Mounting Toroidal Inductors with Common Materials
This information is reprinted by permission and is a copy of a post made to the rec.radio.amateur.homebrew Usenet newsgroup by Roy Lewallen, W7EL. Securing the toroidal inductor in a VFO is very important and Roy reviews tests of various materials which can be used to do the job. He refers to the materials as coatings and these materials are used to affix and immobilize the inductor to the PC Board. In addition, two other related experiments are discussed which are very interesting.
Toroidal Inductor Measurements by Roy Lewallen, W7EL December, 1998
GR 1606-A impedance meter. Stray series impedance was removed by initial calibration. Stray shunt capacitance was separately measured and removed mathematically. Repeatability is within about +/- 2%. No attempt was made to establish accuracy.
A home-made fixture was used. This is simply an air-variable capacitor which the inductor is connected across. Coupling into and out of the fixture with signal generator and oscilloscope is done with very small capacitors. The Q is calculated from the center frequency and 3 dB bandwidth. Use of frequency counter, built-in oscilloscope voltmeter, and switchable 3 dB pad allow repeatability of about +/- 2%. Checks against a commercial Q meter show agreement within a few percent. All measurements were made at 10 MHz.
Experiment 1: Coatings
Several identical inductors were fabricated, then coated with various compounds. Inductance and Q were measured before and after coating. The inductors were wound on Micrometals T-50-6 cores (Carbonyl SF material, relative permeability 8.5) with 25 turns of #22 wire. This just fits on a single layer.
Notes (by coating type):
All inductance measurements are in microhenries Unless otherwise noted, coatings covered the entire core and winding, extending more than a wire diameter beyond the outside of the wire, and the center of the core was filled. Inductances within about 0.3 uH and Q's within less than about 5 should be considered equal.
The uncoated control inductor inductance was measured several times over several days to establish repeatability. Results were 280, 281, 283, 286 and 284 uH. (The apparent trend is interesting, but not conclusive.) Devcon Duco(R) Cement, allowed to dry for 24 hours. Although a generous coating was applied, the dried coating was less than a wire diameter in thickness, and the center of the core wasn't filled. Dap Dow Corning 100% Silicone Sealant, Clear, allowed to cure for 24 hours. This is standard acetic-acid (vinegar) curing RTV. Stanley All Purpose GlueSticks, claimed set time 25-30 seconds. These are nearly clear, translucent white, and not tan or brown colored. This was some stuff I got surplus. It's a black, sticky, rubbery compound something like Coax Seal, but may be of entirely different composition. It's in the form of a thick tape. It's soluble in naphtha (and the solution dyes everything black it gets on), but acetone doesn't touch it. Household canning paraffin. (I don't know what it's called in Britain, but I'm not referring to the liquid -- kerosene to us -- you call paraffin. This is a common wax made from petroleum.) GE Silicone II Household Glue & Seal, Clear. This is a non-acetic-acid curing RTV. Allowed to cure for two days. Still just a little soft even after this much curing.
Additionally, following the measurements of the inductor which was coated with sealing tape, the coating was removed and the was winding cut off. After further cleaning the inductor was rewound and the measured inductance was 2.51 and the Q was 277.
I had made some measurements years ago, but couldn't locate the results. I recall that standard RTV was poor (lowered Q noticeably) but that an industrial non-acetic-acid curing RTV was good. The results with standard RTV in this test were striking. Either a) the formulation of standard RTV has changed over the years, b) there are major differences among brands, or c) my memory is faulty. I recall from my earlier tests that epoxy was quite poor, but this has to be qualified after the experience with RTV. There's a huge number of different types of epoxy, and some may be much worse than others. I might test some in the future, but didn't during this test.
I intend to test Q-dope in the future, but didn't have any on hand.
Of the materials tested, both types of RTV stand out as having a negligible effect on inductor Q. Hot melt glue and paraffin have a small enough effect that they should be tolerable for many applications. Duco cement seriously degrades Q, even in a much thinner layer than the other coatings. The "sealing tape", tested out of curiosity, shows just how great a degradation can be caused by a poor coating. None of the coatings made much of a change in apparent inductance. This implies that the reduction in Q is due primarily to dielectric loss rather than simple increase in capacitance due to the material's dielectric constant.
Note the difference in inductance and Q between the coated and then coating removed sealing tape inductors, which were wound on the same core. Apparently physical differences in the windings (perhaps such as tightness and conformance to the core, or uniformity of turn spacing) are a major contributor to differences between inductors. All the cores used in the test were ordered at the same time, so they may have come from the same batch and have relatively little variation. On the basis of just the comparison between the inductor coated with sealing tape and then the same core cleaned and rewound, it's entirely possible that most of the variation between inductors in this experiment is due to winding differences. The variation might be less if smaller wire with less stiffness is used.
Experiment 2: Turn Spacing
I believe it's well established that even a partial second layer can greatly reduce the Q of a toroidal inductor. But I had recently heard that optimum Q is achieved when the first layer isn't quite full, but rather has about a 30 degree gap in the winding, to reduce the capacitance between winding ends. To test this, I wound 23 turns of #22 wire on a T-50-6 core, and measured the Q with the turns pushed close together to make a gap (of about 30 degrees), and then spread to evenly distribute the turns completely around the core. Measured 10 MHz Q's were 272 and 284, respectively. (Inductances were 2.21 and 2.11 uH.) This one test doesn't by any means exhaust all the possibilities of core geometries, permeabilities, number of turns, and wire size, all of which may play a role. But if there's any advantage to leaving a gap, I believe it would be a small one. And in at least one case, it's slightly better not to.
Another test was run using 10 turns of the same size wire on the same core. With the turns pushed together (the winding covering less than half the core), Q was 213, L was 841 nH. With the turns spread evenly around the core, Q was 209 but the L had dropped to 505 nH. I reasoned that a more fair comparison would be with a winding of about the same inductance as the original close-spaced one. This required 15 turns when distributed around the core, and resulted in a Q of 257 and L of 924 uH. Here again, the Q is best when the turns are evenly distributed. Note that type 6 powdered iron has a very low relative permeability (8.5), so results might be different with higher-permeability materials. However, this is the material I usually use for high-Q inductors at HF, so I'm most interested in how it's affected.
Experiment 3: "Regressive" Winding
In the past, I've found what I thought was a moderate improvement in Q by "regressively" winding an inductor. To do this, you wind half the turns in the normal manner. Then you pass the wire through the hole, but to the opposite side of the inductor (with it ending up beside the first turn), then completing the winding from the vicinity of the first turn back toward the origination of the crossover. The result is an inductor with the two leads coming from points directly opposite each other. Stray capacitance is allegedly reduced by keeping the ends of the winding apart. An inductor wound in this manner with 25 turns of #22 wire measured Q = 285, L = 2.48 uH. This Q is on the high side, and the L on the low side, of the uncoated inductors measured in Experiment 1. I didn't try comparing with a standard winding on the same core, since the same number of turns wound on the same core at different times (e.g., inductors 6 and 11 in Experiment 1) were shown to come out differently from each other. My conclusion is that any Q improvement due to "regressive" winding is slight. Another claim for "regressive" winding is that it eliminates the "single turn" effect of toroidal inductors. A normal toroid will couple into its surroundings as though it consists of a single turn the size of the core. In the "regressively" wound inductor, there are two half-turns in opposite directions, so coupling should be reduced. I haven't tested this in any way, but it may be an argument in favor of the method. The relatively long wire of the crossover turn would contribute to coupling, however.
I'll undertake more experiments and measurements as time permits. I'd love to hear from anyone who has made either supporting or contradictory quantitative measurements.
Roy Lewallen, W7EL