Re: RFI ghosts
When I was in college writing the executive for the Dartmouth time sharing system we had delivered an RCA RACE unit for storage. I was told to hook it into the system as a low-priority task when I had time. So I did but whenever I configured the unit into the system (via a configuration punched card) the system would always give an immediate "transfer timing error". The field engineers would run their diagnostics at night and never saw that fault. Eventually we decided that we should get together and figure out what was going on. I knew exactly which card in the card deck (the configuration card) the system would fail so I didn't bother to load the rest of the cards after it. When the system read the configuration card everything worked perfectly and the system asked for the remainder of the card deck!?
I was about to repeat the experiment with the complete deck when the field engineer (who was very good) stopped me and reached over to the card reader and pulled back the hopper and said, " do you hear that whine?" He then ran his diagnostics with the card reader on and pulled back the hopper on the card reader when the diagnostics were running. Instant transfer timing errors.
It turned out that the field engineers ran their diagnostics from tape with the card reader off as it was too noisy. Investigation showed that the speed of a brush vacuum motor was controlled by SCRs and someone had put capacitors across the motor windings to "eliminate" noise. When the SCR fired it put the capacitor across the utility supply line causing 70 volt spikes. This got into the peripheral wires and caused the errors.
We quietly removed the capacitors and things worked better. The RACE unit never did work reliably - shooting out springs and crumpled cards but at least we solved one problem.