Capacitors: Can’t Live With ‘Em, Can’t Live Without ‘Em
And there's another repair in the books where the failure point was a $0.10 capacitor. I looked at waveforms around all other electrolytic capacitors on the board and did not find any showing significant noise across their terminals, which lends some credibility to my hypothesis that the single capacitor failure may have been an isolated material, manufacturing, or handling defect. Then again, the other capacitors aren’t being fed from a charge pump at 2kHz. The main charging supply rail capacitor is powered from a 300kHz flyback circuit, the 12V rail is powered from the battery/charging rail by a linear regulator, and although I didn’t find it, the micro-controller must be powered by the 12V rail through a linear regulator as well. I must admit I did not really hunt for the 5V as all I needed was to measure it and verify that it looked good. If a supply rail looks good near its load, it will usually be at least as good closer to its source.
While scratching my head about how the negative output spikes appeared to cause the micro-controller to reset, I decided to have a look at what sort of waveform the micro-controller receives from the LM358 by simultaneously probing both of the LM358’s outputs and the micro-controller ADC input. Holding three probes with one hand so I have the other free to take pictures/pause the oscilloscope can be challenging. When I probed the ADC pin prior to replacing the cap, the input signal looked similar to the chopped up LM358 output. Because of that, I did not bother saving it. After the repair though, everything made a lot more sense: the second LM358 channel is used in conjunction with the diodes around it to simulate an ideal diode bridge and present a 0-5V rectified input to the micro-controller. This eliminates the need for the micro to have an ADC capable of dealing with negative voltages. Unipolar chip pins rarely appreciate being pulled more than 500-700mV below ground as this may cause their transistors and diodes to uncontrollably conduct when they aren’t supposed to. Depending on the chip and circuit design, this can lead to issues ranging from random malfunctions to catastrophic failure.
There you have it: if you happen to own a circa-2005 APC BX/RS/XS 1000/1300/1500 UPS with a chronic clicking behavior, you now know that C41 (or whatever the negative supply bypass capacitor near the op-amp handling line voltage sensing might be labeled) should be high on your suspect list. I know I opened this story by saying that I was highly optimistic for a simple definitive fix, but I still imagined I was going to find more than one failed capacitor involved.
What am I going to do with my newly-restored BX1000 when I already have more battery backup than I know what to do with? I have at least two ideas, but they will have to wait until I have more test equipment and parts to experiment with. If there is something in particular you would like to see me do, though, let me know in the comments!