In the previous post, I took a look at the nearly pristine condition Apple Macintosh PowerBook 100 and accessories which were donated to me for some edu-tainment. While it was pretty to be able to look at a piece of history, somehow, I felt that wasn’t enough. I wanted more, and I’m sure that my friend would have appreciated if I could at least do some data recovery from it, even if it was to say there wasn’t anything of value left on it.
A little bit of looking around made me realize that this wasn’t necessarily going to be an easy project. Such old machines are very likely to have a hard drive failure, leaky/bad caps all over their internals, faulty SCSI cables, delaminating LCD displays, problematic power supplies, and the potential to refuse to boot without a functioning battery. Aside from that, there were claims that there were more malfunctioning units in the wild compared to working units, and an opinion that the quality of the “cheaper” 100 was not up to scratch compared to their American counterparts. Even with the service manual in hand, there were many warnings from others that connectors and cables are fragile, easy to damage, and difficult to replace.
It was at this point, I faced a multi-part dilemma – do I:
- Assume the unit is bad in all probability and refuse to do anything with it, and keep it as a nice display unit never to be powered up?
- Assume the unit is bad, and open it up right away to check for faults first, then decide where to go from there, with the possibility of accidentally consigning the unit to death if a cable is torn and also damage its aesthetics?
- Assume the unit is bad, open it up, inspect it and do a pre-emptive strike and replace all of the capacitors, even though there’s a possibility this will kill the unit if I make a mistake and damage its aesthetics?
- Assume the unit is probably functional, power it up anyway, and risk collateral damage in the case the out-of-spec components cause overstress of the remaining components?
It was a tough one to address, so even though I got the unit home on the 6th, I didn’t actually do anything “major” until the 9th.
Sorting the Batteries
Before I even began to think about any action, the first thing I had in my mind was to prepare the necessary prerequisites to keep the machine happy.
The easiest part for this was the replacement of the PRAM/PSRAM backup batteries, as some information indicated that some models may have trouble booting or with display in the case of bad or missing PRAM batteries. The CR2430 cells inside the unit were all expired, so I hopped down to my local element14 after placing an order for three Multicomp CR2430 cells at AU$2.21 + GST each, to be picked up the next day. These were the cheapest cells, and should do, but I wouldn’t leave them in the PowerBook 100 over extended periods in case they leaked. A more expensive alternative from a more reputable name might be justified in that case (e.g. Renata at AU$8.63 + GST each), but I wasn’t too enthusiastic that it would work.
The bigger issue was the primary battery for the laptop. As far as I have searched, there is no one-for-one replacement, and there’s no direct equivalent either. There have been numerous hacks using R-C car batteries and wood blocks or hollowed out aftermarket replacement batteries which were easier to dismantle. The fact of the matter is that the “claimed” 7.4V is actually the “topped up” voltage of the 6V SLA battery, and the charge characteristics of lead acid (constant voltage) aren’t suitable for using Ni-CD/Ni-MH cells with. The specific voltage in question isn’t even a good match to Li-Ion, despite 7.4V being a Li-Ion voltage, the actual fully charged voltage of a series string of two is 8.4V which is potentially a little too much.
If anything, I would rather do too little, than to do too much and hurt the unit. I decided to see if there was any chance of reviving the two packs I already had.
I measured the voltage across the first pack – 38mV. I measured the voltage across the second pack – an even worse 8mV. They were beyond dead. I decided to see what would happen if I applied the 7.4V full charge voltage across them for a day. The news was uninspiring. The current into the cell began at 0.1uA and only increased to 0.5uA after a full day. That’s miniscule. Once the voltage was removed, the battery read 138mV but quickly collapsed. It’s good to see the cells failing so inertly – it seems they’re pretty much “open circuit”.
Sorting the Power Supplies
The two power supplies looked visually good, but my gut feeling are that old switchmode supplies are trouble. Instead of ruthlessly plugging them in, I decided to hook them up to a pure-sinewave inverter through a variac and under the supervision of my Tektronix PA1000 power analyzer.
I gingerly raised the voltage on the first power adapter until we reached the “US” standard 120V. This should be enough, and wouldn’t stress the primary side capacitor as badly as say 230V would. The PA100 showed it was consuming roughly 0.5W from the primary side, so it probably was functioning in some capacity. I checked the output of the first unit and sadly, it only read 0.882V. This was way out of spec, so I sat the supply to the side.
I tried the second unit, and it was a little better. Consumption was the same, but the unit put out 8.2V. This unit was a little high, as the service manual expects 7.9V at the most. It’s not lethally high in my opinion, so there is some hope for this one. The voltage is likely to fall under load anyway.
I just hoped that nobody had damaged the power socket on the motherboard, as that was apparently a common source of failure which could lead to the need to tear it apart and creatively repair or replace the motherboard.
The Decision is Made: Power Up
After thinking about it for a while, I made the decision to power the unit up. It would be the easiest route to assess the viability of the machine without cosmetic damage, and seeing as some other malfunctioning PowerBooks didn’t suffer irreparably from being run unstably, I thought I could take the chance. I wasn’t expecting it to work.
I decided to prepare a camera for the occasion, seeing that it might be the last time the unit powers up – it could die right afterwards. Things could fail, explosions could happen. Nobody knows with this sort of gear just what will happen.
As I didn’t have a suitable battery, and some owners have said their units won’t run without one, I felt like my chances were poor. It was still worth a try though, and if the unit crashed when the hard drive tried to spin up, I knew where to go next. I plugged the second power supply unit into the laptop with the battery bay empty, and turned the power supply on.
The screen flashes once. Then it stops for a brief moment. Then it flashes again. And then it stops. All the while, the speaker clicks in time, but there was absolutely no response from the hard drive.
I am bemused. Was this the curse of trying to power up a PowerBook 100 without the battery?
A Short Breather: A Rash Decision
I felt a little disappointed, but not entirely. I stopped the camera, and sat down to have a think. The screen backlight did engage briefly, and the speaker clicked. It was trying to come to life, but something was stopping it. My immediate suspicion was that the second power supply was also defective.
At this point, I could have ordered a 2.5mm plug from element14 (such as the Cliff Electronics DCPP2-FLANGE for AU$4.16 + GST), but it was a Friday and I would have to wait until Monday to have it. My patience wouldn’t stretch that far – I had everything set-up. If it was to run without any major repairs, I told myself that today was the day.
I took the first power supply, and a pair of scissors, and cut the cable. Yes, I cut the cable. I felt immediately bad about the decision, but I felt it was the necessary thing to do. I didn’t feel like I would ever repair that power supply, and I didn’t feel like it was a critical loss having a second aesthetically identical unit with a better output voltage. But with this loss, comes the second opportunity – the one to revive the PowerBook.
Another Attempt is Made
The cable had two conductors internally – one coloured black, and one coloured white. Which one do you suppose is the tip, carrying the positive connection? I wasn’t going to leave this to guesswork, or to chance. I measured it. It was the black coloured lead. Very unexpected.
To power the laptop, I chose the best in my arsenal – the Keysight Technologies E36103A. A truly lab-grade power supply costing AU$1276 + GST with 0-20V and 0-2A capacity that I won from a review-and-keep, I wasn’t going to chance it with the less regulated Manson supplies I had. At least with this supplying the PowerBook, there should be no blown fuses and no chance for overcurrent.
With the wires attached to the binding post, and the output plug firmly seated into the laptop, the supply was set to provide 7.5V at 2.06A (the extra 60mA was the extra capacity the supply had – I decided to give it just a little extra as the OCP current limiter is very fast-acting by default, and I wanted the drive to spin).
I enabled the supply. This time, the screen flashed and then sat there, quietly, almost as if dead. That was different. Maybe this means we’re ready to go. I set the camera up and …
… eureka. The unit lives. Unexpectedly, the unit lives.
The unit had a total of 4Mb of RAM indicating a 2Mb PSRAM extension was installed. The hard drive was a replacement upgrade – it had a 120MB capacity and was partitioned using a third party OnTrack Disk Manager software, as the Apple HD SC Setup doesn’t recognize non-Apple drives. There is a line across the screen and minor row/column interference as was normal with passive matrix displays. The keyboard keys all seemed to work, and the trackball tracked pretty well, without any major flaws.
While the unit is alive, we really are on borrowed time. Every minute from now is precious, as the unit could stop working at any time. If there’s one thing I didn’t want to do, that was to power down in case it would not come back up again. I immediately called my friend to let him know of the news, and to start taking a quick look of the documents.
I made a decision that backing up the unit was historically desirable on the account there were a few installed applications and documents. The easiest way would be to take out the SCSI hard drive and image it on my Linux recovery box. However, this would not be an option as I don’t have a 2.5″ SCSI drive connector/adapter nor do I have any internal SCSI cabling and adapters.
The next best option might have been to boot the unit in SCSI Target Disk mode, and try mounting it over SCSI. This would need an HD-30 lead, which I do have. Alternatively, hooking up a ZIP 100 SCSI drive to the lead would make for an external drive of a decent speed. Maybe if I had hacked the serial port, I could use a terminal software to send out the drive contents by serial, although so slowly that the unit might fail by the time it completes. I wasn’t familiar enough with Mac terminal software, and the resource fork issue will come to bite me if I didn’t covert everything to something else (Binhex, AppleDouble).
Unfortunately, to attach SCSI drives requires powering down the machine. Also, as it turns out, I forgot to attach the floppy drive lead as well. I decided to power down the unit (reluctantly) rather than leave it running while I tried to sort this part out. I really didn’t expect to get this far.
The next blow to me was that my search for the SCSI lead proved unsuccessful, and it also occurred to me that I might need to install software to format and mount the ZIP disks as well which would complicate things. I was sure I kept the lead with the rest of my SCSI gear, and I did find the network adapter, but I couldn’t find the lead! Arrgh.
Backup: The Slow Way
Luckily (or unluckily) for this PowerBook, I wasn’t the sort to give in. I decided I would take the time to back it up the slow way. The saviour? A treasure trove of high quality teflon coated 3.5″ high density floppy disks which I salvaged from the uni.
To be precise, forty one of them (the last box has an extra one from another box). These were all pristine, sealed, unused floppy disks. I wasn’t going to use anything less than the best – who knows just how long the floppy drive would last either. As each of the disks were IBM formatted, it was a slow march to format and verify each disk, then copy the contents across. If you saw a post on Facebook about me staying up all night and dealing with floppy disks … this was the reason!
Because the internal hard drive was a bit full, I started off by copying off some data directly to the floppy disks, write protecting the disk and then deleting them from the hard drive. As MacOS has a verify after write as an automatic procedure, I had some confidence data would not be lost. This continued on until most of the smaller “sub-floppy-sized” files were taken off the system.
To archive the applications and other files larger than floppy sized, I needed to employ some software. Luckily, the computer had StuffIt Lite installed, which could compress files into .sit archives and split them to disk size.
As a result, I decided that each program will be in its own archive, and that would be split across several disks. This avoids the archives spanning too many disks and potentially failing if any disk became unreadable.
However, to think that this was going to be straightforward would be underestimating the situation. As expected, the laptop wasn’t always going to co-operate.
Within the first 30 minutes, the screen became very difficult to read, as each halves’ matrix controller seemed to have lots of interference issues. It would spontaneously revert to the cleaner display (above) but not for long.
The other issue was the system occasionally “bombed” out the F-line errors and finder crashes. I bumped up the supply voltage by 0.1V (as above) to compensate for a little bit of cable loss (could have probably done more) and it seemed to help things along a bit.
With such a system, archiving data presents a potential danger – the system may corrupt data written to floppy disks, there may not be enough power for proper operation of the drive, or the power may be too unstable to the CPU/RAM resulting in computational errors in creating the StuffIt archives. I had a fear of this, so I decided to create a moderately large archive, and then verify it on the PowerBook 100 just to see if it was okay. Thankfully, it seems, the display might have been unhappy but the unit was stable enough to proceed.
During the procedure, three disks had issues – one formatted and then had “re-verifying format” as 12kB of bad sectors were marked out but the disk remained usable. Another had formatted with bad sectors the first time, but formatted without bad sectors the second time. After trying to fill the disk, I was informed that the data could not be verified after it was written, indicating that the verify process found bad data. A third disk failed the initialize procedure entirely and was spat out at “re-verifying format” probably implying a bad sector in the system area. I suppose, after all these years, it’s possible that the media has degraded slightly, or more likely, the drive might have slightly dirty heads or insufficient write current to overcome the pre-formatted pattern that has been stored for years.
The back up took a total of 10 hours exactly as timed by the onboard RTC. The PowerBook 100 valiantly held in for all 10 hours, despite the awful screen corruption, suggesting it’s really not in too bad of a shape maybe with the exception of the LCD drive circuitry. All the user accessible data was archived to floppy disk, and no data was lost despite the last recorded use being in 1995 (21 years ago). The drive did sound a little worse for wear on shutdown, with a loud grinding noise, but I suppose most 2.5″ SCSI drives of the era are no longer with us anyway.
Recovering the Backup
The next step was to read out all the disks. I could have done it during the backup and passed a single floppy back and forth, but seeing as I probably don’t have many chances to use the disks, I’d rather just write them once and store them as a backup copy.
Because I was only interested in the data and there was no authenticity or copy protection to worry about, I was going to avoid the hassle and slowness of dragging out my Kyroflux and 3.5″ drive. Instead, I oped to run two USB floppy drives, namely my Mitsumi 2x and the Teac (HP) regular USB floppy drive. This is possible because high density Apple disks use MFM formatting which is compatible with the IBM format in a way. With two drives operating in parallel, it shouldn’t be too hard to read it all back.
As it turns out, I was wrong. Something about those Apple written floppies flummoxes both USB floppy drives, reporting bad sectors in exactly the same places – one or two per disk. At this point, I was sweating on the possibility that some data was lost, and by Disk 29, I decided that I’d just get out the Kryoflux and change over.
It was a good thing I did, because the Kryoflux combined with a Sony floppy drive read out every faulty disk and every disk from 30 to 41 with no faults whatsoever. It really is a versatile and important tool for reading out floppy disks.
With all the data in one place, I decided to make a new hard drive image of about twice the size using hfvexplorer, and move all the files from the disks to the new volume. Booting my System 7.5.5 disk image in Basilisk II, I was able to rejoin all the StuffIt archives and extract them with zero integrity issues and reconstruct the file system in the same order as the laptop.
I fixed the bootability of the new drive image by using a boot disk, after which a replica of the system of the laptop was running inside Basilisk II with a colour screen and twice as much RAM.
It worked so well that I was able to access the SoftPC (MS-DOS) virtual machine that was on the laptop to edit some olfactometer calibration data that was stored:
Talk about a whole load of fun. I’m running MS-DOS editor, inside SoftPC running on System 7.1, running inside Basilisk II emulator, running on a Windows 7 host. The data, and all the apps were saved.
Examining the Faulty Power Supply
This begs the question of what happened to the first power supply. I wasn’t going to just throw it away … so I thought I might as well open it up in case it helps others with their problems. The easiest way, it seems, is to pry with a flat-blade screwdriver where the cable exits the case. By prying along there, the case splits at the glued seam (with a varying level of neatness).
At the first pry, the brown colour of electrolytic capacitor fluid was visible, and the cause of failure obvious. This oily substance has a moderately bad smell, and basically tells you that the electrolytic capacitors that should have been sealed have degraded and the electrolyte has been lost. This, in turn, means the capacitors no longer function as they should, and more importantly, may mean that adjacent parts are coated in this electrolyte and corroded. The tops of the heatsinks can be seen to have some corrosion, no doubt, due to the fluid.
To remove the board from the case requires undoing a screw on the PCB near the output cable. As the screw is probably the low point and flexes the PCB downward, it’s also a natural place where the fluid has accumulated, thus the screw has been rusted dramatically.
Once extracted from the case, we can see its full glory. The unit is made by Sony, and the “switchable” power inlet plug is removed once the halves are separated. There is a suppression capacitor hooked right across the back of the pins. The unit has a primary side fuse of T1.25A rating. Part of the board’s insulation is made up of a L shaped plastic shield. There is a vertical segment of PCB as well, with adjustment trimpots for voltage and current.
The underside shows a clear isolation barrier between primary (the majority of the board) and secondary. The capacitor fluid has corroded joints on the secondary side, near the screw as well as near the middle-bottom right corner. The corrosion isn’t too serious as to be irreparable, although to heat the joints up with a soldering iron would be mildly unpleasant.
Looking at the supply from the side, the isolation barrier can be seen on the vertical PCB as well.
From the opposite side, we can see fusible resistors of higher power rating, capacitors and what appears to be a diode in a TO-220 package mounted on a heatsink.
Looking at the supply from both ends, we can see the majority of the capacitors are on the secondary side. These appear to be Elna banded capacitors – the LongLife branding is one of their lines. All of them appear to be quality caps, but the problem is age. The primary side also has some disc capacitors, filter inductors, bridge rectifier and an “epoxy dipped” package IC. The PCB is dated 27th November 1991.
The rear of the package is marked SH-C1 SONY 10MV1.
The primary to secondary is bridged with a Sharp PC113 photoconductor (i.e. optoisolator). Even inside, just underneath, there is an electrolytic capacitor hiding on the main board.
The other caps hide towards the output, where there is another IC marked C358C, likely a dual op-amp.
For those who might bother to repair the unit, it’s a small mess to clean up with only a few caps to replace. I’ve made a crude electrolytic capacitor map based on the teardown, although some capacitors deep within the unit might be a bit hard to get to, and corrosion might need to be “soldered across”.
Approximate positions and silkscreen numbers are indicated. Sorry if you find the writing a bit hard to read – the usual disclaimer applies – use at your own risk. Note that other failures might also exist, as some of the other types of capacitors not normally known for failure might also fail occasionally (e.g. the suppression type polyester film capacitors).
Going on a gut feeling and with a little bit of preparation, the unit surprisingly lived to start up, and ran for 10 hours providing enough time for the data on the unit to be backed up to 41 floppy disks. The data was successfully retrieved and the system recreated in an emulator.
Failure of the power supply appears to be due to capacitor leakage, which is not uncommon across devices of this vintage. This is a messy clean up and repair, but is by no means impossible to do.
The question now is, having recovered all the data and have the system run for the 10 hours, should I:
- Leave the unit as-is, seeing I haven’t damaged the cosmetic condition, but probably jeopardize any future possibilities of using the unit.
- Open the unit up for an internal examination with risk of damage, but do nothing else to it.
- Open the unit up and do a full replacement of all the capacitors, like for like as best as possible to prolong its lifetime, although in the process, with greater risk of damage and the possibility of no payoff (e.g. something else gets damaged or the SCSI drive gives out which I suspect it will).
- Repair the power supplies as well, but break their casing in the process, with a risk that the repaired supply could possibly fail in a way to damage the unit.
- Invest even more money and time to obtain modernization parts (e.g. CF PowerMonster II) and perform the necessary modifications while it is still available, even though the machine is not of great importance to me at present.
It’s a tough call – I’m not sure exactly what I’m going to do with it now, besides keep it in its present state (the easiest option). I’m still relatively elated that I got it to work … using a trackball is pretty cool.