DEC PDP-8/e Restoration Part 1

Part 1 - Initial Dismantling and Cleaning
Part 2 - Processor Checkout and Debug
Part 3 - Loading and Running Binaries


This pdp-8/e was at one time owned by Hughes Microelectronics (once the UK branch of the Howard Hughes empire and now Raytheon) in Glenrothes where I once worked and was my main lab machine for several years. As mentioned here it was rescued when decommissioned by another employee sometime after I had left and eventually made its way back to me. It originally ran OS/8 from an RK05 removable disk drive which I also have and this is a future goal but the first thing is to get the processor itself going. As of now (July 2014) it is around 40 years old and hasn't been powered up for towards 30 years. Also it has been stored in less than ideal very damp conditions so it seemed a good idea to strip and overhaul it before attempting to power it up again. This is what the machine looked like after a superficial external clean-up and strip-down. The back of the machine has the original configuration listed (strangely only 1 omnibus-worth though the machine has always had 2) but this has obviously been altered over the years. Overall looked in good condition, a testimony to DEC's build quality. Click on any picture for a higher-resolution version.

general view of the 8/e

front view

rear view
note the original configuration label

The machine has an L-shaped cover which slots onto the right-hand side of the chassis and hinges up and over the internals to form an enclosed box. The chassis is all-steel construction measuring around 61cm deep, 48cm wide and  26.7cm high (24x19x10.5 inches) and the machine weighs around 45 kilograms complete. There may be screws or pop-studs holding it down but in this case both of the pop-studded straps were broken from much use and the lid simply hinged over and lifted away. The inside top and side of the lid has a layer of black foam glued to it which is intended to prevent the cards from vibrating around and dampen any resonances. This foam is (in)famous for its habit of decomposing and distributing little bits of itself everywhere so I scraped it off and vacuumed as much of the debris out of the machine as possible. I took photographs to log the original positions of the various cards then removed and stacked them separately for later. Note that some card sets, particularly those such as core memory or rk8e interfaces should be retained as sets - in particular core memory stacks are set up during diagnostics so are best kept together. The machine physical design uses an early form of hardware bus, the omnibus, which carries all of the main signals and into which all of the cards are slotted. The cards can be very tight in the omnibus especially if they have not been removed for some time and if they are handled too roughly it is possible to break some of the omnibus mouldings; it's necessary to ease them out with firm but gentle side to side movements. Some of the card sets for instance processor, memory, rk8e/rk05 interface have additional over-the-top connectors (H851) which carry inter-board signals which are not distributed via the omnibus and these fit just as tightly - they are identical and there is no need to identify them separately. I vacuumed the omnibus backplane thoroughly to remove black foam debris and dirt - quite surprising just how much had collected in over the years.

inside view with the top cover raised

side view of omnibus and cards
note the top connectors coupling some cards

omnibus with cards removed
note that the front panel occupies a card position
note also the M935 connectors coupling the omnibus modules

The front panel assembly is retained by 4 crosshead screws behind the sides. There are 3 screws each side and the extreme upper and lower are the ones to remove. Before removal the display function switch knob needs to be removed - it is retained by 2 small hex screws and the shaft has no locating flats. There is no need to try to remove the switch levers. The plastic front panel is simply slipped into the frame guides so a hand should be held underneath to prevent it from dropping out as the frame is removed which it does easily. There will likely be a Velcro pad gripping the back of the panel (this machine certainly had one) and it will need a light tug to separate it from the electronics card then lifted over the switch paddles. The programmer's panel card is fronted by a laminated metal bar which is intended to collimate the light from the panel bulbs. It is held in by 3 (or 4?) screws - the two remaining screws at the sides and a plastic screw through the centre of the card. There is a further plastic screw towards the left of the pcb which has had its head removed to make disassembly possible and there is no need to disturb this. Probably this is a factory bodge - I have never seen one with a complete screw here and it would certainly be a nuisance to remove. Both of the plastic screws carry a tubular metal spacer between the front panel and the card with insulating washers which are easily lost. There are 2 spade connectors at the left side of the pcb which need to be removed - yellow wire on top, blue wire on bottom (but the pcb is marked if you forget). The programmer's panel card itself is plugged into the first slot on the omnibus and had likely never been removed before as it was extremely tight.

front panel and surround

bulb guide mounted on programmer's panel card
note Velcro patch to retain front panel

programmer's panel card removed from omnibus

The omnibus backplane has 48 power and ground lines and 96 signal lines all common to all omnibus slots. The omnibus itself is retained by 4 hex bolts at the ends and located over threaded posts on the chassis (note there are other bolts mounting the omnibus sections to the pcb which should not be removed). The hex bolts were only moderately tightened - I don't know whether this was usual or the result of vibration, but there is no real need for over-tightening as they carry little load and there is a risk of damage. This 8/e has 2 omnibuses linked by 2 x M935 omnibus connectors but it is equally common to have only 1 for smaller builds. Power to the omnibus is carried by wiring harnesses from sockets on the power supply to spade connectors on the end of the omnibus, there are separate sockets and harnesses for each omnibus. There's no real need to note these wire colours as they naturally fall into place but it's good practice. Note that each omnibus is comprised of 2 halves mounted together on a pcb. This is flexible and moderately heavy so needs to be handled gently to avoid generating any bad joints. There were further sheets of black foam loose but sticky underneath the omnibuses - don't know why - possibly further sound damping - also requiring scraping off and vacuuming.

1 omnibus with power wiring disconnected

plan view of omnibus

this omnibus was built in 1973

The next step is to try out the power supply - the pdp-8/e uses the H724 which is an extremely heavy (around 20 kilograms) linear supply mounted along the left side of the chassis. It is key operated and a 3rd position on the keyswitch disables most of the front panel operations for security purposes. (The key was originally manufactured by Chicago Lock, now CompX, as ACE number 7654321:5173757 and known by DEC as number XX2247 - it was used in many DEC machines). The front panel spade connectors (blue/yellow wires) were taped up to prevent accidents and the omnibus harnesses removed (very tight plugs, probably never been removed before!). Note that there are 2 jumpers (one each for live and neutral) on the back panel which need to be in place to power the supply up. The supply has an over-temperature thermal trip which is resettable by a pushbutton on the top if it has been triggered or mechanically disturbed. For the first trial I powered it up from a variac to bring it up slowly as it had been sitting for so long (though I remain to be convinced that this helps significantly with rebuilding the electrolytic capacitors however it perhaps minimises any resulting explosions ;o)). The power supply is enabled via a 24 volt relay (K1) so unless this is shorted out it isn't possible to bring the supply up completely slowly. In this case I brought the variac up enough to latch the relay then quickly reduced it again to as low as would keep the relay latched. First observation was that one of the cooling fans wasn't turning although it spun freely when coaxed round. This is a real nuisance because the designer of this power supply had a bad day and buried the fans inside the bowels of the supply making replacement a right royal pain-in-the-ass! Meantime I loaded the +5v and -15v rails with appropriate headlamp bulbs to give them around 50% load and brought the supply up over a period of about 24 hours. Since it's a linear supply it works fine from zero to maximum load but it makes sense to measure performance with a representative load. There were no surprises and the rail voltages seemed just about right (only the +5v is well regulated, the +/-15v are supposed to be balanced but are not so well regulated and the rest doesn't matter much) and low hum levels on the 'scope so possibly only the fan to sort out.

While I was working on the power supply I found that switching off the PDP-8/e frequently tripped the RCD on my workshop supply. No doubt the mains filtering was designed long before such mains protection devices became common and there's the occasional surge to ground. Luckily my workshop was wired with this in mind and has a secondary supply which bypasses the main RCD and has its own local RCD. This supply can be conveniently used for these older machines and if and when they trip it it is easily to hand to reset.

check the US colour code for UK mains wiring

power supply in situ - fan (intake) side

power supply in situ - business side

DEC's documentation is legendary. I have a fair number of original manuals but in any case there is a substantial number on Al Kossov's 'bitsavers' site - there are 1 or 2 mirrors but try  or google for 'bitsavers'. The manuals cover all aspects of the PDP-8 - circuit diagrams, theory of operation and troubleshooting as well as installation and routine maintenance. Thanks to the popularity of the PDP-8 and the efforts of a large number of enthusiasts much of the original documentation and software is still available today and though inevitably some has been lost for ever, from time to time an announcement in the newsgroups signals another discovery.

Very annoyingly, the power supply access cover is held by lots of small countersunk screws some of which are hidden behind the sides of the machine chassis so the supply needs to be removed if it is to be worked on. The power supply is held into the chassis by 2 screws at the side in front and 2 screws at the bottom at the back. It is extremely heavy and a snug fit but comes out OK with a little sliding and coaxing. The grill at the business side is held by lots of the aforementioned small screws which can all be taken out straightaway as they don't hold anything else in place. The fans and heatsinks form a single assembly which is held in place by the 8 screws holding the fan guards. It's safe to just take them out as the assembly is snugly located and not about to go anywhere. I turned the power supply fan-grill-side downwards to work on it and got some pieces of packing ready so that things can be moved around later without straining the wiring. There is a 10-way Molex connector at the top of the assembly which needs to be unplugged - very tight and awkward to lever out. I found it necessary to lever out the extreme right-hand heat sink as shown because the wiring was too tight and I didn't want to try to unsolder things down in the guts of the thing. (The wiring is heavy duty and well wrapped around and soldered - good engineering practice but difficult to undo!) With this held out of the way the rest of the assembly could be jiggled out, clearances are tight. There are 2 little pcbs located just underneath the assembly - they are plug-in and I removed them first in case of damaging anything accidentally if something gave way suddenly.

The fans themselves are mounted directly onto the heatsinks and there are 6 screws now which hold the heatsinks to the fans (the two bottom centre mounting holes are unused). Some are slightly awkward to get at through the transistor wiring and you need to be careful not to lose any of the non-slip washers or spacers amongst the wiring. Things now look as shown with heatsinks and fans separate. Somewhat surprisingly I found that one of the wires to the 'faulty' fan had corroded through inside the insulation (ptfe I think) and the fan itself was OK. I checked it out fairly thoroughly as this isn't a job that wants doing too often. I had already bought a Papst equivalent for the job (115v, ball-bearing, approx 20w) so it will do as a spare though these fans are seemingly quite long-lived.

power supply dismounted - business side

access cover removed

part of heat sink assembly removed
due to tight wiring

heat sink and fan assembly removed

fans dismounted from heat sink assembly

power supply reassembled - fan (intake) side

Reassembly is a "simple" reverse although the mounting screws are very awkward to refit, access being at an angle through the transistor wiring. Note that some of the fan screws are held by spire nuts and some by threaded spacers. It's fairly obvious which go where. The fans assemble to the heatsinks and the whole assembly slides back into place taking care not to strain or trap the wiring. Lastly the  fan-guards are bolted back in place to secure things. At this point I had to replace the right-hand heat sink. This isn't easy due to the way it's mounted and it needed some grease and a good deal of leverage to force it into the bottom plastic mounting clip!

Inside the supply I found that most of the screws and nuts were rather slack, some very slack, no doubt as a result of years of vibration. The main electrolytic bank's grounds are distributed via some of these screws and thick chunks of busbar so it's a good idea to go over everything for tightness before running the power supply up and checking/resetting the voltages. Mine were good enough so I left well alone. Fans are fine - not unduly noisy (not that you'd hear them because of the roar of the huge airflow). I reattached the wiring harnesses and measurements at the output ends on approximately half load at full (220v-240v) input were:
+5v rail - +5.03v
+15v rail - +14.74v
-15v rail - -14.91v
+8v  - +9.72v
DC Volts OK - +4.94v
which seem OK so I left well alone.

After several hours on load the voltages hadn't moved much and the heatsinks were barely warm so all seems well. The fans were slightly warm to the touch but both were the same so this is probably normal (?). Visually none of the main capacitors seemed to be bulging or leaking so there is little else that can be done at the moment. Time to move on to reassembly.

Part 2 - Processor Checkout and Repair

Page last updated by on 16th January 2015