Jim Williams was always intrigued by RMS measurement. Implicit computing RMS-to-DC converters like the old Analog Devices modules, and later the monolithic devices, were interesting to him from a business standpoint. These things were about as complex as a dual op amp, yet sold for a lot more than a dual op amp. Try as he might, he could not match the performance at the same price point by the time all the temperature compensation was done. He admired ADI's ability to price it just below the equivalent discrete circuit function and get better performance.
The other way to measure RMS voltage was the thermal method. One of the instrument manufacturers always used thermal RMS measurement in its products and achieved higher performance than the implicit computing chips (at a much higher price, using custom silicon and exotic proprietary packaging). From time to time, Jim would propose ways of making a thermal RMS detection IC, and would call me to bounce ideas around. We agreed that the biggest issue was the packaging. He kept asking me if ADI was planning to make a thermal RMS IC, and I was always evasive.
One year at Jim's annual ISSCC Analog Dinosaur party, I wandered into his home lab and noticed he had one of the newer ADI RMS chips under the microscope and had been tracing out the circuit. I decided to have some fun and added a few random connections here and there on the reverse-engineered schematic so that his attempts to simulate the circuit would have issues and slow him down a bit. A few weeks later I called to tell him what I had done. He took it well, but I knew he would get revenge one day.
Partly in response to that episode, Jim decided to take a shot at making a thermal RMS-DC converter. The trick really is in the packaging. While most IC chips are packaged in such a way as to minimize the thermal resistance between the chip and the package (and off to the ambient air) so the power dissipated in the chip does not cause the various junctions and passive components to heat up and drift, a thermal converter needs thermal isolation to work. After all, the idea is to use the signal voltage to heat something up enough to make a thermal feedback loop with another thing.
The lateral thinking arose when Jim and Bob Dobkin realized they had the circuit already done: the LM199/299/399 "Super Zener" voltage reference, originally developed at National, and later second-sourced by LTC. This device achieved extremely low temperature drift by using a heater resistor, feedback amplifier, and diode sensors to heat up the die to an elevated temperature (around +85°C) and keep it there, regardless of the outside environment. The packaging of the LM199 included a plastic thermal insulator and special die attach to isolate the die from the package, minimizing the amount of power needed to raise the chip to +85°C.
It was simple to rewire the die to turn it into a thermal RMS detector… and it worked. There was more work to do on the die-attach technology to get enough thermal isolation to achieve good dynamic range. Months later, we called it "bubble-gum," since it was literally full of bubbles.
Jim submitted a paper on the chip, which became the LT1088, to the 1986 International Solid State Circuits Conference, and it was accepted. When the paper ("A 25 MHz Thermally Based RMS-DC Converter") was presented at the conference, Jim was quoting the crest factor performance of the device. He stated in his presentation that it was 1 percent accurate up to a crest factor of 10, but the published text said 9. Lew Counts, who was responsible for a lot of the RMS-DC converter designs at ADI, took Jim to task, demanding in the open Q&A session to know if the real spec was CF=9 or CF=10, and declared loudly that "9 is not equal to 10." In the later LTC AN-22 App Note, Williams referred to this incident in Footnote 5, calling it the "Counts Theorem."
In both the ISSCC paper and the App Note, more hidden mischief is included. Jim acknowledges the contributions of several people, including among the real names a mysterious co-worker named "Repus Renez" for assistance. Spell it backwards, it was none other than the "Super Zener" IC. And in the ISSCC paper he also credits a "D. Grant… for inspiration." This may be the only time a competitor was credited in an ISSCC paper.