An DEC MM8E core memory module 4K memory for a PDP-8/E computer Part of the Core Memory pages by Douglas W. Jones THE UNIVERSITY OF IOWA Department of Computer Science

Contents

• Overview
• Documentation
• Provenance
• Condition

This module contains 4k 12-bit words. The module consists of 3 boards 10.436 inches (265 mm) wide by 8.5 inches (216 mm) from edge connector to the flat backs of handles (excluding the actual handle). The boards are 0.056 inches (1.42 mm) thick and spaced on 1/2 inch (12.7 mm) centers. The dimensions are specified on pages 8 and 9 of Digital's 1972 Logic Handbook, excepting the board spacing which can be inferred from the dimensions of the H803 connector block on page 379 of the Handbook. Deduct 1/16 inch (1.59 mm) from the board height where there are no handles (determined by measurement of an actual DEC board).

The outer two boards contain addressing and read-write electronics, while the middle board contains the actual core memory. The memory is arranged as a single core plane containing 12 patches of 64 × 64 cores, for a total of 49,152 cores. Each patch is further divided into two sub-patches, each containing 64 × 32 cores. The entire core plane is glued to the surface of the middle printed circuit board.

The core plane occupies an area 6 inches wide (152 mm) by 3.6 inches (91.4 mm) high. Each patch of 4K bits is therefore 1.5 inches (38.1 mm) wide by 1.2 inches (30.5 mm) high, with sub-patches 0.6 inches (15.2 mm) high. The cores themselves are 0.020 inches (0.51 mm) in diameter and 0.005 inches (0.13 mm) thick with a 0.012 inch (0.31 mm) hole. Core measurements were taken with a long-focal length measuring microscope, as shown in the microphotograph to the right.

The X and Y addressing wires run the full height and width of the core plane, passing through multiple 4K patches as they do so. At the edges of the plane, a subset of the X and Y wires are brought out and soldered to printed-circuit-board traces, while the other wires enter a bundle where they reenter the core plane. Each individual X and Y wire begins at an edge of the plane and then traverses every 4K patch before exiting the plane at the opposing edge.

Each 4K patch has a single pair of sense-inhibit wires, each of which traverses half of the cores in the plane, but because they are tied together at one end, they form a single sense line that visits every core in the plane. These enter and leave the plane at the two lower corners, and when not in the plane, the pair is twisted, with all of the twisted pairs glued together into a rope of sense-inhibit wires going out to the inter-board connectors.

A close look at the close-up photos of this core plane reveals an interesting feature, little black blobs hanging from the edges of each 4K patch of memory. These are splices made to the sense lines during manufacture. Evidently, stringing the sense lines was sufficiently difficult that breakage during manufacture was a commonplace.

The card-edge connectors on the bottom of each board plug into the backplane of the computer, whle the connectors on top are used for interboard connections. The two outer boards, of course, make connections to the backplane's address and data lines, along with timing lines. The board supporting the core plane itself only connects to the ground lines on the backplane. The interboard connector blocks, DEC part number H851, are simply short segments of backplane with two backplane slots each.

The G227 Memory Select board reads the 12-bit memory address from the backplane and (in conjunction with some diode arrays on the memory board itself) translates the address into pulses on exactly one of 64 X drive lines and one of 64 Y drive lines. This takes a fair amount of power, as evidenced by the heat sinks on 4 of the transistors on this board.

The G104 Memory Sense-Inhibit board contains the sense amplifiers needed to extract data from the memory during a read operation and also the logic to drivers to inhibit writing for those bits that should be set to zero during a write cycle. All connections to the data bus are on this board. Again, high power required to read and write core memory, as evidenced by the row of 12 large 16.9Ω 1% power resistors across the center of the board.

The sense-inhibit board contains the decoder for the 3-bit memory field, inhibiting all activity on the MM8E memory subsystem if the field is not selected. This allows up to 8 4K memory modules to be installed on one system.

Documentation

For a detailed explanation of the operation of the MM8E memory system, see Section 4 of the PDP-8/E Maintenance Manual, Volume 1. The 1971 and 1973 edition contain essentially the same material for the MM8E.

The Maintenance Manual states that the MM8E has a cycle time of 1.2 microseconds and 1.4 microseconds. Figure 3-4 of the maintenance manual suggests that instruction fetch is a fast cycle, while execute is slow and some indirect memory references require a slow cycle. Slow cycles are required for any operations that involve the possibility of a read-modify-write cycle, as opposed to a read-refresh cycle.

Provenance

Michael L. Ardai found this memory module in a junk pile in Massachusetts and sent it to me.

Digital Equipment Corporation parts from that era were routinely stamped with a variety of quality control stamps, many of which are cryptic, but most boards are dated. Here, we have found the following date stamps:

• G104: JUL ?? 1971 (partly obscured by remains of a sticker)
• G227: APR 21 1972

The G619A Planar Stack Board has the usual illegible DEC quality control stamps, plus an elaborate label from the OEM, Ampex:

There are two identical Ampex warranty seals on two of the 8 screws holding the plastic cover over the core plane:

The seals were broken to take some of the close-up photos of the memory plane shown here.

Condition

When Michael Ardai sent this memory module, it was obviously damaged. The connections to each of the 12 sense-inhibit line are made through 4-pin black boxes along the top edge of the card. 2 of these boxes were missing, and one was hanging by one wire.

The surviving broken black box was sanded down in order to determine what was inside. It contained a ferrite transformer core wound with two 10 turn windings. Comparable ferrite cores were located and wound to fabricate replacements. These are visible to the left end of this photo, with the remaining functional black boxes extending to the right. The result worked and remains functional, although the replacements are obviously somewhat more fragile than the expoxy potted transformers they replace.

Of the 8 nylon screws that held the plexiglass cover over the core plane, 4 snapped when they were removed. It appears that nylon screws, when screwed sufficiently tightly into their nuts for a half century, tend to stick tightly enough that they break instead of unscrewing. Fortunately, exact replacements or these 2-56 nylon screws are widely available.