The System/360 power supply restoration project

 
I recently visited my friend Robin Whittle*, and in his backyard discovered a power supply unit from an IBM System/360. It was just standing out there in the weather, rusting up and collecting dirt and leaves.

My recent enthusiasm for preservation of historical computer hardware prompted me to ask if I might have it, a request that was granted. There is not much left these days of the 360 series of machines, and this power supply may well be the ONLY part that remains of the particular machine that it came from. So I thought it would be a worthwhile, interesting and satisfying exercise, to take this sorry piece and restore its former dignity.

The size of the unit is 150x150x515mm (6x6x21"), it weighs about 15kg (33lbs), and its nominal output is 12 volts at 18 amps. It was built about 1968. 

After performing some useful function in the role it was designed for, the System/360 that housed the supply was scrapped and the unit surfaced in Sydney, where it was bought by Robin. Around 1981/82 it was once more gainfully employed powering "chaser" light displays in front of the musicians in "Dean Richard's Hot Half Hour" - a 10 to 12 member Big Band which used to play regularly at "Inflation" in Melbourne, amongst other places. At some point the supply failed, resulting in it's retirement to Robin's back yard, where it lay waiting for a warm and caring home.

The design is a straightforward linear regulator. The output of a large transformer with a centre tapped secondary is rectified by two stud mount diodes and filtered by 3 medium sized electrolytic capacitors. The regulator section is mostly contained on a single SMS card, apart from the seven series pass transistors which are mounted an a large aluminium heatsink. Some of these supplies also had a second SMS card to provide crowbar overvoltage protection, but this one doesn't. See the Schematic diagram below.

The regulator circuit requires a bias supply of a few volts higher than the output voltage. The bias supply on this unit is provided (somewhat inefficiently) by the voltage drop across a large power resistor (0R1 90W). This is in contrast to some other SMS power supply designs that instead had a separate small secondary winding and rectifier/filter for the bias supply.

Click the images below for enlargements
The unit as it was when I acquired it. 
(it was dirtier than it looks in these shots).
The innards.
The seven IBM type 108 germanium power transistors and heatsink. This is actually a giant Darlington transistor, the rightmost transistor was driven by the the regulator card, and it in turn drove the other six in parallel.
Here the 0.1 ohm, 90 watt power resistor that provides the regulator bias supply is visible.
The schematic diagram of the power supply.

 

The dirty bits after the unit was disassembled.

 

And then after a good clean up.

Notice that the power transistors have been replaced with cleaner ones. (One of the others was dented, all were dull.)
Now the supply is not totally original, but the replacements are still authentic parts from the same era.
Interestingly, all 14 transistors were very closely matched, they all had a current gain of around 21.7, to within 1% (@IC=1.85A).
The electrolytic capacitors were gradually acclimatized to the electrostatic stress that is their lot in life by subjecting them first to 1/3 of their rated voltage, then 1/2, 3/4 and finally full rating, stopping for 15-30 minutes at each point.
This is supposed to "re-form" the dielectric oxide layer. This could be a load of crap, but it didn't hurt. 
They were all in excellent form. They didn't leak DC once charged, and they self discharged at a very low rate.

The heatsink was reversed to hide the nicks and scratches. This put the emitter terminal bolt in the way of the base bus strip, shown here lying next to the heatsink. It was replaced by two lengths of heavy copper wire, that were bent around the terminal bolt. The six output transistors share the current equally because of resistors on each emitter. These resistors are actually short, equal lengths of extremely stiff resistance wire.

The reassembled unit, ready to go. I haven't yet dared to power it up.
Performance specs to come when I do.

 
*Whittle, Robin: Wizard of electronic musical instruments, Kombi aficionado and world authority on "The Gentlemanly Art of Spanking the Woman you Love". Robin's other interests can be explored at First Principles

 
 
My RetroComputing home page Intabits home page Send EMail