Very high voltage, series connected, electrolytic capacitor bank

[Mads Barnkob]

I got 35 of these BHC 1500 uF / 450 V electrolytic capacitors that all comes from the 3x 30 kVA UPS that I took apart a while ago, as these come from a continues duty equipment they are properly to worn out that a little fun could be had with them.

I want to try all 35 capacitors in series for a 42.8 uF / 14000 V bank, when fully charged that would store 4194 joule and a charge of 0.5 coulomb, if we compare it to all them in parallel, that would be 21 coulomb instead, so the higher voltage might actually be good on the capacitors when discharging them, assuming there will be a much lower short circuit current alone from the "high" ESR.

A single capacitor has a ESR of 7.1 mOhm, so all 35 in series would have a ESR of 2.485 Ohm, short-circuited at 14000 V that would result in just under 6 kA short-circuit current, which is about 1/10th of what my other 4000 joule single capacitor can deliver in short-circuit current.

Do anyone of you have experience with making a extreme series connected capacitor bank like this?

[Mads Barnkob]

The 450 VDC capacitors has a surge rating at 495VDC for some 1000 non repetitive pulses, so a balance between that and not burning too much power in the balancing resistors themselves, I chose 82K resistors. The discharge curve can also be seen in the attachment. The 100 Watt burned from the load balancing and discharge resistors does that discharge after charging has to happen fast, but at 400 VDC across each capacitor, it will be at its 4kJ charge. The high losses is the price to pay for the increased safety margin on a safe level of maximum voltage across each capacitor.

For each capacitor I will add a red LED in series with a 120K to 220K resistor, this corresponds to approximately 4 to 2 mA through it at 450VDC, ~2 to 1 watt. This is only for charged indication on each individual capacitor.

I will have a analogue meter across the entire bank to measure voltage, a long string of resistors for the voltage dividers adequate voltage rating is necessary.

[Mads Barnkob]

All the capacitors have been soldered together in series and have their balancing / discharge resistors mounted. My biggest worry about this simple construction is that it will blow apart at the first possible chance, a better way to sure the capacitors and also to mount resistors/LED would be to make some PCB strips that mounted in all 4 snap-in pins of the capacitor.


The complete bank measures 48.17 uF, so that the capacitors are still above specification is not a bad sign on their condition and that gives me an extra 500 joule, so the total charge in the bank at 14000V will be 4.7 kj.


LEDs with 630K current limiting resistor in series have been mounted.

[futurist]

Looks very nice, why would soldering like this be a problem? I did similar, worse job for my MMC caps, without additional wire, just bending the leads and soldering them

[Mads Barnkob]

This is intended for pulse discharge applications, so the theoretical 6000 Ampere peak discharge current (that is only the ESR, not external circuit) will do its job on any part of the circuit, weak soldering joints or pieces that can not withstand the electromagnetic repulsion between conductors can be ripped off and send flying in a spray of sparks

[Mads Barnkob]

The first charge test was done with a OBIT that can output 10kVAC at 0.023A, half wave rectified that would give me roughly 14kVDC, charging the bank through 3x47K power resistors to limit the current, I managed to charge the bank to 8kVDC before it would stall.

I could not move the kVDC meter due to the bank being charged, so I am not sure if the remaining 6kVDC was across the charging resistor, but that would indicate that there is something wrong with the ESR of the capacitor bank itself.

Next test I will try to omit the charging resistor and just bring the voltage up slowly with a variac.

Enjoy the 1.5kJ bang at the end

[Mads Barnkob]

Great work Barnkob!

Thank you, as soon as I have part 2 of the Ikea induction stove published, within this week, I will resume testing this bank at full voltage Just needed the work bench cleaned for that!

[Mads Barnkob]

I had time to set up the experiment once again, a clean work table was needed!

I tried with the original OBIT but was only able to push it to 9 kVDC without the charging resistor (was 8 kVDC in previous try), I also tried with two larger OBITs in parallel and was still only able to push it to 10 kVDC.

I did not capture any screenshots from the oscilloscope, but the peak current is rather low, as it is expected from the relatively high ESR from a electrolytic capacitor used in a pulse discharge like this. It was around 4000-5000 Ampere peak when charged to 10 kVDC.

/>
I will have to do some more tests to find a way to charge this to its full potential of 14 kVDC. Which I really had expected that the 10 kVAC OBITs should have handled just fine.

[Mads Barnkob]

With a rearranged test setup, but the same two OBIT power supplies in parallel, I have been able to charge the bank up to 13 kVDC when being patient, but charging to 12 kVDC is fast and easy, when getting above this, the balancing resistors simply eat too much of the charging current.

At 12 kVDC the bank holds 3.5 kJ and that is also satisfying enough, however the long delivery time does make the explosions rather dull and the server motherboard might have too many parallel lines that dissipate the energy rather than getting some good damage to the traces.

Considering that the estimated short circuit current would be around 6 kA due to ESR, the measured results of 7.36 kA is fine, with error margins these two numbers are practically the same and thus they perform just as well as first expected.

This video is the conclusion of this experiment, it has now been salvaged and made room on the bench for the next project