Looks like I'll be getting a Mega ST4 soon

[PhilC]

That looks like bit 4 of that bank is stuck high or bit 20.

Have a look at the ram page of the schematic for that board and it should show you which ram chip is which bit.

[rubber_jonnie]

Thanks @PhilC I'll take a look.

[rubber_jonnie]

Going back to this earlier post in the thread: https://www.exxosforum.co.uk/forum/viewt ... =40#p27737 it looks like bit 4 of bank 0 is the problem, which is U44, which is giving a different result with my logic probe on Din than the other chips.

Interestingly U64, which equates to bit 4 of bank 1 is also giving the same logic probe result on Din, so since I have spare chips I'll replace U44 first, then if things are still off I'll replace U64 as well.

I do hope that I don't get any more RAM chips dying as I have a really limited stock of replacements, and certainly not enough to replace every chip the machine has.

It is beginning to feel like this was a Friday machine!

[PaulJ]

Interestingly U64, which equates to bit 4 of bank 1 is also giving the same logic probe result on Din, so since I have spare chips I'll replace U44 first, then if things are still off I'll replace U64 as well.

Rj, strange that the b4 looks suspect in each bank. Maybe compare the resistance of b4 and another bit when measured to +5 rail. This machine if a car would be referred to as a lemon.

[rubber_jonnie]

@PaulJLemon indeed. I was just about to get into making some moves and swapping out the RAM because I can and its not like I can go out and do things at the moment, and I knocked my favourite soldering iron off the bench and broke it 😭

Have glued it all up, and I have a backup to use if it doesn't work, but not happy right now. At least it still seems that replacement irons are still available, so if it is broken I can get one and carry on using the old base.

I'll give your suggestion a go, see what happens too.

[rubber_jonnie]

Ok, so after a bit of digging, I discovered that the error messages are interprted as follows:

0000000000010000 0000000000000000 000008

0000000000010000 - Value written to RAM

0000000000000000 - Value read back from RAM

000008 - Memory location tested.

So this test doesn't relate to the BIT value in the bank, but it looks like the beginning of RAM possibly, so I will start taking a closer look there.

The LS244s have been ruled out as removed and refitted after being tested with my TL866II+.

Out of curiosity, can anybody recommend a replacement for the TC511000P ram chips used in this model of Mega? They are hideously priced per chip these days and not so easy to find. Think I'll have to have a delve around UTSOURCE to see what I can find there.

[PhilC]

It's still bit 4 though. Remember it can fail due to faulty ram or bad tracks on the motherboard.

May be worth you metering out the mad lines etc to each chip, which is quite easy as I'm sure you know as they are more or less wired in parallel for each bank.

Edit:

I may have a few spare chips lying around that you can have, they will be used though.

[PhilC]

Just a quick thought.

Check the ras and cas lines to each of the 4 banks. I've noticed that bank with failed chip will usually not pulse these lines but a good bank will have pulses ras and cas.

(That's if I remember correctly :lol: )

[rubber_jonnie]

It's still bit 4 though. Remember it can fail due to faulty ram or bad tracks on the motherboard.

May be worth you metering out the mad lines etc to each chip, which is quite easy as I'm sure you know as they are more or less wired in parallel for each bank.

Edit:

I may have a few spare chips lying around that you can have, they will be used though.

Getting a different result now:

Bits_8_and 10.jpg

Bit's 8 & 10 now, this after replacing the bit 4 chip in both banks. I've tried piggybacking on both banks (I think Bank 0 is the problem), but no different. I also buzzed out the address lines and they are all ok. Will check the RAS/CAS lines as suggested.

[rubber_jonnie]

So delving into the service manual for the Mega, and thinking out loud here for the benefit of others, I've come to the conclusion that the the info I got here wasn't quite correct: http://92.222.181.53/viewtopic.php?f=15&t=21681#p193222

Whilst what he says makes sense for the normal RAM diagnostic, since you get a value written followed by a value read, as per the service manual:

If an error occurs, the error code is displayed, followed by
the address, data written, data read, and the bits which did not
agree. E.g.: " R2 45603E W:603E R:613c bad bits: 1,8".

It is clearly not the same as a dead machine RAM test.

Looking at other documentation, and what @PhilC posted here: https://www.exxosforum.co.uk/forum/viewt ... =29&t=1956 it does look pretty clear that the error report is actually the following format:

Error High Bank (2049-4096) Low Bank (0-2048) ADDR

I1 RAM/data line error (D15)0000000000000000(D0) (D15)0000010100000000(D0) 000008

Based on the service manual data, the error above is for bank 0 as I had thought and suggests that reading right to left gives the faulty bits of 8 & 10 as I suggested in my last post. Using the following:

Bank 0:
U40= bit 0 0
U41= bit 1 0
U42= bit 2 0
U43= bit 3 0
U44= bit 4 0
U45= bit 5 0
U46= bit 6 0
U47= bit 7 0
U48= bit 8 1
U49= bit 9 0
U50= bit 10 1
U51= bit 11 0
U52= bit 12 0
U53= bit 13 0
U54= bit 14 0
U55= bit 15 0

This tells me that U48 and U50 are bad, so that's what will get changed next, or at least it will when my order of 18 pin sockets arrives. In the meantime I'll either get the logic probe (Great for very quick tests) or the scope onto the RAS/CAS lines, see whats going on there.

I have tried piggybacking with good chips, but it didn't work, but on a positive note, the LS244s and LS373s have all been verifed as working as they have been removed and socketed.