A friend of mine who I’ve done several repairs for in the past recently trusted me with a rather rare, and very poorly, Acorn A4 laptop.
The A4 was an early laptop and the only one that Acorn ever made – it was effectively an Archimedes 5000 squashed into a portable case, including an 1800mAh NiCad battery pack, ARM3 CPU @ 24MHz, 1.6MB 3.5″ FDD, RISC OS 3.1, either 2MB or 4MB RAM, and an optional 60MB 2.5″ IDE internal HDD. This one is the high-end model with 4MB RAM and a 60MB HDD, which would have cost an incredible £1699+VAT when it was new.
Apparently the A4 wouldn’t boot or display anything on-screen, which I confirmed on its arrival – the power supply seemed OK as the PSU and power LEDs worked, and the battery LED flashed as this was removed. I also noticed that the NiCad battery pack had started to leak, but thankfully the internal NiCad RTC battery had apparently already been removed.
The A4 suffers from two major problems: the NiCad RTC battery on the mainboard physically leaks a corrosive alkali which can cause significant damage to the mainboard; the SMD electrolytic capacitors on the mainboard have a habit of failing and physically leaking corrosive electrolyte, which can not only prevent the system from working, but can also cause damage to the circuit boards and the components on them.
I must therefore emphasise: if you have an A4 in original condition, it needs to be serviced! The original RTC battery needs to be removed and the original electrolytic capacitors need to be replaced. These systems are dying, day by day.
The first step was to disassemble the unit and check over everything inside.
Disassembling the Acorn A4
The A4 is easy to dismantle with basic tools: the battery pack unlatches and slides out; the Econet expansion cover at the top-left of the keyboard flips out; the two display ribbon cables under this can be disconnected by lifting the locking tabs and gently lifting up the cables; three Philips screws at the rear of the case hold the display assembly in place; three Philips screws on the underside of the case hold the keyboard assembly in place; with the display removed, the keyboard cables can be unclipped in a similar manner to the display cables, again taking care not to damage them.
The DC-DC converter can then be removed, this is connected to the mainboard via a pin header and simply lifts out; the FDD RF shield also lifts out; the FDD is held in place with four Philips screws on the underside of the case, then it lifts out – it is connected to the mainboard via a ribbon cable which can also be removed in a similar manner to the display and keyboard; the HDD cradle is held in place with one Philips screw, and the HDD is connected to the mainboard by an IDE cable which can be unplugged.
The mainboard and rear case piece are held into the case with three Philips screws on the rear of the underside of the case; the mainboard is held into the rear case piece by the hexagonal screws around the D-sub ports, which can be removed using a hex driver; the cables that connect the mainboard to the case can then be unplugged, taking care to note where they go and in what orientation.
Mainboard Rebuild
Aluminium electrolytic capacitors are commonly used for filtering, smoothing, and decoupling in both high- and low-voltage electronics. They typically comprise aluminium windings which are coated with a liquid electrolyte, which can dry out over time (negatively affecting the performance of the capacitor, often causing them to fail dead-short), or leak out and cause corrosion to the PCB and surrounding components.
There is only one production variant of the A4 mainboard, whose electrolytic capacitor values and locations are available in the technical reference manual – it’s fairly tricky to recap as it has eighteen SMD electrolytic capacitors and the pads/traces are quite delicate.
The original parts can be removed using a hot air rework station with kapton tape and aluminium foil to protect the surrounding areas, or by carefully twisting them off using needle-nose pliers (this technique may not be suitable if the pads are damaged, as they could delaminate from the board).
The pads can then be cleaned up using new leaded solder and either desoldering braid or a desoldering station. Finally, the board should be thoroughly cleaned to remove any leaked electrolyte and leftover flux, using isopropyl alcohol and ESD-safe brushes. I also gave the board a run in my ultrasonic bath.
The new parts can then be fitted. The originals were aluminium electrolytic capacitors with a liquid electrolyte, however a good equivalent replacement is tantalum electrolytic capacitors – these use a solid electrolyte, meaning that they will not physically leak.
When substituting electrolytic capacitors, the capacitance needs to be the same, and the voltage rating can be the same or higher (within reason) – when you’re going through all this effort to recap something, be sure to use high-quality replacements.
I couldn’t find any commercially available tantalum capacitor packs for this version of the A4, so I just made up my own by noting the specifications of all of the electrolytic capacitors on the board, and ordering a set of high-quality known-brand parts.
When fitting new electrolytic capacitors, you must take care to ensure that the value, voltage rating, and orientation of the new capacitor are correct – electrolytic capacitors are polarised, so must be installed the correct way around, else they’ll get hot when powered on (and probably explode). The polarity is marked on the case: for aluminium electrolytic capacitors, the negative side is usually shown by a white stripe (for through-hole) or a black bar (for SMD); for tantalum capacitors, the positive side is usually shown by an orange or white bar (for SMD). This catches a lot of people out!
You can’t always trust the orientation markings on the PCB silkscreen (if it even has them, not all boards do), as sometimes mistakes were made in the design from the factory (take the PCB layout of the audio circuit on the Commodore CD32, for example), so care must be taken to match the orientation of the new capacitor with the original. Make sure to take lots of “before” pictures for reference.
The leakage around C134 was particularly bad, and the capacitor and both its pads fell off the board after being barely touched – I couldn’t get enough purchase on its vias to fit wire bridges and replace the pads directly, so I just mounted the replacement part directly to the other side of the vias on the underside of the mainboard, reinforced with some epoxy.
I also took this opportunity to replace the PRAM battery, which had already been removed – I cleaned up any remaining alkali corrosion using white vinegar and a toothbrush, then built a CR1220 coin-cell adapter using a through-hole CR1220 battery holder, a CR1220 3V coin cell, and a Schottky low-drop signal diode to protect the new battery against back-charging – the original PRAM battery was a 1.2V NiCad cell, but the RTC IC used by the A4 (PCF8583) supports 1.0V-6.0V battery input, so a 2.8V equivalent is fine.
I wanted to try making a custom PCB for this, but didn’t have the time – if anyone else would like to give this a go, the lead spacing on the battery vias is approx. 13mm.
I also cleaned all of the sockets and controls using contact cleaner.
I reassembled the unit for testing, and after a CMOS reset (holding DELETE at power-on to clear the PRAM to its default settings) it seemed to boot up OK into RISC OS 3.1 in ROM, however the HDD wasn’t detected despite it spinning up.
HDD Replacement
The original 60MB Conner 2.5″ IDE HDD in the A4 is notoriously unreliable, and suffers from sticky head stops – it’s possible to fix these temporarily by opening up the drive, but you risk data loss by doing so, and it’s not a permanent solution.
For this machine, I wanted to move to a solid-state solution – it’s not necessarily that simple though, as the A4 can be quite picky when it comes to IDE drives, due to its IDE interface hardware and the buggy version of ADFS that shipped with RISC OS 3.1.
I tried a number of combinations of both SD card and CF card to IDE adapters and different cards (manufacturer, speed, capacity), but had no luck – using a self-extracting !HForm v2.48 I always got disc error 20 with the SD cards and disc error 23 with the CF cards.
Based on advice from Phil on the Startdot forums, I fitted a CF-IDE adapter and Transcend 2GB 133x CF card. I also burned a 27C512 EEPROM with Phil’s Wizzo v315 A4 expansion ROM image using my TL866II+ programmer, and fitted it in place of the original expansion ROM, which is used by the A4 to load additional packages such as the battery manager/controller – the Wizzo ROM also includes IDEFS, which is much more reliable and compatible than the RISC OS 3.1 ADFS.
To configure IDEFS, you need to open the command prompt using F12, enter “*Configure FileSystem IDEFS”, “*Configure Dir”, and “*Configure IDEDiscs 0” – this should make IDEFS the default file system and allow automatic detection of IDE drives.
You can then load !IDEFormat to format and initialise the CF card as follows:
- Copy a self-extracting SparkFS archive and an IDEFormat v3.53 archive to a 1.44MB 3.5″ PC floppy disk from a Windows PC using a suitable 3.5″ to USB adapter.
- Boot the A4 with the disk inserted.
- Configure a 1024KB RAM disk & set it as target (“*RAM” via the command prompt).
- Set the SparkPlug archive filetype on the disk to “BASIC”.
- Double-click the BASIC icon on the disk to run – it should extract itself to the RAM disk.
- Double-click the !SparkPlug icon on the RAM disk to run – you should then have a SparkPlug icon appear on the taskbar.
- Drag the IDEFormat archive on the disk to the SparkPlug icon – the !IDEFormat application should then become available on the RAM disk.
- Run !IDEFormat by double-clicking the icon – you will then be able to add and initialise new drive partitions. RISC OS 3.1 does not support partitions greater than 512MB capacity, and large file allocation units of greater than 1024 are very inefficient, so I created and initialised four partitions of approximately 480MB each, named HDDX – I also loaded some useful tools onto HDD0.
Battery Pack Repair
I had previously noticed that the NiCd battery pack had started to leak – this comprises twelve Varta 1.2V 1.4Ah sub-C (SC) NiCad cells in series (14.4V nominal 1.4Ah), with a thermal fuse in the centre to disconnect the battery chain if this gets too hot, and two thermistors to provide temperature feedback via the battery connector.
Nickel-Cadmium batteries aren’t commonly used in electronics any more for health and environmental reasons, but you can still buy them for some applications, such as in power tools. It would have been possible to re-pack the shell with twelve sub-C 1.2V 2200mAh NiCd batteries, but they’re quite expensive and still have limited capacity and longevity. The owner was happy to run the A4 from the power supply instead of on batteries, so I decided to just gut the battery pack and use it as a dummy.
The pack is simply clipped together, and the cover comes off quite easily.
You should be careful handling and disposing of NiCad batteries and their leakage, as cadmium is a toxic heavy metal and can be absorbed through the skin.
I removed all of the cells and cleaned the interior using white vinegar.
At this point, as I had some spare Samsung 2600mAh 3.7V lithium-ion 18650 bare cells lying around, I thought I would try doing a bit of experimentation with fitting these to the battery pack without requiring modifications to the computer itself.
Disclaimer: lithium-ion batteries can be extremely dangerous, especially when mistreated (i.e. charged incorrectly or overheated), so make sure you know what you’re doing before messing about with them – this is not a guide!
I do not accept any responsibility for the actions of my readers, so attempt any of the following at your own risk – do not attempt to modify or build any battery charge circuit unless you are suitably competent and aware of the significant risks involved.
NiCd cells are generally quite simple to charge – the original batteries seem to be float charged with constant voltage directly from the PSU (unregulated 21Vdc-26Vdc, minus a diode drop so approximately 20Vdc-25Vdc) with a variable current (either a low-current trickle charge or a higher-current fast charge, depending on the temperature and voltage of the batteries and the actual current going into them).
However, lithium-ion cells are trickier to safely charge, whether they are bare cells or cells with built-in protection – they have more complex voltage and current control requirements that need to be managed properly by a dedicated charge controller, multiple cells need to be properly balanced, and bare cells like these ones need to be properly protected (i.e. overcharge, over-discharge, etc).
With yet more help from Phil (thanks Phil!), I therefore decided to use:
- An MPS EV26124 evaluation kit PCB based around the MP26124 charge control IC designed specifically for 4S lithium-ion packs (four 3.7V lithium-ion cells in series (14.4V nominal, 16.8V maximum)), with 10-24Vdc input range.
- A 4S lithium-ion protection PCB with charge/discharge balancing, overcharge and over-discharge protection, and short-circuit protection.
The EV26124 board was, from the factory, too large to fit in the battery pack. I had to remove the large pins for the inputs and outputs as well as the configuration jumpers, and I even had to cut the board down to size using a Dremel – I wouldn’t recommend doing this on just any PCB as you’re likely to cause damage or inter-layer shorts if you don’t know what you’re doing, but this is a 2-layer PCB with the major components at its centre, and its copper layouts are available in the datasheet for reference.
I then added insulation between the EV26124 board and the protection board, and secured and connected them back-to-back, with the battery side of the EV26124 board connected to the charger side of the protection board.
I configured the EV26124 board for 4S operation by soldering the “4CELL” jumper, and set the enable polarity jumper “EN” to ground to permanently enable the charger (as the enable line is pulled low). The charge current setpoint on the board can also be changed by modifying resistor R4 (default 200mΩ 1% 2W 2512 SMD) – this is 1A by default, which I left as it was (approx. 0.4C for the 2600mAh cells that I’m using).
I fitted two sets of two-way 18650 battery holders, one into each side of the battery pack, which seemed to fit perfectly – I soldered the series strings together including a 3A 102C thermal fuse in the centre of each pair of batteries, added a set of flying leads for ground and the B+ of each cell for connection to the balancing board.
I cut out some of the plastic in the centre of the pack to accommodate the new charge PCBs, then roughed up the plastic where the battery holders would sit with sandpaper to improve grip, and hot-glued the battery holders securely into place.
I tested & kept the original thermistors which are used by the A4 charge management IC to monitor battery temperatures & automatically stop charging if they get too hot.
I then wired up the battery holders and pack edge connector to the charge boards as per the below connection diagram, including the balance connections to the B+ of each cell, a 1N5822 40V 3A Schottky diode across the input and output of the EV26124 board, and a 3.15A slow-blow fuse in a 5x20mm fuse holder on the pack input/output.
I also secured the original thermistors & new thermal fuses to the centre of each pair of cells using electrical tape, being careful to leave as little of an air gap as possible.
With the battery pack rebuilt, I tried charging it up using a current-limited bench DC PSU set at 21Vdc output – the power and charge LEDs on the charge board lit up, which the EV26124 datasheet shows is as expected when the cells are being charged. The pack seemed to be drawing approx. 150mA at 21Vdc, or 3.15W. I left this set up until the pack was fully charged (charge LED turns off), monitoring it carefully whilst running to make sure that the cell temperatures and voltages looked OK, and that the cells were being properly balanced.
Now that I was happy that the pack was working as expected outside of the machine, I put it into the A4 with the PSU disconnected, and switched the A4 on – it booted up OK. I left the A4 on until it flagged that the battery voltage was low, and it cut out about half an hour after – it ran for several hours on the single full charge.
Parasitic load can confuse chargers by bogging down the battery voltage and sapping the charge current, so if possible the A4 should be switched off whilst charging, allowing the cells to reach their target voltage & current saturation point unhindered.
I then connected the PSU, and the A4 immediately started fast charging the pack – this continued for several hours, until the A4 dropped back to trickle charging for some time, then finally indicated that the pack was fully charged. When I turned the machine on, the state-of-charge indicator LCD on the front edge of the A4 read 60%, and the indicator bar on the desktop taskbar showed around 3/4 full – this seems to be because the pack isn’t getting fully discharged by the A4, so its full capacity is not being used (see potential improvement points below).
After a few more discharge/charge cycles, it still seemed to be working well.
The battery light turns solid amber when the battery is being fast charged, solid green when the battery is being trickle charged, flashing green when the battery is charged, and flashing red when the battery is nearly fully discharged – this also shows a pop-up warning on the screen recommending that you plug in a PSU.
The battery indicator on the taskbar and an LCD on the front panel shows the approximate state-of-charge of the battery, as well as an arrow for whether it is charging or discharging.
While this seems to be a working solution, further improvements could be made:
- It would be possible to use lithium-iron-phosphate (LiFePO4) cells instead of lithium-ion cells, as these are more stable and easier to balance, particularly at low charge and discharge currents – these are also available in 18650 packages, but would require a different BMS board designed for LiFePO4 batteries, and would require more cells to reach a nominal voltage of 14.4V (LiFePO4 cells typically have a 2.5-3.6V operating range with a nominal voltage of 3.2V, so a pack comprising five cells would have a 12.5V-18V operating range with a nominal voltage of 16V).
- Phil recommended using a low-power comparator and a suitably-rated MOSFET to automatically switch in the output from the balancing board when the charge voltage is disconnected, rather than using the Schottky diode – this would still have a small diode drop from the MOSFET, but may be more efficient.
- It would be possible to design a custom PCB to fit the A4 battery pack, based on the MP26124 reference design – this could also include the fusing and comparator/diode, and cell protection and balancing onboard too, maybe even additional thermistors. A similar charge board is already available on GitHub.
- As it currently stands, the A4 automatically cuts out the battery pack at exactly 14.0V, presumably to prevent over-discharge of the original NiCd cells – I’m not sure yet whether this is a hardware problem (i.e. the battery voltage feedback is incorrect, which is confusing the A4 and it’s cutting the battery out too soon) or whether this is a firmware feature which is included in the version of the battery management software on the Wizzo v315 A4 expansion ROM image that I’ve fitted. What this means is that although the cells are being fully charged to 4.2V as expected, they are only being discharged down to approx. 3.55V ((14.0 + 0.2) / 4), whereas they can be safely discharged down to 2.75V. As a result, a large portion of the usable capacity of the cells is not being utilised. This requires further investigation.
Reassembly and Testing
Reassembly of the A4 is just the opposite of its disassembly.
Once the A4 was reassembled, I also fitted a replacement for the missing brightness control knob, which the owner provided for me.
After its repair, the A4 seemed to work great!
One thought on “1992 Acorn A4 Laptop Repair, Restoration, & Modifications”