Considered Opinions

Charlie Ellenberger

At a time when there was no such thing as a semiconductor memory, electronic memory devices were constructed from millions of tiny magnetic cores with wires running through them in long strings. On either power supply rail of the current driving circuitry for each string of cores was a “high-side” switch and a “low-side” switch, requiring a PNP “core driver” and an NPN “core driver” bipolar transistor to drive the magnitizing current of the cores. We had been making transistors that exhibited the very low “storage-time” required^[i]… the major component of the time, typically in nanoseconds, needed for the transistor to switch off, by gold-doping the Collector to minimize the space charge there with recombination centers^[ii]This method was later replaced by what is called a Schottky Clamp, across the Collector-Base junction, which prevented the device from going into current saturation, but that technology only became available later with the advance of integrated circuits..

At the time, the conventional wisdom was that it was not possible to gold-dope a PNP junction transistor because the gold would find its way into the Base area and kill the carrier lifetime there, thereby limiting the h_fe, ^[iii]The h_fe, also referred to as β, of a bipolar device is the factor by which the base current is multiplied to produce the amplified Collector current of the transistor.. Since nobody had told us at the time that it was impossible, we did it anyway and it worked just fine.

Later on, when the first P-channel MOSFET came out of the development lab^[iv]The new devices, known as 0049, would ultimately be assembled in matching pairs into 6-lead TO5 style headers and sold as “analog-switch-signal-choppers” to General Dynamics for its space program. in Palo Alto, it was transferred to our production department in Mt View because it used the same doping impurities and similar diffusion profiles as the “core drivers” we had been making. That was a great opportunity for us^[v]At the time, I was the day-shift foreman on the PNP wafer fab line at Fairchild, Charlie was the Supervising Engineer for the Department. to get in on the ground floor of a whole new technology – one that would ultimately open up one of the few remaining closed windows^[vi]One other such closed-window was the inability to make a normally-off, enhancement-mode, N-channel device because of Q_ss, but that is a story for another time. to Complimentary-MOS Integrated Circuits that dominate the high-tech world of today.

The new device was unique in those early years of the industry, because it was not possible to grow or deposit Gate oxide in such precise thicknesses to maintain the required threshold voltage needed. Instead, it had to be grown thermally and each wafer individually “etched-back” by hand to the desired thickness. This was done by dipping wafers in HF^[vii]Hydrofluoric acid was the only available method for etching the oxide at that time. for a few seconds, then rinsing and examining each under a microscope.

As I recall, that optimum thickness was around 1,100 Anxtrom Units^[viii]… a unit of length equal to one hundred-millionth of a centimeter, 10⁻¹⁰ meter, used mainly to express wavelengths and interatomic distances. Whatever thickness it was, the standard for measuring it was to view the wafer at 600X, through a metallurgical microscope, under incandescent light, and match the color of the interference pattern reflecting from the oxide layer to the color of Charlie’s eyes.

And so, was the state of the art of semiconductor technology in 1965.

By: Jim
Written: March 2022
Published: March 2022
Revised: March 5, 2023

Reader feedback always appreciated^[ix]… thoughtful commentary perhaps more so than shallow thoughts