2008-08-20
Antenna Experiments are Great Teaching Aids
With the QRSS beacon now working pretty well it is down to getting the antenna system in better order. My limited space gives me little chance of a full-size dipole or vertical, so loaded compact antennas have been the mainstay of my station.
The Loaded Dipole
The 30 meter compact dipole is only about 5 metres long. It is currently erected sloping from the guttering to the balcony railing. Each arm has a 2 metre radiator then a ~27 uH loading coil wound on a piece of PVC pipe using 1.8 mm magnet wire, followed by a roughly 1/2 metre capacitive loading element which is rolled up at the ends to roughly tune the antenna. Fine tuning is achieved using the matching unit at the feed point.
The Matching Unit
The matching unit is a conventional Z-match topology.
The coil is 16 turns on a T120-6 core. The load winding is 6 turns over the main resonated winding. The coax centre feeds into the resonated winding via a 6-210 pF polyvaricon (both gangs together in parallel) at a tap 5 turns down from the "hot" side. The cold side is returned to the coax braid. Another polyvaricon is used to resonate the main winding, also using both gangs paralleled.
Achieving a match is an iterative process, but quickly finds very good return losses. I typically just hang the impedance bridge off the feed point while tuning up. No explicit balun is implemented, but current balance in the dipole appears pretty good (based on fluorescent tube experiments - see below) - that said no attempt to measure braid current has been made but it doesn't seem significant.
The Vertical
As detailed previously the 3 metre aluminium radiator used for the 80 metre beacon is matched simply with a polyvaricon (120 pF gang only) varying its effective capacitance as seen by the 7.8 uH loading inductor. The inductor is wound on a piece of ~1" OD PVC conduit.
Once the capacitance of the whip was tuned out it appeared as almost exactly 50 ohms resistive. This took me by surprise, I expected a much lower value. At the time I blamed excessive ground losses, but even then that didn't completely make sense as on 80 metres the loss wasn't anywhere near that bad and the increase in radiation resistance with the higher frequency couldn't explain the high feed point impedance either. It worked, so I choose to ignore the odd result... Big mistake, science 101; when your results and model disagree check both!
Antenna Interaction
While initial QRSS transmitting experiments were carried out on the dipole (the vertical matching not yet attempted), all the recent transmissions have used the vertical. The signal seems better in practice, but I was curious how much the two antennas interacted, especially after the odd feed point impedance measurements of the vertical. When I switched over to using the vertical I simply disconnected the feed line to the dipole and plugged it into the vertical in close proximity. With the Z-match still tuned-up at the feed point all seemed fine, but I got a very good lesson in electromagnetics when I checked the current distribution using a fluorescent lighting tube!
With the vertical energised with only 1.5 Watts from the beacon TX I could easily light the tube in proximity to both dipole antenna loading coils. Clearly the antenna was parasitically absorbing a lot of the near-field of the vertical, no doubt re-radiating it and distorting the pattern of the vertical.
Experiments have just begun into the nature of the interaction, but detuning the Z-match or disconnecting it reduced the antenna current to negligible levels. Unfortunately this causes the feed point resistance of the vertical at resonance to drop enormously and offer a very poor match to the feed line. Remote signal levels seen in Grabbers are inconclusive, signals seemed to improve immediately after the change, but have since dropped off and now appear much worse! Only careful experiments with switching the resonance on and off could really tell for sure if the pattern produced by the interacting antennas is "better" for some destinations or not. Experiments such as this are complicated to implement because of the interaction itself and how much it changes the feed point impedance of both antennas. For an "apples to apples" comparison the antenna configuration in use must match the TX properly, accept all its power and not change the power the TX can generate (i.e. offer the same load).
Another complication was noted, the change in return loss of the vertical caused the beacon TX internal temperature to change, resulting in a visible drift of the beacon frequency by several Hz. A similar effect was noted when a freak sleet storm detuned the antenna enormously by covering it in icy slush. This is a form of drift I didn't predict after having gone to quite a length to minimise direct load pulling. I did have second thoughts during assembly about putting the amplifier inside the same box because of its heat generation, I had hoped it would just help keep the beacon above room temperature - which in fact it does quite well - but variations in dissipation with load condition is now something else to consider. Nature is often a very inconsiderate teacher, but a completely consistent one at least. No doubt people in the space program and other mission critical engineering applications have similar battles with what is totally obvious *after* it happens.
Fixing the Vertical
With the dipole no longer significantly interacting with the vertical I measured the input resistance of the vertical + loading coil at resonance as almost exactly 16 Ohms. An L-network was designed to match this to the 50 Ohm transmission line. The series inductance of the L network is small enough (370 nH) to absorb into the existing 7.8 uH loading inductance, leaving just the 458 pF shunt capacitance to be added. A 470 pF silver mica transmitting cap was soldered across the feed point allowing an excellent return loss from 50 Ohms to be achieved.
The feed point resistance once resonated on 80 metres was 22 ohms (must recheck this value my technique and instrumentation has improved since then). With a 500 milliohm radiation resistance on 80 metres this puts the antenna efficiency at about 2.3 % or about 16 dB down on a lossless 1/4 wave vertical. A similar calculation for 30 metres where the radiation resistance is about 4 ohms gives 25% efficiency or 6 dB down. 6 dB is one S-point - the loss is fairly trivial on 30 and not a massive deal on 80 either despite what might be called woeful efficiency.
How valid is this loss figure? If we take the 16 - 4 = 12 Ohm loss resistance for 30 metres and assume half of it is from the inductor finite Q then the Q figure that would give 6 Ohms loss is only 83 (XL ~= +j500). One hopes the inductor Q is better than this... Lets assume the inductor Q is at least 150 which seems reasonable, this is a RL of 3.3 Ohms, leaving the losses to be dominated by the ground loss, about 10 Ohms of it. Does 10 Ohms ground loss seem reasonable? The balcony railing is Stainless Steel which means it has a higher resistivity than Copper, but also a deeper skin depth as a result. Physically it isn't large enough to be a "good" ground-plane, so 10 Ohms does sound fairly reasonable IMO.
For 80 metres where the XL is 1.4 kilo-ohms we have about 10 ohms more apparent loss. An inductor Q of about 117 gives 12 ohms of loss to add to the 10 ohms of estimated ground loss (likely not the same on 80). The geometry and construction of the 80 metre loading inductor could easily see its Q in this region. Measurement of the actual inductor Qs at the frequency of operation would be instructive to get a better estimate of ground loss, but it is obvious that the antenna efficiency is likely dominated by the ground losses on 30 metres and even the best efforts to improve loading coil Q on 80 won't improve the overall system efficiency all that much.
Other Notes
Because these are narrow-band devices and matching techniques, the return-loss seen on your instruments will rapidly tell you if you have a harmonic energy problem. A good return loss peak (or null if you prefer) will be almost impossible to achieve unless the transmitter is fairly spectrally pure.
My fluorescent tube ionises more easily than an NE-2 bulb - this was surprising. However its ionisation hysteresis seems larger. This is probably predictable as it is filled with an Argon/Mercury mixture rather than Neon and the pressure is likely quite different, but an interesting observation none the less.
The impedance of the 3 metre vertical is much lower on 30 metres than on 80, hence the Q is smaller and tuning up is easier and less touchy. Also the RF voltage produced at the same power level is smaller so Polyvaricons are quite suitable in the matching network. This also means my usual method of using an NE-2 bulb hanging off the mast clamp to tune for maximum output doesn't work. :( One of the big advantages of QRP is that you don't need insanely good components to withstand the voltages and currents associated with QRO operation, and you can't really burn yourself even from high-Z points.
Antennas aren't magic. I know that is stating the obvious, but like most Hams I think I am a little in awe of how a piece of metal the right size and shape can transduce electromagnetic energy with space. The laws of physics still apply and even your basic AC circuit theory is usable in the antenna domain. Radiation can make it seem like it is different, but if you model radiation much the same as loss (an extra resistance) it all works and makes sense. Antennas can be extremely confusing at times, and it is all too easy to omit the major radiator from your model! If experience has taught me anything about antennas (or RF in general) it is that you have to measure everything you can, confirm your assumptions and test your model. The more I work with antennas the more I am driven to build test equipment, especially impedance bridges, RF ammeters and (calibrated) field strength meters. To me this is what Ham Radio is all about, a technical hobby of life-long learning about the electromagnetic force of nature.
Leave a comment on this article.
Attachments
| title | type | size |
|---|---|---|
| Z-Match Circuit Source | application/postscript | 10.743 kbytes |