From: Paul
Date: Sun, 31 Dec 2000 16:35:27 +0000
Subject: [TSSP] Progress report
I've been quite busy with the modeling software over the hols, quite a bit has got done, so here's a summary. The new capacitance determination program is almost working properly. There are a couple of mathematical bits that I'm stuck on, eg the calculation of self-potentials for the curved surfaces of toroids, but rather than get held up on that, I've put in some temporary calibration factors to hide a residual 1% error in external cap and rather more in internal. Some inaccuracy in determining the end capacitance of secondaries has been cured by increasing Cext resolution from 32 to 500 steps. I'll do a similar thing with Cint in the near future, which should remove the need for a cal factor. The modeling code has been set to run at a reduced resolution so we are trading precision for throughput. Accuracy is now only about 2% at f1, 3% at f3, and 4% at f5, but it does mean that a resonator can be fully modeled in about 1hr on a P500 - that includes working out the cap matrix. Both cap matrix and transmission line simulator element sizes are now around 1/500th of the coil length. As a result, my cluster machine can now process around 280 systems per day, which means that we can soon start to generate a database of coil performance, albeit at modest precision. About 30 days or so will give us statistics on around 10,000 'virtual' coils, chosen to span the space of typical Tesla configurations. I've also added in modeling of primary windings and their coupling, so there is now an option in the model software program to feed via primary rather than base - I'll add in a center feed for the bipolar configur- ation in the near future. Q factor estimation is still poor, I think, although we're a little short on reference measurements. If loss factors are adjusted to align model Q with measured Q, then the input impedance, Q and voltage gains are commensurate with one another, which means that the software is correctly implementing the lossy equations - its now just a case of getting a reliable proximity loss estimate. I'll just review the model comparison results for the various systems that I think of as a 'reference' database. Quality results on these are essential, for two reasons. a) They will be required to 'stitch' a virtual coil database to real coil results, which should take place at a fair number of check points in the configuration space. And b) In order to persue checks on some of the predictions in pn2511, we must have accurate values for the equivalent reactances. A prerequisite for this is an accurate match on the first three odd quarterwave resonant frequencies, f1, f3, and f5. Lets see how well we're doing... First, my coils: |pn1: h/d 1.36 bare | f1 150.7kHz 150.2kHz -0.3% | f3 360.0kHz 362.0kHz +0.6% | f5 543.0kHz 548.7kHz +1.0% | |pn2: h/d 2.76 bare | f1 90.9kHz 91.4kHz +0.6% | f3 213.0kHz 216.8kHz +1.8% | |pn2-t: h/d 2.76 loaded | f1 66.7kHz 65.9kHz -1.2% | f3 193.3kHz 194.2kHz +0.5% | f5 307.0kHz 295.1kHz -3.9% >From Terry Fritz: Some ambiguity with the inductance of this high loss coil. |tfltr: h/d 2.92 bare | f1 148.4kHz 146.5kHz -1.3% | f3 353.4kHz 349.6kHz -1.1% | f5 513.8kHz 502.0kHz -2.3% | |tfltr45: h/d 2.92 loaded | f1 97.9kHz 95.3kHz -2.6% | f3 321.4kHz 315.7kHz -1.8% | f5 490.2kHz 478.9kHz -2.3% >From Malcolm Watts: Results inherited from Q factor tests. |mwa1-1hd0: h/d 1.00 bare, high elevation | f1 600.0kHz 667.1kHz +11.2% | |mwa1-4hd0: h/d 4.00 bare | f1 224.0kHz 225.5kHz +0.7% | |mwa2-4hd0: h/d 4.00 bare | f1 220.0kHz 225.4kHz +2.5% >From Mark Rzeszotarski: Two problematic sets of data for high elevation coils, |mz1: h/d 6.00 bare | f1 885.0kHz 1048.9kHz +18.5% | f3 2338.0kHz 2525.0kHz +8.0% | f5 3436.0kHz 3547.5kHz +3.2% | |mz2: h/d 6.00 bare | f1 645.0kHz 705.0kHz +9.3% | f3 1627.0kHz 1696.9kHz +4.3% and two examples from a set of 5 readings just received, mz3012-1: h/d 3.18 bare | f1 647.8kHz 696.2kHz +7.5% | f3 1575.4kHz 1656.0kHz +5.1% | f5 2264.1kHz 2369.4kHz +4.7% | |mz3012-5: h/d 3.18 bare | f1 665.9kHz 725.5kHz +8.9% | f3 1591.1kHz 1697.0kHz +6.7% | f5 2277.9kHz 2398.9kHz +5.3% these coils demonstrate our continuing inability to master the strange spacetime distortion present in Mark's basement. More on these in another post to follow. >From Marc Metlicka: Two interesting coils with large h/d |mm1: h/d 8.72 bare | f1 455.5kHz 436.1kHz -4.3% | |mm2: h/d 10.11 bare, turns guessed as 1700 | f1 276.9kHz 279.0kHz +0.8% | f3 711.8kHz 705.9kHz -0.8% a third with large inductance |mm3: h/d 4.63 bare | f1 61.9kHz 63.1kHz +1.9% | f3 157.9kHz 154.6kHz -2.1% | f5 229.7kHz 218.1kHz -5.1% | f7 294.4kHz 273.0kHz -7.3% | f9 355.6kHz 323.1kHz -9.1% and a fourth with a medium h/d |mm4: h/d 6.72 bare | f1 237.0kHz 255.1kHz +7.6% these promise to be a useful set of coils. Inductances need to be checked. Others: |thor: h/d 3.75 loaded | f1 65.5kHz 64.2kHz -1.9% | f3 222.8kHz 220.7kHz -0.9% | f5 346.3kHz 337.8kHz -2.5% which has a large discrepancy in secondary inductance. We could do to extend this database a little by collecting measurements on more reference systems, and to refine some of the existing results. More to follow later, Regards, -- Paul Nicholson, Manchester, UK. --
Maintainer Paul Nicholson, paul@abelian.demon.co.uk.