Dimond Rings and Read-Only Ropes

Dimond Rings and Read-Only Ropes

October 11, 2012 1:54 pm 5 comments

From 1969 to 1972 I worked for a computer memory company out in Phoenix, called Quadri Corporation.  One of its major product lines was rope (sometimes called core-rope, transformer, or wired-contents) read-only memories.

This was back before semiconductor read-only memory technology had been perfected, so designers were seeking ways in which data could be permanently stored and quickly accessed.  The basic idea was to create a transformer with a multi-turn sense winding for each bit, and a wire for each word in the memory.

Crossbar 5 AMA Translator

Crossbar 5 AMA Translator

Word wires went through the transformer for each bit that was to read a ONE for that location, and bypassed each transformer that was to read a ZERO. Readout was accomplished by pulsing the desired word wire, and detecting which sense windings developed an induced voltage in them.  Note that the magnetic cores didn’t store information;  they just helped couple magnetic flux from the word wires to the sense winding.

Although none of the engineers told me this at the time (in fact, they may not have known), I later discovered that the original idea was due to Thomas L. Dimond (that’s correct; no A), a designer at Bell Laboratories, who created the Dimond Ring Translator for Bell’s #5 Crossbar Switch.  It was used in the 5XB’s Automatic Message Accounting system to determine the phone number of a subscriber making a long distance call.

There were ten (one each for 0-9) rather large magnetic toroids for each of the four digit positions in an exchange number.  When a telephone subscriber was “provisioned,” a wire connected to the switch terminal used (up to 10,000 per exchange) was threaded through the four toroids corresponding to the subscriber’s phone number.  When a toll call was initiated, that wire was pulsed and the calling number was recorded.

Test ROM for Apollo 11 Lunar Module

Test ROM for the AGC

The majority of rope ROMs were used to store microcode for computers and their peripherals, or in character generators for displays.  For example, Quadri built a character generator ROM for CBS Laboratories first vidifont machine, used to write characters on television images.  The technology was also used to store the programs used in the Apollo Guidance Computer for the moon missions.

Access times for rope ROMs could be made in the 100-150 nsec range, and of course the memory contents were incorruptible, which made them highly desirable.  Some rope ROM makers used E or U shaped magnetic structures into which the woven word wires and the sense windings could be placed, with an I shaped “keeper” installed afterwards on the open end of the E or U to complete the magnetic path.  This allowed automated weaving of the wire bundle, but suffered from erratic sense voltages due to the difficulty of maintaining positive contact between the keeper and the main magnetic core.

U core with keepers

U core with keepers

Quadri preferred using magnetic toroids to avoid this problem, but constructing them was labor intensive.  First the multiturn sense winding had to be hand threaded through the core, then the word wires had to be woven by hand into the right pattern.  This was done by (primarily female–leading to the nickname “little old lady,” or LOL memory) operators using fine wire and tiny elongated bobbins, guided by “stringing sheets” marked with which cores stored a one for each word wire.

Needless to say, the process was also quite error prone, so after a rope was wired, the first one was subjected to multiple manual reads for comparison with the source documents containing the desired program.  Once determined correct, this golden master was used in an automated comparator to test all subsequent units in the production run.

Continued in Part 2: Jacquard Looms and Rod ROMs >


 

5 Comments

  • I was privileged recently to tour the Capehart Collection, a private telecommunications museum featuring old telephones, old switchboards, and old central office equipment. Besides the wealth of telecommunications history, Don Capehart’s memories and anecdotes contributed greatly to the tour. Among his artifacts he had a Dimond Ring Translator.

  • John Chiu

    Great info. For the AGC rope memory, was the sense winding a single turn or a multi-turn structure. From the photo, it appears only one turn was used. The output must be in the “dirt” – low amplitude near noise level. Either the design engineer use high current pulses in word line or high permeability ferrite and not powder iron cores. Can you comment.

  • Phil Ryals

    I’ve never seen an AGC memory up close, but I’d be willing to bet that the sense winding had multiple turns. There’s just no way to find the signal from a one-turn primary winding without many turns in the secondary.

  • I agree, Phil. I cannot see a way to do this where the secondary (sense) winding does not have considerably more than one turn.

    • John Chiu

      From what was published by MIT Instrument labs, the rope ROM in Apollo program did not use linear transformer technique. To maximize flux change or voltage change, the core (square hysteresis) was switched as in the erasable core memory module.

      To make a long story short, the Apollo rope ROM used one turn as sense winding. One turn sense winding or sense loop is a lot more manufacturable so I can see why Apollo team chose this approach. Please correct if I have erred in assessment.

      What was really interesting is the fact the Apollo rope rom implemented the address decoder with additional single turn windings in the same core!!! None of the expensive Fairchild NOR IC gates were used.

Leave a reply