I used madmanguruman's recommendation of a Murata power supply.
The only difference is, I used the existing sense lines from the oscilloscope board for the sense lines back to the small 8-pin connector. The reversed power switch is not bothersome.
I did try to troubleshoot the existing supply, but it was less cooperative than I liked. The PSU has a lot more parts than the example schematic for the 3842A (presumably for PFC) so there could be other failures.
My notes on that journey below.
Apply some 63/37 tin-lead solder to the pads before you try to remove parts. This technique is particularly effective on the metal frame. The existing lead free solder has a high melting point and is difficult to rework otherwise, the tin-lead trick makes it much easier to melt. Make sure that the solder flows for this trick to work. Alternatively, hunt down some indium based solder.
1) It seems that units running on 220 - 250V might be more vulnerable to failure. The main cap is rated 400V, and can see 350V in some countries. Couple that with the local heat, and life could be short. Also, the 7A fuse is overrated for 250V, meaning more energy can let loose in a failure. Finally, some 250V countries (UK, for example) have very low impedance and can provide a lot of destructive current in a half cycle.
2) Capacitor ripple current should be lower at 250V, but I think the higher voltage offsets any advantage of the lower current. A higher voltage electrolytic cap always improves life.
3) The failure appears to cascade like this:
- Main capacitor 100uF 400V cooks down to almost no capacitance
- Switching FET shorts out under voltage/current stress caused by high resistance C1
- High current flows to ground through 0R11 2W current sense resistor on top side of board, which connects the drain of the switching FET to ground. Note that the fuse could blow here, limiting the damage to the fuse, the transistor, and the cap. There's no thermistor on the TI sample schematic, but this part is also at risk in the Agilent PSU. , - 0.11 ohm resistor blows open circuit. The drain of the main switching FET now rises to the voltage across the cap, 350V in 250V regions and 170V in 110V regions. The gate, now part of a shorted blob, follows.
- With the FET drain path to ground open, the 350V now flows from the FET gate through the 2R2 gate drive resistor and into the output stage (pin 6) of the switching PSU controller 3842A. Therefore, the 3842A does not survive if the 0.11 Ohm resistor blows. The 2R2 resistor also blows.
- there is also a 2K2 resistor that connects to the sense input on the 3842A. This resistor can also blow, if the shorted internal 3842A drive path goes open circuit. The high voltage will flow to ground through the sense circuitry.
- There's a final path to ground through the optocoupler phototransistor, but these transistors usually have a quite high current rating and seem to survive.