From: "Dave Babcock" <dave@kaleidosoft.com> Date: Sat, 9 Nov 2002 01:12:44 -0800 Subject: IBM 1620 Group: Fw: IBM 1620
To all, Wayne contacted three other members of the original 1620 design team and one has already sent me more information. See below. I responded to his email with answers to his questions and more questions for him. Joe, there's a reference to another patent we should get. I'm in the process of trying to organize the information we have and where there are big holes in our knowledge of the 1620's history. We have people who can possibly answer them. Please send me suggestions for what to ask them. Thanks, DaveB Wayne Winger contacted me about your history project on the 1620 CADET. I got on your website and was dumbfounded to find you had so much information already - it never had occurred to me to look for 1620 data on the web! I'm Bob Jackson - find I'm listed on your website, but as Wayne points out in his emails to you there were several others in Poughkeepsie working on the project, some of whom he has identified. I joined the project to work on the engineering functional specifications along with a woman engineer - Anne Deckman. The two of us had competed a first pass at the specs when she was killed in an auto accident (we had been working O/T and she was on her way back after dinner - it was devastating to all of us!). After Anne's death I continued on the project and finished writing the specs, including descriptions of the console operation, completing it just before the machine was announced October 22, 1959. Your website lists an R F Dean, Endicott Advanced Systems Planning, as co-author of the 1620 functional specs. I've never heard of him (nor has Wayne) and no one from Endicott was involved with us. I wonder if he was involved with the 1401? You should research this and correct it because it has to be incorrect. I was not personally involved in developing the original table lookup idea for the "Add" instruction and I'm not sure where it came from but would guess it was Jim Brenza's. But I was involved with the work defining the table lookup "Multiply" instruction. The acronym CADET I think came from "Computer with ADvanced Economic Technology" but some wag in San Jose said it was "Can't Add, Doesn't Even Try" and the name stuck. I presented a paper at the Western Joint Computer Conference at the Jack Tar Hotel in San Francisco May 3-5, 1960 entitled "A Built-In Table Lookup Arithmetic Unit". Co-authors were James G. Brenza, William H Rhodes, Wayne D. Winger and me. I can send you a copy if you are interested. Quoting from the "Conclusions" of the paper I gave: "Elimination of the expensive adder circuitry heretofore required was made practical by using many of the components needed for other operations. The low cost, high performance core storage provided the table area and speed required for efficient arithmetic operations. The MARS array and registers, too, were required for other uses by the instruction set. That being the case, a few extra components and a logical sequence provided the same final result as a conventional adder, but at reduced cost". "Multiply" has a US patent # 3,049,295 "Multiplying Computer" issued August 14, 1962 with Rhodes, Brenza, Winger and me listed as co-inventors. It's a huge document - I would think you could get a copy of it from the US Patent Office if you'd like to have it. If a application for a patent for the "Add" instruction was ever filed, it never issued as far as I know. The 1620 that was announced did not have a "Divide" instruction. Can't remember why, but maybe it was to save memory space. Again quoting from the "Conclusions" of my paper: "No division hardware was provided because the use ratio of such an instruction to others was found to be small. Since other features of the 1620 provided convenient subroutine programming, division was included as part of the general program package of mathematical subroutines ... included with the system." I seem to remember that it was done with successive subtractions. Never-the-less, Bill Florac and I defined a hardware "Divide" instruction and obtained US Patent #3,239,654 "Dividing Computer" issued March 8, 1966 with William A. Florac, Winger and me as a co-inventors. I think San Jose eventually added a hardware "Divide" instruction, but I don't think it was the algorithm Florac and I defined. The Poughkeepsie Development Laboratory took the machine through announcement on October 22, 1959, but it was released to San Jose as Wayne pointed out. San Jose had a competing proposal (don't remember it being called CADET), but our machine won out. Wayne, Bill Rhodes and I (maybe others) were present at a meeting John Hannstra (sp?) General Products Division President(?) in White Plains when the decision was made. Hannstra said something like "the San Jose version is top of the line and not expandable, while your proposal has all kinds of expansion capability - never offer a machine that cannot be expanded". It was difficult for us to give the machine to San Jose, but we understood the reasons and it really didn't belong in Poughkeepsie where the large mainframes were being produced. I don't remember any competition with the 1401 since they were aimed at business applications while CADET aimed at the scientific. The San Jose group made several changes which puzzled me. If you look at the console "Memory Address Register" (right-most display just before the "Memory Address Register Display Selector" knob) in the announcement photos you will see that the memory addresses are displayed just as you would write a number (i.e.: 12,345) but San Jose turned the display 900 so that the tens-thousands digit is at the top, thousands digit below it, hundredths digit third line down, etc. This is the way I remember the machine being shipped. That never made sense to me and I wondered if the Human Factors engineers ever were involved. Also in the announcement photos the Control Keys and Signal Lights were on the same panel as the other displays, but on the shipped machine they were part of the desktop. I remember someone in higher management (maybe Hannstra) not being convinced that we could make core memory function in this small machine. So another engineer was loaned to the group, Gerry Ottaway, to devise a backup proposal using magnetic drum memory. The design never got very far since the core worked fine. But it was a cliff hanger just before announcement. As Wayne said, the full working model was submitted to the Product Test Lab which had the final say on whether the machine would perform as designed. The core memory failed repeatedly. At the last minute our people found that the fan in the gate was malfunctioning causing spurious pulses to be picked up by the core. After that we were home free! Many of the announcement photos were taken here in Poughkeepsie showing an professional model and me at the console in a posed engineering environment. I can send these to you if you are interested. Do you have the IBM 1620 Data Processing System Reference Manual A26-4500-2, 71 pages, dated July 1961 - much of the text is what Anne Deckman and I wrote. That's it for now. Let me know if you have questions - I'll try to answer them.
Go to messages for November 2002 or latest