Pulse caps in series or parallel?

[Felix M.]

Hi all,

A while ago I got my hands on five Electronicon power film capacitors, each ratet for 1,5kV @ 350uF. I didn‘t find any datasheet for this specific type but one with very similar specs is ratet for smth around 10kÂ.
I want to get as much ‘‘bang“ as possible out of this bank and I‘m wondering if I should wire them parallel or in series…

Does it even make that much of a difference?
A parallel configuration would allow for high currents but the voltage might not be high enough to reach them, where as a series connection would have the voltage but not the current capacity because each cap can only do 10kA?

What would you suggest?

Greeting

[MRMILSTAR]

It depends. What is your application?

[Felix M.]

I want a system with the capability to release as much power as possible in the shortest time possible.

[klugesmith]

I want a system with the capability to release as much power as possible in the shortest time possible.

Stored energy is invariant, so power is proportional to the inverse of discharge time.
Parallel or series banking makes no difference to power or speed, though the optimum load R and L values differ by a factor of 25.
If you give details for a load, then we can help you decide whether parallel or series bank will serve you better. 

[edit] For example, if you want to make a can crusher, series bank is the way to go.   Parallel bank would have too much capacitance and not enough voltage to match a practical can-crushing work coil.    For the series bank, equal division of the charging voltage can be insured by having a bleed resistor across each capacitor (and not trying to have a charge time much shorter than the bleed time). 

[edit edit] You mention datasheet for a similar capacitor type, giving maximum current "something around 10 kA".
Please understand: that doesn't mean the capacitor is guaranteed to deliver that much (or more) when short circuited.  Typically it can deliver a lot more.
The limit means that if your circuit allows more current to flow, the capacitor might be damaged because you are abusing it.
Can you share the datasheet with us?

[Felix M.]

I want a system with the capability to release as much power as possible in the shortest time possible.

I have many applications in mind, can crushers included. More generalized, I want a cap that can deliver high pulse currents even when connected to higher resistive loads.
Normally I'd go for paralleling the caps but my concern is that I might not be able to reach high (enough) currents as 1,5kV is not that high of a voltage.
Even 0,1Ω would limmit the current to 15kA which could probably be done by one cap. Of course that's still a lot of current but the point is that even only the resistive part of the load limmits a lot. If I'd connect them in series for a voltage of 7,5kV such problems wouldn't occur but then the current would be limmited by the capabillity of one cap in terms of current.
I think those caps can tank some punches as I'm only using them in a pulsed application and they don't have to stand any rms current. I would tend to wire them in series to avoid the resistance of connections, busbars, etc. as a limmiting factor.

Here is the datasheet. The one marked has identical proportions in terms of size, it's capacitance is a bit bigger (30uF)
DC_Link_Caps.pdf

[MRMILSTAR]

There are constraints for coin shrinking, can crushing, and disk launching. First of all, you need sufficient energy. Decent coin shrinking requires about 7000 joules. Can crushing with a good theta pinch requires about 1000 joules. Disk launching requires about 500 joules. It has been found by several experimenters that a resonant frequency of about 8 KHz is ideal for all three of these use cases because it provides a sufficiently fast rise time but not too fast. This, combined with the small number of windings required on the work coil to provide good coupling, constrains the capacitance to be in the range of about 50 to 200 uF. So taking all of this into account, you would use a capacitor topology that satisfies these constraints.

A higher resistance load is going to require higher voltage than a low resistance load to generate the same energy. A series configuration may get you that higher voltage but may reduce your capacitance below what is ideal. You have to trade off all these requirements against your application.

[Felix M.]

Going for a series config would result in 70uF. I guess that‘s fine considering what you said.

[MRMILSTAR]

Whatever your topology is, it has to deliver sufficient energy for your application according to

E = 0.5 * C * V * V

where E is in joules.

[davekni]

Decent coin shrinking requires about 7000 joules.

It has been found by several experimenters that a resonant frequency of about 8 KHz is ideal for all three of these use cases because it provides a sufficiently fast rise time but not too fast.

My coin shrinker uses 3000 joules, 14uF cap rated for 20kV, charged slightly over 20kV to reach 3kJ energy.  Shrinks US quarters from 24mm diameter to ~16.5mm diameter:
    https://highvoltageforum.net/index.php?topic=1793.msg13556#msg13556
However, to get results with relatively low energy requires fast discharge.  Around 17us discharge in my case, well above 8kHz resonant frequency.

Your 2kJ cap set could possibly shrink coins to a reasonable extend, but only if you far exceed 10kA cap rating, to 40kA or 50kA.

[Felix M.]

The caps have a pretty low ESR, ESL which allows huge pulse currents way beyond the rated 10kA (probably even beyond 50-75kA). That‘s why a series connection is probably better because this current can be delivered at the according voltage due to i=u/r. Higher currents drain the caps faster and produce more intense fields. That’s benneficial for for example coin shrinking or disk launching. I‘ll put them in series with aluminium bussbars soon and do some test.

[TizianoBll]

I'm not a big fan of connecting them in series. Series connection makes charging and discharging much more difficult/dangerous.

You'll need to connect an equalizing resistor for each capacitor (to ensure equal charging voltage).

During a pulse, capacitors with lower capacitance will charge at a negative voltage (this may reduce pulse life)

Applying a shorting bar will be much more difficult since you'll need to individually short each capacitor and each one could sit at a different potential making the operation riskier (in parallel they're all connected to the two same terminals).


Higher voltage could help with what you are doing, if series connection is necessary, I would limit the series voltage to a few kV’s. Maybe two banks in series, each bank composed of multiple caps in parallel?

[Felix M.]

It's a pitty that I have 5 caps on hand. If I had an even number I would do mby 3s 2p but it's hard to do as 5 is a prime number and doing such a mixed connection with 5 wouldn't really make sence.
The issue of equal charging is very important but with a given leakage current, the ballancing resistors can be determined. They also discharge the caps over time which is good. For charging up to 7,5kV I'm planning to use the rectified output of an oil burner ignition transformer that can do 7,5kVac. (Voltage can be measured with a voltage divider)
Of course this voltage is way more dangerous as it allows 5 times more current to flow than with 1,5kV if the resistive load is the same. (I'm pretty sure that 1,5 and 7,5kV have the same effect on a humans wellbeing when connected accross the body  )

I still think a series configuration will be more suitable for my purposes allthough I agree with you about the safety part.

[davekni]

For 350uF at 1.5kV, series is the only reasonable way to connect, at least for the common uses of coin shrinking or can crushing or disk launching.  Exploding wire might be better with parallel.  Parasitic inductance of interconnect and discharge switch is already critical at 70uF and would be much more so at higher capacitance.  Presuming good cap quality, should be close enough in value to have little issue with reverse voltage.  Film capacitors won't mind a little voltage reversal (unlike electrolytic).  Equal resistors across each cap is critical.  Keeps voltage balanced and bleeds off any residual charge.

Yes, safety is critical, at 1.5kV or 7.5kV.  Sufficient energy to be lethal either way.

The caps have a pretty low ESR, ESL which allows huge pulse currents way beyond the rated 10kA (probably even beyond 50-75kA).

10kA rating has nothing to do with what current you can get from the caps.  Rather it is the current that you may draw without damaging caps.
However, since your pulse discharge use won't be 24 hours per day, you probably have margin to abuse the caps some.  20kA might last as long as you would want.  At 75kA, cap life is likely to be quite short.  (Note: I haven't looked at the specifications for these specific caps.  Not sure if 10kA is a spec for repeated exposure or for occasional one-time events.  In general, DC link caps are not at all intended for pulse discharge use.  So even 10kA full discharges may limit cap life.)

[klugesmith]

Early on, I made the same comment about 10 kA in datasheet:  it means that if circuit allows more current than that, a single shot might damage the cap.

The cited datasheet is for "DC Link" capacitors.   
Underlined is a 380 uF 1500 volt unit, whose main max-current spec is 60 amps RMS.  That's tolerable forever (with adequate cooling).
Equivalent series R is 1.8 milliohms and L is 60 nH.   sqrt(L/C) is 12 milliohms.   sqrt(LC) gives self resonant frequency around 33 kHz.


With a perfect short circuit on one capacitor, we'd get damped oscillatory discharge with peak current of about 100 kA (delivering zero power to load).
.   
We'd have the exact same shape, timing, and capacitor abuse if parallel or series bank were short circuited with negligible outside R and L (which is impossible).
Felix could explore discharge dynamics with a simulator like LTSpice.
Load resistance needs to be large compared to internal resistance and interconnect resistance, if we want most of the stored energy to go to the load.
Series bank of 5 would have equivalent R of at least 9 milliohms and equivalent L of 300 nH, which are in the ballpark of external parasitics in a good can crusher.

[klugesmith]

Even 0,1Ω would limmit the current to 15kA which could probably be done by one cap.

0,1Ω load on series bank would be equivalent to 20 mΩ on a single capacitor.  That's very close to critical damping, if interconnect and switch R and L are negligible (not very realistic).  Peak current 45 kA, with peak 900V at load (single cap) or 4500V (5S bank).

0,1Ω on single cap, or 0,5Ω on series bank, or 20 mΩ on parallel bank, might be more like what Felix had in mind. Capacitor's internal resistance and inductance aren't greatly reducing current from the simplistic model.

[Felix M.]

Depending on whatever load, a series resistor might be a good idea. I‘m thinking about a shunt or maybe even a properly sized inductor like it‘s used in defibrillators. If the external resistor needs to be larger than the combined ESR, would a current shunt with 0.1ohm be enough? Combined with the given ESR of 9mohm, current would be limited to almost 69kA (not including the inductive part). This of course is only the case when the cap is directly shorted which would result in 6,9kW dissipated across the shunt. If there‘s an additional load tike a coil for disk launching, etc. it would again be a different story.

Unfortunately I currently don‘t have too much time as school is quite time consuming but I might be able to get the cap ready this weekend. The bussbardesign will also add further inductance.