By the late 1960s, MOS had already made it. The instability problems that plagued MOS transistors had been licked and many manufacturers offered p-channel and n-channel ICs. Then one day at RCA Laboratories in Princeton, N.J., one of the engineers had a thought. Wouldn't it be neat if we could create complementary n-channel and p-channel transistors on one chip? Such a chip, he reasoned, would chew up very little power. It would consume power only in switching between the n-channel and p-channel devices.

It wasn't easy. The problem lay in fabrication because you would have to make the n-channel and p-channel devices on the same substrate. So it was necessary to dig a deep well for the n device and deep diffusion was tricky. Then, in 1972, along came the electron-beam process, which eliminated that problem.

Al Medwin, who made the first CMOS products, was at RCA's Somerville, N.J., facility, which manufactured semiconductors and, more important for this story, employed managers transferred from the tube-production plant in Harrison, N.J. His earliest CMOS ICs included some dual flip-flops, some AND and some NAND gates. RCA called the technology COSMOS, for complementary symmetry metal-oxide silicon. Others called it CMOS.

Medwin thought there might be a great future for CMOS ICs. CMOS, he recognized, was a unique device. It was almost a perfect switch and consumed incredibly little power, so you could pack large numbers of circuit elements on a chip. For engineers like Medwin, whose background lay in vacuum tubes, there was another allure. You could use a single positive-voltage supply, like the B+ voltage in vacuum-tube circuits.

Medwin's earliest devices used a +15-V supply, then quickly came down to +5 V. He and his colleagues never dreamed that CMOS circuits would one day operate from a 1-V supply, but they were hoping to get down to 3 V so that CMOS could be used in watches. It was easier, though, to use a simple converter to boost the watch-battery voltage to operate the CMOS circuit.

The main emphasis, though, was on digital circuitry, Medwin remembers, because CMOS was such a nifty switch. The immediate market he and his colleagues went after was the handheld calculator, a large-market digital application. They paid rather little attention to analog applications because digital was a natural for a great switch.

RCA management didn't provide much leadership, but it didn't interfere, either. Well, not much. At one point, Electronic News, a leading newspaper at the time, published a front-page story about this wonderful new technology and featured Medwin and Art Lipschutz—with their photographs. A rebuke came down from on high: RCA's engineers were not to be subjects of publicity.

A genuine IC

At the start, Medwin had 13 transistors on a chip, a genuine integrated circuit. A man with a vision, Medwin wrote an article for the in-house magazine in which he pictured a day when we might see as many as 1,000 transistors on a single CMOS chip. He remembers that people stopped him in the corridors and laughed.

It is likely that those same people laughed at the thought that transistors might some day replace vacuum tubes. In the early 1960s, RCA itself predicted that transistor sales would rise to some level, after which vacuum-tube sales would decline. RCA hired two individuals, Howard Bierman and George Rostky, to write a book that would, without bias, evaluate the relative merits of transistors and vacuum tubes. The conclusion was a draw. Each had strengths and weaknesses.

Despite RCA's prediction that vacuum-tube sales would reach a peak, then decline, management was amazed and delighted to note that tube sales kept rising. It was wonderful. It took some time before RCA management realized that the reason its tube sales were rising was that the other vacuum-tube companies were going out of business.

Nevertheless, management recognized that the potential of CMOS was great, and there was government support. In time, it was CMOS and only CMOS that made possible handheld calculators, digital wristwatches, laptop computers and a host of other products that were later taken for granted.

Despite the pressure, RCA kept running into production problems but Al Medwin found out why. He knew that a static charge could kill the gate on a CMOS device, so he provided detailed instructions on how to handle the chips in production and shipping. Yet the company kept encountering failures in the field.

Checking up, Medwin found that the people in production and in shipping weren't following instructions. Handcuffing themselves to ground was just too much trouble—they never had to do anything so silly on the vacuum-tube lines.

Management couldn't understand the need for all that extra effort either. Managers were mostly old vacuum-tube people and they couldn't make the transition.

Says Medwin: "No tube company ever made a really successful transition to solid state."

The Century of the Engineer: Misunderstood Milestones