BigRAM A2630 Instability

Caution: Non registered users only see threads and messages in the currently selected language, which is determined by their browser. Please create an account and log in to see all content by default. This is a limitation of the forum software.


Also users that are not logged in can not create new threads. This is a, unfortunately needed, counter measure against spam. Please create an account and log in to start new threads.

Don't Panic. Please wash hands.
  • Hello, I have just attempted to install my new BigRAM A2630 on my unmodified, 25MHz A2630 with 4 MB RAM. I am having stability issues with the board installed. I get the following:


    - First time boot after card being out of the system often boots fine to workbench but eventually locks up with power light flashing

    - Subsequent power cycles either get into a boot loop of flashing screens, flashing power light, or solid power light and nothing at all

    - Removing the BigRAM makes the machine boot perfectly fine again

    - Leaving the BigRAM in the machine with power off for a 15 minutes or more allows it to boot again with the same issues above


    My first guess was something to do with caps and indeed I have read in a previous thread that this is likely due to some PSU caps, specifically the 5V rail. Could someone please advise me exactly which caps I should look at replacing and a recommendation for modern parts to replace them with?


    Many thanks.

  • Most stability issues do indeed stem from power supply problems, but given that Commodore has used a number of different PSU chassis in A2000 computers, it's down to "open the unit and write down what values are required".


    However, there's a lot of things you can do before that: The data bus load on the 2630 can be reduced dramatically by removing the FPU, and (hoping that at least part of the ZIP ram is in sockets), removing 2MB of the local memory. Further, it's been reported that mask set F91C of the 68030 CPU is much more stable with this speed of memory.

  • Sadly the 4 meg of ZIP ram is soldered. I think Commodore was running out of parts when they made my A2630 - instead of resistor packs, there are individual discrete resistors sticking up from the RP pins! No sockets for me.


    I'll try pulling the FPU.


    Can you explain what "mask set F91C of the 68030 CPU" means?


    Finally, I'll take a look at the PSU. If there is an issue there, I'd like to resolve it in any case even if the FPU delete fixes things.


    Thanks.

  • Can you explain what "mask set F91C of the 68030 CPU" means?

    It's the die revision (silicon revision) of the CPU. We've had a few cases where this instability was only fixed by using the latest mask set of 68030 CPU.

  • OK, I pulled the 68882, reinstalled the BigRAM and ran Amiga Test Kit. The machine had been off overnight so it started up and ran fine, at least initially. It did the first pass of the RAM test but failed with the exception below. Soft rebooting caused the software failure so it's safe to say it's still unstable.


    The 68030 appears to be 59C74N, if I'm reading this correctly. Is it worth hunting down a F91C 68030 to replace this? Or should my next step be the PSU review?

  • Is it worth hunting down a F91C 68030 to replace this?

    Yes, it is. Depending on where you are on the planet, I could ship one - I just saw that one of the ACA1234 prototypes has the exact thing installed in a socket, and I don't need that prototype any more. Let me know where you are, maybe a simple chip-for-chip swap is already the solution for you.

  • Thanks for the very kind offer. I don't see a way to DM using this BBS - if there's a way, please DM me, otherwise if you'd like to contact me via email, let's see if this is a viable approach.

  • Two things:


    1. I've been searching for a replacement 68030 mask code 01F91C on eBay and most seem to be fake. If anyone has a pointer to a real one, I'd appreciate it. Question for the group, I'm pretty sure it's OK to buy a faster part, say a 33 or 50 MHz part and use it at 25MHz but you never know with Commodore things, will there be a problem? Also I assume that plastic versus ceramic, it doesn't matter right?


    2. Since I started testing the BigRAM, I suspect that I put a lot of stress on the PSU 5V rail. The power LED is flickering now, even with the BigRAM out. I put a voltmeter on it and it's fluctuating from 3.5 to 5V. I take it that this is a sign that the PSU is failing?

  • Plastic versus ceramic doesn't matter @25MHz. You can always clock faster units slower. Only problem is when you go below 6.25MHz - at least that's where 68020 CPUs are spec'd at the low end. Not a question here.


    I take it that this is a sign that the PSU is failing?

    Sounds like it, yes. I'm surprised that the machine even runs if you measure drops to 3.5V!

  • Further on the voltage situation, it might be localized. Putting a voltmeter on a drive power line shows a stable 5V. However, the power LED shows significant fluctuations and the mouse is no longer working. I guess I'm going to have to pull out the schematics and figure this one out. BigRAM project now on hold until I sort out this mess...

  • I have had to order a number of parts to work on these issues and they all arrived today. Progress so far:


    First thing I did was replace the 68030 CPU with a nice, used ceramic-packaged 01F91C which I bought from a reputable seller. This had an immediate effect of being able to boot the computer from the Buddha IDE / SD adaptor with the BigRAM board installed. I was also able to boot into AmigaTestKit v1.18 on an ADF image using my GoTek external drive. By the way, I'm now suspicious of the original chip in my A2630 - the FPU is ceramic, why would the CPU be plastic?


    After changing the CPU, running the memory test got me optimistic but... after about a minute, it crashed with illegal instruction exception much like above. I rebooted a few times and tried different memory tests - including just the chip RAM - and eventually got an exception each time.


    Next, I recapped the power supply. This wasn't too hard, the board is all through-hole and very easy to work on. Other than a sense of accomplishment of not breaking it, there was no improvement to the memory tests. Illegal instruction crashes continue.


    I'll note that although I can boot into the Workbench, eventually corruption starts creeping in and strange things happen. I tried running Basilisk II with MacOS because that's an app that can allocate a fair amount of memory, however eventually everything crashes.


    The blinking power light is still going on. I followed the schematics and changed a transistor that controls it with no change. I haven't done it yet, but next up is changing the 74F08 that is next up the chain. I don't think I should be concerned about this as it relates to the BigRAM as it's just the LED; the 5V rail is rock solid.


    So that's where I am. I now have a ITX power supply and I'm waiting on an adaptor cable for the Amiga to try out a completely different power supply.


    In the mean time, is there anything else that I should be testing or doing?


    And is there a chance that the BigRAM might actually be faulty here? How would I know?

  • The blinking power light is still going on. I followed the schematics and changed a transistor that controls it with no change. I haven't done it yet, but next up is changing the 74F08 that is next up the chain. I don't think I should be concerned about this as it relates to the BigRAM as it's just the LED; the 5V rail is rock solid.

    THe LED can be controlled with software, and when a "Guru meditation" happens, the LED blinks intentionally. There is no need to look for a fault. As for the "rock solid" - I'd like to see a ripple measurement before I accept that. However, with new caps on te PSU, I don't think you'll get much improvement from a different PSU. Instead, I'd double-check grounding: Is the wall socket a 3-prong with separate grounding? Does the A2630 have the back plate tightened to the A2000 case?


    In the mean time, is there anything else that I should be testing or doing?

    Yes: Please remove the FPU; it loads the data bus a lot, and therefore adds to the load that the 4MB on-board ram causes. This will ultimately cause data to arrive "late" at the 68030. Removing the FPU will speed things up.

    And is there a chance that the BigRAM might actually be faulty here? How would I know?

    RUnning a memory test is definitely the right thing, but you have to give it the environment that has a chance of working. With the top-notch-capacitive load on the data bus, there's a major showstopper in your system that will give a false result. Rest assured that we test each and every unit before it gets it's warranty ID.

  • Regarding the LED, it's not flickering due to software on startup or when the audio filter is changing mode - it's flickering all the time. That's why I suspect the U303 74F08 has failed because that's what controls it. I initially suspected the Q302 transistor but changing that didn't fix it. It's a big job for a minor issue so I need to find the time to tackle it. I've attached a scope reading of the LED 5V line (flickery!) and of the 5V rail of the PSU (rock solid!)


    Back to the issue at hand, I have removed the FPU and have been running the memory test all the way through (takes a long time with this much memory!) and so far, no crashes. The FPU indicates MC68882RC25A, 1C12R, QENW9212 on the copper part and MSIA GKNWQ-5 and D22873-101 on the ceramic part.


    What does this indicate, that my FPU is bad? Or do I need to replace it with a specific mask like I did for the CPU?

  • What does this indicate, that my FPU is bad? Or do I need to replace it with a specific mask like I did for the CPU?

    It's not bad, but it's loading the data bus quite heavily. It's an A-Type, so it may be an older mask. Unfortunately, there is no open list of FPU masks. Neither from Motorola, nor Freescale, nor NXP. So it's only educated guesses; if you can find an FPU that has a "C" at the end, it may indicate that it's the latest mask, which also indicates the smallest die-shrink and thus the smallest capacity on the data lines.

  • I've searched and I can't find much evidence that a MC68882RC25C or any 68882 with a C at the end exists. Is there something else I should be searching for? Perhaps a known-good mask code?

  • Like I wrote, there is no public information about the masks. Also, I wouldn't look for a 25MHz C version, but a higher-rated one, as the faster ones are more likely to have a higher die-shrink level.

  • The last reply was more than 365 days ago, this thread is most likely obsolete. It is recommended to create a new thread instead.