Burroughs B7800 Mainframe

My involvement with Burroughs and the mainframe worlds was from late 1985 through to 1990 when by employer ended its usage of large mainframe systems and replaced it with various minicomputers and migrated the legacy databases. There I labored under the steady and true gaze of the great Mr Bruce A. Seaman one of natures gentlemen and the greatest engineering manager that I have had the privilege to work for.

In that time I performed engineering maintenance functions on the mainframe and its peripherals. Sadly very little of it remains, no documents from that time have survived and only a few relics and only one museum piece remains from that era.  This was well before the days of the web, near infinite disk storage and digital cameras , so all thats left are rapidly fading memories.

Memoirs of a mainframe engineer,
my time as a Burroughs B7800  "field" engineer at Monash University

My souvenier of the Burroughs B7800 CPM. In this design, every register and flip flop was brought out to the front panel. Each LED had about 16 different meanings depending on the switch settings. The display was allways on and looked magnificent. You would walk amongst rows of these panels, it was like the imagined space ship main battle station of the Death Star.  It was a great crowd pleaser at open days and I am sure that this display made the job of Burroughs salesmen so much easier. As an engineering tool, it was usefull but not essential. We had a B800 mini computer whose only job was to read the front panel and tell us that led 27, row18, panel 2 was off. We would walk around and sure enough the led was in the wrong state. Very powerful, wise and all knowing it made us engineers feel!

The above image shows a plane of B5500 magnetic core memory, image courtesy of Mr Alan Thorne who actually used it ! See the irregular spacing of the rather large toroids, for this core plane was hand threaded. The higher density core plane, elsewhere was machine threaded and is more regular.
These core planes were made in Hong Kong, presumeably by the poor folks that had escaped from communist China and would do this highly detailed and tedious work for nothing more than a beating, a handfull of rice and a vague promise that things might get better. Click on this image to get the full resolution.

The console from a Burroughs Disk Pack Driver Controller. There were two redundant controllers in every B350 controller cabinet. The controller was a B700 minicomputer in its own right. The console could be inserted and removed while online. It was removed to discourage "ge-finger-poken".  I used these regularily to perform disk pack maintenance after a special controller program had been loaded either online or from a Kansas City standard cassette. There was a hard wired writable control store ram diagnostic in this device.

This card was one of about a thousand that came from the memory cabinets. The memory chips are intel  4kbits by 1 .  Our Burroughs was maxed out with 8 Megabytes.  The Burroughs had very good ECC and could tolerate a small number faulty chips. We spent a good portion of our time replacing many of these rather forgetfull memories.

Core memory plane from a Burroughs DCP, one of about 20 such boards in every DCP

 

click and zoom in on this image

A PCC cabinet line driver interface card

Console card from the maintenance processor

This mainframe relic is on public display in biulding 28, the Mathematics Biulding at the clayton campus of Monash University under the stairwell at the main entrance. Be sure to visit it if you have  business at Monash. Its usually open during the annual open day as well.

A view of the Burrough B7800 mainframe and backplane. The backplane was a robot wire wrapped marvel and maintenance nightmare. All done with twisted pair wire, all colored alike. We we had to perform modifications ( hardware patching! none of this download new firmware nonsense) we used yellow wire wrap wire to help remind ourselves where and what we had done. Wire wraps could be up to four levels deep. Dead wires were left in place because it was impossible to disentangle them. Worst faults were intermittant ones whereby insulation would be cut through by adjacent sharp wire wrap pins. The small floating PCBs are clock distribution modules.

The awesome backplane. Source of many intermittant faults and surely the last word in heroic complexity.

Some views of the PSU  There were about 40 of these in each mainframe cabinet. They took a raw DC input of 120Volts and a pioneering switchmode power regulator transformed it to 4.8V or -2.0V. The two coils visible here are current shunts for automatic short circuit constant current mode running. You cannot parallel voltage sources, they must have some resistance and these coils of bronze wire provided it.
The primary DC supply was derived from switched SCR rectifiers that were directly connected to large 3 phase step down transformers and these were direct on line.  Everytime the lights blinked we moaned to ourselves about our fates and attended to repair 3 or 4 of these units when the output transistors had been popped by a surge or other transient. We became very good at replacing power transistors in a hurry. A tripped breaker at the front told us the unwelcome news. Often the machine would continue to run with up to six dead PSUs.

     

A sadly empty card cage

The above two images are from one of the hundreds of circuit cards from the Burroughs B7800 CPM and IOM mainframes.  The chip family is the little emulated and not lamented CTuL family that was made on special order from Fairchild and ITT. All the boards were totally wire wrapped with the ICs soldered in place, the chips were never socketed. The wire wrap was two deep per post. The close up shows a clock distribution amplifier, every card had at least one of these. We had to perform chip level repairs on these boards. There were no schematics.  What we had we the so called LOCAL books (Logical and Chip Analysis) which had a typographical tabular convention that specified the logic equations for the board and the complete wire map in tabular form. From test vectors injected by the B800 mainenance processor the machine and the board called be put into a state one clock cycle before the error vector. We traced the faulty logic state back from the front panel, to one or more cards and finally to the chip or wired OR signal. Some horrible wire OR signals had over one hundred terms and could take days  of very pernikity probing and thinking. Intermittant faults were common. When working this way with a suspect board out of the card cages on an extender we would be exposed to the full blast of the underfloor cooling and this was highly unpleaseant, the set point temperature was 12 degrees C.  Cooling failures were potentially diasterous and one of my very first projects was to construct an overtemperature detector and remote shutdown controller.

CTul  was somewhat like ECL ( see my ECL rant   ) but yet completely unlike it. It was a strange RTL-like logic with open emitter output. It was incredibly power hungry, a chip left powered up without a hurricanes' worth of cold air blasting over it could get too hot to touch. I dont believe that anybody other than Burroughs ever used this logic family. They choose to use this logic family very early and only gave it up when the A10 series mainframe was released make of TTL.  The logic family was quite fast. The B7800 master clock ran at 8Mhz.

There were only SSI functions available in the family, but it was sort of TTL compatible and the B7800 used TTL gates for things like bipolar memory. Logic supplies were 4.75 volts and negative 2Volts.  The CTuL gates did not have gain, and that meant that after a logic signal had propagated through 3 levels of logic a buffer or amplifier would be required to regenerate the logic level.  
CTuL first appeared in 1965 and the images below are from an ancient Fairchild manual. It predates TTL by about 10 years. I suspect it was developed to get around Motorala ECL patents.

Historic CTUL data sheet from Fairchild

electrical compliance plate from a type 235 quad density disk pack drive unit. The power consumption was no engineering overestimation it was quite real.  Most of the power consumed was utilized in spinning the 12 fourteen inch diameter platters of the disk pack at mains synchronous speed, the power consumed was mostly the induced drag in air. The air was given a significant centrifugal acceleration and this facilitated both cooling and cleaning the air gap space. The suction induced at the core of the pack would help suck air through the absolute filter that required changing every six months or so.  This particular drive unit had 170 Megabytes of formatted space.

Burroughs chip house code to commercial equivalents list

These pages are burroughs house part number codes to commercial equivalents. I prepared this list sometime in 1987 to aid myself in sourcing replacement commercial parts. Burroughs then liked to charge $50 for a 7400 quad nand gate because it had their  house code on it...and off course, your mainframe wouldnt work unless it had the genuine burroughs part...wouldnt it now!
I have provided this list as a service to hobbyists and radio hams that may come across discarded computer boards packed with assorted TTL chips that are useless unless you can find the commercial part numbers. The list was constructed largely from schematics, order books and sometimes chip manufactures would print the burroughs house code on top of the chips but also print the commercial type numbers underneath!  This list is not available in machine readable form.

Memoirs of a mainframe engineer,
my time as a Burroughs B7800 mainframe engineer
 

Here is my complete scan of the B6700 systems architecture manual

Homepage