The University of Iowa's DEC PDP-8

A Tour of the System

Part of the UI-8 pages
by Douglas W. Jones
THE UNIVERSITY OF IOWA Department of Computer Science


thumbnail image  
The system.


What follows is a guided tour through the parts of the PDP-8 computer originally purchased by the University of Iowa psychology department in 1965, serial number 85. Click on any of the thumbnail photos for a larger view. This tour serves as an inventory of the parts of the system and an index into documentation for those parts.

The Computer Rack

thumbnail image  
The computer rack.
The computer is mounted in a DEC CAB-8B relay rack, missing the table. The missing table would extend 18 inches from the cabinet and be 31 1/4 inches wide — the 22 1/4 inch width of the cabinet plus a 9 inch overhang on the right side. Black and white photos suggest the table was white. None of the DEC catalogs seem to offer the possibility of buying the machine without the table top. The 1967 Logic Handbook lists the rack alone at $1000. The User's Handbook lists the empty cabinet weight as 225 pounds.

The front rails of the rack are drilled and fitted with 10-32 threaded inserts (probably Rivnuts) on alternating 3 inch and 2.25 inch centers. These spacings are compatible with standard 19-inch relay racks, allowing mounting equipment that is a multiple 5.25 inches in height. 5.25 inches is 3 standard rack units, a unit of measurement dating back to the 1930s.

The side skins of the rack are easily removed. They are held on by gravity and friction. To remove them, just lift up (hard) and pull away from the rack.

At the bottom of the rack is a Caravel intake fan sitting atop an air filter assembly. Comair Rotron still makes the Caravel line of fans.
115 VAC 50/60 CPS 53 WATTS

The filter itself had the following glossy paper labels on the side, but these fell off as soon as the filter was removed from the assembly for cleaning:


Madison Wisconsin
OR SOAPY WATER · · · RINSE · · · LET DRY · · · ·
U.S. Pat. 2,865,466 · Pat. Canada 1959


Research Products Corporation is still around, doing business as Aprilaire and they still sell RP Super Filter Coat Adhesive.

PDP-8 Computer

thumbnail image  
The computer
on slides.
The computer is mounted on slides within the rack. The two halves of the backplane of the computer are mounted on hinges on the top half of the slide-out framework, while the front panel hangs below. Under the center rear of the welded steel frame is a difficult-to-read aluminum label carrying the machine's serial number:

d i g i t a l

At the bottom left of the rosewood veneer front of the computer just above the front panel there are two yellowed pieces of paper taped to the front. They contain typed text, octal memory dumps of a modified version of the RIM loader and a tape copy routine:

 7756- 6014 6011 5357 6016
 7762- 7106 7006 7510 5374
 7766- 7006 6011 5367 6016
 7772- 7420 3776 3376 5356
 7776- 0000 5306
  200   7200  6016  6026  6011
        5203  6021  5205  5200

The computer came with the following hardware documentation:

The User's Handbook lists the combined weight of the computer and power supply as cabinet weight as 250 pounds.

Front Panel

  thumbnail image     thumbnail image  
Front panel
The front panel is suspended from the bottom front of the slide-out framework. It contains lights that indicate the state of every bit in the CPU, and it has switches allowing entry of addresses into the program counter and storage of data in memory. There are two locks on the front panel, one controls the power, the other (panel lock) disables the switches to prevent tampering with running programs.

Large parts of Section C of the User Handbook explain how to use the front panel.

Memory Half Backplane

  thumbnail image   thumbnail image   thumbnail image  
Memory half backplane
With the computer pulled fully forward on its rack slides and the lock on the center top opened, the left rosewood-colored formica door swings to the left to expose the wire-wrapped interconnect of the memory half backplane of the computer. This holds not only the memory, but the console teletype interface, and some other options.

There is a paper label just to the left of top center when seen from the wire-wrap side of the memory half backplane. The label is a pre-printed form with typed in identifying numbers, and it is starting to peel off. The label reads:

08-291-1000 Type 184A

The back (hinge-side) vertical rail of the frame has an aluminum sticker on its upper inside face -- almost entirely obscured by the pins of the wire-wrapped backplane. This reads:

d i g i t a l

The black box filling the center of the upper third of the memory half backplane holds the actual core memory of the machine, 4K words of 12 bits each. The memory box has two labels on its top side, a larger aluminum label giving the serial number and a smaller paper quality-control label:

MODEL:         SERIAL: 51-238-71
BY         DATE 10/25/5

The most likely interpretation of the handwritten date on the quality control label is October 25, 1965, but the writing is cramped enough that there is some doubt about the day.

The theory and high level operation of the core memory is explained in Chapter 4 of the Maintenance Manual:

  thumbnail image  
Map of memory
half backplane
The black box holding the actual core memory is attached at the top center of the outside face of the memory half backplane. This is surrounded by support electronics, mostly packaged on flip-chip modules. To the left of the memory are 12 sense amplifiers to convert the faint output pulses from the core memory to standard logic levels. To the lower left are the inhibit drivers, with their inhibit drive resistors on the upper right. The memory select line drivers are on the lower right. Click on the thumbnail image for a map showing the board layout.

The memory half of the backplane came prewired for three options that could be installed by merely plugging in the appropriate boards.

The Teletype Control - installed
This is actually two devices, an asynchronous receiver and an asynchronous transmitter sharing a common clock to set the data rate. Today, we could call it an asynchronous serial communications port. It was a "standard option", in the sense that while it might have been possible to remove it, this was not done. It is well documented:

The Type 188 Memory Parity Option -- missing
This option added a 13th bit to each word, allowing detection of memory errors.

The Type 183 Memory Extension Control -- missing
This option added a rudimentary memory management unit that enlarged the memory address from 12 bits to 15 bits using a pair of 3-bit registers, one to extend the program counter, and a second used only for indirect addressing.

Max Dietrich, who worked on this machine between 1966 and 1968, recalls being asked, at the end of each fiscal year, what to spend the surplus grant funds on. He always asked for memory. Unfortunately, the cost to add the first 4K additional words was $13,000, which was far more than the grant surplus.

The Memory Utilization Module List in chapter 10 of the Maintenance Manual gives the detailed map of the memory half backplane, identifying what board plugs into what slot and (on a coarse scale) what data flows through that board. This is a giant fold-out page just under 2 by 3 feet.

CPU Half Backplane

  thumbnail image     thumbnail image  
CPU half backplane
With the computer pulled fully forward on its rack slides and the lock on the center top opened, the right rosewood-colored formica door swings to the right to expose the wire-wrapped interconnect of the CPU half of the computer's backplane. This holds the 12-bit program counter and accumulator, the link bit (a single bit of condition code), the arithmetic logic unit, and all of the control circuitry for instruction execution.

There is a paper label just to the right of top center when seen from the wire-wrap side of the CPU half backplane. The label is a pre-printed form with typed in identifying numbers, and it is starting to peel off. The label reads:

08-291-0100 Type PDP8-85

Note that (unless this is a coincidence) the serial number of the machine is embedded in this label.

The back (hinge-side) vertical rail of the frame has an aluminum sticker on its upper inside face -- almost entirely obscured by the pins of the wire-wrapped backplane. This reads:

d i g i t a l

  thumbnail image  
Map of CPU
half backplane
Seen from the outside, the 13 double-height R210 boards on the upper right side of the CPU half backplane hold the 12 bits of the accumulator, (AC) and link (L) bit, along with associated multiplexor logic. The 12 dougble-height R211 boards below these hold 3 registers each, along with associated logic for the memory buffer (MB), memory address (MA) and program counter (PC). The CPU control logic fills the entire left side of the top 4 rows of boards. Click on the thumbnail image for a map showing the board layout.

The CPU half of the backplane came prewired for two options that could be installed by merely plugging in the appropriate boards in the bottom two rows.

The Type 182 Extended Arithmetic Element - installed
This option added multiply and divide instructions to the instruction set of the CPU along with a 12-bit multiplier-quotient register (MQ). The list price for this was $3,500.

The Type 189 Analog to Digital Converter - missing
This option added a single channel low-performance analog-to-digital converter to the machine. Although this was operated by IOT (input/output transfer) instructions, conversion operations stopped the CPU and took over the MB register for use as a scratchpad.

The KR01 Auto Restart Option - missing
This option added the ability to receive an interrupt on power fail, allowing the interrupt service routine to save the processor state in core memory (which is non volatile). On power up, a complementary interrupt would occur on restart, allowing the computer to pick up where it had left off, as if nothing had happened.

The Processor Utilization Module List (UML-E-8P-0-19) gives the detailed map of the CPU half backplane, identifying what board plugs into what slot and (on a coarse scale) what data flows through that board. This is a giant fold-out page just under 2 by 3 feet in size included in the Maintenance Manual.

Type 708 Power Supply

  thumbnail image     thumbnail image  
Power supply
The Type 708 power supply is mounted to the rack below the computer and behind the front panel. It is inaccessible until the the computer is pulled out of the rack on its slides, exposing the front panel of the power supply directly behind the front panel of the computer, and exposing the top of the supply.

The power supply front panel has the total-hours meter for the machine, reading 26879.0 hours when we received the machine. That comes to almost 1120 days or a bit over 3 years. If used for 40 hours per week and 50 weeks per year, this translates to over 13 years of productive use, roughly 1966 to 1979. (The machine was almost certainly used more in its early years and less in its later years.)

This panel is only intended to be accessible when the machine is undergoing maintenance. It has a power outlet where a technician can plug in small equipment, an auxiliary on-off switch, a meter to monitor the supply output, and a knob for the variac that controls the marginal check voltage.

All power supply outputs are somewhat regulated by the use of a constant voltage transformer. Banks of large electrolytic capacitors are used to eliminate ripple. The capacitors were sized to allow continued operation during a power outage lasting 1/20 of a second. The marginal check and memory inhibit outputs of the power supply are more complex, with linear regulators. The supply has the following outputs:

 Output   Regulated?   Capacitance 
-15 no 0.210 ~24 logic
+10 no 0.105 ~12 logic
+/- 0-15variac 0.035 2 marginal check
27-37 linear0.034 2 memory inhibit
27-37 linear 1.5 memory read/write

In restoring the power supply, we found C15 was bad (the final output filter capacitor on the memory inhibit supply). After the bad capacitor was removed, we found that it had a date code reading 6532. 4-digit date codes are frequently composed of a 2-digit year code followed by a 2-digit week code. If so, this capacitor was manufactured during the 32nd week of 1965 -- some time in early August.

The details of the power supply indicators and controls are given in the Maintenance Manual.

The internal workings are documented here, with a focus on the relationship of the power supply to the core memory:

The schematics for the basic supply and regulators are also given:

David Gesswein has extensively documented his experience repairing the power supply for his PDP-8 computer:

Rear Plenum Door

  thumbnail image  
The plenum door
The French doors on the back of the DEC CAB-8B relay rack are cosmetic. With them open, a full-width door is exposed, hinged on the right, with a latch at the top. This door is a steel frame with tapped inserts on 2 inch centers, filled from top to bottom with rack-mounted aluminum panels. Note that the 2-inch hole spacing is not compatable with standard rack-mount equipment, and the rear panels are 16 5/8 inches wide instead of the standard 19 inches. Equipment mounted on the rear door cannot, therefore, be mounted on the front side of the rack.

Type 834 Power Control Panel

thumbnail image  
Power Control
The plenum door panel directly behind the power supply is the Type 824 Power Control Panel. The plexiglass cover over the panel carries this label:

  d i g i t a l  
POWER  834

The outside face of the plenum door has an on-off switch. Inside, there is a power contactor Behind the plexiglass, there is a relay and several terminal strips.

The Analog to Digital Converter Rack

  thumbnail image     thumbnail image  
The ADC rack
The ADC rack is a variant of the DEC CAB-8B or CAB-3 relay rack, without table. It does not precisely match the drawings of either of these, but as with the uncommitted areas of the computer rack, the front rails of this rack are drilled and fitted with threaded inserts for rack-mounted equipment that is a multiple of 3 standard rack units in height. The right side of the rack carries an aluminum University of Iowa inventory sticker at the top:


A-D Converter-Multiplexer

  thumbnail image     thumbnail image  
The ADC/multiplexor
The top of this rack contains an analog-to-digital converter paired with an analog multiplexer allowing the selection between multiple analog inputs.

The top front of the ADC rack contains a control panel with a 12-bit display labeled A-D BUFFER and a 6-bit display labeled MULTIPLEX ADDRESS. To the right is a rotary knob labeled WORD LENGTH adjustable from 6 to 12. The legend on the left reads:


The list price for the Type 138 Analog to Digital Converter was $4,500 and the list price for the Type 139 Analog Multiplexer was $3300. Under the indicator and control panel, behind the French doors on the front of the rack, there are three mounting panels making up a wire-wrapped backplane. The top two panels have identifying stickers on the inside face of the left mounting plate, when seen from the rear. These stickers can only be read with some of the boards removed from the backplane. the topmost mounting panel has the following aluminum sticker:

d i g i t a l
EQUIPMENT   138E-14 

The second mounting panel has two paper labels and an aluminum label sticker, applied one under the other:

08-291-2600 Type 139

AD Converter
08-291-8100 Type 138E

d i g i t a l

These stickers confirm that these two panels are backplane of the ADC-multiplexer advertised by the control panel at the top of the rack. The wire harness linking the control panel to these boards is further confirmation. There is documentation for this device in:

Unknown Logic

The third panel of the wire-wrapped backplane at the top of the ADC rack has a paper label on the outside of its right mounting plate, when seen from the rear. This reads:

  d i g i t a l  
 MOUNTING   1943

The Type 1943 mounting panel was a generic backplane segment. The logic handbook gives a base price of $111 for this panel. The mounting panel includes 8 Type H800-W connector blocks, each able to hold 8 single-height or 4 double-height flip-chip circuit boards, using wire-wrapped interconnection with 24-gauge wire.

The wiring on it appears to be hand done with 30-guage wire, the wrong gauge for these blocks. There is a bundle of green wires from the left rear of this panel to the control panel immediately below. From the labels on that panel, we conclude that this is a locally designed and built controller for a long-gone magnetic tape drive.

Max Dietrich, who worked on this machine between 1966 and 1968, recalls that the lab found a rack-mounted magnetic tape drive somewhere and attempted to interface it to the PDP-8. At the time he left, the interface was not yet functional. This is probably the remains of the interface for that drive.

Unknown Control Panel

  thumbnail image     thumbnail image  
An unknown panel and Tally reader
Directly below the French doors on the panel and above the paper tape reader, there is an obviously home-made control panel that is wired to the unknown logic directly above. The labels on this panel give a hint about its function: switches and buttons on the left marked WRITE PERMIT, END FILE, REV; lights marked READ, WRITE, REVERSE, FORWARD and finally, on the right, a switch marked ON and OFF. This is consistent with what would make sense for a conventional magnetic tape drive, suggesting that the unknown logic and this control panel were all associated with such a drive, now long gone.

Tally 424 Paper Tape Reader

  thumbnail image  
The Tally reader
By default, PDP-8 computers were sold with just one input-output device, a Model 33 ASR Teletype. This included a low-speed (10 byte per second) paper tape reader and punch, but this was so slow that many PDP-8 systems were equipped with high-speed paper-tape handling equipment. This machine was equipped with a Tally high-speed (60 byte per second) paper tape reader. The reader is mounted on a panel just under 8.75 inches high, which is 5 standard rack units. This makes it an awkward fit on a rack designed for 3-U panels. Max Dietrich, who worked on this machine between 1966 and 1968, recalls this reader being added while he was there.

Looking at the back of the Tally reader, one striking feature is the startling number of differentials, 3 in all. When the switch on the front panel is on, the drive motor runs continuously, spinning the differentials. Clutches on one differential allows the capstain to advance the tape in either direction, while brakes on the other two differential engage to turn the take-up reels if there is too much slack in the tape.

The Tally high-speed paper tape has two labels on the back. The smaller appears to be from the vendor, printed in red ink on aluminum and stuck to the center right top of the panel; the larger label to its right is from the manufacturer, printed in hard-to read negative, blue on aluminum:

 SERIAL NO  351/0811  

                      24   PULSE VOLTS
  50   COIL OHMS  

This paper-tape reader cost $825 in 1964, according to the A Fourth Survey of Domestic Electronic Digital Computers. An abridged Tape Reader Manual - Model 424 is available on-line. As of 2013, replacement manuals are commercially available for $35.

The fanfold tape bin below the reader does not look like the fanfold tape bins sold by Tally. It is largely held together by pop rivets, and was probably made by the psychology department's machine shop.

Oscilloscope display and other I/O

  thumbnail image     thumbnail image  
The scope display interface
The French doors at the bottom of the rack hide a multi-function two-panel backplane that, like the ADC-multiplexer backplanes, has factory identification stickers on the inside faces of the left mounting plates, when seen from the rear. The topmost of the two panels has two paper labels above an aluminum one that read as follows, although the left end of the middle label has been torn off:

08-291-3400 Type 34D

//-/91-2500 Type 151

d i g i t a l

  thumbnail image  
Map of 34D
Scope Display

The Module Utilization List for the Type 34D Oscilloscope Display documents the display interface as part of a multiple-device backplane wired for the following bundle of devices: The Type 350B Incremental Plotter, the Type 34D Oscilloscope Display, the Type 750C High Speed Reader, the Type 75E High Speed Punch, the CR01C Card Reader and the PC01 (a variant high speed reader-punch). The labels on this device hint that the Type 34D Display was offered bundled with a real-time clock and perhaps other devices, but inspection of this backplane shows that large parts of it conform to the documentation for the above bundle.

From this, we infer that we have the following device interfaces, plus others that seem to have been added locally and may require reverse engineering to work out what they do.

Type 34D Oscilloscope Display - installed
This includes two 10-bit digital to analog converters for the X and Y deflection inputs to an oscilloscope, plus an intensify (Z) output and one bit of input from a light pen. This is sufficient to allow 1024x1024 point-plot graphics. The Tekitronics RM-503 oscilloscope was the recommended output device (cost in 1965, $625) but it should work equally well with many others. Max Dietrich, who worked on this machine between 1966 and 1968, recalls that the display they used was a Tektronics storage scope -- in that era, the Tektronics Type 564 (or the rack-mounted RM564) was the only likely candidate in that era.

The Oscilloscope Display Type 34D is documented in the following places:

Type 750C High Speed Paper Tape Reader - installed
As built, this supported photoelectric readers operating at 300 characters per second. The modules in the backplane match the map match those listed in the Module Utilization List but there is locally added backplane wiring. The software interface is documented:

Type 75E High Speed Paper Tape Punch - installed
This supported a Teletype Model BRPE High-Speed Tape Punch, able to punch paper tape at 63.3 characters per second. The backplane wiring appears original, but needs close checking. The software interface is documented:

Aside from the I/O bus slots, There appears to be no factory original wiring on the bottom row of this pair of mounting panels. As it came to us, there were loose wires dragging on the floor from this section, and the homebrew wiring here was done with 30-gauge wire, while DEC used 24-gauge wire.

The switches in the blank panel at the very bottom of the rack are further evidence of the extent of local modification: The are both labeled ON and OFF. The left one, in red, says MUST BE ON FOR, with the remainder missing. and OFF. The right one, in green, says MUST BE ON FOR DR RANDALL S PROGRAM . The wiring to these switches appears to be entirely missing.

Rear Plenum Door

  thumbnail image  
ADC back door
As with the computer rack, opening the French doors at the rear of the ADC rack exposes a full-width rear door that is not compatible with standard relay-racks. There is considerably more hardwasre mounted on this door than on the back door of the computer cabinet.

Unknown Interface Panels

  thumbnail image  
The interface panels
The top panels installed in the back door of the ADC rack appears to be the analog interface panel for the ADC. These panels began life as standard 2-unit high by 19-inch rack panels. The gaps between them are the result of the nonstandard spacing of the holes in the rear-door mounting rails. These panels carry a variety of different types of plugs, including binding posts, banana jacks and, BNC connectors. The whole thing screams "I was built by grad students."

Unknown Power Supply

  thumbnail image  
A power supply
The topmost of the two power supplies mounted on the back door of the ADC rack is built on an original DEC panel, but the workmanship makes it clear that it is home-made. The layout and internal wiring of the supply are neat, but the wiring is entirely uninsulated, and the labels on the outputs are done with dymo label tape.

Type 779 Power Supply

  thumbnail image  
Type 779 supply
The lower power supply on the ADC rack carries a paper label as a DEC product:

  d i g i t a l  
POWER  779

The type 779 power supply is documented, in passing, in the Oscillosocpe Display Control 34D Instruction Manual. This references additional documentation in the DEC Systems Modules Catalog (C-100). A schematic for this supply is included in the PDP-7 Maintenance Manual.

Teletype Model 33 ASR

  thumbnail image  
The Teletype
The default input/output device that came with every DEC PDP-8 computer Model 33 ASR Teletype. This was the first device ever sold that supported the ASCII character set (albeit an upper case only subset). It is entirely non-electronic, using mechanism, not electronic logic, to encode keypresses in binary and to decode binary data for printing. It operated at 110 baud, using an asynchronous serial communications protocol with 1 start bit and 2 stop bits per character. This meant that it could handle 10 characters per second. The interface was a 20 mA current-loop interface, not the RS232 interface that dominated later serial devices.

The Teletype label on the lower back, just to the right of the power cord, reads, in negative, black ink on aluminum:

           MODEL 33   CSAM
SUPPLY  INPUT:  115 V.  60~ A.C.  4 AMP.  MAX.

Under the cover, on the right side of the typing unit chassis, just below the carriage motion belt, there is a label giving the rebuild date of this unit, 10/30/74. It is printed in negative, in bright yellow ink on aluminum (a combination that is not necessarily easy to read):


The Teletype ASR 33 contains the following subsystems:

The following documents came with this Teletype:

The teletype manual states that the Teletype should be lubricated with KS7470 oil. This is an old Bell System part number, and it is long gone, but it appears that several white mineral oils are close to exact substitutes for it.

DEC routinely modified Teletypes used with DEC computers. They disabled the "Who are you" answer-back mechanism, and they included hardware flow control. These changes are documented in the Maintenance Manual.