BTW, contrary to what was said in the video, the A2630 can run with the host 68000 removed, if you switch the card to "German/early A2000" mode, where one needs to remove the 68000 for all accelerators anyway. This is done by setting JP302 on the card. This is even mentioned in the manual of the A2630, Page 13, Appendix A. I just tested this on my A2000 + A2630, and it works. MIne is a late model with 2 layer PCB, just as Adrian's. (Also posted pictures on Twitter as proof.)
That would be a proper fix here. He used trash socket replacement, I would never use that here. If his issue was bad traces, it would not likely fix it's self with light pressure.
The /RESET pin on the 68000 (pin 18) might be worth checking out as it must be one of the few pins that, when disconnected, wouldn’t affect running once booted.
Yeah, that's the first thing that crossed my mind as well. The wedge not working when placed on the lower end of the CPU vs the middle also strongly suggests /RESET.
Yea, I'm going to guess it is either the trace or via between the CPU and RP101, reason being it's on the same side as the battery, and also in the area where pushing down on the board helped.
I was going to message that it might be a fine SHORT between pins, you wouldn't notice that with a continuity check and pressing down and distorting the Mainboard might be enough to separate it and allow it to boot. Once it's booted the short isn't enough to stop it running. Just a theory.
The only things I'd worry about with this "fix" is the speaker moving on it's own and shorting something. It might seem firmly wedged, but moving the system around and bumping it through normal use could nudge it ever so slightly each time. And the second is that mechanical stress from flexing like that could exacerbate issues in the future and just create even more work if you try to permanently repair the board. But the thing I'm most leery about is flat pins in round sockets. Try replacing that socket with a dual wipe one and see what happens.
As a wedge, you could use a wine cork, as it's slightly flexible and maintain pressure like a spring , and you can cut it to the length you need. As an added bonus, you will have some nice wine sent coming out of your Amiga when it heats up :D
When I was a kid we had a Spectrum ZX48K that would constantly crash. My dad fixed it one day, he jammed a wine cork between the case and the Z80. He told me it was popping out of the socket when it warmed up.
@Mr Guru I agree, I've worked in electronics for 25 years and I often hear/see people using turned pin sockets for DIL package ICs when all they needed to do was use a decent quality dual wipe socket. ICs never sit in a turned pin socket correctly causing contact issues, stray capacitance, risk of bending legs and an absolute pain in the backside if you ever want to remove the IC.
This would be a great amiga to use a replacement pcb on. It'd be time consuming to swap everything over, but then you wouldn't have to worry about flaky traces.
Also I think in the case of the Amiga, the boards aren't some marvel of engineering. All the secret sauce is in the custom chips. As long as you're running real chips, you're running a real Amiga.
Don't get me wrong, finding the exact reason why it's flaky will make for a great video, but I also love build videos. It's like adult Lego Mindstorms. There's something satisfying in seeing a board be built from the ground up and work. Also, like car restoration, I'm a firm believer in usability over originality. The Amiga wants to run and be played with, not sit on a shelf and look pretty.
@Mr Guru I guess you nailed it. And, it's August, it's hot, and Adrian doesn't like it too much to spend at least 10 hours to make a new pcb complete! Yup
I am a noob, and... I have bought an Amiga 2000, with battery damage, of course, and it black screens. Now, the other day, it showed WHITE once when I booted it. So this video IS actually useful. It indicates my socket, on the CPU or ROM is marginal. So Ill try opening it up again, and try pushing down on chips. If I get more life out of it, its kind of diagnostic as to WHERE the main issue is, eh!
I used to have similar problem on my A3000. Common battery leak around Paula chip destroyed bunch of stuff. The computer would not start until I pressed down the Paula chip. It would work fine for 30 min, but then it would freeze, and I had to press it again. I fixed it by replacing the socket and reflow solder on most pins around the damaged area. So far so good. I do still have a slight problem with VGA output only, the screen would flicker randomly until the A3000 warmed up.
I just refurbed an Amiga 1000. I can whole-heartedly agree with your sentiment. They really had a lot to learn about case design. Those things are a PAIN to work on.
@Mr Guru I've done a bit of 3D modeling and then had a service manufacture it from the files, so I get that sometimes you make really "safe" moves to ensure that it works, to avoid having to go through the cost and time of an iterative loop. That's probably what they were going through -- low budget, not much time. Still: The back panel assembly order is a chore. The little plastic clips that hold screws for the front panel are really awful. The screws that are meant to hold the memory expansion to the shielding around the edge connector on the front barely even thread. (At least mine do.) While we're on screws, the number you have to remove, for example, to get to the ribbon cable on the floppy drive is just insane. And speaking of the floppy drive -- the ground wire (which is already a bit of a kludge for a drive in a metal shell) secures to a bolt going through the motherboard and the RF shield around the edge connector, and terminates in a nut that is hidden underneath the bottom RF shield plate. You have to remove _the entire board_ from the case, and then remove the shield from the bottom of the board, to fasten that bolt once you loosen it at all. It's a hack-job of a case design. Which, it may well have been exactly that. People complain about the Power Mac 8500/9500 cases -- which, admittedly, are also a royal pain because you have to disassemble everything to get to anything. But man, that Amiga 1000 case takes the cake. It _looks_ like it'll be a walk in the park, but noooooo. :-D You're gonna be screwing and unscrewing parts of it over and over again on reassembly, until you memorize the order of pieces in the puzzle.
I run Amiga of Rochester and have fixed a ton of 2000s. I always sand the area around cpu. I'll be working on one soon. I love my a1200, best Amiga to me
I had a music synth which kept having bootup problems. I removed the PLCC chips, cleaned all the legs and socket pins and it never failed again. The A500 and beyond used them a lot, so that's always one thing to check. I had one of the PLCC custom chips burst once, the top cracked and let out magic smoke.
No shame in finding a workaround, even a mechanical one. I had the same issue a while back. Took some looking, but eventually I found a trace nowhere near the CPU that had a nice gouge taken out of it. One bodge later and it was good-as-new. Good luck!
The fact that it's on startup might indicate an issue with the reset/halt circuit, preventing the 68K from ever starting. Adding a separate circuit right at the CPU might make it work reliably by bypassing it.
Welcome to Adrian’s Jury-rigged Digital Basement! As creative as me using a paper clip to replace a broken clip attaching my accelerator cable to the carburetor on my 1986 Cutlass Salon in the late 90s. :)
There's one of those square PLCC chips in a socket further under the drive assembly. Try removing that, give it a blast of deoxit, reinstall it and see what happens.
Good suggestion. Those are susceptible to board flex since they squeeze the chip on all 4 sides. Failure to post should be traceable if you know what the boot sequence is, which roms are accessed, stack ram, cpu reset lines, etc. Should be able to narrow it down.
I was so glad to see you seat the power supply connector fully at @11:13. I noticed it was sticking up a bit... and I think you might have bumped it as well when you got your first successful boot while pressing the CPU.
And the metal part of the speaker will act as a heatsink. Winning! One other thing. I don't believe the device on the board is a flicker fixer but rather a DKB Mega-A-Chip which gives you an upgraded Agnus facilitating the 2MB of chip RAM you saw. The 2000 otherwise would only have 1MB of chip RAM.
I think the problem your having is due to using those strips of machine sockets because if they are not all level with each other and you put in a cpu with so many pins, I believe some pins don't necessarily make good contact. If you change it to a dual wipe socket then it has the ability to grab much better.
The theme of this video seems to be "My CPU don't jiggle jiggle, no sweat, I need to do a wiggle wiggle, for sure, it is really silly silly, you know, bending down and pushing, you really have to see it" But sometimes, if it works, it freakin' works, and even if it's silly, you have to do it. Old machinery is tricky, and requires... special care, even if it means just pushing down and bending.
9:45 - I did that to an AM386DX40 - used a dowel and jammed it under a long expansion card. The board was pretty warped for some reason. It worked for a few years. I still have the board, but it no longer POSTs, I/O reader shows some basic activity. IIRC I deduced that the controller IC doesn't work anymore, probably from broken traces in the intermediate layers.
I had a similar problem with my a600 and I found cracked resistor that pulls IPL signal to HI (if I remember) worth to check passive elements for cracks, but it's mostly impossible to find one without moving each one.
The issue of it working at boot when you flex the board, and continuing to run after boot, even if you remove the flex, strongly suggests the rest pin. Furthermore Don't discount the sockets... I built a HighDEF NES mod into a top loader NES a few years back. It was frequently intermittent, and the seller of the mod ended up swapping from a machine pin socket to a dual wipe style socket, cause apparently the machine pin sockets he had included would just sometimes be unreliable with some IC's pins.
Dice make a good implement for blocking like that. They're cheap, can be glued together and ground down, and if they get loose in the machine, they won't cause any shorts. The hack is a terrible one in the sense that it may accelerate the corrosion damage due to mechanical stress, but if it already wasn't working, you were going to have to bodge those lines anyhow. I really think the only way you'll ever be able to trust this machine is to install a new PCB, or at least a less damaged one. I don't even think this hurts its vintage value, the magic of the Amiga is in the custom chips, not the PCB.
If/when this fails again you've got a few useful clues as to where to look. Given the flex of the PCB it looks like the faulty trace is on the top (flex pulls lower traces apart and pushes top traces together). Also with it being at start up something to do with reset and/or initial ROM chip-select perhaps? Assuming you've got the time (yeah I know) a bit of multi-meter continuity testing _might_ find it fairly quikly... Totally understand the bodge fix tho. Sometime you just want to get things going.😁
A2000's were terrible for reliability I found myself. Had 3 of them over the years and they all had issues where they would crash or not start up due to movement or thermal cool down/ warm up cycles. The last 2000 I fully recapped everything including all accelerator board and PSU, cleaned thoroughly the connectors inside including Agnus and it still had issues where it would crash or not start up at all the next day. Drove me insane! The board was in pristine condition too with no battery leakage at all. I love my Amiga's to bits but never again will I get another 2000.
Another very good video! It looks like it will be some months before you re-visit this computer. When you do it may be worth using a device other than just a digital meter to bell out the traces. In industry we retained some moving coil meters for checking continuity because digital multimeters pass such ridiculously low amounts of current, so even a whisker of metal will measure as being OK. I think I've seen that you have a bench power supply which has a constant current function. It's probably a poor idea to try out Amps, but possibly passing 50mA or 100mA between test points would be OK and measuring voltage drop between points on each trace using your digital meter would highlight any dodgy joints or part-eaten buried traces. Only use very low voltages of course, possibly set the supply to 0.4V open circuit? Check what voltage TTL chips will stand if connected in reverse; less than a diode drop should be OK. Reversing the flow of current and comparing voltage drops for forward and reverse flow might highlight problem traces and/or trace/pin joints. You'd be looking for anomalies so a bit of experimentation and note taking would be useful. Perhaps practise on one of your boards that is already consigned to Silicon Heaven. Good luck anyway!
Great Suggestion to be honest and very thorough with the possible outcome. Using a Shunt with low resistance inline to pickup Voltage drop would also be possible, as any additional resistance due to bad traces will screw up ypur voltage readings.
I've never used an old speaker like that, but I've used a lot of other non-standard bit's and bob's in similar circumstances. Not all have worked, but sometimes the MacGyver spirit is necessary and sometimes ya just gotta do what ya gotta do.
Okay, quick suggestion that may be a solution in regards to corrosion: Dilectric grease. It's recommended to prevent oxidation/corrosion on terminals that are prone to conditions such as snow, rain, battery acid, etc. Might I suggest that on the repaired sockets before repopulating them?
If you are pushing the 68000 down into the socket, bending the board downward than I would think the problem would most likely be on the top of the board since the top would experience contraction, bringing connections together, while the bottom of the would expand. Just a thought
I've had similar issues repairing battery damaged PC boards. The issue is usually bad pads on SMD chips that make a connection when flexed/pushed on. Very frustrating because they test OK on continuity tests because the force of multimeter probing connects the trace's.
It will bend the board, but I think the only long-term fix is to replace the PCB, so anything he does in the meantime to squeeze some more use out of it is fine as long as the custom chips aren't damaged.
I had 386 DX/40 with a similar "mod" - the board was already warped, but I managed to get 3 years out of it with a wooden dowel stuck between an AWE32 (or other long card) and the CPU. It eventually died. I still have it in my pile of dead AT boards, but I fear whatever caused the board to warp took out the traces within the board itself.
I fixed in a similar way an apple iBook G3 with the known problem of cracked solder joints under the GPU: just a shim between the iBook undercover and the Gpu so that it is pushed to the pcb and it works like a charm, no more crash ever since.
I had a 21” Sylvania monitor I used with my PowerComputing Mac clone that had an intermittent issue like this, so I had a plastic crochet needle wedged in a vent hole!
If this is the Commodore A2630 (68030 @25mhz) accelerator card, the system will work without the 68000 on the motherboard. It's one of the few, if not the only accelerator card that can do this. Had problems with my A2000 aswell, kicked out the 68k on the mainboard, ran fine ever since.
Always remember then admitting that they repaired returned machines and then sold them as new and not refurbished. Just one of the many clues that Commodore were never going to last.
Maybe in-house, but even they'd know this was likely to come apart in shipping, and that the loose part could do further damage. Sending a machine to anyone like this has such a high probability of being DOA that I can't see them letting this leave.
I did something similar to that in my 486 PC back in the early 90s. It was a Packard Bell desktop that had 3 ISA slots on a riser card so they were parallel to the motherboard and the CPU socket was just in front of those slots. I had an AMD 5x86 CPU in a 5V to 3.3V adapter in the socket and a heatsink & fan on top of the CPU so the entire stack was quite tall. The sound card w/ CD-ROM interface was super long and was in the top ISA slot and cleared the CPU stack, but my USR 28.8 modem was also fairly long and touched the top of the metal fan shroud on the CPU stack. I wedged a small plastic spacer between the modem and the one side of the fan shroud to keep things from shorting out. It wasn't a pretty solution but it did work! Luckily the SCSI card I had in the bottom ISA slot for my ZIP drive was really short so there were no clearance issues there.
I used this connector with round pins for a ROM on my Amiga 3000 and it was causing the black screen problem. ROM chips I used have had flat pins, like the CPU have.
As a guess I would say pushing down stretches the bottom layer of the PCB and compresses the top layer, so I would be looking at the top layer for a break. It could also be a break in a via, ie, it needs more solder in the THT holes so that you can see solder on top and bottom. You should also remove every component in the affected areas because you never know what's under them if you don't.
@Mr Guru I was thinking about that, ie, machined headers use a wedging press fit that wears with multiple insertions, ie, it gets looser. Wiper is spring loaded so has constant pressure, but because it's not a press fit it can suffer capillary moisture attraction. Dual wiper is better because if one side corrodes it still has the other, and I've seen on EEVblog multi pin press fit connectors with no solder that are very reliable. Me personally I like wiper, but I guess both have pros and cons.
@Mr Guru I just thought of another problem with machined pin headers. If you have a dimensionally long IC, like the one with the problem, and your PCB is bent, eg, a dip in the centre, the end pins will have the correct press but the middle pins won't. I suppose it's rare, and more likely you soldered the headers in not flat, but I guess it can happen.
This reminds me of the time when I was in high school and I '"fixed" the loose connections on a computer lab PC's VGA port by shoving a folded-up piece of paper in-between where the monitor cable was plugging in and whatever was plugged in next to it. (People kept bumping the back of the PC and that cracked some connections on the VGA port.)
I had a C64 with no intermittent audio out of the RF connection. I had to shim the RF RCA jack to make the connection solid. I later on got the SAMS book for the C64 and made a DIN cable to send the audio to my stereo receiver.
I had a bad CPU socket (corrosion) and my A2500 was a no boot. However with the same Commodore CPU card you have, it would boot. Did not seem to mind the CPU being dead.
Not craptastic, Adrian! There is no shame in using the "wedge as repair" technique. In 1987, I shoved a piece of foam into my HP 11c scientific calculator. A couple of leads on the surface-mounted main chip had come loose and I only knew through-hole soldering. It's still working!
Love your A2000 coverage. Your repair, "Speakers volumes". Lol 😁👀👏 Always found the A2091 the best scsi card. Too many issues with all the others. Oktagon was a good runner up.
The ROM of the A2091 is pretty much garbage. :) It needs the ROM 7.0 version to work with 040 accelerators at all, and also has various other limits. Also fails to use DMA with 32bit Fast RAM. On Aminet there are various patches for the ROM to make it work better.
Dear Adrian, maybe I'm very wrong but, please check all vias in that area, I remember CuriousMarc have the same problem with an IBM and after lots of test, even they develop a special bios to check things, they found broken via just inside. They have to do the same as u did to make it work. (Sorry my poor english, cheers from Argentina!)
Here in the normal world, you can just buy the metric screws by the kg at any hardware store. However, the PC screws have to come out of a PC. I've scrapped many PC cases just to get some spare screws.
That solution is not so horrible for a temporary fix at the household level. 👍 In the context of this channel it brings a very funny note of humor. 😊 And I'm sure in real life many of us have encountered solutions as crappy or crappier than that, 🥶 (even at the enterprise technical service level). 👋
It seems that as the circuit is functioning, heat expands it, and makes a better connection. Testing the continuity during a cold stress (compressed air), should show the defect.
As to the SCSI drives, recheck the transfer speed mask and that the correct drive metrics are selected (use a more recent OS, HDToolBox, and FastFileSystem).
I've got a front loader Nintendo, that I HDMI converter, that just don't work when I've screwed it back together. Finally I figured it that it was caused the motherboard being slightly bent on the back where the cartridge connector goes and that caused the connector to get slightly bent. Using the console like that works just fine, but when I screw it back together, the cartridge holder pushes on the connector in such a way that some of the pins get disconnected, causing issues related to those pins.
I'm not familiar enough with those SCSI cards but the random failures suggest a lack of termination power and/or a lack of termination resistors at the ends of the SCSI chain (or an inadvertent terminator in the chain but not on the end). IIRC termination is set by software on the SCSI2SD.
Short is most likely under the CPU slot not the 68000. You seem to be putting presure on the CPU expansion slot rather than the 68000. Cleaning debris has fallen down on the socket pins. Also need the v7 roms for scsi card, same as the A590.
LOL at the very start of the video for a second i thought that was the Amiga display and i went to go look up when the 2000HD was released and just about had a heart attack before zooming in seeing what was actually happening there :P
Maybe the plated through holes for the socket or some Vias that are underneath of the socket are not connecting from top to the bottom of the motherboard.
This may be a small/petty thing, but thank you for not saying “yep, it’s a 2nd channel video” this time. We know it’s a 2nd channel video, we clicked on it lol.
Pretty sure you've got a dry joint somewhere else on the motherboard - when you press down on the CPU the board is flexing a lot and probably causing the dry joint to move and make a connection, time to break out the magnifying glass or microscope and search around for that telltale crack in the solder on a joint.
I'd check those resistor packs under the drive by the CPU...perhaps they are not properly connected and they are not pulling up/down certain signals that might need be. Also check any other surrounding componenets.
Your fix reminds me of putting a piece of cardboard under the laser on my Playstation back in the day to keep help it read disks better. Janky, but functional. Talking of having too much fast RAM, my A1200 will crash on booting if I plug in my CD-ROM drive with more than 4mb of fast RAM installed.
Queue the torches and pitchforks! 😜 At least you admit this is a terrible "fix". But a full video tracking the problem and actually fixing it would be great.
I have no idea if any are available, but maybe you might want to consider getting another motherboard and just trashing the one that's currently in the computer.
In terms of reliable vs. unreliable Amigas, there is of course the A500 Plus model too which suffers onboard clock battery troubles, as well as having owned one (which died when Amigas were near worthless and cheap to pick up), I've seen myself a few externally nice ones on ebay from optimistic sellers with "amazing condition, perfect example!!!" style descriptions along with a high asking price, and then seeing the green death around the expansion bay pins which they remain silent about, yeah, those batteries suck, but those sellers suck too...
You'd know better than I would, but my hunch is that it's not the board like you're confident it is. I think you need a proper socket for it instead of those pin headers. Whatever is going on seems to be affecting the reset line.
How many winced every time Adrian said “bridge card” instead of “bridge board”? 😉 Kidding aside, I concur with others that you are likely dealing with corrosion that has seeped into the pc board itself. Time to look around for an aftermarket motherboard to which you can transfer all the custom ICs.
I wouldn't fancy a magnet potentially coming loose and possibly interfering with magnetic storage, maybe put blob of hot glue at the top point and the chassis for piece of mind?
Nothing wrong with a bodge, time constraints and all! That said, the first thing I'd do would probably be to use hot air to just reflow all the joints on the back side en masse just in case it's a bad joint. Might not necessarily be next to the CPU socket either just a quirk of the particular way the board gets flexed when you press there? I don't think it would take too long, thin layer of flux on it all, wide nozzle, fume extractor and off you go. But it sure could be a broken trace too. Another commenter mentioned DIP chips in turned pin sockets being a bad idea, not sure if you've heard of that causing issues before? Anyway great video as always! This machine reminds me of my childhood Atari ST which had to be vigorously twisted diagonally to make it start up. It was probably just a chip needing to be reseated but I didn't know anything back then :D
@Adrian's Digital Basement ][ This seems very much like a cold joint problem to me. Perhaps you just need to go over all the connections around the CPU with a soldering iron to reflow and solidify them. I would do this before removing the solder mask like you were suggested. EDIT: Don't worry about the AC noise, we can barely heard it. Not a problem at all. EDIT2: The SCSI to SD thing could be a voltage dip problem. Those adapters require a fair amount of power to work right. You may wish to check PSU power output and voltages connections. That system might be suffering a "brown-out" situation.
It sounds like it may be time to run the board through the dishwasher if it is corrosion. I'm sure that would get deep into wherever the corrosion is and neutralize it.
![Adrian's Digital Basement ][](http://yt3.ggpht.com/ytc/AIdro_k7CrQRe71qN6c0Ua7A1P_I8KkxjqLa70u0zYeCvdvnKw=s48-c-k-c0x00ffffff-no-rj)
BTW, contrary to what was said in the video, the A2630 can run with the host 68000 removed, if you switch the card to "German/early A2000" mode, where one needs to remove the 68000 for all accelerators anyway. This is done by setting JP302 on the card. This is even mentioned in the manual of the A2630, Page 13, Appendix A. I just tested this on my A2000 + A2630, and it works. MIne is a late model with 2 layer PCB, just as Adrian's. (Also posted pictures on Twitter as proof.)
That would be a proper fix here. He used trash socket replacement, I would never use that here. If his issue was bad traces, it would not likely fix it's self with light pressure.
@Mr Guru But the problem was nothing to do with the CPU socket.
@Mr Guru And you know the problem was the socket? How? Magic? Please stop with the condescending remarks.
The /RESET pin on the 68000 (pin 18) might be worth checking out as it must be one of the few pins that, when disconnected, wouldn’t affect running once booted.
It is the reset line on the cpu, a friend had the same problem and spent 2 weeks on drags.
Yeah, that's the first thing that crossed my mind as well. The wedge not working when placed on the lower end of the CPU vs the middle also strongly suggests /RESET.
Yea, I'm going to guess it is either the trace or via between the CPU and RP101, reason being it's on the same side as the battery, and also in the area where pushing down on the board helped.
I came to make the same comment so +1
I was going to message that it might be a fine SHORT between pins, you wouldn't notice that with a continuity check and pressing down and distorting the Mainboard might be enough to separate it and allow it to boot. Once it's booted the short isn't enough to stop it running. Just a theory.
The only things I'd worry about with this "fix" is the speaker moving on it's own and shorting something. It might seem firmly wedged, but moving the system around and bumping it through normal use could nudge it ever so slightly each time. And the second is that mechanical stress from flexing like that could exacerbate issues in the future and just create even more work if you try to permanently repair the board.
But the thing I'm most leery about is flat pins in round sockets. Try replacing that socket with a dual wipe one and see what happens.
As a wedge, you could use a wine cork, as it's slightly flexible and maintain pressure like a spring , and you can cut it to the length you need. As an added bonus, you will have some nice wine sent coming out of your Amiga when it heats up :D
When I was a kid we had a Spectrum ZX48K that would constantly crash. My dad fixed it one day, he jammed a wine cork between the case and the Z80. He told me it was popping out of the socket when it warmed up.
@Mr Guru I agree, I've worked in electronics for 25 years and I often hear/see people using turned pin sockets for DIL package ICs when all they needed to do was use a decent quality dual wipe socket. ICs never sit in a turned pin socket correctly causing contact issues, stray capacitance, risk of bending legs and an absolute pain in the backside if you ever want to remove the IC.
This is what I thought but slight splay of the legs would prove that. Also the board flexes a lot so it could be a fault nowhere near the cpu.
This would be a great amiga to use a replacement pcb on. It'd be time consuming to swap everything over, but then you wouldn't have to worry about flaky traces.
Normally I'd call bunk, but yeah, this sounds like the right time for a new pcb.
I thought the same thing.
Also I think in the case of the Amiga, the boards aren't some marvel of engineering. All the secret sauce is in the custom chips. As long as you're running real chips, you're running a real Amiga.
Don't get me wrong, finding the exact reason why it's flaky will make for a great video, but I also love build videos. It's like adult Lego Mindstorms. There's something satisfying in seeing a board be built from the ground up and work. Also, like car restoration, I'm a firm believer in usability over originality. The Amiga wants to run and be played with, not sit on a shelf and look pretty.
@Mr Guru I guess you nailed it. And, it's August, it's hot, and Adrian doesn't like it too much to spend at least 10 hours to make a new pcb complete! Yup
I am a noob, and... I have bought an Amiga 2000, with battery damage, of course, and it black screens. Now, the other day, it showed WHITE once when I booted it. So this video IS actually useful. It indicates my socket, on the CPU or ROM is marginal. So Ill try opening it up again, and try pushing down on chips. If I get more life out of it, its kind of diagnostic as to WHERE the main issue is, eh!
A similar problem on a similar machine made me buy a milliohm meter. At the lowest setting you can watch the resistance change from heat on a trace.
@Mr Guru Wow are you going to put this rubbish on every single reply?
All of the A2000 design originated in the Braunschweig office, so not surprising that every screw is metric.
I used to have similar problem on my A3000. Common battery leak around Paula chip destroyed bunch of stuff. The computer would not start until I pressed down the Paula chip. It would work fine for 30 min, but then it would freeze, and I had to press it again. I fixed it by replacing the socket and reflow solder on most pins around the damaged area. So far so good. I do still have a slight problem with VGA output only, the screen would flicker randomly until the A3000 warmed up.
I just refurbed an Amiga 1000. I can whole-heartedly agree with your sentiment. They really had a lot to learn about case design. Those things are a PAIN to work on.
@Mr Guru I've done a bit of 3D modeling and then had a service manufacture it from the files, so I get that sometimes you make really "safe" moves to ensure that it works, to avoid having to go through the cost and time of an iterative loop. That's probably what they were going through -- low budget, not much time.
Still:
The back panel assembly order is a chore.
The little plastic clips that hold screws for the front panel are really awful.
The screws that are meant to hold the memory expansion to the shielding around the edge connector on the front barely even thread. (At least mine do.)
While we're on screws, the number you have to remove, for example, to get to the ribbon cable on the floppy drive is just insane.
And speaking of the floppy drive -- the ground wire (which is already a bit of a kludge for a drive in a metal shell) secures to a bolt going through the motherboard and the RF shield around the edge connector, and terminates in a nut that is hidden underneath the bottom RF shield plate. You have to remove _the entire board_ from the case, and then remove the shield from the bottom of the board, to fasten that bolt once you loosen it at all.
It's a hack-job of a case design. Which, it may well have been exactly that.
People complain about the Power Mac 8500/9500 cases -- which, admittedly, are also a royal pain because you have to disassemble everything to get to anything. But man, that Amiga 1000 case takes the cake. It _looks_ like it'll be a walk in the park, but noooooo. :-D You're gonna be screwing and unscrewing parts of it over and over again on reassembly, until you memorize the order of pieces in the puzzle.
I run Amiga of Rochester and have fixed a ton of 2000s. I always sand the area around cpu. I'll be working on one soon. I love my a1200, best Amiga to me
I had a music synth which kept having bootup problems. I removed the PLCC chips, cleaned all the legs and socket pins and it never failed again. The A500 and beyond used them a lot, so that's always one thing to check. I had one of the PLCC custom chips burst once, the top cracked and let out magic smoke.
No shame in finding a workaround, even a mechanical one. I had the same issue a while back. Took some looking, but eventually I found a trace nowhere near the CPU that had a nice gouge taken out of it. One bodge later and it was good-as-new. Good luck!
The fact that it's on startup might indicate an issue with the reset/halt circuit, preventing the 68K from ever starting. Adding a separate circuit right at the CPU might make it work reliably by bypassing it.
I was suspecting the power on/reset circuitry as well. Only used once, usually.
Welcome to Adrian’s Jury-rigged Digital Basement!
As creative as me using a paper clip to replace a broken clip attaching my accelerator cable to the carburetor on my 1986 Cutlass Salon in the late 90s. :)
I see how it is! Doing Unspeakerable things to an Amiga 2000 🤓
There's one of those square PLCC chips in a socket further under the drive assembly. Try removing that, give it a blast of deoxit, reinstall it and see what happens.
Good suggestion. Those are susceptible to board flex since they squeeze the chip on all 4 sides. Failure to post should be traceable if you know what the boot sequence is, which roms are accessed, stack ram, cpu reset lines, etc. Should be able to narrow it down.
I was so glad to see you seat the power supply connector fully at @11:13. I noticed it was sticking up a bit... and I think you might have bumped it as well when you got your first successful boot while pressing the CPU.
And the metal part of the speaker will act as a heatsink. Winning!
One other thing. I don't believe the device on the board is a flicker fixer but rather a DKB Mega-A-Chip which gives you an upgraded Agnus facilitating the 2MB of chip RAM you saw. The 2000 otherwise would only have 1MB of chip RAM.
Love this channel. I just came across a few old machines including an IBM ps/2 m30 and a IBM ps/2 M77. All have issues, but I’m excited to dive in.
I have an A2500 that's still plugging along. The battery leaked a few years ago, but I was lucky enough to catch it before it did any real damage.
I think the problem your having is due to using those strips of machine sockets because if they are not all level with each other and you put in a cpu with so many pins, I believe some pins don't necessarily make good contact. If you change it to a dual wipe socket then it has the ability to grab much better.
The theme of this video seems to be
"My CPU don't jiggle jiggle, no sweat, I need to do a wiggle wiggle, for sure, it is really silly silly, you know, bending down and pushing, you really have to see it"
But sometimes, if it works, it freakin' works, and even if it's silly, you have to do it. Old machinery is tricky, and requires... special care, even if it means just pushing down and bending.
9:45 - I did that to an AM386DX40 - used a dowel and jammed it under a long expansion card. The board was pretty warped for some reason. It worked for a few years. I still have the board, but it no longer POSTs, I/O reader shows some basic activity. IIRC I deduced that the controller IC doesn't work anymore, probably from broken traces in the intermediate layers.
I had a similar problem with my a600 and I found cracked resistor that pulls IPL signal to HI (if I remember) worth to check passive elements for cracks, but it's mostly impossible to find one without moving each one.
The issue of it working at boot when you flex the board, and continuing to run after boot, even if you remove the flex, strongly suggests the rest pin. Furthermore Don't discount the sockets... I built a HighDEF NES mod into a top loader NES a few years back. It was frequently intermittent, and the seller of the mod ended up swapping from a machine pin socket to a dual wipe style socket, cause apparently the machine pin sockets he had included would just sometimes be unreliable with some IC's pins.
I've had a problem like this before, and I just wedged a block of wood. It worked for many more years.
Dice make a good implement for blocking like that. They're cheap, can be glued together and ground down, and if they get loose in the machine, they won't cause any shorts.
The hack is a terrible one in the sense that it may accelerate the corrosion damage due to mechanical stress, but if it already wasn't working, you were going to have to bodge those lines anyhow. I really think the only way you'll ever be able to trust this machine is to install a new PCB, or at least a less damaged one. I don't even think this hurts its vintage value, the magic of the Amiga is in the custom chips, not the PCB.
Maybe use a heatsink. At least at that point it will help keep the CPU cool at the same time.
I just wanted to write, Adrian forgot to put thermal grease under the speaker cone 😄
this probably would work also, but wouldn't need to cool it with the accelerator card being the used cpu
If/when this fails again you've got a few useful clues as to where to look. Given the flex of the PCB it looks like the faulty trace is on the top (flex pulls lower traces apart and pushes top traces together). Also with it being at start up something to do with reset and/or initial ROM chip-select perhaps? Assuming you've got the time (yeah I know) a bit of multi-meter continuity testing _might_ find it fairly quikly... Totally understand the bodge fix tho. Sometime you just want to get things going.😁
A2000's were terrible for reliability I found myself. Had 3 of them over the years and they all had issues where they would crash or not start up due to movement or thermal cool down/ warm up cycles.
The last 2000 I fully recapped everything including all accelerator board and PSU, cleaned thoroughly the connectors inside including Agnus and it still had issues where it would crash or not start up at all the next day. Drove me insane! The board was in pristine condition too with no battery leakage at all.
I love my Amiga's to bits but never again will I get another 2000.
Even though you didn't use any duct tape, Red Green would be proud of you for this janky fauxpair.
not the "right" fix but it was the right fix at the moment.
Another very good video! It looks like it will be some months before you re-visit this computer. When you do it may be worth using a device other than just a digital meter to bell out the traces. In industry we retained some moving coil meters for checking continuity because digital multimeters pass such ridiculously low amounts of current, so even a whisker of metal will measure as being OK. I think I've seen that you have a bench power supply which has a constant current function. It's probably a poor idea to try out Amps, but possibly passing 50mA or 100mA between test points would be OK and measuring voltage drop between points on each trace using your digital meter would highlight any dodgy joints or part-eaten buried traces. Only use very low voltages of course, possibly set the supply to 0.4V open circuit? Check what voltage TTL chips will stand if connected in reverse; less than a diode drop should be OK. Reversing the flow of current and comparing voltage drops for forward and reverse flow might highlight problem traces and/or trace/pin joints. You'd be looking for anomalies so a bit of experimentation and note taking would be useful. Perhaps practise on one of your boards that is already consigned to Silicon Heaven. Good luck anyway!
His issue is the dodgy socket, not the board
Great Suggestion to be honest and very thorough with the possible outcome. Using a Shunt with low resistance inline to pickup Voltage drop would also be possible, as any additional resistance due to bad traces will screw up ypur voltage readings.
I've never used an old speaker like that, but I've used a lot of other non-standard bit's and bob's in similar circumstances. Not all have worked, but sometimes the MacGyver spirit is necessary and sometimes ya just gotta do what ya gotta do.
Okay, quick suggestion that may be a solution in regards to corrosion: Dilectric grease. It's recommended to prevent oxidation/corrosion on terminals that are prone to conditions such as snow, rain, battery acid, etc. Might I suggest that on the repaired sockets before repopulating them?
If you are pushing the 68000 down into the socket, bending the board downward than I would think the problem would most likely be on the top of the board since the top would experience contraction, bringing connections together, while the bottom of the would expand.
Just a thought
I've had similar issues repairing battery damaged PC boards. The issue is usually bad pads on SMD chips that make a connection when flexed/pushed on. Very frustrating because they test OK on continuity tests because the force of multimeter probing connects the trace's.
Hope it will last a while, but I fear it will just bend the board after a long time, as the material in the board negates the tension.
It will bend the board, but I think the only long-term fix is to replace the PCB, so anything he does in the meantime to squeeze some more use out of it is fine as long as the custom chips aren't damaged.
I had 386 DX/40 with a similar "mod" - the board was already warped, but I managed to get 3 years out of it with a wooden dowel stuck between an AWE32 (or other long card) and the CPU. It eventually died. I still have it in my pile of dead AT boards, but I fear whatever caused the board to warp took out the traces within the board itself.
I fixed in a similar way an apple iBook G3 with the known problem of cracked solder joints under the GPU: just a shim between the iBook undercover and the Gpu so that it is pushed to the pcb and it works like a charm, no more crash ever since.
I had a 21” Sylvania monitor I used with my PowerComputing Mac clone that had an intermittent issue like this, so I had a plastic crochet needle wedged in a vent hole!
Thar title somehow tricked my mind into thinking that Adrian was going to jam an actual Staples Easy Button in there.
You have a brand new a2000 pcb hanging on your wall (although that is a rev 6.2), you could rebuild it using new/used parts. :)
I have a 3d printer which was only showing garbage on its display. The fix was "pressure" at the right position on the main PCB. 😀
If this is the Commodore A2630 (68030 @25mhz) accelerator card, the system will work without the 68000 on the motherboard. It's one of the few, if not the only accelerator card that can do this.
Had problems with my A2000 aswell, kicked out the 68k on the mainboard, ran fine ever since.
The Bridge Card lives another day!
Funnily enough, that's exactly the fix Commodore would have used back in the day.
Always remember then admitting that they repaired returned machines and then sold them as new and not refurbished. Just one of the many clues that Commodore were never going to last.
Maybe in-house, but even they'd know this was likely to come apart in shipping, and that the loose part could do further damage. Sending a machine to anyone like this has such a high probability of being DOA that I can't see them letting this leave.
Are you using the the version 7 ROMS on the a2091a? I can't quite remember if the version is 6.something or 7.something
I did something similar to that in my 486 PC back in the early 90s. It was a Packard Bell desktop that had 3 ISA slots on a riser card so they were parallel to the motherboard and the CPU socket was just in front of those slots. I had an AMD 5x86 CPU in a 5V to 3.3V adapter in the socket and a heatsink & fan on top of the CPU so the entire stack was quite tall. The sound card w/ CD-ROM interface was super long and was in the top ISA slot and cleared the CPU stack, but my USR 28.8 modem was also fairly long and touched the top of the metal fan shroud on the CPU stack. I wedged a small plastic spacer between the modem and the one side of the fan shroud to keep things from shorting out. It wasn't a pretty solution but it did work! Luckily the SCSI card I had in the bottom ISA slot for my ZIP drive was really short so there were no clearance issues there.
Awesome fix! Did you „professionally“ hot glue the speaker so that it does not short anything … or moves?
Nothing crappy about a fix that works reliably and saves you hours of time...
I used this connector with round pins for a ROM on my Amiga 3000 and it was causing the black screen problem. ROM chips I used have had flat pins, like the CPU have.
As a guess I would say pushing down stretches the bottom layer of the PCB and compresses the top layer, so I would be looking at the top layer for a break.
It could also be a break in a via, ie, it needs more solder in the THT holes so that you can see solder on top and bottom.
You should also remove every component in the affected areas because you never know what's under them if you don't.
@Mr Guru I was thinking about that, ie, machined headers use a wedging press fit that wears with multiple insertions, ie, it gets looser.
Wiper is spring loaded so has constant pressure, but because it's not a press fit it can suffer capillary moisture attraction.
Dual wiper is better because if one side corrodes it still has the other, and I've seen on EEVblog multi pin press fit connectors with no solder that are very reliable.
Me personally I like wiper, but I guess both have pros and cons.
@Mr Guru I just thought of another problem with machined pin headers. If you have a dimensionally long IC, like the one with the problem, and your PCB is bent, eg, a dip in the centre, the end pins will have the correct press but the middle pins won't.
I suppose it's rare, and more likely you soldered the headers in not flat, but I guess it can happen.
This reminds me of the time when I was in high school and I '"fixed" the loose connections on a computer lab PC's VGA port by shoving a folded-up piece of paper in-between where the monitor cable was plugging in and whatever was plugged in next to it. (People kept bumping the back of the PC and that cracked some connections on the VGA port.)
A tall stack of sticky notes would be a prime candidate for wedging
Have you tried a Blue SCSI on the SCSI card? I would have used a piece of scrap wood as a wedge. As a temporary fix, of course.
I had a C64 with no intermittent audio out of the RF connection. I had to shim the RF RCA jack to make the connection solid. I later on got the SAMS book for the C64 and made a DIN cable to send the audio to my stereo receiver.
I had a bad CPU socket (corrosion) and my A2500 was a no boot. However with the same Commodore CPU card you have, it would boot. Did not seem to mind the CPU being dead.
Not craptastic, Adrian! There is no shame in using the "wedge as repair" technique. In 1987, I shoved a piece of foam into my HP 11c scientific calculator. A couple of leads on the surface-mounted main chip had come loose and I only knew through-hole soldering. It's still working!
Love your A2000 coverage. Your repair, "Speakers volumes". Lol 😁👀👏 Always found the A2091 the best scsi card. Too many issues with all the others. Oktagon was a good runner up.
The ROM of the A2091 is pretty much garbage. :) It needs the ROM 7.0 version to work with 040 accelerators at all, and also has various other limits. Also fails to use DMA with 32bit Fast RAM. On Aminet there are various patches for the ROM to make it work better.
Dear Adrian, maybe I'm very wrong but, please check all vias in that area, I remember CuriousMarc have the same problem with an IBM and after lots of test, even they develop a special bios to check things, they found broken via just inside. They have to do the same as u did to make it work. (Sorry my poor english, cheers from Argentina!)
Here in the normal world, you can just buy the metric screws by the kg at any hardware store. However, the PC screws have to come out of a PC. I've scrapped many PC cases just to get some spare screws.
That solution is not so horrible for a temporary fix at the household level. 👍
In the context of this channel it brings a very funny note of humor. 😊
And I'm sure in real life many of us have encountered solutions as crappy or crappier than that, 🥶 (even at the enterprise technical service level). 👋
Bodge installed. Job done.
It seems that as the circuit is functioning, heat expands it, and makes a better connection. Testing the continuity during a cold stress (compressed air), should show the defect.
As to the SCSI drives, recheck the transfer speed mask and that the correct drive metrics are selected (use a more recent OS, HDToolBox, and FastFileSystem).
That was awesome. I'm relieved to see that I;m not the only one who has done fixes like that. :D
This fix is quite something to speak-er of!
I've got a front loader Nintendo, that I HDMI converter, that just don't work when I've screwed it back together. Finally I figured it that it was caused the motherboard being slightly bent on the back where the cartridge connector goes and that caused the connector to get slightly bent. Using the console like that works just fine, but when I screw it back together, the cartridge holder pushes on the connector in such a way that some of the pins get disconnected, causing issues related to those pins.
I'm not familiar enough with those SCSI cards but the random failures suggest a lack of termination power and/or a lack of termination resistors at the ends of the SCSI chain (or an inadvertent terminator in the chain but not on the end). IIRC termination is set by software on the SCSI2SD.
That socket matches the quality of the board too...if you haven't replaced it I would swear that it was the culprit.
Short is most likely under the CPU slot not the 68000. You seem to be putting presure on the CPU expansion slot rather than the 68000. Cleaning debris has fallen down on the socket pins. Also need the v7 roms for scsi card, same as the A590.
Whatever works, works imho.
LOL at the very start of the video for a second i thought that was the Amiga display and i went to go look up when the 2000HD was released and just about had a heart attack before zooming in seeing what was actually happening there :P
Maybe the plated through holes for the socket or some Vias that are underneath of the socket are not connecting from top to the bottom of the motherboard.
@Adrian, the Screws with which for example CD-ROM Drives are fixed into a PC are also metric - M3.
Personally I dislike machined sockets, nothing like a good double wipe socket. Looking forward to you getting a Coleco Adam!
If I were to venture a guess, maybe the reset line trace is broken? It would explain why it comes on and keeps working even with the bending removed.
13:50 - Stick it in an oven for 20 mins, no? I did that to a 486 board 5-6 years ago. Still works to this day.
This may be a small/petty thing, but thank you for not saying “yep, it’s a 2nd channel video” this time. We know it’s a 2nd channel video, we clicked on it lol.
Pretty sure you've got a dry joint somewhere else on the motherboard - when you press down on the CPU the board is flexing a lot and probably causing the dry joint to move and make a connection, time to break out the magnifying glass or microscope and search around for that telltale crack in the solder on a joint.
I'd check those resistor packs under the drive by the CPU...perhaps they are not properly connected and they are not pulling up/down certain signals that might need be. Also check any other surrounding componenets.
...at least you know where the fault is! If only all computers could have a quick fix like this all the time just to get things going for a bit
Your fix reminds me of putting a piece of cardboard under the laser on my Playstation back in the day to keep help it read disks better. Janky, but functional.
Talking of having too much fast RAM, my A1200 will crash on booting if I plug in my CD-ROM drive with more than 4mb of fast RAM installed.
If it works, who cares? Its yours and you did enough already to bring it back. After that battery mess, its a miracle.
Queue the torches and pitchforks! 😜 At least you admit this is a terrible "fix". But a full video tracking the problem and actually fixing it would be great.
I have no idea if any are available, but maybe you might want to consider getting another motherboard and just trashing the one that's currently in the computer.
In terms of reliable vs. unreliable Amigas, there is of course the A500 Plus model too which suffers onboard clock battery troubles, as well as having owned one (which died when Amigas were near worthless and cheap to pick up), I've seen myself a few externally nice ones on ebay from optimistic sellers with "amazing condition, perfect example!!!" style descriptions along with a high asking price, and then seeing the green death around the expansion bay pins which they remain silent about, yeah, those batteries suck, but those sellers suck too...
Sticking the speaker in there is like using duct tape on anything 😄 It'll probably work for years now.
Sometimes half-ass is better than full-ass. Gotta admit that "fix" would keep me up at night. :)
You'd know better than I would, but my hunch is that it's not the board like you're confident it is. I think you need a proper socket for it instead of those pin headers. Whatever is going on seems to be affecting the reset line.
It ain't a stupid fix if it works. The only weirdness I would feel is keeping a **magnet** anywhere close to operating floppy and hard disks.
@9:57 REALLY freaked me out until I realized it was probably your video processor thingy :D
Heh!! Yeah the Elgato capture device seems to do that for some reason when there is no input. You get flashes of whatever was last input.
How many winced every time Adrian said “bridge card” instead of “bridge board”? 😉 Kidding aside, I concur with others that you are likely dealing with corrosion that has seeped into the pc board itself. Time to look around for an aftermarket motherboard to which you can transfer all the custom ICs.
Going dim to bright, other times just stays dim.... It's like he's describing me 🤓🤣
I wouldn't fancy a magnet potentially coming loose and possibly interfering with magnetic storage, maybe put blob of hot glue at the top point and the chassis for piece of mind?
if the board is a multi-layer board, you may have a plated hole via that may be cracked.
Nothing wrong with a bodge, time constraints and all!
That said, the first thing I'd do would probably be to use hot air to just reflow all the joints on the back side en masse just in case it's a bad joint. Might not necessarily be next to the CPU socket either just a quirk of the particular way the board gets flexed when you press there? I don't think it would take too long, thin layer of flux on it all, wide nozzle, fume extractor and off you go.
But it sure could be a broken trace too. Another commenter mentioned DIP chips in turned pin sockets being a bad idea, not sure if you've heard of that causing issues before?
Anyway great video as always! This machine reminds me of my childhood Atari ST which had to be vigorously twisted diagonally to make it start up. It was probably just a chip needing to be reseated but I didn't know anything back then :D
@Adrian's Digital Basement ][
This seems very much like a cold joint problem to me. Perhaps you just need to go over all the connections around the CPU with a soldering iron to reflow and solidify them. I would do this before removing the solder mask like you were suggested. EDIT: Don't worry about the AC noise, we can barely heard it. Not a problem at all. EDIT2: The SCSI to SD thing could be a voltage dip problem. Those adapters require a fair amount of power to work right. You may wish to check PSU power output and voltages connections. That system might be suffering a "brown-out" situation.
It sounds like it may be time to run the board through the dishwasher if it is corrosion. I'm sure that would get deep into wherever the corrosion is and neutralize it.
Why would running it through a dishwasher neutralize corrosion? That doesn't sound right to me.
We need a T-shirt with this as the craptastic logo!