A4000 fastram address error

[Castellen]

Start by checking there is address bus data on all 32 lines at U890 (Ramsey).

At any one of the fast memory SIMM sockets, check there is multiplexed address data (row/column) on all 10 of the fast RAM address lines.  There should always be activity on these lines due to the constant refresh cycles, regardless of any system activity.  Looking at the signals with a 1uSec/division timebase, you should see something like 4 active low strobes on each line.

Would also be worth checking that all 4 column address strobes at the SIMM sockets have activity.  Those 4 signals are common to each SIMM socket, so doesn't matter which socket you check.

Could also be a defective U890, though it wouldn't be my first guess.


What behaviour do you see when running the system normally, with standard ROMs?  Are you seeing something like it running OK with chip memory only, but you get some kind of failure (e.g. yellow screen) when trying to boot with any fast memory fitted?  If you try to boot with no startup and report the system memory (using 'avail'), do you see the correct amount of fast memory reported, even if it isn't usable?

[Castellen]

If I only have chipram installed everything is fine. It boots up and the diagrom happily runs the chipram tests, including address error checks.
If I boot from normal kickstart with fastram installed the power led blinks and the screen stays black.

That's what I'd expect with an address error, so DiagROM has picked up the error correctly.  The standard ROM only does a quick write+read test, which passes in this case, then adds the detected memory to the system.  When the system tries using the memory, it fails catastrophically due to the address problem as it ends up with data for one address overwriting a different address.

Are there any other tests I can run to narrow it down before bringing out the big guns (the oscilloscope) which I'm not 100% sure how to use yet... Haha

Not really.  And unless you have months of time to waste, I'd not recommend the shotgun approach of 'replacing every component until it works'.  You really need to know what you're looking for with a fault like this, it's not a great 'my first oscilloscope' job.  You might like to give it a try, there's a fairly low risk of damaging something.  The trickiest bit is the fact that half of the important things you need to measure (at U890) is underneath the CPU board, making physical access difficult.

Another option is to send the board to me in NZ for repair.  I'd suggest checking freight prices first, international freight prices have quadrupled in recent years.  The shipping weight of the A4000D main board is around 1.3kg.  The shipping carton size would typically be around 30 x 40 x 10cm.