Spark gap as a snubber (TIG inverter)

Unfortunately spark gaps aren't "designed to extinguish the arc" - they simply exploit the physics of a gas discharge through a gas.

The process of breaking down a spark gap starts with the appearance of a large potential difference between the electrodes. This persists for an indeterminate time (a microsecond?)until a free electron shows in the space between the electrodes.

If the electric field is high enough, the electron will be attracted to the positive electrode, and accelerate to a speed where it can undergo an in-elastic collision with a gas molecule, knocking off a second electron and creating a positive ion, which goes the other way. This produces an avalanche of electrons going one way, and positive ions going the other.

When the positive ions get to the negative electrode, they hit it hard enough to knock off secondary electrons, which proceed to generate more positive ions ....

This is a "glow" discharge. The voltage drops at the electrode surfaces are of the order of a few hundred volts.

If the current density at the cathode is high enough, the surface will get hot, get soft, and distort under the electric field to emit electrons by a process which is a mixture of thermionic emission and field emission.

This is an "arc" discharge. The voltage drops at the electrode surfaces are an order of magnitude lower than in an arc discharge.

The glow-to-arc transition can be fast - sub-microsecond - if the current is high enough.

The arc-to-glow transition occurs if and when the current density at the surface drops to the point where the surface can cool off to a temperature where it is solid, and the termionic electron emision stops.

The glow discharge stops at the point where there are so few electrons in the discharge that the statistical nature of the avalanche process allows it to self-extinguish - the was only one electron in the gap, and it didn't gnerate any descendants.

This is a rather low current density.

--------- Bill Sloman, Nijmegen

You would get a lot more predictable results with a gas tube arrestor. At least, the weather (especially humidity) wouldn't alter its response.

Here are some examples from the Digikey catalog: (bottom of page)

Some are even in stock. ;-)

You would have this on the output of an --uh -- Tungsten Inertgas welder designed to sustain an arc (once struck) across a gap between tungsten and metal in inert gas?

Brilliant! Well, it will be for a very short while anyway.....

They extinguish when there is insufficient power to keep them going. This works for brief transients; we used them in traffic controllers as partial protection against nearby lightning strikes that induced spikes in powerlines and other wiring. If there is enough power to sustain the plasma, a spark becomes an arc --as in a Jacob's ladder. The voltage necessary to sustain it is much lower than the voltage it took to initiate it.

See

Note how the voltage drops on the gas tube (green plot) microseconds after initiation.

A better bet for you would be a large MOV -- and you need to commutate your IGBT's quickly. Use the IGBT's to chop, but don't try to use PWM (off time) to regulate the current. Use your welder's controls to control welding current. You can certainly vary the duty cycle of the opposite polarities, you just need to minimize "off" time. Use the welder's controls to regulate welding current.

If you used a 200 volt MOV, at 200 amps it would dissipate 40KW when snubbing -- but if your commutation time is only a microsecond then the energy per commutation is only 40 millijoules. For 300 commutations per second that's about 12 watts. A couple of large MOV's should be easily able to handle that.

I am curious...

For high current snubbing, would it be possible to make a spark gap from 4.8mm tungsten rods, with a certain gap between them, so that it would spark at, say, 600 volts.

If so, then would this simple thing be a better snubber for complete bridge turnoff at full power, than huge diodes, caps etc. Note that intermittent, short duration turnoff during switching can be taken care of with a small capacitor. The above applies to one of a kind conditions that cause complete turnoff of the bridge under load, an unusual situation.

Using a spark gap strikes me as a very foolproof approach.

Any thoughts?

I have 10 4.8 mm tungsten electrodes (cost $19) and could use one for this purpose.

Thanks. It looks just like what I need.

Do you think that it is, in fact, a solid idea and the right way to go, as opposed to having a monster capacitor, big ass wires, big diodes etc.

I really like this concept, to be honest. Seems like there is very little that can go wrong with a spark gap in inert gas.

Don, I am confused. I would think, naively, that these arrestors are designed to extinguish the arc once voltage drops back to operating range. If they start an arc and do not interrupt it, what is their value? A clarification would be appreciated.

For example, think of a tig torch and HF arc starter. I can hold my torch so far from work, that arc starter creates sparks between electrode and work, and yet the main arc does not start.

That's why I am curious as to what these surge arrestors do once they spark from a surge.

I'll defer to others on RC snubbers for IGBT's.

Not off hand. I'd have to look at specsheets for rated power.

Can't say without a datasheet.

I see now. Thank you.

I agree with you regarding the use of a varistor.

My plan is to have a RCD circuit that would handle regular commutation events. I calculated that a 4 uF capacitor should handle these events just fine and limit the circuit voltage to 200 volts.

Then the varistor would only conduct during much higher spikes other than those that occur during regular commutation.

Does that make sense to you?

If so, what kind of capacitor is appropriate. I have looked at some C-D capacitors, but want to hear your opinion.

Don, do you have a suggestion regarding a particular varistor?

I have some 14N391K varistors. Are they usable for this application, perhaps paralleled?

As always, I appreciate your input.

It kinda does, and 4 uF in series with maybe 0.75 ohm sounds like a reasonable value though I still defer to others with more IGBT experience. What does RCD mean, BTW?

Again I defer to others on that, but I've had best luck with polypropylene caps. My experience may be dated -- last time I did any of this sort of work was nearly 10 years ago already. Time flies when you're having fun -- I retired in 1999 (grin!) Now I get to play in my own lab/shop every day (when I'm not gone fishing) and do what *I* wanna do! Tonight I was sticking surfacemount parts on a board on the bikelight project for my daughter, tomorrow I plan to start learning how to weld 4130 moly with TIG and O/A for building bike frames.

I was actually sticking an SOT-23 transistor and a 1206 resistor where there are no board traces -- this is a patch. It really is possible to "glop wire" with SMT parts, #30 wire, some wee dabs of epoxy and a stereo zoom microscope. It ain't pretty, but it works.

I can only say "probably" based on their physical size. If they don't get hot, they should work OK.