[The Signal Path] snagged a fancy Rohde & Schwarz vector signal generator that can go up to 3.2 GHz, but sadly it wasn’t in working order. It powered up and even put out a 1 GHz signal, but the amplitude output was very wrong. Interestingly relative changes to the output were correct, it was just that the absolute output amplitude was off by quite a bit and changed with frequency. That started a detective job which you can follow along in the video below.
The instrument is pretty high-end, and did not report any problems even during self-check. This implied that all the internals were probably good and whatever was wrong probably lay close to the output. The service manual’s block diagram wasn’t terribly useful, especially given that all the processing portions appear to work well.
This is a good excuse to just look at the inside of this well-built instrument. Given that the output amplitude changes with frequency, a good guess can be made that the issue will involve some component or components that involve a reactance either on purpose or due to some failure mode.
The obligatory cat didn’t seem to have much of an opinion besides finding the instrument a satisfactory perch. Tracing through the output shows some fast diodes to sample power and trigger a reverse power protection relay. Those diodes were easy to see through their transparent cases via optical microscopy.
As it turned out, the MOSFET controlling the relay was always holding the relay off, and replacing the MOSFET restored normal operation. We don’t think of a relay as a capacitor, but a tiny relay in the off position does exhibit capacitance at these frequencies. Had it been on as it should have been, there should have been no attenuation of the output. But because the failed MOSFET kept it open, it provided a capacitive path that reduced the amplitude of the output signal.