EII+HD no longer sees second RAM bank

[Emu276]

I have enjoyed the ability to load two different banks into my EII+HD (Bank A and B) for many years without problems. I simply press "0" twice on the keypad to switch between them as designed. Today I was working on some repairs inside the instrument, replacing a RAM socket (IC14) that had a corroded pin, as well as repairing a corroded pin on one of the Z80A CPU sockets that I believe was preventing the EII from booting.

Once I completed the repairs I was happy to see that the EII+ booted fine, but now I cannot switch between Bank A and B! When I load a bank into Bank A it plays fine, but when I press "0" twice the display does not switch to Bank B. Instead it just stays on Bank A. For example if the display reads "P01 Piano" and I press "0" twice it reads "P   Piano"...so instead of switching banks it simply doesn't display the "0"s because there is no "00" preset in the bank. (It think this is normal behavior on a standard EII, yes?)

Two additional interesting behaviors:

1) When I run the Memory Test disk, the test proceeds through Bank A all the way to Segment F and then instead of proceeding with Bank B the display goes completely blank and it hangs there. (Normally the test would proceed through Bank B and then display an "all memory is OK" message).

2) When I power on the EII+ and it boots the OS from floppy, I notice that it loads faster than I am used to. The familiar series of  mechanical sounds of the drive reading from the floppy is shortened, as though the EII is now no longer loading some information that it normally should.

Here are my thoughts: It seems the second bank of 512K is no longer being properly detected by the machine, so it boots as though it is a standard EII  (and not an EII+). Could it be that the shortened boot time is because it is not taking time to load a second empty bank from floppy to Bank B?

What could cause the machine to no longer see the second bank? This particular unit is a late model EII+HD so instead of a RAM daughter board it has M41256PP-15 ICs installed.

I would greatly appreciate any tips on troubleshooting this further.

Thanks!
Howie

 

[Emu276]

Ok here is some additional info after troubleshooting: To rule out the possibility of one or more bad DRAMs causing the loss of Bank B access I systematically removed DRAMs one row at a time. I thought the OS was loaded into the first row IC13-20 but this is not true on the EII+HD because I can completely remove the first row and the machine still boots fine and loads/plays sounds fine. I determined that the OS is actually loaded into row three IC51-58. When that row is removed the machine won't boot.

The rows three and four are the complete Bank A memory, including the memory for the OS as I mentioned above. With just those two rows populated with RAM the instrument behaves just like a normal stock (unexpanded) EII. When I run the Memory Test disk with just those two rows installed it passes the Bank A tests fine, and then loops back and starts testing Bank A again. This is interesting to me because it seems it is not even looking for the Bank B RAM so the test does not crash.

If I have rows two, three and four occupied with RAM I get the same result.

When I fully populate all four rows (all 32 filled RAM sockets) the Memory Test crashes after finishing Bank A.

I have swapped the DRAMs between the rows and determined there is no evidence that any of the individual RAMs are bad. So I am now focused on understanding why the machine is failing to properly access rows IC13-20 and IC31-39 on power-up. I believe the machine is detecting them but cannot properly read/write to them. So perhaps there is an IC that handles memory addressing that has failed (or now has a cracked solder joint).

I'm looking closer at the schematics to see which ICs may be involved, but I would certainly appreciate any advice on this!

Howie

[dfiorucci]

Check the jumpers at the bottom of the board to make sure they are getting connection.

[Emu276]

Thank you for that suggestion to check the RAM jumpers. I confirm the jumpers are still connected - I get continuity with my meter. I assume this jumper setting is correct, since it was set this way years ago when I first got the EII. See attached photo.

[dfiorucci]

Yes that is correct.
I'd go back to IC14 and check where things where corroded and make sure there was not damage to the board itself there from the corrosion.

[Emu276]

When I repaired the sockets I cleaned where the corroded socket pins were and made sure the traces/pads on the PCB still looked good before proceeding. With the ICs reinstalled in the new sockets I get good continuity with my meter between IC pins and checkpoints on the PCB.

A very interesting thing happened this morning when I powered on the EII+HD after it had been off all night. The very first boot from floppy made the full set of mechanical seeking noises that I was used to hearing, and I could successfully switch between Bank A and B and could load sounds into them properly! The sounds played fine. I was very excited because I thought it had been a loose connection all along that had now resolved (maybe due to PCB flexing). However, within two minutes the machine froze, and then when I turned it off & on again it went back to the problem I had before . And that is how it remains now.

I have already reseated all the ICs on the digital board by full removal and re-insertion (I wanted to do this to inspect for any other corroded sockets) so I am pretty sure it is not a problem of an IC needing reseating. The reseating was done last night before the one successful boot this morning.

What do you think? Is it likely any of the ICs related to DRAM access could fail intermittently such that it works occasionally? I would think that once an IC fails it simply won't work properly at all...?

Or am I looking at the dreaded case of a tiny cracked trace / solder joint somewhere?

Thanks,
Howie

[dfiorucci]

I know this is going to sound strange but check the calibration on your pitch and mod wheels.
If they are out of tolerance it can cause some weird behavior and even fry one of the chips.

[dfiorucci]

It's IC27 that can get ruined by out of tolerance values for the pitch and mod wheels.

[Emu276]

Thanks for the suggestion to check the wheel calibration. I saw that note in the service manual.

I have found now 50% of the time when powering on it successfully boots and allows Bank A and B switching, but it is interesting that I can load two banks at first and both sound fine, but if I leave it on Bank B after a minute the machine freezes. At the moment it freezes if I have notes held down there is a sudden blast of distortion and the notes sustain forever. However if I have two banks loaded but keep it on Bank A the instrument will play fine and not crash at all.

Therefore since the memory access to Bank A remains fine I think the problem is isolated to components (or traces) that are specific to Bank B and not something shared by both. But I don't know why I can load Bank B and play it fine for a minute before it crashes, though!

One thing I will also do is reinstall all the screws holding the digital board. Up until now I have left them out to make board removal easier, but maybe if the problem is a cracked trace then tightening the board can reveal that.

Thanks again for all of your help!

[dfiorucci]

You might check the connection between the digital and analog board before you button it up.
Also re seat all of your cable harnesses.
Hope you find the issue soon and get it back up and running.

[dfiorucci]

Another thing you could do is run the debugger.
You would need the debug IC along with a USB 2 serial cable like this one
https://www.newegg.com/grey-c2g-6-ft-cable-connectors/p/N82E16812196147?Item=N82E16812196147

It can give you a better understanding what is going on with the memory.

[dfiorucci]

RAM Address control is stored in a set of latches for upper and lower address bytes check IC's 94, 95, 96 and 97.
This is listed from the service manual on RAM Address Control.

[Emu276]

I'm happy to say it appears the EII+HD is working now!

I did three things:

1) Reflowed solder on the pins of the new sockets I installed for the Scanner Z80 CPU and DRAM IC14
2) Swapped DRAMs IC13-20 (row 1 that holds Bank B) with IC51-58 (row 3 where the OS loads as part of Bank A)
3) Swapped the two identical SRAM ICs between the Z80 Main CPU and Z80 Scanner CPU

I normally wouldn't do so many things at one time because it's hard to know what actually fixed the problem, but I was getting very tired of doing thing methodically and felt these were the 3 likeliest targets. I was able to load Bank B and it has not crashed for several hours now, and thankfully neither does Bank A. The Memory Test disk gets through both banks and displays "Memory Good".

I had previously swapped the DRAMs around, so I don't think doing it again changed anything. And I had previously tested the new socket connections so I don't really think it was that either.
My feeling is it has to do with the CPU SRAM that I swapped. Maybe if one of those is malfunctioning it impacts the Main vs Scanner Z80 differently so the bad IC can still work alongside one but not the other...?

I may never know what exactly the problem was, but for the moment all seems back to normal, and if it stays this way that's what I care about. I haven't closed everything up yet...will wait and see if things remain good for the next 24 hrs.

Thank you for all your help. Your comments kept me motivated to sort this out.

See the photos - these are sights I like.

Howie

[dfiorucci]

That's great glad you got it working again.
My money would be on re-flowing the solder as what the solution was.
FYI the Z80's are readily available just in case you need to swap one in the future.
Hopefully not.