Computer Models - Re: chokes

["Tesla list" <tesla-at-pupman-dot-com>]

Original poster: "Terry Fritz" <twftesla-at-qwest-dot-net>

Hi Matt,

        You bring up a very important and critical point that is near and dear
to, at least, "my" heart :-)

Tesla coil computer models are only as good as their ability to predict or
reproduce actual measurements.  The paper at:

http://hot-streamer-dot-com/TeslaCoils/MyPapers/modact/modact.html

had a tremendous amount of work behind it (two years) in not only getting a
computer model to match actual measurements but finding ways to even measure
those voltage and currents to verify the models...  Of course, I didn't write
fancy papers up on all the dismal failures :-)))

Recently Paul's secondary modeling project:

http://www.abelian.demon.co.uk/tssp/

Went through the same modeling and actual measured verification process... 
There is data there from 13,500 coil configurations that took 420 processor
days to compute.  That is many more times the number of coils we all on this
list have ever actually built let alone measured!

http://www.abelian.demon.co.uk/tssp/vsd/

You may remember all the discussions of what the voltage profile along the
secondary looks like over the years...  Paul's models say:

http://www.abelian.demon.co.uk/tssp/pn1710/

and the actual measurements matched (after the fact):

http://hot-streamer-dot-com/TeslaCoils/MyPapers/NSVPI/NVSPI.htm
http://www.abelian.demon.co.uk/tssp/pn2510/

Funny that the real answer was far different than we ever thought :-))

When you have a realistic computer model, then wonderful things are possible. 
LTR coils were first predicted by computer models...  The computer said to do
"this and that" and something wonderful will happen!  Sure enough, the computer
predictions were right and LTR coils were born and well understood even before
they were first built.  Many people swear by the seat-of-the-pants trial and
error methods, but the armchair computer punchers won that one ;-))  Computer
models and formulas used to be scoffed at since they were so very poor. 
However, they have come of age and can now do very well.

Non-linear things take more time, but in coiling, the models seem to like the
chaotic stuff.  Matching various caps with various gaps and all can get really
messy firing going on.  But if you step back and compare that to what really
happens.  It is the same...  Programs like MicroSim have the advantage in that
they are time-iterative so nasty multi-order problems tend to work out very
well.

Ed's mention of NST secondary modeling may at first seem daunting to us that
know what it takes, but we are pretty resourceful :-))

Cheers,

        Terry

 

At 06:51 PM 10/4/2001 -0400, you wrote: 
>
> In a message dated 10/4/01 2:16:08 PM Eastern Daylight Time, tesla-at-pupman-dot-com
> writes: 
>
>
>>
>>
>>    One useful contribution to general knowledge here would be for someone 
>> to go to the trouble of trying to model the outer layers of an NST 
>> secondary and then determine current and voltage distribution during a 
>> discharge.  Can surely be accomplished by modern modeling techniques, 
>> but I don't know how to do it. 
>>
>> Ed 
>
>
>
> After twenty years as a systems simulation and forecasting engineer in the
> pipeline industry, I am somewhat aware of the strengths and weaknesses of
> simulator packages. They work quite well for deterministic steady state
> systems and quasideterministic periodic dynamical systems and poorly for
> chaotic dynamical systems. I also happen to believe (and it IS a belief) that
> chokes probably have too many disadvantages that outweigh any possible
> benefits. I also believe that simulation applied to chaotic dynamical systems
> is, as one on my numerical methods professors used to say, "The fine art of
> pushing a dead mouse through a maze and carefully recording what it does."
> Put in more prosaic terms, the region of stability of the model is never
> exactly coincident with the region of stability of the system it represents.
> Usually, simulations are only tested against reality when they say that
> something will work, not when they say it won't. 
> In chaotic systems, non-linearities are as likely to cascade as they are to
> die out, and this is where second-order approximations can run into trouble,
> still leaving room for experimental work to verify the models. 
>
> Matt D.