Ultimate 1Mb Upgrade

[flashjazzcat]

Good point. I will provide one with the next release. :)

[Jacques]

On 3/10/2021 at 1:02 AM, Stephen said:

Nice and tucked into the ribbon cable.  Question though - will this cause any issues if I were to plug in a newer model U1MB with the built in resistors?

 

Are you sure resistors connect to pin 7&8 (RD4&RD5) of the MMU? Looking at the picture I'm not so sure... 

https://web.archive.org/web/20170709155909/http://spiflash.org/node/18

 

Edited April 24, 2021 by Jacques

[+Eyvind Bernhardsen]

I've just found out that my (original Candle) Ultimate1MB is what has been causing SpartaDOS instability in my 1088XEL. The symptoms I'm seeing are that SpartaDOS occasionally crashes almost immediately on boot and is very unstable even if it makes it to the end of the autoexec without hanging, but on other boots it would be perfectly capable of running stably for hours. Warning: wall of text imminent. Don't worry, there'll be a TL;DR at the end.

 

Since I've been using that machine mainly for playing games (which never crash) from the SIDE loader, fixing it hasn't been a high priority. When I finally did run XRAM on it, I found out that it was only showing 768kB of extra memory. Odd. I tried to jiggle a few things, reflowed all the dodgy-looking solder joints on the 1088XEL, but nothing helped. I managed to get XRAM to show 1024kB of memory once, but then almost all of the formerly missing banks had errors instead:

 

 

This made me curious enough to correlate the bank numbers, and I found out that all of the banks where bit 1 is high and bit 6 is low were either missing or failing (possibly coincidentally, that's the bit combination that enables the XEGS game ROM).

 

Now, as far as I know it would be very hard to cause this error by bodging a solder job. One specific combination of two specific bits written to a specific memory location causing memory to be mismapped somehow, with the system otherwise apparently stable... that seems more like a software problem. Or a firmware problem, possibly on the U1MB. It does seem to explain why SDX was sometimes unstable though: when it decided to use the full 1024kB, the failing banks would corrupt something important.

 

Fortunately, I had a way of testing if the U1MB was the problem: I have another U1MB (Lotharek version) in my XEGS. Side note: I used to have the Candle U1MB in that machine, and it had no problems with memory then. I put the Lotharek U1MB in the 1088XEL, ran XRAM, and got 1024 functioning kB. Yay! - except that, for slightly complicated reasons, I'd prefer to use the Candle U1MB in the XEL, and I haven't tested if it still works in the XEGS.

 

That brings me to the TL;DR: will flashing the CPLD repair my U1MB? This post suggests that it might: 

 

 

...but digging out the hardware and VMs (and skillset) I'll need for that operation from cold storage will take a while, so I thought I'd ask if anybody had seen this problem before, or has any better ideas about what could be wrong.

 

Thanks for reading to the end! Have a ?, you've earned it.

[+skr]

Do you have a Maxflash 8mbit Flash Cartridge? Not the programmer, just the cartridge. That will allow you to flash the CPLD for the U1MB with your Atari. No need for any other hardware.

[flashjazzcat]

You can't flash the CPLD with a MaxFlash cart; you can use it to flash the PLCC ROM chip. To reprogram the CPLD, you need a Xilinx platform cable and a copy of Lab Tools (free download).

 

As to whether the CPLD has actually gone bad: I'm not sure. I have indeed experienced one Incognito and one U1MB which went bad and recovered after reprogramming of the CPLD, and in one case, the symptom was the lower half of extended memory being mirrored in the upper banks in 1MB mode (and the net result was a UFLASH ROM update wiping out the main BIOS).

 

The only other possible cause - since we have ruled out the host machine - is a dry joint on the CPLD itself, but I would have expected this to flag problems long before now on a ten year old board.

 

[+Eyvind Bernhardsen]

Thanks for confirming that my analysis makes at least some sense. I’ll start digging through my boxes of hardware.

 

[+skr]

29 minutes ago, flashjazzcat said:

You can't flash the CPLD with a MaxFlash cart; you can use it to flash the PLCC ROM chip.

 

Oh, I mixed up the names again, I guess. Then of course, it doesn´t help. I just remembered, that I had a faulty U1MB where I could flash that removable Chip (PLCC it must be, right?) with the Maxflash.

[+Eyvind Bernhardsen]

On 4/27/2021 at 11:19 PM, Eyvind Bernhardsen said:

Thanks for confirming that my analysis makes at least some sense. I’ll start digging through my boxes of hardware.

Digging through my boxes I found both the programmer and Dropcheck's MMU adapter (needed to safely use the old U1MB in a real Atari), but reprogramming the CPLD didn't help and I get the same memory errors when I put that U1MB in the XEGS, so I've probably given one of the DRAM chips the old electric finger.

 

Surprisingly the 576KB mode does seem to work reliably, so I'll stick with that until the postal service has recovered enough from COVID that it's safe to order a new U1MB.

 

Thanks for the help, guys

[bwmott]

I'm installing an Ultimate 1MB upgrade from Lotharek in my Atari 600XL; however, I'm running into a potential issue with checking the system's RAM. The 600Xl has been previously upgraded to 64K RAM (very similar to the 600XL 64K upgrades discussed in various threads on AtariAge). I've successfully tested the 64K RAM using the built in OS memory test, Super SALT, and Shoestring's Atari 8bit RAM test (v1.6) without the U1MB installed.

 

Once I install the Ultimate 1MB everything seems to work fine and the built in OS memory test passes without any issues; however, if I try the Super SALT memory test it reports errors. I flashed Shoestring's Atari 8bit RAM test to one of the OS slots in the U1MB and used it to test the 64K RAM as well. It consistently reports that U11 is bad; however, replacing that chip with another one has no effect (nor swapping U11 with U12). I've also tried multiple power adapters to try to rule that out as an issue.

 

Overall the system seems to function fine with the Ultimate 1MB installed; however, I'm concerned that something isn't quite right since some RAM tests are failing. In looking at the 600XL schematics, it seems like most of the signals for U11 and U12 are shared (except for the data lines), so it isn't clear where the fault might be. Any thoughts on what might be causing the Super SALT RAM check (and Shoestring's Atari 8bit RAM) to report failures when the U1MB is installed?

 

Thanks for any pointers!

 

-- Brad

[_The Doctor__]

On 5/22/2021 at 10:12 AM, Eyvind Bernhardsen said:

Digging through my boxes I found both the programmer and Dropcheck's MMU adapter (needed to safely use the old U1MB in a real Atari), but reprogramming the CPLD didn't help and I get the same memory errors when I put that U1MB in the XEGS, so I've probably given one of the DRAM chips the old electric finger.

 

Surprisingly the 576KB mode does seem to work reliably, so I'll stick with that until the postal service has recovered enough from COVID that it's safe to order a new U1MB.

 

Thanks for the help, guys

the ribbon cables etc on these are often a source of weird problem as well, make the the ribbon and header connection are good, if it's and old old old board, make sure whatever fixes have been dones, 2 resistors or some such thing... if they weren't done... then you might be able to get it replaced or some other mercy worked out.

[bwmott]

17 hours ago, bwmott said:

I'm installing an Ultimate 1MB upgrade from Lotharek in my Atari 600XL; however, I'm running into a potential issue with checking the system's RAM. The 600Xl has been previously upgraded to 64K RAM (very similar to the 600XL 64K upgrades discussed in various threads on AtariAge). I've successfully tested the 64K RAM using the built in OS memory test, Super SALT, and Shoestring's Atari 8bit RAM test (v1.6) without the U1MB installed.

 

Once I install the Ultimate 1MB everything seems to work fine and the built in OS memory test passes without any issues; however, if I try the Super SALT memory test it reports errors. I flashed Shoestring's Atari 8bit RAM test to one of the OS slots in the U1MB and used it to test the 64K RAM as well. It consistently reports that U11 is bad; however, replacing that chip with another one has no effect (nor swapping U11 with U12). I've also tried multiple power adapters to try to rule that out as an issue.

 

Overall the system seems to function fine with the Ultimate 1MB installed; however, I'm concerned that something isn't quite right since some RAM tests are failing. In looking at the 600XL schematics, it seems like most of the signals for U11 and U12 are shared (except for the data lines), so it isn't clear where the fault might be. Any thoughts on what might be causing the Super SALT RAM check (and Shoestring's Atari 8bit RAM) to report failures when the U1MB is installed?

After some additional testing, I noticed that SpartaDOS X was enabled on the U1MB while running the RAM checks. After disabling SpartaDOS X all of the memory tests are completing successfully, so it looks my U1MB upgraded 600XL is fully functional. Looking forward to playing around with the upgraded machine some more.

 

-- Brad

[ugliestgolf]

how do i identify which jed is on the u1mb? mine says 0.0 in the menu.

 

flashed to v3 file posted in this thread still says 0.0

 

how do i know if its working?

 

[flashjazzcat]

All release JEDs for U1MB have the same revision info and should report the same ((1.0, but a bug in the BIOS fixed in the next release means they show 0.0).

[ugliestgolf]

thanks fjc

how can i tell im on v3?

just if the basic slot selection works?

e.g. if i can boot into altirra or assembler its v3

 

if it ignores and boots to Basic C its v2?

[flashjazzcat]

2 minutes ago, ugliestgolf said:

how can i tell im on v3?

just if the basic slot selection works?

Yeah: that'll confirm it.