ACA1233n-new: MapROM unstable

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,


    ACA1233n-new 55mhz

    Warranty code: GT4C/


    I've had the card for two days, and have tested it carefully. Only one thing is not working: ACAMapROM (as well as MapROM function in ACATool) causes immediate and severe instability. Typically the machine will crash immediately, but sometimes it lives long enough to invalidate the HDD, scramble the screen graphics, or bring up a Guru trap in Workbench (perhaps several seconds.) A simple reboot does not cure this, a power cycle is needed (I assume MapROM is reset-persistent.)


    I've tested in the following manner in an attempt to rule out any software interference:

    ACA1233n card factory default options only

    ACATool and ACA1233nLib from Wiki (latest, not beta or alpha versions)

    Cold boot with no startup-sequence, then invoke "ACAMapROM i" or ACATool Maprom

    Tested both before and after setpatch

    Both with internal ROM and Kickstart images

    MMUlibs are installed to the OS, but are not loaded in my testing (disabled startup-sequence, cold boot, no use of CPU FASTROM, MuFASTROM or any other MMU-activating commands.)

    Have tested with two different PSU (see below)

    Have tested by pulling card back on the connector slightly a few times

    RAM tests so far in MBRTest-2 have passed every cycle (tested with MMU disabled, of course)


    A1200 NTSC US

    Motherboard rev. 1D.1

    E123C and E125C were initially present, but have been removed with no change in behavior.

    E121C and E122C were unpopulated from the factory.

    Power Supply: Have tested with A500 4.5A brick, and with AmigaStore Spain brick.

    HDD: 2.5" IDE CF card adapter on very short cable. Have also tested with an Elbox EIDE'99 buffered adapter which works excellent in conjunction with IDE Speeder.
    Kickstart 3.1.4 EPROMs installed (internal MapROM tested, and also tested with 3.0 A1200 kickrom image file)

    AmigaOS 3.1.4.1


    Finally.. using CPU FASTROM or MuFASTROM works perfectly. No issues mapping internal kickstart ROM to 32-bit RAM when using the MMU to do so, and dramatic performance improvement in opening folders confirms this is working.


    I'd like to add that aside from this, the card has been functioning without any other flaw. Not one single guru or any observed issue outside of testing with MapROM. Sysinfo reports more than 11000 stones, and data burst mode is confirmed to be stable. IDE speeder at maximum setting works excellent if used in conjunction with buffered interface. No anomalies noted with TGTest in AIBB. WHDLoad compatibility has been perfect, even where my previous M1230XA accelerator failed. I get nearly 5MB/sec IDE performance. It's a remarkable achievement. With such stable behavior, I would not suspect a hardware fault.

  • We have two more support cases here in the forum, and already managed to reproduce the error on one of our machines. Since it only happens with a small part of A1200 computers, and the problem only occurs if the card runs at full 55MHz, but not at lower speeds, I Have made a CPLD update that slows the card down in the reset-phase. This fixes it for our test-setup, but it has not been verified in the field yet (first card is yet to arrive back here for the update).


    I believe what happens is that Gayle sees the reset and activates it's internal OVL function. Now the CPU fetches the reset vector from $0000.0004, jumps to the mapped Kickstart and executes the code so fast that Gayle doesn't manage to switch off it's OVL function fast enough. This results chipmem to not be available when it should. In simple words: THe ACA1233n-55 is just too fast with MapROM enabled. Please note that this is just my current hypothesis; it's not been verified yet.


    Finally.. using CPU FASTROM or MuFASTROM works perfectly.

    Are these reset-resident? Because if they aren't, you don't get to the point where Gayle gets in the way, as yes, the MapROM function of the ACA1233n is reset-resident.

    and with AmigaStore Spain brick.

    Please retire that; it's no good for an Amiga. It uses a MeanWell chassis that has already been shown to not be suitable in other support threads here. Further, the wiring of this PSU is chosen to actually cause maximum voltage drop on the input coil of the A1200, causing more trouble, including latch-up of CPLDs in modern hardware. The original PSU is what you should use, and after all these years, all you need to do is to exchange the output electrolytic caps.


    and also tested with 3.0 A1200 kickrom image file

    Please note that the current ACAtool has a bug in Kickrom image file handling. We already have an internal fixed version, to be released either tomorrow, or early next week.


    I would not suspect a hardware fault.

    It kind-of is, as we need to have your card back here for the CPLD update. If you are willing to test this, please ship it back ASAP. If you prefer to wait until others have confirmed that this "slowdown in the first second"-approach fixes the behaviour, you can wait and watch this thread in the German part of this forum.

  • Hello Jens.


    Thank you for attending to this. I plan to send to card back ASAP.


    In reference to the linked thread on the German side, here is the stenciling on top of my Gayle:


    CSG

    391424-02

    AA-GAYLE-R5

    copyright CBM 1992
    4992 50023A


    Quote from Jens

    Please retire that; it's no good for an Amiga. It uses a MeanWell chassis that has already been shown to not be suitable in other support threads here. Further, the wiring of this PSU is chosen to actually cause maximum voltage drop on the input coil of the A1200, causing more trouble, including latch-up of CPLDs in modern hardware. The original PSU is what you should use, and after all these years, all you need to do is to exchange the output electrolytic caps.

    I'd like to ask about the PSU situation. I had previously retired the A500 PSU due to noise with a very large peak-to-peak. I was experiencing some chip memory errors in conjunction with an accelerator I was testing for the A500, and I can see wavery interference in the RGB video as well. So I scoped VCC to VDD at the 68000 socket and saw high p2p noise there. Here's what it looked like on the scope. Replacing the capacitors in the PSU made no difference. I tested it briefly with this situation, but I think it's not safe to use.


    The Amigastore PSU showed a much smaller p-2-p, and I am getting more than 4.9v at the CPU, so I started using that instead. The memory issues didn't occur with this PSU.


    I also have the parts to build a PicoPSU, which is an increasingly popular option. Do you happen to have any advice with regards to modern PSU replacements, in the event I can't resolve the A500 PSU noise? There appears to be a vacuum in the availability of high-quality PSUs for Amiga.


    Thank you,

  • I have put the card in the post to you today. It's been in use on a daily basis, and no new issues have been noted aside from the MapROM issue. I've removed the Amigastore PSU from service and replaced it with a Commodore A500 PSU bought online and then recapped.


    Thank you,

    Rob

  • Hi Jens,


    I also have an ACA1233n-55 ("new"), which is exhibiting similar problems. As I have modified the card by installing an FPU & oscillator, I'm wondering if you're still willing the install the CPLD update & extra voltage regulator? The problems were already present before installing the FPU.


    Thanks,


    Alain

  • Hi,


    I will of course do that modification for you, just like I did for other customers who installed an FPU on their own (see the German part of this forum). Just send the card, and we'll take care of the rest.

  • Hi iComp developers! I just received my new ACA1233n-55 (new) card yesterday. I was testing it until now. I found the following issue: 55MHz (which is 60.5MHz in case of mine according to SysInfo) combined with MapRom always causes crash. Constantly, almost immediately, software error, graphical glitches, guru meditation sometimes as well, ending up always with simply hangup freeze. OK I accepted this fact, MapRom is not compatible with the fastest speed. I tested thoroughly on 26MHz, works perfectly fine with MapRom (3.1 from my ROM chip), MapRom from file (3.1.4). If I activate the 55MHz after booting it into MapRom, instant freeze again. OK then I wanted to test it without MapRom, because I don't insist on 3.1.4. Complete 3.1 reinstall, the system is stable. But when I launch an application that changes the screen resolution (MUI, or I go to different resolution or color depth) the screen goes glitchy again. Sometimes colors disappear, or scrambled lines appear (can see the content, but very ugly). If I go back to 26MHz: everything is awesome. Screen mode changes handled perfectly (MUI, etc). I can go to interlaced modes with best results ever from my flicker fixer, very stable screen. But my question is, if this card is so unstable with 55 (or maybe 60.5) MHz, why is it advertised like it. I would appreciate your help.

  • But my question is, if this card is so unstable with 55 (or maybe 60.5) MHz, why is it advertised like it. I would appreciate your help.

    You always need to look at the complete system: Not just the accelerator, but also the A1200, the expansions in there and especially the power supply. Please list the full system, so we can dig deeper. Rest assured that each card is tested for stability before it gets it's warranty ID.

  • I have no expansions attached (RTC module on the card only from iComp). Power supply: Original Amiga 1200 25W, P-NO. 391029-03, Model: PSM-601-2 (5V 3A, 12V 0.5A, -12V 0.1A) The scan doubler, flicker fixer is an external one, so I would no count it as an expansion. My 1200 has a real HDD installed.

  • The scan doubler, flicker fixer is an external one, so I would no count it as an expansion. My 1200 has a real HDD installed.

    Both consume power, and most likely on the 5V rail. Especially a mechanical HDD may cause additional ripple.


    Are both the PSU and the main board re-capped?

  • Hi Jens! Thank you for your answer. I don't know the background story of this Amiga and the PSU, but I assume none of them were recapped. Probably this is the culprit. What I did now for testing: I replaced the phisical HDD to a CF IDE adaptor. The glitches do not happen upon screen mode changes of workbench, but the screen is still more flickery than it was on 26MHz. MUI and other apps which are switching screen modes run into glitches, but I appreciated that it happens after the a longer use. So after it was turned off for a while it is stable and gets less and less stable. Probably the old capacitors are doing this. With this CF to IDE solution I could not manage to map the rom from file though. It always give me red bootscreen. I copied over multiple times from trusted source. Strange that it was working with the real HDD.

  • I tried with the copy of the OS3.1 and 3.1.4. Both the same red results. Somehow it can't read from my CF to IDE adapter which has a CF to SD adapter in it. I did put back my trusty old HDD, it works perfectly copied from the same floppy image (I use Gotek) with OS3.1 and now I'm installing 3.1.4 again but with lower CPU speed 27MHz. No problems at all. I ordered a better PSU for it, because I don't want to recap this old PSU (I don't have the tools for it). I don't want to advertise here (you can delete this if you want), just asking for your opinion: the Amiga PSU Duo from C64PSU.com is the chosen one. I read a lot about it yesterday, and seams to have very good ripple handling, focused on 5V mostly.

  • I ordered a better PSU for it,

    The model you're mentioning uses a MeanWell chassis, which is not suitable for an Amiga. I recommend to cancel the order.


    Meanwell specifies 100mV ripple, but doesn't live up to this spec (I measures up to 120mV on a sample that a customer sent me). The original Commodore specification requires 50mV or less.

  • Hi Jens! It arrived and it is not a MeanWell inside but custom made. (I can see through the vent holes.) PSU works really well, stability highly increased. After a while it started to glitch again, but when I lifted the Amiga up letting more air in, it stabilized. I experimented more to add a passive heat sink with adhesive thermal tape. The results are great. The Amiga is fully stable now. I tested it with a game with intensive 3D rendering so it heated up quickly. Now with the heatsink on it does not overheat anymore and MUI screens, games are without graphical glitches. I am very happy about the card afterall. I think it is a great product, just needs some cooling on 60MHz. I hope this helps others with same issues.

  • (I can see through the vent holes.)

    No, you can't. The MeanWell type number is really small and in different places of the chassis, depending on the person who did the final test at the factory. Or maybe the sticker was even removed by the PSU vendor because they know that I'm recommending to avoid their gear.


    After a while it started to glitch again, but when I lifted the Amiga up letting more air in, it stabilized.

    A clear sign of improper voltage regulation, but just an overvoltage setting. It is correct that too-low voltage is causing CPUs to crash, but a too-high setting will increase heat buildup, and cause crashes again. That's what I believe is happening here. I know that the Polish vendor you've linked to is using the MeanWell chassis with adjustable 5V rail, and their pre-setting is fairly high in order to hide the cable drop and filter loss. Additional heat may be caused by excessige Ripple/Noise from the power supply.


    just needs some cooling on 60MHz.

    It should not. During development, I've let the card run with a piece of styrofoam on the CPU and other chips, so heat is barely dissipated. Only when I managed to adjust timings to a spot where this worked for hours, I was satisfied with it. However, I did this before I started in-depth investigations about power supplies. When I did those measurements, it turned out that the A600 PSU I was using in the setup was about the only PSU that came close to the original Commodore specs: Very good voltage regulation and low ripple. See all the measurement data in the Seminar info that I published here.

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