The Transistor, Part 3: Endless Reinvention | Hacker News Reader

The Transistor, Part 3: Endless Reinvention

technicshistory.wordpress.com

155 pointsHooke11 comments

This series of longform articles is incredible. The first batch are on the history of the switch and probably add up to a shortish book in length. There's a nice focus on how at several steps there were many who discovered a new innovation, but there were fewer who had a clear grasp of what their innovation meant and could carry the field forward. Can't wait for this author to publish a book.

I'm amazed at how fast things moved in the computer industry. Per the article, FETs only just began to work in 1960, but by 1964 Seymour Cray at CDC was completing the 6600[1], a "supercomputer" with massive numbers of them.

[1] https://en.wikipedia.org/wiki/CDC_6600

I'm pretty sure the 6600 was done in 10K ECL, which was a bipolar technology, not FET. In ECL, the transistors are biased such that they never saturate, which decreases switching time at the expense of power consumption. Also, the inputs are configured as a diff-amp, which helps switching speed also.

Some time after the 6600 (1980's) I was a logic designer working in 100K ECL and compatible gate arrays (144 transistors per chip). Each gate array required 5 Watts of cooling. We had one 20,000 BTU Ton water chiller for the CPU, and that was the design-limiting constraint.

I have next to none experience with ECL chips, but recall in the 80s the 11C90 ECL divider which was the only affordable choice if one needed a prescaler for his frequency counter to measure frequencies of several hundred MHz. That chip ran hot compared to other TTL ones, but a few years on, probably late 80s, I used a 8 pin DIL ECL chip for a PLL (can't remember the name) which I recall running much cooler than the 11C90. They probably improved the technology before it got surpassed by more modern ones.

If large numbers of FETs are your thing, you might be interested in The Beast with a Thousand JFETs [1], a discrete audio amplifier dissipating about 140 watts per channel and producing output power of 70 watts with a 2 ohm load.

Other articles/designs by Nelson Pass are certainly interesting as well. I'm not much of one for audiophile stuff, but he has a knack for interesting designs.

[1] http://www.firstwatt.com/pdf/art_beast.pdf

I'm always amazed by the story of Shockley (repeated here). He was surely a smart guy, didn't he at any point stop to think that his own behaviour might be the root cause of his problems, and at least not in his best interests?

The more successful someone is with a behavior set X, the less likely they are to feel that they should change anything about X. Today’s best example of this is Trump.

That should hardly be that surprising or unfamiliar here on HN. Epithets like 'sperglord' didn't come from nowhere - the rigid unbending nerd on a righteous self-destructive crusade on a selfish or minor point whose folly is obvious to anyone but themselves. Shockley is particularly striking because he cost himself so many billions of dollars, but there are a lot of failed startups or projects with similar tales.

The photo of Shockley's lab in Mountain View is at least two teardowns out of date now. The small plaque visible in the picture is gone in the latest construction, but in its place there is a larger plaque, and several huge, towering sculptures of electronic components. It's a pretty neat display.

This series was very well researched and written. He is starting a new series on the history of the Internet. I'm really looking forward to it.[0]

[0] https://news.ycombinator.com/item?id=17322090

holy... fantastic material

Reply on news.ycombinator.com