[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: ScanTesla V7.61 - V7.62 ;-)



Original poster: Jim Lux <jimlux@xxxxxxxxxxxxx>

At 04:51 PM 6/7/2006, Tesla list wrote:
Original poster: Vardan <vardan01@xxxxxxxxxxxxxxxxxxxxxxx>

Hi.

At 08:22 AM 6/7/2006, you wrote:
At 11:11 PM 6/6/2006, Tesla list wrote:
Original poster: "dest" <dest@xxxxxxxxxxx>

Re:

In the case of tesla coil modeling (with sparks), we're somewhere in between. We have equations that are "physics derived" in that they have a basis in the physics underlying what's going on (i.e. omega = 1/sqrt(LC)), and we have others that are essentially empirical models of observed behavior: 220K resistors in series with wires.

Of course, "my" little model is only a "tiny" step. I suppose one day there will be a giant 3D E-field model that will take into account all of Paul's work, Bart's work, and other people's work who might not even be born yet, and model streamers at the sub-atomic level. I might be at the "sub-terrainian" level when that happens ;-))

But "right now", there is the model of "1pF/foot+220K" and the dynamic (the "first" dynamic) model, at least for coiler's... If anyone has any better one, I have been waiting for a "year" to hear about it :D ScanTesla was designed for new dynamic leader models and how they effect the system and streamer length. "Not having" a model became sort of a problem... So "I" made one ;o)) Like it, or not... If anyone has a better one.....

That's probably "good enough".. One might be able to come up with something like a segmented transmission line with voltage sensitive switches, but somehow, I don't know that it would give a "better" result than what you're getting now. There was a discussion of such a TL based model, oh, 5 years or so ago, I think.



As is widely acknowledged, the actual modeling of sparks, in physical terms, is poorly understood.

They are actually very well understood!!!

Hmm... steady burning arcs, perhaps. propagating leaders from a single impulse, somewhat less so. Sparks from an HF AC source with low impedance, I think not. There's a lot of stuff with charge flowing into and out of the channel, etc. that is not particularly well understood, at least beyond fairly straightforward phenomological descriptions. And when you get into something like an actual 3D model that deals with forking, etc....

That doesn't stop us from using an approximation that is based on observed behavior and which allows better prediction of actual coil behavior. The fact that the model resembles current thinking on what's actually happening with the spark is just a manifestation that such a model is computationally or conceptually efficient.

It is "better" than the "old one", vastly far from "perfect" ;-)

A lumped order of magnitude model for a spark/streamer is a heck of a lot better than NO model, or modeling it as a fixed R and fixed C. A time varying model that matches observed behavior for actual sparks is a logical next step.

I am sort of amazed at how long and successful "1pF/foot+220K" lasted and still rules today (five years...)!!! You will not find any DRSSTC calculations or any modern "models" that don't use it!! Obnoxiously, it seems to still work well... I forget the post where I spent 30 minutes "making it up"... I am not sure "dynamic" modeling makes a giant difference, but it is a step forward.

I suspect that things like construction variability, or just the random variability of sparks in general, might be greater than the difference from using the static RC model.

What a dynamic model might help with is a better understanding of BPS influences, or, "why does the spark go *there* and not *here*" kinds of questions.


While it would be nice to have an actual model of actual behavior, we're a ways from that.

Consider just making a "static" 3D E-field model of a coil with a streamer. Time goes up about 800X in E-Tesla. If we use the "dynamic" time frame of ScanTesla it is 10,000X more... But that would all "depend" on a "dynamic leader model" :o))) Without "that model", a Giga-Giga-Giga-Hz CPU of computer power is no more use than a space heater...

It's not that bad. You'd do some sort of variable gridding, and only compute the fields at high resolution in the area of the spark channel. However, this greatly increases the complexity of the modeling. And, as you say, it's still gonna take a lot of processor crunch.

That said, lots of processor crunch is available. If someone were to be bold and make a TCsim@home screensaver..... Why should the aliens and genes get all the resources.



Jim