I work at a National Laboratory, and was asked to put together some special instrumentation which needed three PMTs (photomultiplier tubes). This little side project had zero budget, so I proceeded to gather the necessary parts from the plentiful store of used equipment we have available.
The PMTs are exquisitely sensitive light detectors, but each one runs at a slightly different optimum value of high voltage. Since these were used tubes, it was necessary to "characterize" them, running a plot of their output vs. high voltage input to determine the best operating point.
I set up a rack of test gear and proceeded to characterize each tube, powering them with a small high-voltage supply that had recently been taken out of service from another instrument setup that was no longer needed. The supply had a rather small analog meter for reading the output, too small to trust for reading the values that were going into my spreadsheet. It did have a rotary range selector and three digit turns counting dial, however, and experience had shown that these worked well. The tubes all responded reasonably within expectations, and soon I had the response curves I needed.
I obtained two more high-voltage supplies and set up all the tubes, carefully setting each supply to the optimum operating point. To my surprise, only one tube was reading back: the one connected to the original high-voltage supply. I swapped cables a few times and found that all three tubes did indeed still work, but only when connected to the first supply. I measured the output of both of the other supplies and found that they were spot on, exactly the values requested by the range selector and turns counting knob. The coaxial high-voltage connectors I was using made it impossible to directly measure the output voltages when connected to the tubes, but the supplies should have had ample current capability to drive these loads.
After multiple checks to make certain I was seeing what I thought I was seeing, I decided to step away and come back to it later. I was thinking that perhaps I could find a coaxial "T" that would allow me to measure the supplies under load.
The following morning, a thought occurred to me. The only supply I hadn't measured was the one that was working! I fired up the rig when I got to work and found that, indeed, that first supply had a problem which caused it to deliver about 60 percent more voltage than was being requested. The operating points I had measured were actually way too low, and the voltages the other two supplies were providing were insufficient for the tubes to operate.
So the supplies that were "not working" were fine, and the one that was "working" was broken!
The main confounding factor was not having data sheets for the tubes, which meant their approximate operating voltage was unknown. Also, the misbehaving power supply had recently been in service but was not calibrated there. It was simply run up until the system began working without noting the knob settings. Hindsight is always 20/20.
This entry was submitted by Todd Johnson and edited by Rob Spiegel.
Todd Johnson worked at Fermilab for 29 years, first as an accelerator operator and then as the operations specialist for the recently retired Tevatron.
Tell us your experience in solving a knotty engineering problem. Send stories to Rob Spiegel for Sherlock Ohms.