Bob Pease: His last challenge, Part three--The precision resistor
Editor’s note: Edwin Pettis, recounted the following story for EDN readers regarding his experience in working with Bob Pease on a very difficult piece of a system design that included precision resistors which needed to be ultra-stable over temperature, including their own self-heating temperature. This system design, an ultra-stable voltage reference for a high power constant current source, is described in Part one and Part two of this series.
Here is what Pettis said about the following story about his experience in working with Bob Pease in his last challenging design effort, “I suppose you could call these the ‘highlights’ or ‘low lights’ depending on which point in time you were at. There are a number of first attempts which never had drawings and were dropped for various reasons, one of which was the TCR problem, but the requirement for such insanely low warm up drift, from cold to hot, killed most of the early prototypes anyway.
The precision resistor design effort
The other day I found a drawing that Bob had sent me of the rough schematic of the DAC control of the CCS circuitry, Bob had said they were having quite the time finding components good enough. Bob seemed a bit frustrated in that he couldn’t seem to find out some of the relevant information needed about the ‘load’ that this PS-6 was going to drive, ‘they’ (the Russians?) seemed to have only let bits and pieces trickle out after Bob complained long enough.” Pettis continues his story:
It was December 17, 2009 when Bob contacted me about a resistor (Bob referred to them as the “big hairy resistor”) he needed for a project, the rough first specifications were:
0 to 4.4 amps variable, long term stability, 100 PPM (preferably 50 PPM) drift from cold to hot operating conditions, tolerance of +/-0.1% or +/-0.2%, 5 watts minimum and 1/4Ω.
The crux of the problem was the variability of the constant current source; the ‘resistor’ had to hold it initial value within 50 PPM from cold start to full load. Initially, Bob thought he could go buy a hundred ¼ watt, 24.9 Ω, 10 PPM TCR, 1% metal film resistors, parallel them with Kelvin connections and a “huge fan”. There are problems with this approach, the TCR, even averaged with 100 resistors was very likely to be too high, even with the fan and there are real problems with paralleling resistors in a Kelvin configuration, particularly with low ohm resistors. After some debate we decided that we could go with a 0.2 Ω resistor (Feb. 3, 2010) for current sensing because Bob was confident he could keep the electronic part of it accurate and stable enough to work over the required range. At this point the parameters had not been completely specified for the PS-6 and would not be for some time come.
A single resistor could not be made that would hold all of the anticipated parameters (we were guessing at what some of them might be until we had real specs in our hands). Another specification which was also troubling was the overload condition, while Bob worked on handling that part of the problem in the electronics, I also found the overload to be a significant problem, particularly since it was required that the current sense resistor maintain its original value as closely as possible after a current fault, which Bob calculated to be 20.98A, duration was unknown at this time. These units were to run 24/7 and would not be shut down except in the case of a fault.
I made several single element resistors of 0.200 Ω and 0.250 Ω values, potted them and began running tests on them over the next few weeks as I waited for more input from Bob. During this time, I found that these potted resistors would not meet some of the specs. I had acquired several samples of precision metal film power resistors in a TO-247 case with Kelvin leads to test as well. Initially, they had some promising specifications. Testing indicated that they could hold the required TCR if the lowest TCR available was used and sufficiently high air flow on the heat sink was used. Initial tolerance was acceptable but further testing showed that while short term stability was adequate, the initial value drifted further than was preferable, month to month and long term drift was unacceptable. Performance under fault conditions, even of relatively short duration, caused irreparable damage and changes in the resistor’s characteristics. Bob was seeing similar results although at a different time.