Thousands of brave young heaters help win WW2
Heater.
Posts: 21,230
Perhaps you have all seen this story of the Colossus computer already but I have only just discovered it. http://www.youtube.com/watch?v=rjfGAGWNU8c
Comments
One thing I've always wondered about those massive early computers was how they could connect all those wires and not make a wiring mistake? I have trouble connecting wires to just a Propeller chip and sometimes connecting something to the wrong pin!
And how they would know if one of those "valves" (or tubes) was no longer working? (I've read that some of those early machines could only operate for a couple of hours before a tube went out and it needed service (not a problem with the colossus as mentioned in the video).
Tubes are much more reliable when you never turn them off. Like light bulbs the heaters in tubes suffer from huge current in rush when they are switched on from cold and the thermal cycling kills them prematurely.
Besides I don't think tubes were as unreliable as people say. I was brought up with tube radios. I have seen many examples that have worked for decades without issue. We were always testing for dead/soft tubes and I inly remember finding a couple.
At that time it was possible to have something called 'tube insurance', so my father didn't actually have to pay for all the replaced tubes. It shouldn't come as a surprise that 'tube insurance' was a concept with a short lifetime..
And as a 7-year old I was fascinated by those tubes the TV service guy kept changing, I used to watch when he opened the back of the TV and my interest in electronics was awakened: I wanted to learn how to repair electronics, and that's exactly what I did. And that lead me to microelectronics, and computers, and programming and everything that I have done in my working life.
-Tor
A very well designed tube in low-power service might have a MTBF of 20,000 hours. More than that was very hard because the coating for dull emitter filaments would wear out gradually with continuous use. The most frequent mode of tube failure was in fact emission drop rather than filament burnout for this reason.
And it was not unusual for those early computers to have 10,000+ tubes. There were other factors; machines with multiple mercury delay line memories had to have the delay lines in temperature controlled ovens so they'd all have exactly the same propagation speed to sync output. It was a common ritual for the machine to be serviced in the early morning and to run fine until early afternoon, when you'd schedule less important work because the glitches would start to show.
As for how they serviced those beasts, the techs were intimately familiar with their ins and outs and had special diagnostics which were necessary in the course of building the machine components to tell whether they were working at all. This was all baked in at a low level in ways nobody would bother with today. A lot of hardware was in those machines just to make sure the real hardware was working; a parity bit and error detector was a bit of an extravagance but necessary if you couldn't really trust your memory to reliably return what had been written.
Those early early computers had 1,000 to 2,000 valves, but later machines got more complex. Trust Wiki to have a list of them:
http://en.wikipedia.org/wiki/List_of_vacuum_tube_computers
Check out especially the AN/FSQl-7 with its 60,000 valves. Part of the reason for it having *so* many valves was failure remediation so that a failure wouldn't bring the ongoing operation to a screeching halt.
NUSSE from Norway is missing
http://users.abo.fi/atorn/History/Page14.html
This could of course be because the CPU is an English APEXC CPU, but the rest was designed and built in Norway. The English CPU designer called it APENC.
(1000 tubes, 512 x 32bit word memory, paper tape input/output, dedicated printer. )
Ran the game of NIM in 1953.
It could add, subtract and multiply.
A typical multiply took .24seconds, and a divide .27seconds(using a subroutine)
They had problems with it being unstable at certain times of day... When a nearby nail/bolt manufacturer turned on and off the furnaces.
Long calculations therefore ran at night when the power was more stable.
Tommy Flowers was a London boy with a Cockney accent, the son of a bricklayer, working for the Post Office on telephone switch problems. He was not an academic at Oxford or Cambridge. Basically his social position in class ridden Britain at the time meant he had little sway with what went on. He just had to shut up and go back to work at the post office.
It was a disaster that way.
To get a feel about what a miserable situation that was watch this presentation by Captain Jerry Roberts. One of the code breakers at Bletchley. At over 90 years old he is still campaigning for proper recognition for Alan Turing, Bill Tutte and Tommy Flowers.
It's too bad Wikipedia doesn't have some kind of mechanism for members of the public to add information like that. Some kind of open-access group moderated editing process perhaps. I bet it would grow like barnbusters if they implemented something like that!
Back when the Wikipedia project was started I hoped it would grow to store 'all knowledge', a bit like the 'Library' of David Brin's "Uplift" books. And at one point everything went well. But now? I have basically stopped adding info to Wikipedia. There's a group of half-official people hunting information that isn't "noteworthy" and removing it. There are some great, valuable, correct articles about some old microprocessors, for example, which are marked for deletion by those people because the info/processor is not "noteworthy" enough, or there's info too close to so-called 'self-research', or both. People have to fight to keep real information in Wikipedia. I just don't dare add much anything anymore. There's still lots of info there, but the number of contributors have falled dramatically from what I can tell. Just the other day I read about information being removed from Wikipedia because these pundits don't think information about areas in Africa isn't interesting. All because of their own ignorance, of course. And talking about ignorance, they also delete information about old electronics because they believe it must be wrong. They're incompetent and don't know it. But what I dislike the most is the whole concept of "noteworthy".
Regarding old computers, I dedicate 100 percent of my Wiki editing efforts into dedicated wiki sites, not Wikipedia anymore.
-Tor
Actually, tube insurance has been taken to the level of fine art by Siemens, Philips, and GE. It's called glassware coverage for your X-ray tubes which range in price form $3k for a low end X-ray tube for a rad room or portable unit to around $250K or more for a high end CT tube.
FF