The case of the resistor turned strain gauge
For you younger engineers working on today’s modern data acquisition systems, back then there was only one ADC for all of those parameters. The front-end programmable-gain amplifier was a classic differential amplifier constructed of discrete components that consisted of a pair of mirrored transistors, each driven by constant current sources made from a pair of current mirrored bipolar transistors. The gain resistor shared between the input transistors was a thermally stable Kelvin wire-wound resistor.
The day before we were going to fly Ship 780 for the historical first flight, all of the analog parameters assigned to a particular RMDU began to drift with temperature. Not wanting to swap out the ADC module (ADP-M), which would mean recalibrating 1000-plus measurements, I pulled the unit and took it to the lab to see what the problem was.
After running some informal thermal tests with a hair dryer (aerospace talk for "heat gun"), I verified that it was indeed the ADP-M causing the problem. After disassembling the module and putting some extension cables between the board so I could get a scope probe on the front end, I verified my first suspicion that was where the problem was. The Kelvin resistors had the lowest thermal coefficient of any resistor available at that time (and probably still today), so I didn’t suspect it as the troublemaker. I replaced the transistors and examined the PCB traces under a microscope for cracks. After reassembly, there was no improvement. I pulled the Kelvin and put some test leads on it and heated it. Again, no change.
I was fiddling around with the Pomona clips while they were attached to the resistor and noticed a resistance change. It was a repeatable phenomenon. The Kelvin resistor is a small square block with the leads exiting on the bottom like a radial lead electrolytic cap. If I pulled on the clip leads in opposite directions, the resistance changed by a significant amount. The resistance returned to its original value when the tension was released. It was acting like a strain gauge.
Nobody in my group had ever seen anything like this before. As I mentioned, the Kelvins were some of the most stable resistors on the market. Out of curiosity, I called the manufacturer, which responded that it had never seen this either.
This particular resistor had developed a microscopic defect somewhere in the wire used to fabricate it that was acting like a repeatable strain gauge. I concluded that the thermal expansion of the fiberglass PCB was stressing the “strain gauge” gain resistor and causing the amplifier gain to change with temperature. After I put in a new resistor (handpicked to be the same value), everything went fine and Ship 780 went on to make history.