Amiga 3000 zip ram problem

[Castellen]

Does the fast memory work normally again if you remove all of the Zorro expansions?

I gather you've got 32 x 1Mbit ZIPs fitted to the A3000.  Are you saying that ZIPtest indicates that the first three 4MB blocks of fast memory is not working, but the last 4MB is showing as good?  Please post the entire test output from ZIPtest here as it's not obvious what's possibly good or bad, or what aspect of the test fails.


If you briefly pull the _FAIL line low, does the fast memory then appear normally afterwards?  You can do this by briefly linking U190 (real time clock IC) pin 2 to ground (the metal chassis) with the system running, which will force a hardware reset and initialise the fast memory controller (U890) to the correct power-up state.


Failing that, it may be worthwhile collecting more clues using DiagROM and using the 'Slow scan of 16MB fastmem-areas' to see what's going on.  Again, post the test result here.

[Castellen]

I'm somewhat confused as to what memory is fitted during the various tests as there are a mixture of different results.

anyhow Bank0 only or all populated.

Does this mean bank 0 (8 x DRAMs from U850 to U857) are populated, but somehow ZIPtest is showing that 24 of the 32 DRAMs are working correctly, even though most of them are not fitted?


ziptest-bank0.txt contains useful clues, where I gather only bank 0 (8 x DRAMs from U850 to U857) are populated, and the test indicates U850 and U851 (data bits 0 through 7) have problems.

At this point, if you were to exchange U850 (showing as defective) and U852 (showing as good), does the fault indicated by ZIPtest move to U852, or does it remain at the same U850 physical location?  I'm picking the latter.

The test results from DiagROM don't agree with what ZIPtest reports, where ZIPtest suggests data bits D8 to D31 are good, where DiagROM suggests none of the memory is working at all.

I'm inclined to believe ZIPtest at this point.  And if there's no problem with the ZIP memory (i.e. exchanging U850 and U852 doesn't change the outcome of ZIPtest), the common factor is octal transceiver U891.  If you don't have any spare 74F245, then exchange locations of U891 and U892.  Then if you run ZIPtest again with the same bank 0 installed, and the fault moves from U850 & U851 to U852 & U853, then the problem is definitely a defective 74F245 (now in location U892).

The 74F245 will be a DIP20 package, soldered into the board.  You'll need access to a decent de-soldering station to avoid damaging the plated through holes and tracks on the PCB.  New transceivers are cheap and easy to source, e.g. Digi-Key part number 296-3572-5-ND

[Castellen]

I ordered chips, have to wait and practice my soldering skills again on some other hardware.

Any chance to tackle the problem with a digital scope ?
Otherwise I have a cheap logic analyzer 24MHz only ... any chance to down-clock my A3000 to 16MHz instead of 25MHz ?

What was the result of exchanging U850 and U852?  Did the fault result of ZIPtest move with the moved DRAM, or did it stay in the same U850 physical location?

I'm not sure what you've ordered, do you mean you've got new replacement 74F245 ICs on the way?

You could diagnose further with a logic analyser, but it's a very tedious process to set it up then interpret the results.  Sorry, I've not got enough time (or patience!) to describe exactly what to do over dozens of forum posts.

If you have a digital scope, it's easy to see if the octal transceiver U891 is writing data to the DRAM data bus (RD 0:7 on the schematic).  If you're not sure what you're looking at, the correct signal of what to expect should be on U892 (RD 8:15 on the schematic).

[Castellen]

Just checked again, ziptest shows the same output after exchanging U850 with U852.

Great, that likely isolates the issue to something specific to U850 and U851, as ZIPtest specifically shows errors for those, and U852 to U857 as good.

74FCT245

I don't know if a FCT245 will work in that location in place of F245, I've seen problems with exactly that in A4000.  Timing is subtly different between those part variants, and that area of the A3000 is particularly sensitised to timing differences.  That's the reason I suggested the 74F245, the same as the original.

my A3000 Bank 0 showed up one time and system/mouse frozen right after, Workbench bar had 2MB + 4MB

That intermittent behaviour suggests you might be looking at possible IC socket contact issues.  And that could also explain the inconsistent test results you're seeing with ZIPtest.  Column address strobe 0 is the one signal that's common to both U850 and U851.  With the system in the faulted state, check to see if you have a valid address strobe signal at U850 and U851 pin 2.  The signal to expect is a pattern 4 x low-strobes, which should be there all of the time under all system conditions.  I've explained this signal behaviour to death in other threads, start with this one.

If it's missing, then there's a good chance you have intermittent contact issues in the U890 socket (PLCC84 device).  In which case the solution is to replace the U890 socket, though you might like to start with cleaning all of the leads on the existing U890 IC, which might be enough.  Use a single wipe of a fibreglass brush to remove surface oxide, then wipe with isopropyl alcohol on a soft cloth or paper towel.  Be sure to use a PLCC extractor tool to remove the IC from the socket, else you'll likely crack the socket, after which it'll definitely need to be replaced.