Capacitor keeps exploding on nixie HV power supply

Question

I purchased the NK104B HV power supply from here to power some IN-4 nixie tubes. Whenever I turn it on, the 3.3 uF 200 V capacitor explodes. The first few times I used it (for a cumulative ~5 minutes or so) it worked fine. Then, I attached it to an SN74141 nixie driver IC and arduino. I used this to cycle through the digits of the tube; after ~5 minutes of this one of the capacitors began smoking. I turned it off and attached it directly to the nixie tube again, and after a few minutes the capacitor suddenly emitted a large amount of smoke (although the nixie tube stayed on).

I thought this might have been for two reasons. First, I had replaced a missing component (the 200 uF 25V capacitor) in the kit with one of the same specs from amazon. However, I realized the one specified by the kit was actually supposed to be low ESR. Second, I had been running the kit at maximum voltage, and a voltmeter showed something more like 205 V. The blown capacitor was rated for 200 V, so perhaps it was overvolted? I ordered a replacement for both capacitors (getting the exact models on the BoM this time) and tried it again at 160 V. Like before, after a few minutes the 3.3 uF 200 V capacitor blew up (and again the nixie tube stayed on, even though I quickly disconnected the power).

Does anyone have any ideas on what might be wrong? Since the converter worked until I hooked it up to the arduino, I think I might have damaged it. But I'm not sure what components, other than these two capacitors, would be susceptible to damage. Is there an obvious component failure that might be causing overvoltage on the capacitor?

Any idea what might be causing this? Should I just buy a new supply? Thanks for any advice you guys have. Below is a schematic of the HV supply, a schematic of how I hooked it up to just the nixie tube, and a schematic of how I hooked it up to the arduino and SN74141.

This last one is how I attached it to the arduino, except without the coupling capacitor to ground attached to the SN74141 Vcc. Vcc came from the arduino +5V, and I powered the arduino with the same +12V and ground as the HV power supply (both are from a 12V AC adaptor). The power supply had the same setup as the previous schematic, except pin 5 on the supply was connected where it says 180V, and R32 = 10K.

Answer 1

From MC34063 datasheet the 680pF cap on pin CT sets the frequency to something like 45kHz. It's not very accurate but let's go with that.

A 45kHz period lasts ~22µs. With 12V input and 160V output, duty cycle should be 92.5%, perhaps a little bit more due to inefficiency, so let's go with 20µs ON, 2µs OFF...

e=L di/dt, so in 20µs the 100µH inductor powered by... say, 11.5V, ramps up to 2.3 Amps.

Oopsie, that's too much, the current limit sense voltage on MC34063 is 0.3V, and the current sense resistor is 0.25 ohm, which sets the current limit threshold at 1.2A. So, on-time will be determined by the current limit tripping on each cycle, and it will work in discontinuous mode. I don't see any mention in the datasheet about how long it takes for the current limit to trip, but anyway.

The inductor current ramps up to 1.2A. After that the MOSFET turns off and it dumps the energy into the output cap.

Therefore the current ripple seen by the output cap is 1.2A, which exceeds its 62mA ripple current rating by a factor of 18. Also it is rated for ripple at 120Hz, and it is used at a much higher frequency, which compounds things for the worse. It has huge ESR also, probably more than 50 ohms, so when the inductor tries to dump current in it, the very high ESR means voltage will rise above the rating of the cap, accelerating its demise.

It's a big problem with high ratio boost converters. The MOSFET, diode, and output cap have to withstand both high current and high voltage. With a transformer based converter, or flyback, components on the output side would have to withstand the voltage, but they'd get much lower current.

Note the cap will only overheat if the ripple current is high enough. If there is no load on the output, then the converter will spend most of its time sleeping, waiting for the output cap to slowly discharge through the output voltage divider. When it discharges enough, it pumps a cycle to recharge it, and goes back to sleep. So it makes sense the cap didn't pop when the converter was just idling without nixies connected.

A solution would be to use a cap with a much higher ripple current rating.

You could use a film cap, which is going to be huge. Or perhaps a ceramic cap, that would have to be a bit higher in value to compensate for the capacitance loss with applied voltage. Something like 4-5 stacked 1µF 250V X7R. The easiest would be a polymer cap, although it only has 450mA ripple current rating, that's at 105°C ambient. With a more civilized ambient temperature, it should take more. You can always use the 6.8µF version which has higher rating. I wouldn't go higher, as capacitors charged to high voltages are best kept as small as possible.

Another solution would be to use a better designed converter, transformer or flyback, or if it absolutely has to be a boost, something with higher frequency and ceramic output caps.

PS: At this voltage, it is not recommended to check the temperature of the components with a finger.