This project was put together with parts from three different vendors: The chassis, case and Power Supply Unit (PSU) came from one vendor and were originally sold as kits; the motherboard, also originally sold as a kit, came from another; and a New-Old-Stock (NOS) keyboard assembly from yet another vendor. This machine is, therefore, a new SOL-20, born in 2024!
The original state of the chassis, case and PSU can be seen in the photos below.
As you can see, no motherboard was present. Instead, there was an S-100 Single Board Computer (SBC) card, which was removed and saved for some future project. The severely damaged and bowed keyboard assembly was also removed and set aside for future restoration; it wasn't needed for this project, as I used an NOS keyboard assembly instead.
The PSU is a Processor Technology original, but differs from the usual SOL-20 PSU. It seems to be a later, cost-reduced revision, in which the 5V rail is also linearly regulated by an LM323K and there is only one bulk capacitor. The latter is mounted vertically, while the transformer is mounted horizontally, to the side of the PSU's cage. This is the opposite of what one finds in the usual PSU revision. A post at the forum of the Vintage Computer Federation did not lead to any positive identification.
The screw holes on the PSU's cage did not match the pre-drilled mounting holes on the chassis (the respective revisions are clearly not mutually compatible). For this reason, the original owner mounted the PSU in a twisted and precarious manner, which I rectified by eventually drilling new holes on the chassis so the PSU is firmly mounted.
The chassis itself is also peculiar. It does not include the mounting holes for the internal side panels invariably seen in SOL-20 units. Instead, it is screwed directly into the walnut external panels, in places that made it almost impossible to remove the screws. This betrays the fact that it was the original owner who drilled the holes, which were therefore not original. I eventually rectified this by, again, drilling new holes into the chassis. Finally, the chassis had built-in standoffs, which I had to carefully saw away with a flush electric steel cutter in order to fit a regular SOL-20 motherboard.
.png)
All other components of the PSU were replaced with modern equivalent, or higher-rated, parts. Notice that the original owner failed to populate the crowbar circuit altogether, an oversight I rectified. The two main electrolytic capacitors I used are high-temperature, long-life units, and were also hot-glued to the PCB to avoid vibrations. In addition, I have correctly wired the missing UHF cable to the back of the PSU, and used a magnetic support to affix the other end of the cable to the chassis.
.png)
The motherboard was a non-completed kit assembly project by the original owner. Many parts were missing. Of the parts present, some were socketed, some weren't. Most solder joints were poorly made and the 40-pin socket for the left UART was of the single-wipe type, which frequently leads to contact issues. The pictures below show the motherboard almost in its original state, after I'd removed all socketed ICs in preparation for washing, added dual-wipe sockets for the left-most row of SRAMs, and replaced the DC-blocking electrolytic capacitor on the video line. Only after performing these jobs did I conclude that the board really needed washing.
.png)
The motherboard was:
- Washed with distilled water and neutral soap;
- Fully recapped (electrolytic and tantalum capacitors);
- Many solder joints were reflowed with flux and new 60/40 solder;
- All potentiometers were cleaned with electronics detergent;
- Existing sockets were cleaned and lubricated with DeoxIT D5;
- Missing sockets were added and defective old ones replaced;
- Missing ICs were added (only modern, tested parts from reliable suppliers were used);
- All existing ICs were tested off-circuit. Two were defective (a 74-series discrete and the 8080 CPU) and replaced; All SRAM ICs tested OK;
- ICs that operated above 40 degrees Celsius were fitted with a heatsink;
- Power inputs were fitted with bidirectional TVS diodes to protect the board against transient power surges;
- A protective conformal layer was sprayed onto the back of the motherboard, to prevent future corrosion;
- A modern personality module was added, as the original was missing altogether.
.png)
The now-restored motherboard was then refitted into the chassis. The latter was also cleaned, treated for corrosion, and polished after having been re-drilled. The PSU's cage was also fitted into the chassis, this time correctly, with proper alignment and bolts adequate for its weight.
The unit was then stress-tested for several hours in this configuration. Upon further temperature measurements, I concluded that several ICs, despite logically still working, were operating at unnaturally high temperatures and would soon fail. I have thus preemptively replaced those parts with modern equivalents, none of which required a heatsink. The one original UART chip also eventually failed and was replaced with an NOS part. The second, missing UART was added (an NOS part too).
The keyboard, despite being NOS, was also restored. All capacitive foam pads had disintegrated and were replaced with modern equivalents. The contact pads were cleaned and lubricated with DeoxIT D5. One discrete 74-series IC (a 74LS74) was defective and replaced. All ICs were socketed, except the memory (which I did not want to subject to heat stress, since it is very difficult to find an original replacement for it). The 74LS-series ICs that were running unnaturally hot (typically, 74LS-series parts should run under 40 degrees Celsius) were preemptively replaced. ICs that naturally do run hot (older, non-LS-series ICs and the PROMs) were fitted with a heatsink. Although the keyboard's memory doesn't run hot, I also fitted it with a heatsink anyway, to relieve it of whatever thermal stress it does undergo, because of the rarity of the chip. Finally, I sprayed a protective conformal layer onto the back of the keyboard's PCB.
.png)
An original Processor Technology 32KB S-100 memory card was then restored, so to be added to the system. All but three of the original DRAM chips tested bad. I thus preemptively replaced them all with reliable NOS, Texas Instruments parts, tested good, with date codes from late 1983. Most logic chips were also running suspiciously hot and were replaced with modern parts. One of the five PROMs was defective and replaced with a freshly burned NOS part. All PROMs (the new one, plus the four original ones) run naturally hot, well above 50 degrees Celsius, and were thus fitted with heatsinks. All on-board regulators were preemptively replaced with brand-new modern parts. All tantalum capacitors were also preemptively replaced with brand-new modern ones. In the pictures below, the upper row shows the memory card in its original condition. The picture on the bottom row shows the card after restoration.
.png)
Here is the complete system after restoration:
.png)
The case was heavily bent, scratched and marked. So it was straightened, sanded and resprayed. The side walnut panels also had extensive damage, missing material, holes drilled in the wrong places, and multiple cracks. The wood was also extremely dry and brittle. In order to restore the panels, the cracks were seamlessly glued back together and missing wood refilled with epoxy putty and melt wax. The panels were then re-sanded, re-stained where required, and impregnated with linseed oil.
Finally, the paper badge was discoloured and abraded by paper-eating insects. It was thus scanned, digitally repaired, and a new copy was printed out on photo paper, which replaced the damaged original. The transparent plexiglass plate in front of the badge was cleaned and carefully polished with Novus, so to eliminate scratches and scuff marks and recover its natural gloss. The results are shown in the photos below. Notice the reflections of the keyboard on the plexiglass.
.png)
High-resolution, non-cropped versions of the photos above can be found in the Github repository of this restoration.