A 63 volt electrolytic capacitor bank

[klugesmith]

What's a good hobby use for 10 kilojoules of 63 volt aluminum electrolytic capacitors?
I'm here to share some things learned from a "bank" that was saved from scrap bin, when an old storage room was converted to a new lab.

These 14 aluminum electrolytic capacitors came in metal support frames that don't hold them tightly.
Each is nominally 280,000 uF 63 volts, with an individual 4700 ohm parallel resistor and individual wires to ring lug terminals. (10 kJ claim will follow.)
Two screws and two nuts were sufficient to connect the whole set in parallel.


First step on the test bench was to reform the capacitors. a chore that's not only for antique or high voltage capacitors.
(Note the reforming protocols that come with industrial VFD's, for units put in service after more than 12 months of storage.)
https://www.eaton.com/content/dam/eaton/products/industrialcontrols-drives-automation-sensors/variable-frequency-drives/powerxl-dm1-variable-frequency-drives/vfd-storage-and-reforming-capacitors.pdf

/>
Using 0-50V benchtop power supply and 10 ohm series resistor, I progressively set the voltage to 20V, 30V, 40V, 50V over a few hours.
Final current was a steady 0.15A, which can all be attributed to the bleed resistors with 48.5 volts on bank.

Let's measure the capacitance.
With PS disconnected, voltage dropped from 48.48 to 46.47 V in 1 minute, indicating RC product of about 1416 seconds.  If the average R value is nominal then average C is 0.30 F and whole bank is 4.2 F.
I also discharged the bank for about 10 seconds into the 10 Ω 50 W resistor.  Bank voltage fell from 48.5 to 37.1 V.   About 1900 joules made the resistor very hot; I didn't bother looking up energy ratings in its datasheet.
At the end of the session I disconnected PS and monitored the voltage during bleed-down.   Time constant exceeded 2000 seconds (more than 150% of nominal) at lower voltages.  That inspired an exercise (to be reported later) with automatic voltage logging.   Will also describe bringing the bank up to 63 volts, measuring voltage dependent leakage current and capacitance, and dielectric absorption (voltage bounce-back after not-rapid partial discharge).

[klugesmith]

Back to question of practical uses, other than original application and as a teaching opportunity.

To discharge a big C on purpose with a screwdriver would be dangerous and abusive (to screwdriver and capacitor).
Coin shrinking, can crushing, or disk launching with enormous C values at 63 volts is problematic, because we can't make work coils with less than one turn, to get appropriately fast RLC dynamics.  I think the same goes for coil guns -- 63 V is OK, but how could the huge capacitance be justified?

The 14P bank ought to be comfortable with impulse discharges into 15 milliohms, with time constant of 60 ms.   (That's much faster unloading than anyone can get with ultracapacitors.)
Peak current 4200 A, 300 A per capacitor (whose ripple rating at 85°C is 34 A RMS). 1/4 megawatt.

A 126 volt 1 farad configuration (2S7P) would be easy to set up, with incandescent light bulbs as a demonstration load.

Higher voltage series configurations might be useful in something like a disk launcher.   Not really enough voltage or intrinsic speed capability to make an efficient can crusher.  And rapid discharge of HV stack could reverse the voltage on specimen with the least capacitance.   Diodes to protect against that would need suitably high pulse current capability.

[MRMILSTAR]

How about a wire exploder or fruit exploder? There shouldn't be any voltage reversals to harm the capacitors and very high speed shouldn't be required. Use exploding water to make a single-shot water jet is another thought. You can use a very small gauge wire to prime the fruit and water explosions.

[davekni]

Sounds like the supply for a rail gun.  Those require very high current at moderate voltage.

[klugesmith]

This chart shows discharge current pulse measured on afternoon of July 3.    Total energy is just above 10 kilojoules.


The discharge was started by disconnecting the charging power supply.
Load is the built-in bleed resistors: 4700 ohms each, 14 in parallel.
Bank voltage logged every 5 seconds with a Lascar EL-USB-3 voltage logger, and external series resistor to accommodate initial voltage of 63 volts.
Chart Y axis is bleed current in microamps.   X axis is time from logger start in weeks. :-)

Later I'll report on the substantial change in time constant during the pulse, mostly due to voltage dependent capacitance.
Some dielectric absorption was evident in a separate discharge run, where voltage bounced back after "rapid" discharge (10 ohm load).