Amiga 3000 Restoration

So I'd like to apologize right off the bat. I didn't think to take any pictures during this project, as I hadn't expected it to be involved enough to warrant a writeup. But by the end of it I had overcome enough obscure obstacles that I felt it worth documenting, in the hopes of making someone else's job easier. That said, let's begin.

The Amiga was a platform which I hadn't had a chance to play with growing up. I simply didn't know anyone who owned one of the elusive beasts, nor were any of them turning up in the usual places I haunt. So I remained only dimly aware of them (and the huge price tag they usually carried) right up until ~2004 when I suddenly ended up with my first. It was an Amiga 2000 which was quite literally found in an alley beneath a pile of lawn clippings. But I didn't have any of the associated bits and pieces required to make it work, so it sat unused on a shelf for many years. My first real experience was to be with an Amiga 3000.

I had started to see Amiga components coming through Free Geek and the Hackery in 2009. I grabbed an assortment of 3000 parts over a year or two, but nothing was in working order, and an unexpected move meant I lacked both the time and space to give them the attention they required for many months. But in June of 2011 I finally pulled a battered Amiga 3000 and the 2 spare motherboards I had found for it out of storage and got to work.

The assembled Amiga was functionally complete, but only just. Nothing was quite right with it. Nearly every screw had been replaced by whatever the previous owner had had laying around, things were held together with electrical tape, and everything that was capable of rusting had done so. It was a very sad sight. But beneath it all, it was a 25 MHz 68030 A3000 with 2 megs of chip RAM, a 2 gig SCSI hard drive and upgraded kickstart ROMs. So I went to work on it.

The first thing I noticed was that all 3 of my motherboards were covered in corrosion around the area of their Ni-cad clock batteries. Thankfully one of the boards had had its battery removed years earlier, so the damage was minimal on it. The other two, including the one inside of the Amiga, were less fortunate. So before even trying to power anything up I dismantled the complete machine and set about cleaning up the corrosion. Quite a long time was spent scrubbing at that board with distilled water and a toothbrush, followed by alcohol, and finally an unpleasant time attempting to replace dissolved traces with wire.

When the motherboard looked good, or at least less bad, I re-assembled everything and flipped the power switch. Much to my surprise the system actually powered up, and after thinking about it for awhile, the screen lit. But there was a problem. While the system sounded like it was doing all the usual things, you could hardly tell by looking at the monitor. All the screen showed was a bunch of flickering wiggling lines. The battery, it seems, was located directly behind the video hardware, and it should come as no surprise that it was hit the worst. But after several minutes of running the picture on the display slowly cleaned up, until eventually the Amiga workbench could be made out. The system had booted.

Now, I suspect I could have repaired that motherboard if I had spent long enough at it. Perhaps the variable capacitor which controls the scan doubler tuning had been damaged by the corrosion, or something like that. But battery leakage repair is such unpleasant work that I decided to try one of the spare boards instead. So for the second time I tore the Amiga apart.

The two spare boards I had were Rev 9s, meaning they were the final versions produced. The semi-working board in the A3000 however was a Rev 7. Initially this didn't seem to be a problem. I was able to borrow a few capacitors and such off of the damaged rev 7 board to replace ones knocked off of the spare board in handling with no trouble. And the 32 agonizing 512k ZIP chips from the uglier of the two spare Rev 9s moved over to the nicer one with only a half an hour of careful pin straightening and alignment. But then I looked at the ROMs. Both Rev 9 boards had Kickstart 2.04 in the two ROM sockets toward the front of the machine. But the semi-working Rev 7 had Kickstart 3.1 in the BACK two ROM sockets. To this day I'm still not sure what the difference is between these sets of sockets, only that there is one, and that ROMs designed for one set won't work in the other. This shouldn't have been a problem, except for one of the changes between board revision 7 and revision 9. The rear two ROM sockets were missing. In their place was a set of solder pads. And no one knew if the pads were still functional after the changes made in the Rev 9 layout.

So after confirming that rear ROMs definitely would not work in front ROM sockets, I crossed my fingers and reached again for my soldering iron. First I had to clear the pads on the Rev 9 board so that a socket could be installed. My solder sucker made quick work of this. Then I grabbed the Rev 7 board and set about desoldering two of its ROM sockets. This was a bit harder because I was using my little Weller iron since my more powerful iron was still packed away. As such every time I came to a ground pin the thick traces would draw the heat away almost as fast as I could apply it. Much wiggling and fiddling was required to get those sockets out. But finally they came free, and were soon mounted in the freshly cleared pads of the Rev 9. I also noticed pads for a set of jumpers next to the ROMs that weren't filled in on the Rev 9, but were present on the Rev 7. These jumpers were labeled "ROM" and on the Rev 7 were in the "A3000" position as opposed to "A2000". I cleared the pads and soldered in a set of pins in case these needed to be jumpered, but given that the board was presumably working before with A3000 ROMs in it, I left the pins un-jumpered for the time.

I re-assembled the Amiga for the second time using the freshly repaired Rev 9 motherboard. Now the system had 16 megabytes of FAST RAM, 2 megabytes of chip RAM, the upgraded SCSI chip, and the final revision of the kickstart ROMs installed. But I had no idea if any of those components, or even the underlaying board, would work. I had made a lot of unknowns for myself that would be a nightmare if I needed to diagnose a startup error. But luck was with me, and on the first try the system powered up and booted into workbench 3.1. That meant that not only were all of those blasted ZIP chips installed correctly and working, but also that the rear ROM sockets I had soldered in were still wired correctly as well. It may have been ugly, but it was alive!

I set about rummaging through my workshop in search of an Amiga mouse so I could try out my freshly repaired system. And that's when I spotted it, on the top shelf. Another Amiga 3000 I had forgotten about.


So with much cursing I pulled down the second (minty mint!) Amiga 3000 and opened it up. It was Another Rev 9, but the board was also badly damaged by battery leakage. It was also only a 16 MHz '030. So all my hard work had not been for nothing. I stripped both Amigas apart and carefully transferred my nicely repaired and upgraded board into the beautifully preserved case. Now it looked as good as it ran. Better, in fact, because now it wouldn't boot. Oops. A few minutes later I discovered that several of the custom chips had worked their way partly out of their sockets during the motherboard transplant, and a hearty shove on the tops of each had the system restored to functional.

So now we're up to 3 and a half take aparts, and for the third time I very briefly thought I was finished. I should have known better by now.

I was brought back to reality the moment I hooked a SCSI CD-ROM drive into the A3000's external SCSI port. Suddenly my beautiful system wouldn't even power on. It was as though it wasn't even plugged in! The same was true of the external CD-ROM. When it was attached to the system, it appeared to have no power. As soon as I disconnected it, both devices powered up normally. And it did the same thing with the second CD-ROM drive I tried. Obviously something was shorting.

I suspected that the problem was related to the termination power line on the Amiga. I had read an article on the 3000 which mentioned that the termination power diode was sometimes installed backwards at the factory, resulting in no power at the termination power pin. Perhaps my diode was backward, causing the power supply in the CD-ROM drive to short out when it attempted to provide its own termination power. And while I wasn't sure, it seemed plausible that the Amiga's own power supply could be confused by having power rammed into it, to the point of not turning on. The article had suggested a simple test. If +5v couldn't be measured on the term power line, then the diode was backward. I turned the Amiga on and used my multimeter to check pin 25 of the SCSI port. To my great confusion I saw +5v. If the diode were backward I would have seen no voltage at all. So much for an easy fix. I had a mystery on my hands!

But a clue came when I turned the Amiga off and checked each pin of the port for continuity to ground. Now the same pin 25 which had worked while the system was powered on was reading as grounded! That didn't seem right. Wasn't the whole point of the diode to prevent power from flowing back into the Amiga from a connected device? I checked the Rev 7 board I had removed and found that pin 25 of its SCSI port wasn't grounded. So I was right! Something WAS wrong with the term power line. To narrow down the cause I disconnected the internal hard drive to make sure it wasn't somehow shorting things, but the problem remained. It seemed time for another teardown.

Once I had the board out it wasn't long before the problem became clear. While the diode was in fact installed correctly, it had shorted internally. The diode was acting like a simple piece of wire. Current would pass in both directions with no trouble at all. That's why I was able to read +5v when the system was running and ground when it was off. All I had to do was steal the working diode from the Rev 7 board and I was in business.

Once back together both the system and the external CD-ROM drive were powering up as they should. I was another step closer to being able to do a fresh install of OS 3.9 off of the CD. Except while the CD was powering up, the Amiga wasn't able to read any discs I put in it. Wasn't OS 3.1 supposed to support CD-ROM drives "out of the box"? I opened the hard drive utility in the /system/ folder and found that only the internal hard drive was listed. Ahh, but the SCSI ID of the hard drive was 4! The same ID I had my CD-ROM set to. That would certainly account for the problem. So I shut down and moved the ID of the CD-ROM up to 5. Booted again and now the CD-ROM was showing up in the HD toolkit, but still no discs were showing up. Time to do some reading.

After a half an hour of digging through google and Amiga forum posts, I had the beginnings of an answer. It seems that while OS 3.1 comes with a driver for CD-ROMs, it's not actually installed by default. Also, it only support SCSI CD-ROMs. This wasn't a problem for me since I was using a SCSI CD-ROM, but most of the forum posts had I found were from people with Amiga 4000s whining that their IDE CD-ROMs weren't working, followed by a series of misleading replies telling them that OS 3.1 didn't support CD-ROMs. Ahh, the internet! Anyway, I dragged the file "CD0" from the /Storage/DOSdrivers/ folder to /Devs/DOSdrivers/ and rebooted again. Still nothing. Finally I found a forum post saying that 1, the driver wouldn't be loaded on startup unless I edited a config file, even though I had moved it into the /Devs/DOSdrivers/ folder, and 2, the file itself needed to be edited to point to the SCSI ID of the CD-ROM. Seems by default it expects the CD-ROM to be at ID 2 and mine was at 5. So to make things simple I just shut down the machine again and switched the ID of the CD-ROM to 2, then started back up, navigated to /Devs/DOSdrivers/ and manually ran the CD0 file. Ta dah! The OS 3.9 CD-ROM suddenly appeared on the workbench! Now I was ready to create a boot floppy and install OS 3.9. Almost.

When I had transferred my nice shiny Rev 9 motherboard into the better case, I had also brought over the ugly Amiga's floppy drive to go in the second floppy bay. But now neither of them would work when I inserted a floppy disk. I was using the correct double density floppy and the authentic Amiga floppy cable, yet it wasn't working. Further, only one of the floppy drives was showing up in the format tool. I should be seeing two! On a hunch I disconnected the rusty drive and suddenly the first one sprung to life. Ah hah! It seemed as though I had found the problem. Or had I? On a whim I disconnected the nice clean drive and plugged just the rusty one in, and to my bemusement it was now working correctly as well. So either drive would work fine on its own, but they wouldn't work together. Hum.

The clues I needed to fix my floppy problem came from the Amiga 3000 manual. It stated that if you wanted to add a second floppy drive to your Amiga 3000, you had to move the existing floppy drive to the other bay (So it was plugged into the end of the floppy cable with the twist in it, instead of the middle connector) and install the new drive where the old one was. You also had to move a jumper on the motherboard to indicate that there was two drives present. That jumper (J351) was by the ISA slots and was in the awkwardly phrased "NODF1" position. I moved it to "DF1" and fired the machine up. Now in addition to both drives being dead, there was a phantom disk mounted on the desktop. "DF1 Unreadable." I thought some more about what the manual had said about moving the existing drive to the end of the cable. That suggested to me that there was a difference between the drive which shipped in an Amiga and the drive you would buy from Commodore as an add-on. Since both of my drives were factory-installed, it seemed to me that they wouldn't have this difference. And the most obvious difference would be a jumper. So I removed the drive from the middle of the cable and had a look. Sure enough at the back of the drive was a set of tiny jumpers. Both drives were currently jumpered in the "DF0" position. Given the error about DF1 being unreadable, I decided to jumper the middle drive as DF1 and give it a try. The machine fired up and the phantom disk was gone. I inserted a floppy, and the drive light lit up. A moment later a disk icon appeared on the workbench. It worked! I ejected the disk and stuck it in the other drive. Again, it worked! I was very happy. Finally the last bug in the system had been found and squashed.

You'd think I'd have been a little slower to make such a statement by now, wouldn't you? Well, you'd be wrong, just like I was.

It soon became apparent that the rusty drive had a curious problem. If I shoved a disk into it, often nothing would happen. But as I was in the act of ejecting it the light would suddenly come on and it would start to read. I had to hold the disk up slightly to make it work. The problem seemed to be in the little switches which detected that a disk had been inserted. A quick google suggested that this was a common problem with this model of Chinon drive. But I didn't have any replacements, so I decided to try fixing it. I pulled the drive out and unsoldered the little block of switches. With some careful prying the top of the plastic switch box unclipped and the 2 gold springs within fell out. All I had to do was blast the dust which had collected out with a can of compressed air and then give the contact points a rub with an alcohol-soaked q-tip. Then I carefully positioned the springs, clipped the box back together and soldered it into place. Success! The first disk I inserted was read exactly as it was meant to be, as was every one that followed. I had repaired the drive without having to replace anything.

After all that the actual install of OS 3.9 went surprisingly quick and easy. I just had to make an "Emergency-Disk" floppy, boot from it, format the internal hard drive, then start the installer located on the CD. The only snag I ran into was forgetting to check "bootable" in the hard drive partition tool, but it turned out I could go check it after the fact without re-partitioning, formatting and re-installing the whole thing.

And I was done! 4 days, 4 and a half complete take aparts, 4 components desoldered and moved over from parts boards and countless false finishes later, the Amiga 3000 was fully functional.

...Except I still need to rig up a new clock battery.

Page created June 7th 2011

That Christmas magic's brought this tale to a very happy ending