Modelling Parallel (Fan) Dipoles

I've been remiss on blogging recently due to a combination of travel and other matters. As mentioned in my previous article I had to take close to 2 weeks away from antenna work. As of this long weekend I have restarted antenna work.

I was however able to get in several hours of operating during the great conditions the last few days. The conditions were good enough to put several new countries in the log, even those with large pile-ups which are the bugaboo of QRP operating with zero-gain wire antennas. Examples include: TN, TO2TT (Mayotte), FR, TX5D (Austral I.), ZM90DX (Campbell & Auckland Is.)[no, it's ZL]. As always there are the many that got away, that are just very difficult with my small station.

Now back to the present subject. In an earlier article I showed how fan dipoles are easier to model and reliably build when the antenna wires are parallel rather than radiating outward from the centre. Having established that I went on to build a fan dipole for 30, 20, 17 and 15 (plus 10 and 6) meters. That antenna remains up and continues to work well. It just needs some maintenance to prepare it to survive the winter.

Using EZNEC I modelled the basics of the antenna -- though not all of it -- to test its performance with respect to tuning sensitivity, wire placement, etc. Below is the EZNEC view of the feed point of a parallel fan dipole for 30 (top) and 20 meters, which appeared in that article. Wires #4 and #5 connect the antennas, and the source is at the centre of the 30 meters dipole.

This is not the best model for the feed. The reason is that there are a variety of hidden mathematical quirks in NEC2 that afflict short wires, sharp angles between wires and parallel wires. I did not run into those problems in that particular model but I did in a similar model.

This weekend I took down the TH1vn trap dipole and mast to work on them. The mechanical work to add an extended mast for a 40 meters delta loop was straight-forward. Not so for my other, related project to add 17 meters to the this trap dipole.

It is not easy to model a trap dipole with EZNEC (or other NEC2-based software). The model is nonetheless important since there is a potential for differing interactions on each of 20, 15, and 10 meters. There are both mechanical and electrical challenges which I am currently dealing with. When something comes of this experiment I will report back.

The driven element of the trap dipole (the driven element of my old TH6DXX) is quite short. Each ½-element is 3.8 meters long. This is well short of ¼-wavelength on 20 meters (~5.1 meters) due to the load added by the traps for 10 and 15 meters. Of particular concern to me is that this is shorter than a full-length ½-element for 17 meters (~4 meters).

While this is a rat's nest of issues I chose to concentrate on just one for now, since I expect that to be most messy. This is the coupling of the 17 meters parallel dipole to the full TH1vn on 20 meters. It is generally true that the antenna at the lowest frequency is mostly unaffected by parallel wires for higher frequencies, while the reverse is definitely not true. Right now I need to get the 17 meters design figured out.

I added inductors to the centre of each 20 meters ½-element so that it resonated at 14.1 MHz while at the correct physical length of 3.8 meters. I then added 17 meters wires using the feed as used before (and shown above). EZNEC choked on it, unable to calculate feed point impedance.

After some experimentation I isolated the problem to those short vertical wires. Changing the segment length changed how the problem was manifest but did not make things better.

When you see 2 closely-spaced wires connecting sources and loads in an antenna system, what does that make you think of? It is a transmission line. Transmission lines are much easier to model in EZNEC than all those problematic small wires, so that's what I did.

In the above EZNEC view of the modified feed, the source is placed at the centre of the 20 meters element (wire #1) and a transmission line (red square) connects the centres of both elements. This is certainly a less complex model.

Transmission lines in EZNEC have their limitation since, unlike wires, they do not interact with the rest of the model. They are pure and perfect transmission lines that neither radiate nor absorb RF. For such a short section of line (10 cm in this model) the impact should be negligible. It isn't even necessary to get the nominal impedance of the line exact since it's so short relative to wavelength. I got nearly identical results with impedances from 300 to 600Ω. If you like you can always calculate the actual impedance from the wire diameter and separation.

That's one (big) problem solved. If you model feed systems such as this you may want to do the same. You can find similar parallel dipole feeds in a number of commercial antennas such as the Spiderbeam and HEXbeam.

The other problem is the choice of segment length. While it may be difficult to pick out both of the above pictures, there are green dots showing segment boundaries. In the bottom picture this is ~12 cm for both the 20 and 17 meters elements. This is the important point: the segment lengths should be matched as closely as possible for close spaced wires. Do this for any NEC2-based engine.

Even seemingly small differences can, in the right (or wrong) circumstances, cause significant errors in the modelled results. In some of my models the impact was small and in others it was large. If in doubt, modify the segment count in one of the wires and check the SWR, current and far-field patterns. Or do what I do and always strive to adjust the segment count so that the segment lengths match.

Choosing the transmission line feed model makes this task easier since it is nearly impossible in most cases to also get the segment lengths matched as well with the separate wires in the wired feed model. Sometimes that matters, too, but not always.

To summarize, remember these points in the modelling of parallel dipoles.

• Connect the dipoles with transmission line equivalents of the physical wires. If there more than two dipoles add more transmission lines from each dipole feed point to the next.
• Match the segment lengths in each parallel dipole. Closer matches are possible as you increase the segment count.
• Place the source and transmission line terminations in the centre of each dipole. Do this by selecting an odd number for the segment count and connecting  to the middle segment. When you then place the connection (source or transmission line) 50% from one end of the dipole it will be properly centred.

Once I resolve the mechanical issues with adding that 17 meters element and attempt to tune it to operate on all four bands I will be back.