[John L Stewart]

The earlier post on the Collosus reminded me of my experience at IBM.

Before I became an electrical engineer I joined the service crew at IBM. That was in the Summer of 1953. At the time most data processing was done using electro-mechanical & relay logic.

The data was stored on perforated cards. IBM had the patent on electrical reading of the data on the cards. Other people were at a real disadvantage since their cards had to be read mechanically. You older guys out there probably saw lots of these over time. Later there were mark sense cards, which a user could mark with a carbon pencil. Both card types could be read directly by making an electrical contact by the sorter or whatever data processing equipment they were run thru.

Each card had 80 columns in which the holes were punched by both manual & electric punched card machines. Both numeric & alphabetical information could be stored on the card. One hole would store a decimal digit while a second hole in the appropriate place in the same column would be alpha information. End of line & other commands were also possible.

Each of these card types had its own operator for data entry (ie IBM 024). There was a corresponding set of verifying machines (ie IBM 026), again with their own operator.

The data processing equipment included sorters, collaters, tabulators, line printers (guess where LPT came from!) & so on. Arithmetic was performed by the tabulater printer combo (IBM 402 or 405) using what was known as ‘9s compliment’ computation.

None of these could be bought. The customer had to rent. Because of the patent on the electrical card reading system IBM had a captive market, a monopoly.

Not long before, Thomas J Watson, President of IBM had mused that the World would never need more than five or six computers. Thomas J had started out as a hot shot piano salesman in upper New York state, something he was very good at. But I guess his background did not prepare him for the future, much like many of us now.

The first computer if you could call it that, which I had the pleasure to work on was the IBM 602A. It was an electromechanical beast, lots of relays, cams, gears & so on. Compared to today it doesn’t seem like much but it did replace an acre of girls on mechanical calculators. It could add, subtract & multiply & sweat like hell doing simple division. But it could still get the payroll or inventory out on time provided there were no card jams.

Later the IBM 604 showed up. It was electronic & used the 6J6 as a binary flip-flop. Depending on the application there were 900 to 1500 6J6s in each 604. Each 6J6 was mounted in a plug-in assembly, complete with all the R’s & C’s required to complete the flip-flop. That helped a lot when trouble shooting since the whole assembly could be quickly changed. All computations were done in decimal. The binaries were connected in groups of four. So with some feedforward & feedback to the correct place four 6J6s ran as binary to decimal in 1-2-4-8.

Heater supply was thru a regulated PS using saturable reactors. These driven in a NFB circuit by a 6L6 (metal).

Many of these data processing machines used the 25L6 (metal) in the punch card reading circuits.

The Scopes we used were from Waterman, long gone out of business. They had a 3 inch CRT & perhaps 300 KHz bandwidth. Vertical amplifier is a pair of diff amp 6J6s connected in two stages. But they did the job, I think the IBM 604 processor if you could call it that ran only 50-100 KHz. I’ve still got my IBM voltmeter to which I later added a 1.5 volt range. Sensitivity is actually 10K ohms/volt, not bad for 1953. But it looks kind of used now!

Luckily for me I moved on after a year & half to Ferranti where some real research was being done. By then they had discovered the transistor, even though it was a surface barrier type, the SB100. But that is another story to bore all of you.

BTW, there is no test on this!