrofan]

Originally Posted by whitebird

By "piggy backing" another RAM you mean put the new RAM chip over the one that is faulty without disoldering?

That's a way of piggy backing. No, I tell now to put another over a working one. What would happen? Would it recognize another mega?

[whitebird]

No, it will only result in higher currents when the data bus is output. To have more RAM, the chip has to be mapped at different address. Like that both chip receive the same address.

[kipper2k]

Originally Posted by whitebird

So, I bought 1 TMS45160 RAM chip (only 6€) as a replacement for faulty U17. Everything is working fine, the motherboard has again 2MB chip RAM.

Good to know the TMS chip is equivalency for the original A1200 DRAM.

I ploughed that part number into that french website and only found a 16 pin dip. Is TMS45160 the right part number ?

[Retrofan]

Ah, ok. I just read this and was wishing to ask:

"From: MARTYGOODMAN To: DAVIDBOND
PIggy-backing RAM chips involves putting rams chips NEARLY
in parallel … with some kind of enable line held apart for
the two chips so they can be addressed separately.
Back in the days when large size memory chips did not exist,
and even what we today call “medium size” memory chips
either did not exist or were hideiously expensive,
piggybacking RAM chips provided a means to add more memory
to an existing system that did not in its design have provisions
for adding such memory… that is, a system that could ADDRESS
more memory, but which did not have physical space on the board
for adding such extra memory.
The original 4K CoCo 1 could easily be upgraded to 16K, but
beyond that one had to PIGGYBACK two rows of 16K DRAMs.
Just as a goof, I managed to outfit two of my CoCo 3 128K
machines with stacks of 4 piggybacked 4164 chips where they
used to use a single 4464 chip. They both work fine with
the piggybacked chips.
—marty"

Yes, he says that the system has to be able to address more ram.

[whitebird]

The picture do not correspond to the DRAM chip, they show the same 16 DIL chip for many components. That's a bit confusing.

The chip is 40 pins SMC, here is the direct link:

http://www.electronique-diffusion.fr...oducts_id=8426

Retrofan,

Yes it results in more memory, but as written, only if the chips have separate enable lines, which is similar to bank switching, and having separate enable lines is in fact like adding another address bit. But generally it requires some sort of software driver as CPU hardware cannot handle those enable lines by itself.

=> some pins of the 2 chips cannot be mounted in parallel.

[qrulf]

@whitebird: Just of interest, what technique do you use to desolder those 40pin chips?

[whitebird]

As the chip is dead, you have to focus on saving the motherboard so it doesn't really matter if you damage the chip.

I had luck because the dead component was U17, which has one side more accessible than the other DRAMs.

First I cut every pins on one side with a cutting pliers then the chip was rotated 90° and the remaining 20 pins were cut on the other side.

To finish the job, I had to disolder the small pieces of pins that was still soldered on the pcb.

Soldering the new chip was far more difficult because of the pins proximity and difficulty to reach the surface below the pins.

The whole process lasted about 3 hours.

[qrulf]

Ouch, that sounds like a tough operation - but probably a nice feeling afterwards :-)

Thanks for sharing!