Stories
about People and Machines

4 Ton Beast - Daystrom 046 Computer - by Bill Richter (Ham Radio call W5YAR)- sent May 29, 2020

4 Ton Beast - Daystrom 046 Computer My first computer maintenance job (1961) at Union Carbide was on the "Daystrom 046", at the Olefins-2 plant. It weighed 4 tons, had 4k of main memory (magnetic core), and 100k of rotating magnetic drum memory (about the size of a 55gal barrel, driven by a 480 volt, 3 phase, 3hp motor), and about 5,000 transistors. Yet we made some of the purest (99.9%) ethylene gas in world. Direct Memory Access: The VFA analyzers and other instruments automatically went (wrote) directly into dedicated memory locations in the high end of the 4K, therefore no software programming was required to access analog signals, they just appeared in main memory (at rate of 5,000 samples per second). Similarly, control outputs came from unique dedicated memory locations and directly adjusted set-points on the panel-board instruments. This was the first DDC (direct digital control computer in the world). Programming code had to be very tight back then due to limited main memory. Furnace control programs were only a few dozen words long; they simply read a few hard memory locations, did some simple math, and output directly to a unique hard memory locations, which controlled set-points. The math was done by calling pre-loaded math sub-routines. These math sub-routines came from the Daystrom factory, on a roll of punched paper tape, and when loaded, took up almost all of the 4k memory. Hardware maintenance (my job) was another amazing aspect of this 4 ton beast. There were no circuit or wiring diagrams, only several volumes of Boolean logic equations, in Sheffer Stroke notation. Every transistor and diode had a unique logical value in the Boolean equations. Knowing how to read the equations, and how to use of an oscilloscope, along with a few replacement transistors, and a good soldering gun was all that was needed to fix the thing. Here is a link to the specs on the "Daystrom 046" http://ed-thelen.org/comp-hist/BRL64-d.html Union Carbide is listed as one of their customers.... about 9 were built. Salvaged Programmer's Panel (with lights and switches removed) Other oddities with the Daystrom 046 Core Meltdown (almost): Late one night, one of our chemical plant operators called me and yelled "the 046 is on fire, what should we do". Being half asleep at the time, I said the obvious.... "Put it out.... I'll be out there as soon as possible"..... and then hung up. On the 30 mile drive to our plant, I began to worry about my choice of words, as I could only imagine a fire water hose being applied without shutting off power. Luckily, they killed the power and then applied a hand held Halon fire extinguisher. Upon entering the computer room, I saw the magnetic core memory rack, open and slid out on it's rails, with some pretty heavily carbonized circuit boards visible.. There were 22 circuit boards involved, and all looking pretty much destroyed. Why 22 boards? Because Daystrom designed their instruction word with a whopping 21 bits and one parity bit; this seemed like a good choice, I guess, because the 8 bit byte had not been invented yet. Without diving too deep into how magnetic core memory works..... each memory bit is composed of a magnetizable iron core toroid (do-nut), the size of a pin head with 3 wires passing through it, 2 for addressing and one as an inhibit wire. The inhibit wire supplies opposing current (when needed) to prevents a 1 from being written. Anyway.... what happened here was that the master inhibit driver, which normally applied a short write pulse, failed in the ON state, causing the core stack to heat up like a big toaster. What saved the core stack was that the 22 inhibit driver transistors and related circuitry competently burned off the circuit board, destroying the top quadrant, and thankfully caused the current to stop. Fast forward.....Two days latter, working round-the-clock, we rebuilt the 22 inhibit drivers and replaced the errant master inhibit driver. The repair consisted of taking the burnt boards to our machine shop, and with a band saw, cut off the charred top quarter. We then took perforated vector board, epoxyed it to the missing section, and hand (hay stack) wired the replacement circuitry. After that, the "046" ran fine for another 20 years. Hand Cranking the Drum: After many years of service, the motor starting circuit went out in our magnetic drum memory. This was an enormous machine, about the size of an oil drum, and had a 3" diameter shaft protruding out one end. To get it started, we used a large industrial variable speed electric drill, with a rubber lab cork chucked up as a friction clutch, and applied it to the 3" shaft. After a little slippage (and smoke) the drum slowly started turning at which time we applied power. So was computer maintenance in the good "olden days.". Bill Richter ( Union Carbide Corp - Retired) bill333@suddenlink.net

The IBM 650 & early transistors, by Rick Dill Rick Dill - sent September 2019

I was at CMU when they got their first computer, an IBM 650. I'd worked for IBM the summer of 1954 and at RCA Labs the summer of 1955. At IBM as a summer student, they put me in a "new employee" three week class about IBM while the rest of the company went on their "scheduled" two week vacation. It was like drinking out of a firehose, starting with punch cards, meat scales at the beginning, with the details of the evolution of punch cards (aka 1954) and the technologies that followed. I have the hand-out notes and think that someone in CHM has them. I would like the notes back, but CHM needs to understand them. It shows well IBM's path from unit record to computers. 1954, I got to look at the functional layout drawings of the 604. It was a "plug programmed" calculator, but 3 years later when I got introduced to the 650, that very simply was a 604 with I-box (instruction handling set for programming). I saw lots of other stuff as well. At IBM, I was encouraged to "invent" functional electronic devices that did higher level logic than individual transistors. I left with a decimal counter based upon unijunction transistor concepts. IBM never did anything with it but Ian Ross (later head of BTL) built the same thing at Bell Labs and published it at a conference I attended while at RCA. That helped give me confidence that I could play with the "big boys" in the field I had chosen with the Shockley, Bardeen and Brattain announcement of the point contact transistor while I was in high school. NO, I didn't want to start a company, I wanted to invent things like Edison did. As background for this, as a "poor" student needing all the income I could find, I had already been in a part time job for a Physics grad student. My job was to enter numbers from Hall Effect measurements into a Freiden calculator and get him the results he needed. I'd also played in my "kid" life with adding machines and the Freiden. OK, fast forward a little. I abandoned my physics major undergrad and walked across campus to EE, under Rod Williams an incredible communications engineering professor. I got bored because they thought I needed to take undergraduate "Electronics" . Facing a boring and trivial lab report, I just submitted an idea I had for building a degenerate form of parametric amplifier using semiconductor diodes as the non-linear capacitors. Well my office was where some more senior grad students were building these with barium titanate capacitors which had to be in a water bath because their capacitance depended on both temperature and voltage. OK, I build my circuit and published it with my electronics professor, first publications for both of us. The next year I used the same tech and some fun with circuits to build the replica of the classic HP audio oscillator which did a decade of frequency with the turn of a single knob. Published that and was bored, but needed a thesis topic. Well I also needed a thesis adviser because no one in the department was in the field I wanted I went to Williams with three topics ranging from deep physics (which I was poorly prepared for) to understanding the unijunction transistor which had only a primitive explanation of why it worked, to doing more with semiconductor capacitance which I was bored with. He agreed to serve as adviser and turned me loose. I had to make the devices and then figure out how they worked. Shortly afterwards Dale Critchlow who was young faculty at CMU went to IBM for a summer. He is in NAE for his work in DRAM and things like trench capacitor. He talked IBM into supporting three of us. Me, Bob Dennard who you know as the inventor of DRAM but who was just starting with a thesis in magnetism. The third was me already working on theory of unijunction transistors OK, as expected, I had some troubles. I needed some help with getting the physics right. An hour at RCA with Herb Kroemer (nominal inventor of the drift transistor ++++ much more) got me past that. Then I was into non-linear differential equations to follow the electrons and holes and simply didn't have the math tools. Well, CMU (really Carnegie Institute of Technology) got their first computer. A few students got head starts by going tot their local Mellon Bank and programming for their theses on an IBM 659. I had a close friend (that extended through much of my career) who ran over to the new 650 and started programming, first in assembler and then in SOAP which assigned memory locations for memory so that they would come up just after being needed. Karl knew me well. I wanted answers, not adventure with new computer. He directed me to the Wolontis simulator which ran on the 650 (also called BLISS at Bell Labs) It took the 650 and used half the memory to provide a three address computer image. Well with half the memory gone you had 1000 ten bit floating point words. That let me program without all the details like memory management or where the decimal point was. IT WORKED! I later gained a good friend at Bell Labs who had, with a help of a math guy, created the same programs for solving Poisson's Equation, which is core (in three dimensions) in all device simulations today. OK it was NOT EFFICIENT. I used a lot of computer time. I found that midnight to six was easy to get, so at one point Perlis, the guru owning the computer and creating a compiler, "IT". more or less with IBM's FORTRAN. It eventually got folded in by IBM into FORTRANSIT. I didn't use IT because it was in early development and I wanted something stable enough to get me out of school. When I got the IBM support, they wanted me to come to Poughkeepsie monthly to report on my progress. I did that most months and it at least got some people listening to me even they had little deep knowledge to help with. While at IBM, computer time was easy. I just had to find a 650, often with covers off. I could run it until I ran our of personal energy. At CMU the 650 eventually got some core memory, but that was well after I left in Feb 2018. I would love to see the 650 and even better would like to see one run at CHM. It was one of the very first computers to be made in relatively large numbers. Rick Dill This is stream of conscious. Comment or correct please.

IBM 2321 Data Cell from Ignacio Menendez added March 2018

This was a rather unusual random access data storage machine. See https://www-03.ibm.com/ibm/history/exhibits/storage/storage_2321.html and https://en.wikipedia.org/wiki/IBM_2321_Data_Cell I had these interesting experiences with the first 15 IBM 2321 Data Cell Drives, manufactured by SMD, and still on B test in San Jose, circa 1962. We had been working on pickability and strip wear problems for over two weeks on the the final test floor, I happen to run into an old gentleman, by the name of Ollie (?), from manufacturing Engineering, who had been very helpful before, and I related the current problem to him..... I almost got fired for telling this gentleman about it, without permission of the dept. Mgr, or the lead tech.... but the results cleared me, he found that the settings that we were using for the strip housing rotational adjustment, and the cell air adjustments were not the optimal, so he changed the spec.... he also did something for the stripwear, don’t remember exactly what, perhaps allowing the use of the GREEN lapping strips flying for a few seconds, to better burnish the housing. My other experience was with a 2321 that had finished the final test, but when run on the mod30 System, diagnostic routine F650, random seek, it failed with a permanent seek error. I was able to retrieve from the 2841 sense data, the from and to addresses for the servo, and sure enough, it would fail from the CE panel as well, but not anywhere that we would do our seek alternate testing on the final test line.... I wrote a small program on the ‘robot tester’ (modified 1620’s, I think), and ran 4 and 5 subcell moves all around the bin.... This 2321 would fail like crazy at some addresses 90 degrees apart, on these small bin moves.... I took my program and ran it on all the 2321s on the floor, and you guessed it, ALL of the 2321’s failed.... I showed this to QA and they Red tagged all the machines, for which I also almost lost my job, until.... Luckily, Ed Shaules (?) from development Engineering came in, and determined that we needed a more precise Pegasus red tag servo valve, for the blue one that we were using, could not handle with precision these moves on the four roll vane areas of the Hartman hydraulic motor. Anyhow, I hope this was not too boring an old story. Iggy

a COBOL joke - from Keith Falkner - Feb 11, 2014

COBOL joke In 1996, we COBOL programmers were all looking over our shoulders and cringing under the pressure to update masses and heaps of programs that had been handling the calendar year with two digits. As 2000 approached inexorably, we were busily adapting these programs to use four digits instead, and devising algorithms that would reliably infer the first two digits by considering the current year and the given last two digits. Let me tell you, it was dull stuff. But we did it. There was a programmer who rebelled. He was so bloody tired of dealing with trite and sloppy code that he decided to take refuge … in the future! He figured that by 2004, all programs would routinely use four digits for the year, and any bugs left over from the conversion effort would be long since eradicated. He built a suspended-animation chamber, and programmed it to keep him alive, but unconscious for eight years. Unfortunately, his programming skills did not match his engineering skills, and his program was, as we said, not Y-2-K compatible. The result was that he stayed in suspended animation, not for the intended 8 years, but for 8000 years. He emerged in the year 9996! He found a civilization that was still running COBOL programs, but had nobody to maintain or upgrade them, until his reawakening. So he was welcomed most earnestly, and had no choice but to earn his living by converting all the COBOL programs in the world to handle the five-digit year that would soon be upon them. None of the existing COBOL programs was Y-10-K compatible!

Monrobot - one of many "lost" in history - from Donald Caselli - July 6, 2013

Reference - MONROBOT-III, MONROBOT-V, MONROBOT-VI, MONROBOT-IX, MONROBOT-XI, MONROBOT-MU I recently discovered your website was able to find some word references to Monrobot but nothing of any historical significance. You may know that Monroe Calculator Co. of Orange, NJ was the designer and manufacturer of the Monrobot XI computer in 1959. I was introduced to the Monrobot XI in 1965 by way of Peter George Petropolis in NJ who had founded a software company known as Systems Design, Inc. His focus was on programming the Monrobot XI to process back office work for the stocks & bonds industry as well as other financial institutions. One major accomplishment was programming the Monrobot XI to calculate a basis price extending out beyond 40 years. I worked at Monroe in Orange for one year and then joined up with Peter Petropolis and others who were forming Computer Property Corp in Hanover, NJ. Computer Property Corp acquired the entire Monrobot XI line with spare parts, etc. There were 350 installations of the Monrobot XI at the time they took over. ... I would travel to some of the installations over the few years that we operated. I improved the assembler so that source programs could be kept and processed with the 8 channel paper tape and produce object 8 channel paper tape output that would be executable programs. I also developed a dis-assembler that was able to accept an 8 channel object paper tape as input and produce an 8 channel source paper tape that could then be input to the assembler. This was done so that modifications could easily be made to programs that were being sent to Systems Design, Inc. from installations around the USA, and there was one in Toronto, Canada, and I spent a little time with that installation. Two IBM Model B typewriters could be connected. Two IBM 026 keypunch machines could be connected. There was a magnetic card, the size of an 80 column Hollerith Card, that held 96 32-bit words, known as the Monro-Card. I do not know of a physical Monrobot XI in existence. I have some original documents and manuals from Systems, Design, Inc. and some photos, which I posted on flickr.com These photos should be able to be found by using the search keyword monrobot ... There is a computer museum at Camp Evans in Wall Township, NJ which is part of infoage found on the Internet as infoage.org The museum is M.A.R.C.H. known as mid-atlantic retro computer hobbyists. There are other museums at Camp Evans and can be visited at times shown on their website infoage.org In May, 2011 I gave a short presentation regarding the Monrobot XI and distributed a small booklet that I printed regarding a little history of the Monrobot XI. - Cover of a presentation The purpose was to explain that the Monrobot XI computer was able to be installed at a particular place or office and be used to process various data and produce printed reports, checks, etc. instead of sending the data to a data center which would take a few days whereas the Monrobot XI would produce results immediately in an office. British Overseas Airways in New York City and the American Stock Exchange were good examples of successful Monrobot XI operations regarding payroll processing. Of course there were 90 stocks and bonds installations, 35 installations for fuel oil delivery companies and many others. So this was the type of history that I was interested in presenting. ... Donald O. Caselli 900 Darmstadt Avenue Egg Harbor City, NJ 08215 609-685-1528 dorestec@msn.com

Magnetic tape media problems - from Dana Roth - July 5, 2013

A little background from Ed Thelen: When I was fixing computers for General Electric in the early 1960s, one of our plagues to smooth running computer sites were the crappy tape drives and the crappy magnetic tape media. See my tale here. Introducing Dana: In the mid 1960s, Dana worked for Control Data on CDC 6x00 equipment - see his comments here under "More "Cordwood Modules" - MTBF and other tricks ;-)) - from Dana Roth July 5, 2013" The tape certifier had a microscope. New tapes had excess oxide (and other oddities) and needed to be cleaned before certifying. As the tape was being tested, a bad area (from many sources like contamination, bad mylar, etc.) would position below the microscope and could be usually repaired. There was a counter on the tape certifying machine which counted errors. Too many and the tape was sent back for replacement. The issue, of course, was why so many bad tapes. The government site had about 800 tape mounts a day and 25,000 volumes on-site and were serious about data. I understand it to be a lot of remote sensing from satellites keeping track of minerals etc. The 66X tape drives were flying at 200 inches per second, so I became an expert on tape subsystems pretty quickly. I suspect that manufacturing companies sampled tapes to reduce overhead costs. The GOC had a good tape librarian to ensure quality (and beat up on the vendors)! One day CDC flew me down to Dallas TX (on the hottest day in May), Mobile Oil site to work on tapes. They had a quarter million tapes on-site, which looked like a gymnasium filled from floor to ceiling, all seismic data, from all over the world. The tapes were in rough shape (peanut butter on them and the like) and a single process job would run for three days, processing seismic data. A tape would break into the second day and they would have to restart the job. They had three CDC 6000’s in one room, with some special “stunt boxes” that allow for some serious number crunching. We helped develop an algorithm to uniquely address each block on tape (similar to disk) and take checkpoints in case a tape or drive would break. They could then repair a tape and move it to another drive, seek to the broken spot and continue in mid-job, saving days of processing. Tapes had to work, even if bad. In hindsight, maybe that’s why oil is so expensive (a 72 hour job @ $30,000/hr.)! Tape abrasion was a problem and CDC went to glass heads that were pre-grooved so as not to distort the tape passing over the head. On the GE front, one day a computer operator miscued a tape (the system was actually self loading but operators threw the tape on the drive) on a 667 tape drive. He opened the door to check how the tape loading was stuck. A General Electric AC motor capacitor exploded and coated him with PCB oil. In those days, the oil wasn’t well known to be a carcinogen, as it is today. CDC bought him a set of new clothes. I got to change out all the remaining capacitors and had them properly disposed of. What a mess! —- dana

7094 II Move and CoreMemory problem - from John Van Gardner - Mar 17, 2013

When Lockheed won the contract to build the C-5 in October 1965, they knew they were going to need more computing power. At the time Math Analysis had a 7094 they owned and one they were leasing from Greyhound Leasing. The California division had a directly coupled 7094 II that they were replacing with something else. On September 25, 1966 they decided to bring it to the Georgia Division. At this time there was no computer room to put it in and we did not have anyone trained on the Mod. II. Lockheed began building a computer room in an old cafeteria down in tunnel 2 of the plant. I ordered a set of manuals for the Mod. II and spent a whole Labor Day weekend reading them. On Monday we found out there were no 7607 data channels coming with the machine as it had been directly coupled to a 7044. IBM found 2 channels in a warehouse in St. Louis, MO, but they did not have the feature that allowed attachment to a Mod. II. I called Poughkeepsie and had someone read me the wiring list and SMS card list for the change. I recorded the phone conversation on an IBM Executary dictation machine our salesman, Bill Bull, had given me when it broke (just a bad relay). Poughkeepsie shipped me the SMS cards by air. Lockheed put the 7094 II and all the tape drives on their L-100 (Civilian version of the C-130) and flew them to us. By the time the machine arrived Lockheed had built a small section of the raised floor with a temporary ramp in the middle of the old cafeteria in tunnel 2. We positioned the frames and began wiring the channels. It was freezing cold as none of the equipment was powered up so I worked with my coat on. The painters were painting the ceiling with a roller and when I got home I had paint all up my backside. It ruined a good suit. After we got all the construction out of the way everything went as a normal installation. I was able to trouble shoot all the bugs we had and from the day Lockheed decided to bring the machine from California to the day we turned it over to them it was 18 days. I had one interesting bug on this machine about 3 months before it left. They had a program that was a matrix inversion problem. It gave different answers every time they ran it. I found out that if you turned off the overlap switch it would run correctly but slower. Lockheed had a young blond mathematician named Stuart Austin who worked with me on the problem and she found out that one of the instructions in an index loop, that added up a column of numbers, was picking up a bit in the decrement field of the instruction. The 7094 II used an air-cooled split array memory with a cycle time of 1.4 microseconds. I scoped the memory and found that the location of the instruction was being half selected many consecutive cycles during the loop. I talked to Poughkeepsie and they said, that was sometimes caused by a core that had a chip out of it, and it would flip with half current. They said the only way to fix it was to replace the core plane in the array and it would cost $35,000.00 for them to send someone to do it. We knew the machine was due to go out in 3 months and that the program would run if you turned off interleave so I decided to try and fix it. I took a current probe and looked at the current pulse on the half selected driveline and decided to add a resistor in series with the terminator for that line. If I remember correctly I think the terminator was a 250 ohm Non-inductive resistor so I added a 25 ohm in series and looked at the scope. I could not tell the difference in the current pulse. I tried the program and it failed. I had already replaced the 250 ohm resistor just in case so I took the original 250 ohm and added it in series and the current pulse dropped about 10%. The program worked fine. I ran all the diagnostics and bias test and it worked. Well, so much for ohms law. I had forgotten I was trying to send a square wave pulse down a long wire with lots of inductive kickback. The machine ran fine until it left and a 250 ohm resistor saved $35.000.00. Stuart Austin went on to program the CDC computer at the wind tunnel where she met and married Johnny Johnson. They both later went to work for IBM as System Engineers..

IBM - Customer Choice of Colors - from John Van Gardner - Mar 02, 2013

Here is some tourist information about the colors. It all started when IBM offered the customer a choice of color on typewriters. There was set of standard colors some of which I remember: Classic Blue, Flame Red, Garnet Rose, Light Gray, Pearl White , Raven Black, Sky Blue, Sunrise Yellow, Willow Green. In 1971 orders for new typewiters was down and IBM started taking trade-in IBM machines on new orders. These machines were sent to the plant and completely dissassembled and cleaned. All worn parts were replaced and these machines run down the assembly line and adjusted to meet new machine specs. All of these machines were painted Charcoal Brown and sold to employees only at a large discount. I bought an Executive with a wide carriage and carbon ribbon for $220.00 plus $6.60 GA sales tax. IBM deducted $9.44 from each pay check untill it was paid for with no interest charged. A new Executive like mine would have cost $525.00. I still have it somewhere in my attic. In 1955/56 when I was in 704 school in Poughkeepsie they had a computer room on the back of the plant with a 702 that they actually did the plant's payroll and other data processing. It was a nice looking room that they showed potential customers on tour. The whole front was plate glass windows they could look through. There was a desk for a receptionist with a yellow typewriter and yellow telephone. The receptionist had long blond hair like Betty Grable and always wore a yellow sweater. Every day at lunch I would go by there to see if she ever wore anything but a yellow sweater. I never saw her in anything but yellow if she was there. I wondered if that had been written into her job description. When Lockheed at Marietta, GA got their 705 Model 3 they paid $3,000.00 for an RPQ to get their's painted Garnet Rose. That was the first colored system I ever saw. Later on Martin Aircraft in Orlando, FL got a 704 system with each box a different color. When the 7000 series machines came along I was working on a 7094 in a test cell at the Poughkeepsie plant and the next cell had a 7074 that was going to the U.S. Justice Dept. and it's main frame was blue and it had 7340 Hypertape Drives with red and white vertical panels. You could not look over to that test cell without thinking about the American flag. When Customers began to get multiple systems they would get different colors and refer to them as Old Red or Old Blue. Van Gardner from Stan Paddock, March 4, 2013 Back in the late 1960's, I worked for an IBM refurbishment center. I thought it was a bit strange when a complete 1401 system came in covered with wood grained contact paper. A savings and loan company in Los Angeles wanted to spruce up the system as it was visible from the lobby. IBM gave them the ok to adorned the system if they paid to have the contact paper taken off the machine. I pulled my share of the paper off before the machine was sent to be destroyed.

Doing computers in the State of Washington - from Jordan Stedman - Feb 17, 2013

I'm a bit of a dark horse here: I got my start in computing by pinning plug-boards for a still-functional Burroughs E-101 in the fall of 1966 – then writing ML (no SOAP) for an imaginary IBM-650 – all at Lincoln H.S. in Seattle. Then I had a brief chance to write some assembler code for an IBM-1620 at WWSU where my uncle taught. Finding a keypunch brought me into contact with Herb Burke, who then owned the IBM-709 from Lockheed Sunnyvale via the Univ. of Wash., operating it as a small service bureau in Seattle – "Puget Sound Computer Service" – where I worked from age 17 to 25. That machine and its sister from WSU – plus a 7094 that I personally reinstalled back in the '70's – are now part of Paul Pierce's collection. (That's "my" 709 in the top photo on his home page.) We did a few interesting things at PSCS: extended the OpSys and Fortran IV compiler, plotted/drafted maps of defoliant ("rainbow agent") spray missions over Vietnam under contract to the National Academy of Sciences – and resold some used "DecTapes" acquired in a bankruptcy sale (and useless to our purposes) to a couple of young high school students for "a project": the late Kent Larson and a fellow named Bill Gates. From there I went on to an IBM-360 shop (Prodata Systems, Inc) where I wheedled my way into system admin and datacenter manager, piloting upgrades through a National Semi ECL chip-level knockoff of a 3158-3, an IBM-4341/12, 4381/12 and 9221-191, dozens of OpSys upgrades, and build-out of 500-device network using CICS and 3270 terminal technology in both BiSync and SNA/SDLC. We were one of the first implementers of IBM-3380E DASD in Seattle (Wow! 5GB in a large 'refrigerator"!) and an early adopter of 3480s. It was all ultimately ported to a Flex ES emulator running on a Dell server with 120 GB of RAID under UnixWare. I have some interesting stories from my interactions with IBM FEs from that time. – one in particular (Glenn "Jake" Jacobson) actually gave me a "key" to the boxes [key-ring Allen wrench] in testimony of his trust that I knew how to open them up and put my fingers in without breaking anything. He also taught me how to secure and move a string of 3380 DASD and then unlock and restart them without assistance: I paid him back by getting him some interesting employee file commendations and a Christmas bonus. [Interesting story here: one year the company offered this FE a Christmas fruit basket, and he refused it based on strict IBM policy that forbade gratuities. Cup of coffee was okay – but not a fruit basket. So I dug out the name of his manager and I think his manager’s manager, and sent a letter expressing that in lieu of the refused fruit basket we were sending a check in Jake’s name to the local food bank org, Northwest Harvest. About a week or two later, Glenn caught me and said “wow – that letter was WAY better than any chintzy little fruit basket!” I never did learn precisely what he got out of it, but I gather is was manifestly spendable!] I also did some interesting development projects, creating an application infrastructure under DOS/VSE-->VSE/ESA that would now be recognizable as a rudimentary O-O "com-layer" [~20k lines of COBOL and several thousand in HL-Assembler.] Then I took ownership of first a NetWare infrastructure and then Windows NT at 3.5.1-->4.0-->2000-->2003. I was the primary contact and one of the certified employees for Prodata's MCSP agreement with Microsoft for several years. Prodata (f. 1958) was sold in 1997 and subsequently downsized pretty literally into a garage – and I was forced to look for other work in 2010 so I have been taking vendor contracts at Microsoft since. Currently I am working under contract as a member of the MSIT Identity and Access Management – Domain Controllers team – an admin in all AD domains in MS Corp and providing specific support to Microsoft Retail Store. Well, I can't really say much more about that – except hundreds of domain controllers scattered around the world, running on G8 servers each with 2x8 hyper-threaded cores (32 LProcs) 128 GB of DRAM and about net 1.2 TB of RAID-1+0+S storage. This industry has changed a lot in 45 years, but I am still kicking! (and still broke... sigh) Jordan Stedman, Shoreline, WA jlstedman@frontier.com (pref) or jlstedman@live.com http://www.linkedin.com/profile/view?id=16464785 206-227-9988 (mobile)

Coding for a new computer, with no other computer for cross-assembler - from LaFarr Stuart - January 9, 2013

Cyclone, an IAS machine by LaFarr Stuart When I got to Iowa State (a graduate student in statistics), the Cyclone had been running for a year. It was a duplicate of the ILLIAC, using 5-level paper tape, and its' memory was Williams Tubes (41 of them for 1024 40-bit words of memory and one as a monitor that could be dialed to display any bit position of the 1024 words). There was no parity or error checking in the hardware. During the summer, (before I arrived) they replaced the memory with 4 banks of 4096 words of core memory; and 8-bit paper tape I/O. This caused a major change in the address decoding in the hardware, (2 instructions per 40 bit word, each instruction was 8-bit op code and 12-bit address.) The older memory only used 10-bit address, and much of the software used the 2 high order bits, which were ignored by the hardware, as flags or whatever. After the changes the "New Cyclone" had zero software, and the instructions were being being upgraded when I arrived. But three new Flexowriters were working, and the 8-bit paper tape reader and punch were being connected. Getting a computer going, without another computer to make a cross-assembler, run a simulator, and develop software on, was an interesting experience few have had; and nobody will have in the future. Our input/output was Friden Flexowriters that had no electrical link to the Cyclone whose only input/output was paper tape. We did not use Octal on the Cyclone. We used a form of hexadecimal, that did not use letters of the alphabet for digits greater than 9. It used characters like: , . ; for them. Our Flexowriters were slightly customized to have on one key we called "subten" it was a little "10" lower than most letters; and its hex value was 15. The intended output use of that key was to print large decimal values using a power of ten notation.

Young, frisky, and learned to fix the IBM-650 from the manual - from John Falk - July 4, 2012

I, Ed Thelen, paint the environment of the mid 1950s through early 1960s computer era - and maybe sales/service life in general - Sometimes Some Application Engineer was handed a box of manuals and an airplane ticket, say on Friday afternoon. He is told to fly to Chicago to teach a class on XXYYZZ the following Monday morning. Victim - "But I know nothing about XXYYZZ !?!" Boss - "Are you going to waste this weekend ??" A new peripheral is sent to a customer site, with schematics and a VERY preliminary manual. The local FE is told to install and maintain it best he can. I had that happen several times with G.E. - long nights - OK - I was well trained on every system that I was sent out to repair - Nike Fire Control system, General Electric 225 computer system. But not everyone was so lucky - But train yourself to understand and maintain a whole vacuum tube based computer system ??? from John Falk - July 4, 2012 My first assignment as a new IBM Customer Engineer was MGM Studios which ran 24/7. I used this manual [see below] to teach myself how the 650 worked. IBM form 22-6284-1 - IBM 650 Data Processing System - Customer Engineering Manual of Instruction - 5.2 MB .pdf file . A plus about having MGM for an account was eating in the Commissary with the stars and going out in the back lot to watch filming of Ben Hur Yep. and they gave me a parking place in the Exec lot where I could park my plymouth next to Rolls and other cars I didn't recognize. One day I stepped out of the building where Data Processing was and into the little street and Debbie Reynolds almost ran over me in her station wagon. Was all apologetic and dressed in her grundgie clothes. The guards name at the gate was Ken Hollywood and I learned years later that he was brother of girl I went to high school with in Venice. I was 21 and lived 2 miles from MGM and took all night and weekend calls at home directly from MGM. Upgraded the 650 to 4K drum and replaced the 533 with rare 543 reader and 544 punch units. Those were the predecessors of the 1402 read and punch feeds. When I went to 1301 school in San Jose I ended up teaching the 1402 Punch feed part because of the many hours I spent with the 544.

Short Stories I Can Tell - from Jay Jaeger - March 25, 2012

For example, the weekend after I was hired at the University of Wisconsin School of Business as an operator on their IBM 1410 (which they had acquired after the UW Registrar upgraded to a 360), I went in to practice a little. Machine just would not cooperate. Red lights and error printouts on the 1415 Selectric console like crazy. I assumed I was doing something terribly wrong, but I left a note, and headed off to my girlfriend's apartment. I got a call that afternoon -- in fact the machine was really broken, and the errors were all real.      That same 1410 had an annoying trait: when it first powered on, the 1403 printer would show a sync check for nigh unto 10 minutes until it "warmed up". Well "formed up" would be more to the point. One day our CE and the crack CE that knew the 1410 inside and out decided to try and figure out what the heck was wrong. They were at it with their scope on the 1414 I/O Synchronizer that the printer was attached to for hours. Finally they kind of shook their heads and laughed a bit. The 1414 had a bunch of electrolytics with leads that fit over wire wrap pins -- probably several per row of cards (presumably bypass capacitors as a guess). It turned out they had all been installed BACKWARDS -- either on "day one" or as part of some CE change earlier in its lifetime. Oops.      That same 1410 had a special interface built for it so that we could communicate with the UNIVAC 1108 main academic computer on campus. The special interface was hooked into the 1410 as though it were a telegraph. 8^) It used balanced line drivers to get the half-mile to the main computer science building. Bruce Orchard (who passed away a year or so ago) wrote a program to make the 1410 look to the UNIVAC 1108 as though it were a UNIVAC 1004 RJE station. We had far and away the best 1108 printouts in town (the normal UNIVAC printers were drum printers -- wavy lines like you wouldn't believe).      Another related story. Our 1410 CE also did 360/370 work. One day another programmer from our 1410 shop that I was working with to do a student records system on a 370 at the Wisconsin Alumni Research Foundation (WARF) entered the WARF building where we did our work on second shift. The scene was of the CE working hard under a 3330 drive (one of the ones that you thru a switch and the drive slid out like a shelf). Turned out one of the day shift 370 operators (or maybe a programmer -- who knows) and put a TAPE LABEL on a 3330 disk pack, and spun it up. Well, of course that label soon separated itself from the pack -- and took out most of the heads.      (The other 1410 story I already told. The UW Computer Sciences department received a donated IBM 1410 from Oscar Mayer here in Madison. I set the thing up using the CE instructions, and it pretty much ran right off the bat. One of the CEs got the Selectric console working, pro bono. The machine was later stored, and over the winter the temperature in the building dropped due to some HVAC / construction work. I set up the 1410 again in another room the next summer. The machine worked, but some of the wires in the core had apparently snapped, and some storage locations no longer functioned properly.)

The Flash that Snapped the Tapes - from Keith Falkner Mar 24, 2012

Not every photograph of a running computer system is a success. In the mid-1950s, Confederation Life had a fine 704 system on Bloor Street in Toronto Ontario Canada. A professional photographer, using a press camera with an immense flash bulb, took a picture of that system at the end of a long sort run, while four tape drives were doing a high-speed rewind. The strong flash of light confused the tape drives: they interpreted the light pulse as evidence that the beginning-of-tape marker had passed, and immediately put on the brakes that are meant to stop a low-speed rewind quickly. The four tapes snapped, and the photographer was politely ushered out of the room while the operators commenced recovering from the accident.

Printer Tales
worn ribbons - from Ed Thelen Oct 18, 2011

Trying to assure that the correct marks actually get printed on the paper is always a serious problem. One day ( in 1962 ) I got an irate call - The dollar amounts on a long billing run were mostly invisible !!! The printer ribbon on the G.E. printer had gotten worn on one side. (Typically the left side) The worn ribbon had pulled/drifted to the left, leaving the right hand side print hammers with no ribbon. The billed dollar amounts were on the far right of the forms, invisible. There wasn't a thing I could do other than recommend: - more frequent ribbon changes, - expensive to the customer - - manually rewinding the ribbon "left to right" to balance wear - ( a messy tedious process I demoed how to do it - once !! ) - more operator care and observation ( The IBM 1403 printer downstairs had the printer ribbon automatic lateral reposition system and likely would have printed completely across the forms. - another black eye for G.E. )

Extra Circuits - from Hans Franke Oct 14, 2011

Reminds me of a printer we got around 1979. The incident Van [ next letter below ] notes might not have been so rare at all. Otherwise the HFÜ or Hammerflugüberwachung (literally Hammer Flight Control) wouldn't have been invented in the mid 60s. Back then (at least for Siemens) banks where the largest customers for /360ish machines (even before government). Of course they had a need to make sure that every number send out would also end up on paper. So every hammer had not only a coil pushing it, but also a secondary coil where the moving rod created a little blip as feedback. A rather delicate circuit (*1) now compared the timing, signal strength and form to what was expected for a working hammer hitting a typeface thru paper and ribbon. It was a little masterpiece of analog technology. Anyway, long story short, we got the printer in 1979, after years of service at the Munich Tax Office, for free. It's been a great improvement from the 1960s drum printer we used before. We even accepted the 2-5 false alarms per month we got, no matter how careful the printer was adjusted. Hans *1 - you got to keep in mind that 95% of all circuitry in such a printer was analog and you got literally hundreds of pots to fiddle with timing parameters - in fact, it was impossible to adjust it right, since most values influenced each other - I once spend almost a day to adjust one ... Well, OK, to be honest, I wasn't inclined to do any other work waiting for me, but that's a different story Hans' comment about Ed's ribbon story (above) - "Oh, and an addition, the Printer also had a sensor to determine if the ribbon is still centered (otherwise printer got stopped) and a sensor to make sure it's wet (resistance thru the ribbon)"

an IBM 1403 OOPS - from Van Gardner Oct 13, 2011
( the 1403 was probably the world's best "line" printer, BUT ;-))

In the late 1960s when Lockheed was building the C-5 there were over 30,000 people working there. The second most important computer program that had to be run each week was the one that printed the pay checks. The most important job was the report sent to the Air Force to justify each man hour spent on each project . The Air Force had to accept this report before they would deposit the money in the bank to make Lockheed's checks good. It was a very detailed report that listed each empolyee and the hours spent at each cost center by shift worked. The report was printed on full width paper with Shift number printed in the last 7 print postitions at the top of each page. One night after the job had been run and printed one of the operators came screaming into the CE room shouting, "Come look what that 1403 had done to our report". When I looked at it I saw that the F had been left out of each shift number. The hammer bar that fires against the back of the paper to print the letter in front of the paper was broken. This hammer bar was supported by two vertical flat leaf springs and the front one was broken . The 1401 checked that the hammer fire pulse was sent but had no way of knowing the hammer did not move. It was an easy fix to replace the hammer but the whole job had to be reprinted.. Ever since then when I see one of those bumper stickers that says "IT HAPPENS" I think, "It really does". Van Gardner

an IBM 1403 OOPS - from Keith Falkner Sept 25, 2011

Some day I want to ride my motorcycle from Florida to [Computer History Museum, Mountain View] California to see those wonderful 1401 computers in operation. I have greatly enjoyed the simulators I have downloaded and programmed since joining this forum. Here's a simple anecdote that concerns a 1403, which was being driven by a System/360, not a 1401. A man whom I considered clever and insightful was idly looking at a thick report, which consisted of many columns of large numbers. He seemed a bit puzzled, and I asked what concerned him about the printout. He said, "There don't seem to be a lot of sixes in this report." Well, the pages of the report were numbered, so I turned to pages 16, 26, 36, etc. Sure enough, a substantial number of those sixes were incorrect, all the same digit (but I forget what that digit was). As I knew which 1403 had recently been serviced, I went to that machine and stopped it, then tried to corral all the printouts that it had produced since the Customer Engineers had turned it back to us for use. Then I requested the assistance of a Customer Engineer, who studied the print train and reported that it had been incorrectly assembled. I think that we (which means IBM's Toronto Downtown Datacentre) were particularly lucky to catch that error before any erroneous reports reached our customers. My colleagues and I tipped our hats to our perspicacious colleague for a long time thereafter. Techie tip - The printer was a chain printer, with the individual characters as individual slugs of metal. An option for speeding printing highly numeric data was to place multiple series of numbers along with a single alphabet on one chain. This almost doubled the speed of the printer when successive lines were all numeric.

The NUMBERS Report - from Stan Paddock - Sept 25, 2011

A real case from way back when. In a large shop there was a COBOL program that did not terminate correctly. The customer did not complain that there was a memory dump at the end of his report. While the printouts were ok and also the files, the programmers left the program to be fixed at a later date. When they did fix the program, the customer called and wanted to know "Where is my NUMBERS report?"

Reduce the printer load - from Ed thelen - Sept 25, 2011

Long ago - in Silicon Valley - when Fairchild Semiconductor was a big player The mother of my children began programming at Fairchild. Her introduction was to ask internal users if they needed all the copies of the reports they were getting. The universal answer was "YES" even though she could see stacks of computer printed reports gathering dust in corners of offices - seemingly never opened. After rounds of attempts to reduce the huge printer work load for internal reports, management decided to charge the receiving departments by the page for reports delived to them. The printer work load for internal reports was quickly cut in half. ;-)) She was asked to modify some report, and she made and verified the mods. The Programming Standards Committee found that the text of her COBOL comments field in the program started in a non-standard column. She had to wait another two weeks for the program to be re-examined by the Committee. - You might imagine that internal users complained a lot about slow responses to requests for report changes.