followup--transistors, tech journals [telecom]

Right. That is why many academic researchers (including me) issue "lab reports" on the Web. They are a public record that I thought of something.

Excellent detailed historical treatment (IMHO, anyway).

V. unflattering treatment of Schockley by Joel Shurkin, who was at one time Director of News and Pubs at Stanford (or Stanford Med School) -- and who didn't start out intending to write nearly as unflattering a treatment.

Have a copy of one of the earliest editions of this book on my office bookshelves, acquired as a ref book when I had to fill in to teach our undergrad intro to semiconductor devices course many decades ago. It's a *superb* work, IMHO. Maybe I'd better keep my office door locked; just discovered that an amazon used book affiliate also has exactly one copy, listed as above, and priced at $497.97.

The invention of the transistor was a big engineering accomplishment, but what was REALLY big was for folks to decide it was worth doing in the first place.

It's true that the original point-contact transistor wasn't all that useful a device, and it took Fairchild's mesa process to make reliable inexpensive transistors. But that gets into development and not research.

Interestingly, the principle by which the FET operates was demonstrated around the turn of the century, but nobody bothered looking into it any farther.

That makes sense. These days, though, prior art does not seem to be any bar toward obtaining a patent, sadly. This reduces some of the argument for that today.

--scott

the transistor would be a powerful improvement for them. I presume they recognized it would save power and last longer right away.

Both organizations started building transistors to learn how to make them and their characteristics, and then using the transistors in different devices as replacements for vaccum tubes. It took quite some time and research before transistors could function as well as a tube, and then still more time until the price dropped enough.

For IBM, they got their early transistors by cracking open diodes and converting them, by hand, into a transistor. Eventually they got their own crystal developer ovens.

IBM made one foulup: IBM developed powerful machines at considerable expense to make transistors in quantity and gave it to T.I. TI then became very rich on this selling it to multiple customers and a competitor to IBM later on.

Transistors were only half the battle. Manufacturing them in quantity and quality and low cost was the other.

Recall hearing Eli Yablonovich comment in a technical seminar somewhere that "it only took approximately 20 years for the semiconductor industry to learn how to make reliable and usable Si oxide layers on Si wafers".

Note that these histories are... somewhat biased to say the least.

There are still some applications where tubes haven't been replaced with solid state stuff yet. They are shrinking every year, though.

This was a point-contact design. Point contact transistors are interesting laboratory curiosities but not really very practical because they cannot be made reliably. They are basically handmade one at a time.

I'm not sure where this comes from. Was this after the Fairchild process was developed or not? Fairchild's mesa process totally changed the whole notion of what a transistor was.

Absolutely.

--scott

I know of at least one transistor manufacturer in Asia that still hasn't got passivation right...

--scott

You left out the fun part of the story. They developed a fixture where a person would manipulate the contacts using two pairs of gas pliers, while watching on an oscilloscope for transistor action. The person who was most successful at making good transistors this way was a man who was a part-time chauffeur for Thomas Watson Sr.

But whether it was a foulup is open to interpretation. It was a decision of the IBM leaders at the time that they did not want to be in the electronic parts business. Something else it illustrates is the really outstanding prowess if IBM's manufacturing engineers. IBM's success at marketing is widely acknowledged; but their manufacturing engineers were equally potent. For instance, they developed means of stringing magnetic core memory planes by machine, when everybody else in the industry was doing it by hand. Memory became a cash cow for IBM - they could charge industry-competitive prices for it while their manufacturing cost was a lot lower.

This was before the Fairchild mesa process. It was mainly the 2N1301-1306 family of transistors that were used in the IBM 1401 and other small, slow, very high volume machines. Perhaps another way in which it the sale to T.I. was not a foulup was that the Fairchild process pretty well put an end to the market for the 2N130x transistor family in computers, perhaps before the machinery for them was amortized.