Back from the Grave: Acorn A4000 (Battery leak, environmental damage)

Despite its obvious performance advantages against much all of the competition, Acorn redesigned its formerly-Archimedes systems to compete on the new battlefront: price. In 1992, this would be realised with the implementation of the ARM250 - an SoC variant of the ARMv2 chip. While raw performance was slightly inferior to the ARMv3 that had been used since 1990, the removal of many discrete components resulted in potentially significant cost savings

Sadly for Acorn this still wasn't enough to bring them to the forefront. By this point in time, Intel's x86 and IBM's PowerPC had basically claimed a duopoly on both the home and professional computer markets. Acorn would go on to continue their efforts to make inroads with the RiscPC but would not manage to gain the notoriety they deserved until the late 2000s.

 

This particular A4000 was sourced second hand from eBay by my brother in known not-working condition. As is typical it was physically processed and cleaned before addressing the elephant in the room: just about every single Archimedes ever produced, BBC or Acorn branded, has had the persistent memory battery leak, and this one was no exception.

Unlike the earlier A3000, the A4000 had relocated the battery to the rear right corner of the motherboard. This put it away from things like the system ROMS, but unfortunately left it by the relocated keyboard connector, power connector, and several extremely narrow traces related to the keyboard, serial, and parallel ports. Due to obvious battery-related damage, the keyboard connector was removed and replaced, and the battery substituted with a modern supercapacitor.

Continuity testing revealed a number of the nearby traces were no longer making contact, requiring patch wires to be installed. Since this is a much higher frequency machine than anything we've operated on previously it remains to be seen what, if any negative effects may be encountered as a result, but a lack of suitable peripherals makes this both hard to test and largely moot even if it fails.

Unfortunately this system also seemed to have suffered some amount of environmental damage. While we're unsure exactly what, it had left some of the components and solder joints looking worse for wear - though none of them had obviously outright failed based on initial inspections aside from the keyboard connector. As a result this required step-by-step diagnosis, with failures only being guessed at or becoming apparent as other elements came together.

The most obvious failure was the power supply. As a result, for the bulk of this testing a substitute was in use in the form of an old PC-AT PSU. Conveniently the entire Archimedes series could run from just +5VDC, though the other requested voltages (+12VDC and -12VDC) were available in needed.

Initial testing was conducted without the use of a screen. Instead, a CRO was probed into the output of the blue video signal to look for activity. This revealed that the system was operating, but not reliably, as it would frequently boot to a mostly flat signal with only a small spike at the start. Pressing enter would yield additional spikes, pointing to this being the star prompt for supervisor mode - a mode it would enter automatically on failing the POST.

Since we hadn't done it earlier, we attempted to reseat the ROM chips just in case as poor contact could result in such inconsistent behaviour. At this point we discovered that removing them was rather difficult, suggesting that the pins on either the chips or the sockets (or both) were damaged and likely not making good contact. Removing the sockets to replace them, it was discovered that one of the socket pins was either broken or severely compromised with the tip either missing or having been sucked up with the solder, pointing to a very likely cause for an intermittent failure.

Following the replacement of the sockets, the system would no longer boot to the mostly flat line, suggesting it was now either not booting at all or reliably loading the GUI. Pressing F12 (accessing the star prompt), typing BASIC, then CLS would revert the readout to a mostly flat line, so it was a safe bet the GUI was now loading correctly. Now we just needed to be able to see it.

As luck would have it, the A4000 - unlike the earlier A3000 - can operate from just about any standard VGA monitor. If the automatic detection failed, the system could be appropriately configured by holding down specific keys during the boot sequence (I believe this is 3 or 4 on the numeric keypad). Neither automatic detection nor manual configuration make any difference, however, if there is no signal reaching the connector.

Following the signal path using a CRO revealed that, despite their poor physical appearance, the components leading from the CPU to the VGA connector were still functioning perfectly. However, the signal was lost between the PCB and the cable, pointing to the VGA connector itself. Pending the acquisition of a replacement, we simply removed and replaced the solder on the existing connector, just in case, and it seems that this was enough to restore functionality (though the connector will likely be replaced in the long term just in case).

With this done and the system now appearing to operate perfectly, save perhaps the serial and parallel ports, this left just one last repair: the power supply. We've not had a good run on fixing power supplies, with our current rate of success standing at just 33%. Substitution is of course an option (picoATX PSUs have been very, very helpful) but we still wanted to try and retain the original equipment.

CAUTION: Power supply repairs can be very dangerous. While the AC side is usually very quickly rectified to DC and fuses are generally fitted, if anything has malfunctioned (why are you attempting repair if it hasn't?) than this may continue for much longer than expected. Mains AC can kill you, and at the very least can be extremely painful. Capacitor discharge can be similarly unpleasant.

Symptomatically, this power supply was outputting the wrong voltages. Instead of +12VDC, +5VDC, and -12VDC, it was outputting around +6VDC, 0VDC, and -6VDC. As a matter of course the capacitors were replaced but this did nothing to the output. After following the electrical flow on the high voltage side without detecting anything untoward, we turned our attention to the low voltage side.

Using a CRO, we discovered there were small pulses on all of the voltage outputs. The 5v in particular showed more clear evidence of the cause, as it was pulsing both above and below ground level.  As is standard in switch-mode power supplies, diodes would ordinarily be used to determine whether the voltage output was above or below the ground, and none of them should ever change sides.

With the same CRO, tracing back and probing both sides of the 5v diode showed the same pattern, indicating a highly likely failure. Removing it to confirm in isolation showed that it was behaving all the world like a small resistor and would need to be replaced. Due to the high amperage (10A in this case) these diodes are reportedly a likely culprit for power supply failures, along with the main switching transistor and the controller chip, so we will probably start with the low voltage side in future.

With the 5v line now outputting as expected, the rest of the PSU fell into line. Having an accurate reference voltage meant the controller chip could stabilise all three rails correctly, restoring the PSU to full working order. We now have a 67% rate of success for power supply repairs, which is nothing if not confidence boosting.

While it is fair to say that the system isn't back to 100%, it is now operational under its own power. At a later date (pending the discovery of some worthwhile peripherals or just a desire for experimentation) we may revisit the board to repair the broken traces leading to the parallel and serial ports, but for now that is superfluous as we have nothing to connect to them.

And another BftG done and dusted. Following on from the power supply success our next likely candidates are a Mac Classic (which also has some analogue board trouble) and an old Sperry-branded PC-XT system, both of which need to have their power supplies brought back to spec. All going well, I hope boost on that success rate.

Remember that electricity can be dangerous to both property and life.  If you don't respect it you can wind up seriously hurting or even killing yourself or others, and chances are it's going to be painful while it's happening.  Unless you're sure of what you're doing, avoid tampering with any electrical circuit - especially when it's live or mains powered - and always ensure you have a nearby assistant who can contact emergency services.