GAL / JEDEC files for the A3640 card

[Plaz]

Does any one know if the jedec files for the gals on A3640 CPU card exist and where I might collect them?

Plaz

[Castellen]

Yes, they're all available in this archive (http://amiga.serveftp.net/MiscFiles/A3640_PAL.lha).

I can also supply replacement GALs pre-programmed. More information here (http://amiga.serveftp.net/A4000_HardwareGuide/a3640ref.html).

[Plaz]

Awesome! (and fast) Thanks. I just finished reading your page actually, but moved on as it only specifically mentioned U209 for the 3.2 upgrade and I didn't notice the file link. I'm thinking shipping from NZ to the US would cost more than the chips, but what would a whole set of GALS cost me?

The plan is to burn replacements to help me repair my dead cards.

Plaz

[Castellen]

Actually, shipping isn't as bad as you might think. I usually send replacement parts as a DLE letter, which costs about US$1.50 (NZ to international). There's no tracking for a letter, though I've never had one go missing before. Delivery time isn't too bad, usually 1-2 weeks.

The price of the blank GALs is a bit of a killer. You'd be looking around US$11.50 each, I only buy them in small quantities (less than 100 pcs at a time), hence the high price.

The PALs on the A3640 are flash devices, and there's no security set, meaning you could unsolder the devices and verify them, then put them back in circuit as required. Obviously you need a programmer which supports this.

From my experience, it's not that common for the programmable logic to die. Common problems I've seen with A3640s tend to be:
- Intermittent socket contact with crystal oscillator
- 22µF capacitor problems (incorrect polarity, leaking+corrosion)
- Poor socket contact in the 200 pin CPU slot
- Bad 68040

Personally I'd look at these things first before attacking the programmable logic. You could save yourself a lot of time and money.

Other hardware repair info I've written which might help you here (http://amiga.serveftp.net/hardware_repair.html).

[Plaz]

there's no security set, meaning you could unsolder the devices and verify them, then put them back in circuit as required.
I was wondering about the security. Now that I know this, I can test the gals myself. First I have to build the programmer of course. That begins tomorrow. :-) Thanks for the tip.

From my experience, it's not that common for the programmable logic to die. Yes I agree. In my case however I've tested just about every thing else.

- Intermittent socket contact with crystal oscillator
* Socket and oscillator replaced.

- 22µF capacitor problems (incorrect polarity
* All replaced with proper polarity

- Poor socket contact in the 200 pin CPU slot
* Reflowed slolder on each pin

- Bad 68040
* Tried 3 different CPUs

Additional shooting...
Looked for any bad signal levels (~ .5 - 3.5v)
Curve traced the entire board and ....
... found U402 to be bad and replaced.
Reflowed all pins on the CPU slot
Curve tracing also shows address line 3
looking very suspicious. This leads me to
look closely at the GALS since logic levels
and signals look ok at the tristate buffers.

Next is to confirm all traces and GALs

If you have any other ideas, pass them along :-)

Plaz

[alexh]

If you have a v3.0 board the most important thing to perform is the "cut and jumper" patch.

http://wonkity.com/~wblock/a4000hard/a3640ref.html

[Castellen]

Looks like you've covered most of the things I'd have done already.

More ideas:
- Confirm the oscillator module is delivering the correct output, 28.whatever MHz. You'll need a scope or counter to measure this.

- If any of the electrolytics have leaked, carefully check surrounding vias, and that tracks are actually connected to pads.

- If any discreet logic has been exposed to the corrosion from the capacitors, it's worth changing them. The octal latches and buffers used on the board are cheap and easy to get.

I guess saying to check the motherboard clock source jumpers are both in the EXT position is just stating the obvious :-P

[Plaz]

@alexh

Yes, one of the cards is a 3.0 that I performed the jumper hack and installed a new -03 U207 years ago. It worked well in my 3000 desktop for 2-3 years then it croaked one day.

The second card is a 3.1 and had work without issues, so I left it alone 'till now.

Plaz

[Plaz]

@Castellen

More ideas:

I've been down that list pretty well too. I even moved the jumpers around just to tick off the card. :-)

I've bought a few replacement buffers, but have yet to find one that seems to be malfuntioning other than the one chip I mentioned. I do have a scope and I've been testing other chip functions by injecting signals and watching their behavior. The GALS are an unknown as I don't know how the should react with their programming. Now if I had a working card and a logic analyzer, I'd really be dangerous.

It's possible I have 3 bad 040's, but that would be highly unusual. I'd need a known working card to check them in to be 100% sure.

I'm think and hoping that if I can find the reason for these two cards to fail (besides the CPU), maybe I'll identify a common failure in other cards.

Confirm the oscillator module is delivering the correct output, 28.whatever MHz

Did you mean from the motherboad? I was assuming every thing was ok from the motherboard as a 030 card boots ok. But there could be a problem getting 28mhz through to the A3640 logic. I'll try to check that.

Plaz