From: Paul
Date: Fri, 15 Jun 2001 23:01:28 +0100
Subject: Re: [TSSP] Short Coil Experiments
Kurt wrote: > the new measurements of the d=0.385m; h/d=1.15 Thanks. I've run these through the mill to get sk38b50: bare d=0.384m h/d=1.15 sr=0.79 turns=346 w/o mat f1 221.3kHz 230.1kHz +4.0% f3 506.5kHz 557.2kHz +10.0% f5 748.8kHz 882.5kHz +17.9% f7 831.0kHz 1222.5kHz +47.1% without any correction for material dielectric. (Kurt, the f7 measurement is suspicious - I'd expect something around 980 kHz. I see you flagged it as questionable.) The rising trend in errors is clear, and similar to what we have seen on other coils, eg Finn's fh1: bare d=0.160m h/d=0.91 sr=0.86 turns=500 w/o mat f1 340.0kHz 361.8kHz +6.4% f3 725.0kHz 878.5kHz +21.2% f5 1100.0kHz 1355.9kHz +23.3% f7 1450.0kHz 1838.2kHz +26.8% We've seen this pattern on other coils too, but thanks to these readings we begin to see clearly that it is a small h/d thing. Evidence accumulates that the errors occur because we don't take account of the material dielectric. Finn's coil has 5.5mm of PVC (rel.perm = 3.4), and Kurt's is 8mm of solid paper (rel.perm = 3.1). If we take the material thickness as a percentage of the length, we get Kurt's: 0.8/44 = 1.8% Finn's: 0.55/14.5 = 3.8% which we can use as an indication of the potential effect of the former. I put in a rough correction for this, which came out with the right sort of order of magnitude of change, which is encouraging. It also improves the results from larger h/d coils too, where the rising error trend is still present, but less, eg Kurt's Sk-long coil: sk16b50: bare d=0.161m h/d=8.71 sr=0.85 turns=1976 w/o mat f1 152.3kHz 155.5kHz +2.1% f3 387.2kHz 385.8kHz -0.4% f5 564.2kHz 567.2kHz +0.5% f7 713.0kHz 725.8kHz +1.8% f9 850.0kHz 871.8kHz +2.6% f11 973.3kHz 1010.2kHz +3.8% If we put in an approximate correction for the 3mm PVC, we get sk16b50: bare d=0.161m h/d=8.71 sr=0.85 turns=1976 matcor f1 152.3kHz 155.3kHz +1.9% f3 387.2kHz 382.4kHz -1.2% f5 564.2kHz 557.8kHz -1.1% f7 713.0kHz 708.4kHz -0.6% f9 850.0kHz 845.0kHz -0.6% f11 973.3kHz 972.9kHz -0.0% which suggests that the correction is not quite enough. A similar correction to Kurt's h/d=1.15 coil gives, sk38b50: bare d=0.384m h/d=1.15 sr=0.79 turns=346 matcor f1 221.3kHz 225.9kHz +2.1% f3 506.5kHz 525.4kHz +3.7% f5 748.8kHz 810.1kHz +8.2% f7 831.0kHz 1097.5kHz +32.1% which, apart from the doubtful f7, shows a similar improvement. The correction used counts as a fiddle factor, although the order of magnitude is about right. I'll have to do something a bit smarter, but it means writing another laplace solver - about a days work. Fortunately I only have to run it once in order to produce a suitable functional approximation which can then be incorporated into the model. Kurt, you also supplied results for Sk-5cm, sk5b503: bare d=0.051m h/d=8.03 sr=0.91 turns=934 w/o mat f1 979.7kHz 1030.8kHz +5.2% f3 2428.7kHz 2531.9kHz +4.3% f5 5300.0kHz 3707.7kHz -30.0% f7 6924.6kHz 4739.0kHz -31.6% f9 8643.1kHz 5694.6kHz -34.1% Something a bit astray with f5 and above. I get a better match if I interpret the results as: sk5b503: bare d=0.051m h/d=8.03 sr=0.91 turns=934 w/o mat f1 979.7kHz 1030.8kHz +5.2% f3 2428.7kHz 2531.9kHz +4.3% f5 3707.7kHz f7 4739.0kHz f9 5300.0kHz 5694.6kHz +7.4% f11 6605.9kHz f13 6924.6kHz 7502.1kHz +8.3% f15 8384.8kHz f17 8643.1kHz 9272.9kHz +7.3% and with material correction: sk5b503: bare d=0.051m h/d=8.03 sr=0.91 turns=934 matcor f1 979.7kHz 1027.3kHz +4.9% f3 2428.7kHz 2486.4kHz +2.4% f5 3585.5kHz f7 4519.3kHz f9 5300.0kHz 5363.7kHz +1.2% f11 6154.7kHz f13 6924.6kHz 6921.2kHz -0.0% f15 7666.8kHz f17 8643.1kHz 8410.0kHz -2.7% Kurt, perhaps you can have a look for the missing resonances? One last observation, concerning Q factors: The Q of Kurt's sk38b50 is very low, around 100, compared with a predicted Q of well over 500. Is this the sonotube effect we see here. My book on dielectric properties gives the loss factor of paper/cardboard as 0.02, which is abysmal. I'll see if I can use a laplace solution to improve the material correction, but so far it sure looks like it can account for the bulk of the higher mode error. Cheers, -- Paul Nicholson, Manchester, UK. --
Maintainer Paul Nicholson, paul@abelian.demon.co.uk.