Trek 609C-6 high-voltage amplifier/power supply: successful repair

[D Straney]

Just wanted to document this in case anyone runs across a similar problem in the future.  Reverse-engineered schematics of the power board and voltage multiplier board are attached.

The same HV amp I looked inside before (https://www.eevblog.com/forum/testgear/trek-609c-6-high-voltage-power-supplyamplifier-teardown/) ended up being used by someone later, and eventually it stopped putting out any voltage.  I pulled off the cover and saw that (1) the high voltage DC from the voltage multipliers measured 0V, and (2) there was a nice little Lichtenberg figure on the power board that drives the voltage multiplier, reaching from the high-voltage terminal of C10 (clipped off in the photo below).

This seemed like a pretty clear sign that there was something wrong with the power supply, and so I should look at the power board and/or voltage multiplier that generates the +/- 5 kVDC, and could safely ignore the actual amplifier section and control board (with control loop, etc.) described in more detail in the teardown referenced above.

Starting with the power board, I powered the +12V control supply externally to see if it was giving a good gate drive signal to its power MOSFET, while unplugged from the AC line, by applying 14V across ZD3 from a bench supply.  There was no gate drive signal, and it drew an excessive amount of current from the +12V rail (100mA+?).  When I removed the TL494 control chip (thank Zeus for DIP sockets) the excessive current draw went away, so the TL494 seemed to be dead.  I'm guessing that the decades-old PCB material plus however many years of dust (even with a pass with a shopvac by the group using this amp, the board had a bunch of dust visibly stuck to it) allowed the 5 kV supply to reach its sparky fingers out and give the TL494 a zap.

After replacing the TL494, giving it external control power again now showed a reasonable current draw (10-20 mA, if I remember correctly) on the +12V rail, and it generated steady gate drive pulses to the power MOSFET at ~100 kHz.  However, the PCB material obviously couldn't be trusted, even with a more thorough cleaning, so I couldn't power it up in this state.
I bought these parts to recreate the high-voltage sense divider in an "air-wired" arrangement, so that the high voltage wouldn't have to touch the compromised power board:

• Phenolic spacer: Keystone 365
• Quick-disconnect tab: Molex 19705-4201
• High-voltage 25MΩ resistor: Ohmite SM204032505JE
• 4x 120pF 3 kV ceramic caps: TDK CC45SL3FD121JYNNA

(There were more ceramic caps too in case the value needed to be adjusted, but this was the combo I ended up using)
One of my concerns about replacing the high-voltage sense-compensation cap (C10 in the schematic) was that since this design is from The Dark Ages of Ceramic Capacitors, this particular part has Z5U dielectric, which is awful in pretty much every way: varies wildly with both temperature and applied voltage.  The clipped-out C10 measured 448 pF for me at 0V on an LCR meter (meaning the "500M" marking must be for "500 pF" instead of "50 pF"), but I had no idea how much the capacitance would decrease when it had 6 kV across it.
I looked at a similar-looking Z5U ceramic disc cap from Vishay (the 615R series), and the graph in that datasheet showed that the Z5U variant could "typically" expect to lose ~75% of its value @ 80% of the rated voltage (5 kV / 6 kV rating = 83%) = an estimate of ~125 pF in this case.  I used a 2-in-series/2-in-parallel combo of 4 of these 3 kV modern caps (which don't have awful DC-derating!) with a nominal value of 120 pF.

The other engineer designed and 3D-printed a nice little plastic clip to hold the phenolic spacer.  After taking this photo I spread a lot of silicone (non-corrosive) RTV around the junction between the clip and the PCB, and everything seemed pretty mechanically solid.  I'm sure there's much better ways to do it though.


To test the power supply, especially being uncertain about its stability with the possibly-different C10 value, I used a function generator in burst mode to drive an optocoupler connected to the TL494's error amp, so that I could enable the TL494 for only a single short pulse, while capturing the output voltage ramp-up on a scope to make sure it's behaving properly - this is a favorite technique of mine for bringing up new power supplies, as there's so much less that can go wrong in a 10s or 100s of µs period, rather than just turning it on and praying; has saved me from blowing up faulty designs plenty of times.  Anyways, the high-voltage rails showed no signs of working incorrectly, so I gave it longer and longer runs until I was convinced that it could run continous, safely.

Observing the output voltage while giving the setpoint input a voltage step showed both the amp's output, and the internal high-voltage rails behaving correctly.