Simulation of the behaviour of a PCB spark gap

[marc1996]

Hello,

I'm doing an University project about PCB Spark gaps.

My question is:
Someone know if there is any software to simulate electrostatic discharges?
thanks

[T3sl4co1l]

There are a few SPICE models for gas discharge phenomena.

It's not very useful given the complexity of plasma physics, and the extreme variation in practice (breakdown voltage depends on electrode shape, rate of voltage rise, ambient temperature and humidity, UV light or other sources of ionization, surface contamination (adsorbed substances, and dust), and past history, especially where substrate breakdown occurs (carbon tracking).

In short, what would you aim to learn from a simulation, that you can't determine from a pad-and-pen approach?

Tim

[marc1996]

There are a few SPICE models for gas discharge phenomena.

It's not very useful given the complexity of plasma physics, and the extreme variation in practice (breakdown voltage depends on electrode shape, rate of voltage rise, ambient temperature and humidity, UV light or other sources of ionization, surface contamination (adsorbed substances, and dust), and past history, especially where substrate breakdown occurs (carbon tracking).

In short, what would you aim to learn from a simulation, that you can't determine from a pad-and-pen approach?

Tim

Just wanted to test different electrode /spark gap shapes, test how they could distribute the energy that is supposed to be dissipated, how can affect in temperature issues...

Also, the spark gap should be basically for the connection process and not much more. I don't need a really long life protector because the pcb won't get that much ESD's. 

[T3sl4co1l]

Oh, okay.

Then you need a multiphysics E&M and plasma simulator.

https://xkcd.com/1425/ relevant

Tim

[marc1996]

Oh, okay.

Then you need a multiphysics E&M and plasma simulator.

https://xkcd.com/1425/ relevant

Tim

Thank you Tim,

I have tried that before but, I can not simulate exactly the current waveform that I want with those softwares. Do you know if there is any multiphysics software where you can implement electronic components? Because I want to simulate the IEC-61000-4-2 test and there is an equivalent circuit that generates the same current wave.

[T3sl4co1l]

I know Ansys does that, but damned if you're going to find one that handles plasma, too...

Methinks you're very, very quickly running off into either the seven-digit price tag, or the masters-thesis research project.

Tim

[Marco]

I have tried that before but, I can not simulate exactly the current waveform that I want with those softwares. Do you know if there is any multiphysics software where you can implement electronic components? Because I want to simulate the IEC-61000-4-2 test and there is an equivalent circuit that generates the same current wave.

But there is no specified current waveform for IEC-61000-4-2 AFAICS. There is a discharge circuit ... the waveform depends on the load.

Instead of simulating I'd build a test rig to quickly test a gap. PCBs are cheap, ordering one with a 100 different gaps won't cost you a significant amount of money if you have a little patience for the delivery.

[marc1996]

.

[Marco]

I know but, at the end, this only need a capacitor, a resistor (both defined by the test specifications) and the DUT. I thought that maybe there is a way to simulate the current waveform produced by the different test levels depending on the load.

Can't work. You want to simulate how it breaks down, but a current waveform forces it to break down ... the only thing left for the gap to do is modulate its resistance. You would almost certainly be simulating completely unrealistic voltages for many gap designs.

[marc1996]

I know but, at the end, this only need a capacitor, a resistor (both defined by the test specifications) and the DUT. I thought that maybe there is a way to simulate the current waveform produced by the different test levels depending on the load.

Can't work. You want to simulate how it breaks down, but a current waveform forces it to break down ... the only thing left for the gap to do is modulate its resistance. You would almost certainly be simulating completely unrealistic voltages for many gap designs.

I don't know if I understood you correctly but what I exactly want apart of how the shape can affect is how could change the current distribution with different gaps distances.

[Marco]

U=I*R ... the ESD test is fundamentally voltage driven. If you try to make a current driven test you will force the voltage to unrealistic values the moment the gap differs slightly from what the current waveform was measured for. It will have fuck all to do with IEC-61000-4-2 and it will have no practical value.

I doubt you will find any simulation software of useful predictive value, even if you give up on trying to force a current waveform.

[T3sl4co1l]

U=I*R ... the ESD test is fundamentally voltage driven. If you try to make a current driven test you will force the voltage to unrealistic values the moment the gap differs slightly from what the current waveform was measured for. It will have fuck all to do with IEC-61000-4-2 and it will have no practical value.

I doubt you will find any simulation software of useful predictive value, even if you give up on trying to force a current waveform.

The standard describes a system of several meters in length; the risetime of the spark itself is small compared to this length scale.  Consequently, the transient is a propagating wavefront, with well defined impedance: that of the grounding cable, circuit traces and so on.

The schematic typically given (a capacitor, resistor and switch), is at best, a caricature of the real system.  It might be a description of the components involved, but it is not a model of the system!  That would require an RLC equivalent to the published waveform.  At least a two stage RLC network, equivalent to the sub ~300MHz response of this transmission line system.

Tim

[marc1996]

U=I*R ... the ESD test is fundamentally voltage driven. If you try to make a current driven test you will force the voltage to unrealistic values the moment the gap differs slightly from what the current waveform was measured for. It will have fuck all to do with IEC-61000-4-2 and it will have no practical value.

I doubt you will find any simulation software of useful predictive value, even if you give up on trying to force a current waveform.

The standard describes a system of several meters in length; the risetime of the spark itself is small compared to this length scale.  Consequently, the transient is a propagating wavefront, with well defined impedance: that of the grounding cable, circuit traces and so on.

The schematic typically given (a capacitor, resistor and switch), is at best, a caricature of the real system.  It might be a description of the components involved, but it is not a model of the system!  That would require an RLC equivalent to the published waveform.  At least a two stage RLC network, equivalent to the sub ~300MHz response of this transmission line system.

Tim

not sure the exactness of parameters, but a neon lamp seem able to wing it?
checkout the "small" current gg thru the 2 "resistors". there is a HV charge in cshunt parameter
V2 flushed down the "neon" somewhat? "randomly" at 4000A "spark" peak?
if somebody have a better idea about the parameters, pls repost corrections

U=I*R ... the ESD test is fundamentally voltage driven. If you try to make a current driven test you will force the voltage to unrealistic values the moment the gap differs slightly from what the current waveform was measured for. It will have fuck all to do with IEC-61000-4-2 and it will have no practical value.

I doubt you will find any simulation software of useful predictive value, even if you give up on trying to force a current waveform.

Sorry for my delay responding you guys.

I think that my answer can respond to you 3 at the same time.
Using LTSpice that's the circuit that I actually have(attached picture). Everything inside the box is what simulates the waveform of the supposed ESD. This is an equivalent circuit that I found in internet. The simulation is really accurate acording to the supposed waveform. The problem is that varying the distance of the voltage breakdown doesnt change nothing apart from moving the supposed current in the time domain, I mean, when i change the parameter "d", if the peak starts at 5 ns then it will be at 10, but the value is really the same

I don't know if I have to change some parameters of the gap values apart of the breakdown voltage because I do not know at all what they mean.
If someone is more experienced in this software and could help me please let me know.

Also, one member from another forum said me that if I finally get with LTSpice to simulate what I really want, changing the distance value between the electrodes I could simulate an spark gap as an array of micro-spark gap, and I think that is something that could be really interesting.

[T3sl4co1l]

You won't observe anything different because the "spark gap" model is a mere cartoon of the complexity of a physical spark gap.

Tim

[iMo]

Partially off topic: In mid 80ties colleagues of mine did simulation of  sub-micron GaAs semiconductor devices on an Sinclair Spectrum with 48kB ram and Basic. They did PhD with such an gear. Moral of the story - it is not about the gear but how a) good/well you can describe the stuff by the "theory of electromagnetic fields" and how b) good/well you can built a "system of differential equations".
PS: today the diff eq solver for b) is  fortunately ready to get off-the-shelf.. (Spice)

[marc1996]

Partially off topic: In mid 80ties colleagues of mine did simulation of  sub-micron GaAs semiconductor devices on an Sinclair Spectrum with 48kB ram and Basic. They did PhD with such an gear. Moral of the story - it is not about the gear but how a) good/well you can describe the stuff by the "theory of electromagnetic fields" and how b) good/well you can built a "system of differential equations".
PS: today the diff eq solver for b) is  fortunately ready to get off-the-shelf.. (Spice)

Hello imo,

The problem I have is that with spice softwares (I'm using LTSpice), I'm able to simulate the IEC 61000-4-2 test and produce a current waveform really accurate to the real one. But now, after producing this waveform, I'm not that able to check how the spark gap is going to behave depending on the distance. I'm trying to simulate it but the circuit, as other members have said in upper comments, it is not acting as it would in real life.

btw, thanks for your tip!

[licht]

How did you end up simulating the spark gap? There's an LTspice model of a triggered spark gap https://hofstragroup.com/triggered-spark-gaps.html but not sure whether it's applicable to this situation. Were you able to use LTspice in the end or did you have to resort to fancier tools?