What's the best way to trigger these things?

[John123]

When trying to discharge a big capacitor bank into a load using a chicken stick it wastes a lot of energy in sparking, energy which doesn't get dissipated in the load.

Is there some kind of go-to SCR or contactor to reduce wasted switching energy? Assume the load in series is very low impedance, wouldn't it just destroy the switching device too?

[Mads Barnkob]

SCRs or Thyratrons would be the way to go, albeit they can be costly and hard to source where a spark gap can be anything that can conduct a current

The spark losses are determined by the charge in coulomb and thus the lower voltages, the higher spark gap losses/wear. So good and effective spark gaps uses higher voltages. Which is suitable with lower voltage SCRs being more accessible.

I got a paper on spark gap material wear, form, function and losses etc. on my website, download the "Erosion in High Current Discharges" PDF here: http://kaizerpowerelectronics.dk/tools/file-archive/?drawer=application_notes*pulse_power

If you use my capacitor energy calculator, lets take an example of a 50 kJ bank: http://kaizerpowerelectronics.dk/calculators/capacitor-energy-calculator/

Electrolytic capacitor bank at 1000V / 100000uF would hold a charge of 100 Coulombs

Film discharge capacitor at 100kV / 10uF would only hold a charge of 1 Coulombs

So spark gaps do actually become efficient at higher voltages. Downside to SCRs is an uneven hole distribution when it turns on and it can result in hotspots and exploding silicon   

[John123]

I noticed on your website you've come up with a spring loaded contactor, was there a noticable difference in energy transferred to the load when you switched over to this?

[klugesmith]

Question in OP gives no hint about the "large" capacitor banks and the loads.

That covers spaces several orders of magnitude wide
in voltage, capacitance, energy, load impedance, and pulse timing & damping.
Do we all agree that an ideal short-circuit "load" can never receive any power?
That to receive more than half the total energy, load resistance needs to be higher than the resistance of capacitor + switch + interconnection?

No switch type is optimal, or even usable, for all variations.
There are places for SCRs, IGBTs, ignitrons, spark gaps (fixed or electrically or mechanically triggered), and contactors.

[John123]

It's 330v at around 2200uF, so actually not that large compared to the multi-kilovolt monsters on here.

The load could be a coil for launching things or something that's about to get its end blown out.

[ritaismyconscience]

Definitely go with SCR for that one.

[Mads Barnkob]

It's 330v at around 2200uF, so actually not that large compared to the multi-kilovolt monsters on here.

The load could be a coil for launching things or something that's about to get its end blown out.

I agree on SCRs for this voltage. Thyratrons are still for higher voltages and a spark gap would just weld itself shut very fast.

I noticed on your website you've come up with a spring loaded contactor, was there a noticable difference in energy transferred to the load when you switched over to this?

Despite being closed with a 50Hz solenoid, I still think this method is too slow and I would like to make some more designs (some day) from that PDF about erosion I linked earlier.

[rikkitikkitavi]

Spring loaded contacts tend to bounce , which breaks/makes contact very rapidly. But not faster than inductance...

For a < 1200V I would aim for the beefiest SCR I could afford. Or even higher if I found a even beefier SCR.

[davekni]

SCRs etc. have one key limitation to consider, called "critical rate of rise of on-state current".  Spec's are typically around 100A/us, and don't generally increase for large parts.

For single-shot use you can likely get away with 2x rating, so 200A/us for typical parts, especially with high gate trigger current.  At 330V, that's 1.65uH.  As long as your load has at least that much inductance, SCRs will likely work fine.  I typically use BTB16 TRIACs for such.  They work a bit better with negative MT2 voltage, so connect MT1 to the cap positive terminal and MT2 to the load.  Pulse the gate low with >50mA to trigger.  Gate looks like back-to-back diodes to MT1, so trigger with current or voltage through a dropping resistor.  For short single-shot loads like this, BTB16 parts can handle 500A.

If your load is under 1.65uH, you'll have a hard time finding SCR/TRIAC/Thyristor parts rated for high enough current rise rate.  Some parts are rated for 150A/us.  Paralleling is possible too, but requires high trigger current to make sure all parts turn on simultaneously.

For random single shots, mechanical switching is still an option.  It may occasionally require prying contacts apart after they weld together.  Just a couple 12AWG (2mm diameter) wires can work, one against a piece of wood or ... and the other sprung slightly above it at ~90 degrees.  Hit the top wire with a dowel or plastic rod at high (for manual range) speed.  Usually the discharge is complete before any bounce occurs. 

[John123]

Thanks!

[johnf]

Many years ago when I worked for a modem design company we had to test our line interfaces at an accredited lab for lightning surge performance
the test involved a high voltage lightning simulator that applied multiple high voltage pulses to the UUT.
The pulses were sourced from an insulated pendulum that had a brass rod with rounded ends that made contact with with terminals either side as the pendulum swung through. This simulated what happens during a lightning strike by several strikes taking place in rapid succession in the plasma channel.

a very simple switching mechanism that could handle many kiloamps day after day

[klugesmith]

Dave and John have underlined the point that it really depends on quantitative details of load and discharge waveshape.   

Can't do a proper switch selection by saying, broadly, "for launching or crushing or blowing s**t up".  Need to put approximate numbers on circuit's milliohm, microhenry, ampere, microsecond, etc. values.  As mentioned before, if the load impedance is zero then the maximum energy and power you can deliver, with the biggest capacitor and best switch, is zero.

Or you could skip the EE investigation, try something, and see what happens.  That approach has worked for centuries, in disciplines from farming to forging to firearms. Even fusion (witness Castle Bravo test).