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

Re: Similation results predict racing arcs

Original poster: "Gerry  Reynolds" <gerryreynolds@xxxxxxxxxxxxx>

Hi Greg,

I chose 9 segments only because it evenly goes into 36 and it reduced my chances of making an error when positioning the segments using JavaTC :o)))) If you choose more than 9 segments and want to have every combination of coupling, you will need to enter into the simulator the n(n+1)/2 coefficients but you dont need to do as much work finding their values using JavaTC. For example, all pairs if secondary segments with the same spacing between the two segments will have the same coupling regardless of where they are located on the secondary coil. I didnt assume this at first, so I did all combinations but this turned out to be the case.

The rise time of the waveforms from the simulation was 2us and propagation time from bottom to top was 6us. Any signal edge would span 12 inches of coil so a 4 inch segment seems to be a fine enough resolution.

The simulation seems to predict high stress points from the presumed overcoupling of my coil located where my coil actually broke out which was a good correlation so I tried to see if the simulation would predict higher stress points if the coil was miss tuned. It seemed to, although I havent completed this experiment yet. It did show the frequency splitting that coupled resonant circuits can have and this caused me to revisit the tuning a bit closer to make sure I was really on a properly tuned resonant point. I will report more as I finish. Thankyou for the link. I will take a look when time permits. This link points to the work of Paul Nickelson??

Gerry R

Original poster: Greg Leyh <lod@xxxxxxxxxxx>

Hi Gerry,

Some very interesting work in Part2. Was curious about your simulation setup. Why did you settle on nine segments? It seems like a reasonable number, just curious why you chose it. I tried setting up a similiar problem in Simplorer, where I divided the sec into twelve segments. I never managed to get the simulation to produce a believable output waveform however, probably because I was too lazy to develop a coupling coefficient for *every* combination of segments and the primary; for simplicity I ignored any couplings below about 0.05, thinking they contributed little to the total stored energy.

It's interesting that your results seem to correlate well with actual observed racing arcs. I was also wondering if there's any shortcuts to the n(n+1)/2 requirement for the numbers of k's in the problem.

In case you haven't already seen it, you might want to check out the Tesla Secondary Simulation Project at:

An incredible amount of excellent work simulating the complex nature of coupled secondaries has been carried out there.


Original poster: "Gerry  Reynolds" <gerryreynolds@xxxxxxxxxxxxx>

Hi All,

I created a spice model of a distributed coupled secondary using the coupling coefficients calculated by JavaTC. This model has distributed coupling from the primary as well as the distributed mutual coupling between segments of the secondary. I wrote up the results and had Terry host it. The simulation results (using my coil for the model) seem to predict where on the coil the racing arc will breakout from and agrees with my observations.

Just before the conclusion, I stated why I thought the breakout went to the top of the coil - an explaination that I now think is wrong. I have since come up with a simplier explaination, so if anyone wants it, I can email the updates to you. If a lot of interest, maybe I can get Terry to rehost the document :o))


I also reposted the first part with some corrections:


Please look it over and comment. I think we will be able to predict the worthiness of any new primary design.