LaST Memory upgrade fail (1040STFM C070789)

[rmpel]

Hi all.

Today is a sad day.

I now have a broken 1040STFM.

The relevant events and actions taken;

- In preparation of the LaST Memory Upgrade, I desoldered and socketed all RAM on the board. (I had expected a bare PCB, planned on installing pins suitable for round hole sockets myself). At this point in time, no problems.
- I received the upgrade, saw the pins, desoldered them as they would be too high and the square pins are too big for round hole sockets.
- Soldered wire to pin 64 of the MMU (checked 5 times, i got the correct pin), wire routing a little different than instructions as the mentioned via is not open so I took a slightly longer route, wire length is ok, and there is continuity to pin 64 of the socket/mmu.
- put kapton tape on the sockets to prevent shorts
- placed the pcb on the sockets in the correct spot
- used crimp-on dupont pins through the hole and the tape into the socket x34
- soldered the top and clipped off the surplus of the pins
- desocketed the pcb -- very carefully, as 34 pins hold quite firmly
- soldered the bottom side to strengthen the bond (only had to fix one pin that got crooked)
- installed the module (and all this taking careful note of the PSU proximity)
- soldered the mmu wire to the appropriate pad, bridged the 1/4MB jumper.

All is well at this point from an installation point of view, however, ...

Upon boot, empty floppy in drive, I see a short flash of the mouse pointer and the computer resets.
I checked and rechecked all connections, using a multimeter in continuity mode, everything seems fine, but the machine will not boot.

Removed the 1/4 MB jumper, now the machine boots to 20 bombs.
Reinstated the jumper, back to flash of mouse pointer and reboot.

At this point, regret kicked in, and anger at myself for not being content with what I had and ruining a working machine and reverted the machine to 1 MB as before, reinstalling all 32 DRAM chips in the sockets.

Machine boots! however, only 512KB is seen.
Checked continuity on the DRAM chips, one chip did not get continuity on pin 16, fixed that, now all 1MB is seen again, however, the desktop shows corruption and SysInfo will not boot.

RamTest shows no faults, so I think I need a better ram test, perhaps something with a march test?

Video of the "Flash mouse and reset';



Image of the desktop corruption;

atari 1040stfm desktop corruption.jpeg

Any help in any form is appreciated.

Perhaps someone with intimate knowledge of the system could determine by the image which RAM chip/chips are at fault?
( as in; where in memory is the display buffer located, and which of the 4 banks of 256K would that be )

I've read that there is a service manual floating around with "troubleshooting a dead system", which could come in handy.

But I am currently in a state of mind of "doing nothing and asking for advice is better than possibly making it worse".

Thanks in advance to anyone.

Also, maybe someone could help me answer the question why the 4MB PCB did not work. There's not much on there and all datalines have continuity from pins to ramchip, I see no reason for the module not to work ...

R.

[rmpel]

Update;

Because I can't let it go... (story of my life, I won't go home from work until the problem is solved), I just turned the ST on again and ...

The problem is gone.

RamTest and also Yaart both report no memory errors and the screen glitch is gone. And sysinfo no longer crashes on start.

I don't get it.

I mean, I am grateful that it works again, yet... why...

The 4MB module is going on the shelf. Maybe if I ever get a second 1040ST, I'll try it in that one, but for now, I'll be content with the working 1M.

[rubber_jonnie]

TBH you may have introduced problems for yourself by using sockets and round pins.

The upgrade is really designed to be soldered in place so by using sockets you could have introduced marginal connections. It may well fail again if the board is moved or put under mechanical stress and using solder will prevent this sort of thing.

[Badwolf]

Is your MMU strapped in with a spring clip?

BW

[rmpel]

TBH you may have introduced problems for yourself by using sockets and round pins.

Oh, yes, I am very much aware of this, but as stated, I expected a bare PCB, and planned to use fitting pins myself.

I very much blame myself, but tbh, I don't think I'll take the leap and removing all sockets. If I do that and the 4MB module still fails, it is much more work to uninstall and revert back to stock configuration.

Experiences taught me to do as little desoldering as possible on old hardware, which is why I went for sockets.

Hindsight is 20/20.

Round pin sockets versus dual-wiper sockets... the age old discussion which is better. I have had perfect results in C64s with round pins, and I doubt wiper-style sockets would have helped better. The pins were solidly in the sockets before I soldered the pins to the module, so bad connection... highly doubt that.

Is your MMU strapped in with a spring clip?

Yes it is, with a very elaborate one, in fact.

IMG_9508.jpeg

Why do you ask? What secrets do you hold ;)

Thanks for replying, and yes, keep laying it on me, I know I done f'd up, and I need it hammered down to prevent repetition.

[stephen_usher]

The problem with turned pin sockets is that they rely upon the legs and the sockets to have imperfections in their surfaces so as to give an interference fit between the two. The contact areas a tiny but widely dispersed.

Dual wiper sockets use elastic tension in the springs to actively push the contacts hard against the pins. On chips with flat legs this gives a large contact surface area. With round pins the contact area is smaller but the spring tension still makes sure of a good connection.

In the long term corrosion can force the wipers off the legs, but this is less often a problem with dual wiper sockets.

[chronicthehedgehog]

Turned pin sockets are such a pain to desolder when you're a noob like me. I avoid them like the plague :lol:

[Badwolf]

Is your MMU strapped in with a spring clip?

Yes it is, with a very elaborate one, in fact.
Why do you ask? What secrets do you hold ;)

Quasi-random intermittent memory faults can often be laid at the door of the MMU. If it's not a nice firm fit then, especially after doing work on the board, its connections can become intermittant.

That yours is well strapped probably means it's not that. Had the strap been missing we might have had an easy fix for you.

BW

[rmpel]

I See (or IC ;) ... sorry, bad joke...)

I'm currently working up the courage to unsolder all the sockets... (512 pins to desolder) and switch to wiper sockets (as I am still hesitant to solder the 4MB directly in)

Maybe I'll try and find a spare mainboard (or a banged up full unit for parts, where the PCB is ok...) so I can leave this one alone ...

I have a lot of other projects, this one goes on the backburner.

Thanks everyone for your input!

Remon.

[rubber_jonnie]

I See (or IC ;) ... sorry, bad joke...)

I'm currently working up the courage to unsolder all the sockets... (512 pins to desolder) and switch to wiper sockets (as I am still hesitant to solder the 4MB directly in)

Maybe I'll try and find a spare mainboard (or a banged up full unit for parts, where the PCB is ok...) so I can leave this one alone ...

I have a lot of other projects, this one goes on the backburner.

Thanks everyone for your input!

Remon.

If you're going to remove the sockets, it might just be simpler to just remove the only those that need to be removed to solder the upgrade in. It's a lot of work to remove them all and by continuing to use sockets it may get you no further forward.

This upgrade, from what I can see, is designed to be soldered in, so getting bad results from using sockets is something you might expect if not installing as specified in the instructions.