Repair of TTi TSX1820A power supply (spoiler: shorted electrolytic)

[macboy]

Here is my short story of my repair of a TSX1820A power supply, and the unusual component failure that caused it to fail.

I had pulled this power supply out from under the bench after having not used it for probably 3 or 4 years. It seemed to work well, but after about 10 minutes in operation, I heard a slight "bzzzt", then another, than a louder "BZZZT" and I turned my head just in time to see the flash of the internal 10 Amp mains fuse popping. In addition to this, another power supply connected to the same power bar, a Agilent 3620A (dual 25 V, 1 A, linear) also stopped working. More on that one later. My "dim bulb" tester confirmed the TSX1820 had a hard fault, the 100 Watt bulb lit at about half brightness, when I expected it to not visibly light.

The TSX1820 is a 18 V, 20 A power supply which uses a SMPS as a pre-regulator, followed by a linear regulator, to try to take advantage of lower noise and good efficiency. On the mains side, it has a 10 A fuse then some filtering (the usual X caps across the line, Y caps to ground, common mode choke), then a soft-start circuit, and finally, the bridge rectifier and main filter caps. No fancy PF correction here. The caps are a pair of big 1500 uF 200 Volt in series. After the caps, there is a jumper block that can be removed to isolate the SMPS from the ~340 VDC supply, so that each can be tested and/or repaired separately and more safely. In my case, removing this jumper did not remove the fault, so the fault lies in the above described circuit. At least I was not likely dealing with a fault in the SMPS IC, MOSFETs or (much worse) its transformer. I am also fortunate to have a full service manual with schematics, circuit theory, and troubleshooting information. Sadly, they literally don't make them like they used to.

This supply sat around in its sad defective state for a few months before I jumped into fault-finding. It didn't take long to find a shorted TRIAC (in the soft-start circuit) and a faulty capacitor. One of the big 1500 uF 200 V filter caps was a dead short, measuring as a 0.04 Ohm resistor on my LCR meter. I dug around my salvage parts and found a TRIAC and a couple of caps I could use for testing (at just 330 uF/450V they were too small but OK for light load). With these components replaced, the supply powered up just fine and worked well. A couple days later (today), Digikey delivered the permanent replacements. The supply is now fixed and working great.

About that Agilent 3620A: It worked after replacing its mains fuse (1 A fast), but would occasionally blow the fuse even though it appeared to work perfectly fine, and didn't appear to draw excess current even at full load. It turns out that an X2 filter cap across the line was (I assume) partly failed and would sometimes conduct too much, probably whenever there was a slightly higher mains voltage and/or when there was a lot of noise on the line. Anyway, replacing that cap solved the occasional fusing issue and it has worked fine ever since. One plausible explanation is that when the TSX1820's capacitor was internally arcing, it caused significant noise on the AC line, which this X2 cap absorbed, causing it to partially fail.

I wanted to see what was the matter with that failed electrolytic cap, so I opened it up. The photos below speak for themselves. There were two separate spots within the roll which had arc-through. According to Nichicon, short-circuit failure of electrolytic caps is very rare, but this cap was a Philips, so maybe it didn't know that.