Advertisement

The high cost of progress

March 28, 2016

Jack Ganssle-March 28, 2016

For your next design, will you use transistors or tubes? Consider the following ad:

click for larger image
The high cost of progress. (Courtesy of the Transistor Museum. Copyright 2001-2015 by Jack Ward, Transistormuseum.com.) This is a great site for anyone interested in the history of semiconductors.

It’s a June 1961 ad in Scientific American. GE was making a lot of money selling tubes, and transistors were encroaching on that territory. Some marketing genius jumped on the new transistor’s high cost, relative to vacuum tubes, though ignored other expenses like power supplies and fans. The ad claims a single tube could replace 3-5 transistors, which seems a stretch to me as most implemented one or two triodes, tetrodes or pentodes, and each of those was basically equivalent to a single transistor.

It is interesting that the ad specifically mentions the high cost of putting transistors in computers. One of the very first transistorized computers, the TX-0 came out just five years prior to this ad. Reputedly each of the machine’s 3600 transistors cost $200 (in today’s dollars). I have a 128 GB flash drive that probably has a third of a trillion in it and paid $45 for the device. Obviously, memory always has a higher transistor density than CPUs, but the progress is incredible. And today an entire MCU can cost less than the $1.15, even in uninflated dollars, of that single tube.

In today’s light much of the copy in that ad is risible, though at the time the tradeoffs may have been less clear. “The high cost of excessive miniaturization” – gee, build an iPhone from tubes and my back of the envelope calculations suggest it would be the size of almost 500 Vertical Assembly Buildings and would cost $50 trillion, just a bit under the world’s GDP. That might discourage texting while driving, though.

“The over-all reliability of computers powered by electron tubes consistently exceeds the reliability of computers using solid-state devices.” That may indeed have been true though I have my doubts. The TX-0 experienced only 12 transistor failures in almost 15 years of operation. However, in an interview Ken Olsen, one of the TX-0’s designers and future co-founder of DEC, said of the transistors used: “They were so delicate that if you combed your hair and touched one, you burned it out.” Fast forward to today and we’re awash in pink anti-static foam and grounding straps since our components are much much more sensitive to static than in 1961.

“Equipment will meet original performance specifications without costly hand selection of replacement components.” There’s truth in that. I well remember selecting transistors using a curve tracer since their characteristics, even with the same part number, varied widely. And today that’s still true. A typical tube of the era, the 12AX7, had a pretty consistent gain. The gain of a 2N3904 transistor, however, can vary by a factor of three from device to device. We’ve learned to design circuits that can deal with these variations, though in some applications transistor matching is still practiced.

Oh, and that 12AX7? $22 in volume today. The 2N3904 is three cents in thousands.

GE was metaphorically trying to market buggy whips while Henry Ford cranked up his assembly lines. Yet there is an element of truth even today in some of the ad’s copy. And tubes are still not obsolete – your microwave oven has one in it. It would be possible to replace that with transistors, but just as in that 1961 ad, that would be far too expensive.


Jack G. Ganssle is a lecturer and consultant on embedded development issues. He conducts seminars on embedded systems and helps companies with their embedded challenges, and works as an expert witness on embedded issues. Contact him at jack@ganssle.com. His website is www.ganssle.com.

 

Loading comments...