A Type 155 core memory module

Memory for a PDP-5 computer

Part of the Core Memory pages
by Douglas W. Jones
THE UNIVERSITY OF IOWA Department of Computer Science

Overview
Documentation
Provenance
Condition

Overview

The 4K core memory assembly for the PDP-5 is packaged to plug into a System Module backplane; system modules plug into Amphenol 143-22-01 connectors on ³/₄-inch (19.05 mm) centers. The module contains 5 22-pin connectors that mate with the backplane, although the backplane must be modified by removing 6 backplane connectors and replacing two of them with mounting bars to receive the two huge jackscrews that mount the core module to the backplane.

Each System Module backplane segment occupies 3U of 19-inch rack space. It is probable that the core-memory subsystem of the PDP-5 took on the order of 2 backplane segments, but I cannot find documentation for this.

The core stack itself is enclosed in an aluminum box 6 inches (152 mm) on a side, held in place by four 4.5 inch (115 mm) long 1/4-20 UNC studs (comparable to M6×1). The box is mounted on a connector on four 1-inch (25 mm) standoffs; the connector frame is an aluminum casting ³/₄ inch (19 mm) thick 7.5 inches (190 mm) wide, and 4.5 inches (114 mm) high, with ¹/₄-inch square (6.35 mm) bars bolted across the face to restrain 5 22-pin connectors mounted on ³/₄-inch centers. The frame also has 4 alignment pins and two jack-screws on its vertical (short) edges. The jack screws are a total of one foot (305 mm) long, with T-handles at the rear.

The back of the enclosure has 16 Amphenol 143-012-01 connectors on it, two columns of 8 on ³/₄-inch centers. These are 12 pins connectors that mate with 133-012-03 plugs, each of which is mounted on a small circuit board, a Type 1020 Memory Diode Unit. Each of these boards is just under 3 inches (76 mm) wide and 2.5 inches (64 mm) long, with the body of the plug set back from the edge of the board so that board rests against the side of the connector, stabilizing the board when plugged in, and so that the projecting board provides partial protection for the pins on the plug when it is unplugged. The role of the memory diode units is explained in Figure 3-2 of the Maintenance Manual.

The 12 core planes are mounted in 5-inch (127 mm) square frames with 64 X or Y select wires terminating on each edge of each plane, plus 2 sense wires terminating on one corner and 2 inhibit wires terminating on another. Select wires on each edge of the plane alternate so that half of the wires on each side are wired to the next plane up while the other half are wired to the next plane down. Interconnection between core planes is by wire jumpers that are soldered around all 4 sides on 0.1-inch (2.5 mm) centers, so that dissassembly of the planes is all but impossible.

The stack of core planes includes two boards serving as end covers and termination points for the wires, 12 frames, each supporting 4096 cores, plus a thermistor riding very close to the center of the volume. The entire stack is 3 inches (76 mm) thick.

The only way to photograph the core planes without disassembling the core stack was to peer in along the edge, between the grid of wires connecting the adjacent planes. The wider gap between the middle planes gave the best result. Looking at one side of the core stack, the 32 interboard jumpers along each edge of the frame are spaced on 0.1 inch (2.54 mm) centers; since these only connect every other wire in the core plane, the wires in the plane are on 0.05 inch (1.27 mm) centers.

It appears that cores 0.05 inches (1.27 mm) in diameter, 0.01 inches (0.25 mm) thick, and with a 0.03 inch (0.76 mm) hole would fill the available space about as well as the images suggest. Cores 10% larger would be problematic, and cores 10% smaller would be less crowded than what can be seen in the photos.

Documentation

According to the PDP-5 Price List from October, 1969, DEC sold the Type 155 4K memory module for $7500.

Chapter 3 of the PDP-5 Maintenance Manual (F-57 10/64 Digital Equipment Corporation, 1964) gives the overall architecture of the core memory, while detailed schematics are given on later pages. The manual states that one memory cycle (a read followed by a write or refresh) takes 6 microseconds but it goes on to say that 2 of the 6 states during the cycle are not used for memory access, suggesting that the memory could be run with a 4 microsecond cycle in a computer optimized for speed, not cost.

According to the census data in Computers and Automation, Vol. XIII, No. 10 (Oct. 1964), page 48, 55 PDP-5 computers had been sold by Sept. 10, 1964, and 10 more were on order. This is significant in light of the date 9/23/64 on the DEC quality control sticker on the box and the serial number 5 on the EMI sticker.

The Computer History Museum has an almost identical memory module, Catalog Number 102647781, with a quality-control sticker dated 9/24/64, just one day later than this module. The wires visible in the Computer History Museum's photos are not the same colors as the wires on this module, suggesting that the two modules were made by different memory suppliers.

Provenance

This core stack comes from a PDP-5 that was sold as surplus by Utah State University to David Olsen, who scrapped most of the machine, but saved the core memory. He sent the memory to me in spring 2021.

A foil label on the top of the core box says it was made by Electronic Memories Inc. of Los Angeles and that the memory capacity is 4096 by 12 bits. The sticker give a part number, 901952, and a serial number, 5. This low serial number is only sensible if DEC was buying memories from other sources or if DEC was not using memories in the order they were built. Both explanations seem likely. The sticker also says that the core represents SPEC DEC4K-12B which may be a reference to DEC's specification for the memory module.

Below the sticker, there is a quality-control stamp saying EMI 108

On the bottom side of the box, there is a green Digital Equipment Corporation quality-control sticker with the handwritten note Diode OK D H over a large pre-printed O.K. The "inspected by" blank is filled in with illegible initials that could be the D H who checked the diodes or could be R A. Below that, the acceptance date is almost illegible but is certainly 9/23/64.

Condition

This memory seems to be in remarkably good condition, considering its age and its near-miss with the scrap heap. It is, however, missing one diode board. Manufacturing a replacement would be trivial but the missing connector is an expensive part.