Sunday, August 17, 2014

Managing Interactions: Creating an Interaction Model

Modelling interactions among antennas and metal supports is not easy. It is so 'not easy' that it is rarely done by amateur radio operators. Instead most hams proceed by rumour, lore and what they read in articles and antenna books. Nowadays we can add in the advice offered via the internet. Some is true, some is false, and much of it goes unverified for years, even in respected and widely-distributed books.

That is what I used to do for many years. Basically I did what seemed best, getting advice from other hams who really had no better understanding of the subject or I would just shrug and hope for the best. Since any antenna will work, even reasonably well under the burden of serious interactions, there is some truth to a certain silly saying: what you don't know can't hurt you. Although antennas in this situation can be successful they can be made to perform better if attention is paid to interactions.

Modern NEC-based modelling software is a tremendous help. However it remains a challenge to build a model with all those antennas, masts and tower and get useful results out of it. It's an individual choice. Since I am writing these words you can guess which way I lean. Even so I am not a glutton for punishment; I want to get the best results for the least amount of work. That is, I want a model that is simple enough to retain the key elements needed to identify and mitigate unwanted interactions but without worrying about high precision. I focus more on precision for models of individual antennas.

My immediate motivation is the tower I just put up. In addition to a tri-band yagi I intend to put up wire antennas for at least 40 and 80 meters. There will also be antennas nearby on the smaller tower plus mast bracketed to the house. For the present I will ignore these latter items so that I can start with a basic interaction model. Other antennas will be added over time. I will start with the larger tower and the antennas it will support.

If all goes well I will have received the tri-band yagi to be mounted atop the tower in the coming days. It is not a TH3 but is close enough that I used it in my model. To be more precise, I am using the model I developed during my series of articles on high-band yagis. At right is an EZNEC view of this antenna on the tower, along with a prototype loaded half-sloper for 80 meters. This is the first iteration of my comprehensive model.

Since it is difficult to see in the plot I will point out a few noteworthy items in the model. I will elaborate on these further in this article.
  • The yagi has a boom, but the original tri-bander model does not. I omitted this earlier since while the boom does alter antenna tuning by a small amount it is not pertinent to measuring performance in isolation. In a comprehensive model with diverse interactions that is is no longer sufficient.
  • The rotator and mast bearing are not modelled. It is assumed that there is sufficient conductivity between the mast and the tower to exclude those items. There is some risk here since the rotating surfaces are coated with grease and might not reliably yield metal-to-metal contact. Some hams attach a flexible strap between the mast and tower to guarantee continuity for antennas that include the tower, mast and yagi(s).
  • The tower and mast for the yagi are modelled as simple wires. This is not accurate but with some care can be made to be sufficient.
  • Guy wires are not modelled. There are two guy stations and a total of 6 guys. They were designed and built to be largely non-resonant for all antennas supported by the tower. Each guy has 3 sections, separated by insulators: first section is very short (~45 cm), to isolate the tower; second section is 8 meters long (26'), which is near resonant on 17 meters but no other band; the final section is of whatever length is needed to reach either an anchor or a terminal insulator.
  • Cables are not modelled. These will be added later. However I will only add cables in select instances and then only until reaching a common-mode choke, which is assumed to act as an infinite impedance.
  • Every antenna has a source, only one of which will be powered on each run of the model. The other sources represent unused transmission lines that terminate in a short or a finite-impedance transceiver (receiver or transmitter) in the shack. But those will in most instances be excluded from the model.
  • The tower is not connected to the ground. NEC2 does not support wire connections to the ground or wires in the ground. This does not permit modelling of directly-connected ground rods. It is true that at present in its unadorned state the tower is not grounded, but this will change.
Now let's look at a few subjects in more detail.

Yagi model

I have several key objectives in modelling the yagi in the interactions model:
  • Permit a reasonably-accurate top hat model for low-band antennas that include the tower and all its attachments.
  • Retain its performance metrics of gain and directivity (F/B, F/S). I am less interested in retaining accurate SWR calculation since this is less pertinent to interactions; accurate SWR is already designed into the standalone yagi model.
  • Accurately model coupling induced by other antennas.
  • The design should allow the yagi to be rotated to test interactions for when wire antenna elements are parallel or orthogonal to the yagi elements. Since this is a software model we can instead choose to rotate the other antennas if that's easier. Modelling software typically allows selective rotation, but it is up to us to set up the model to make it convenient.
The yagi boom is modelled by 2 wires of the requisite diameter. These attach to the wire representing the mast. If the mast continues higher than the yagi the mast must be made with 2 or more wires. The parasitic element centre sections are split into 2 wires each so that they can be attached to the ends of the boom (you can only connect wires at their ends with NEC).

The driven element does not touch the boom. This represents any yagi with a dipole feed, such as Moseley, Hy-Gain and others. I had to "bend" them so that they pass each other without touching. A short centre wire connects the dipole halves. This is where the source and beta match are attached. The other end of the beta match is attached to the boom as on a typical Hy-Gain yagi.

This last connection is preferred to modelling a virtual short-circuited transmission line since the connection ensures that the driven element is included in the "top hat". It is assumed that a common-mode choke or current balun is at the yagi feed point. (Please note that the Hy-Gain BN-86 is not suitable for this application since it is not a current balun.)


One issue is the attachment of the short-circuited end of the beta match to the boom. Doing so introduces a modest mismatch, raising the SWR to 2 or even 3. You can see that there is excess capacitive reactance in the chart above. However there is a negligible impact on the yagi's performance metrics of gain and pattern. This is acceptable in an interaction model. Just be sure to do it differently in the yagi model itself, or find a way to mimic the impedance behaviour in the model. Although the latter may be possible I am not motivated to spend time on it since it is tangential to my objectives.

Tower model

You can forget about modelling a lattice tower with any hope of accuracy. Even if you took the long, painful step of creating wire for each cross-brace and leg sections between braces of a continuous-taper tower it would still not work. That degree of accuracy in such a complex arrangement cannot be expected using NEC2 or even NEC4. I briefly touched on this subject in earlier articles on 40 meters loop antennas.

A better approach is to model a single wire with a diameter equal to the average width of one face of the (triangular) tower, and as much as (or even more than) 10% shorter than the tower, as measured by W8WWV. Since you can't really shorten the tower at the top you should do it at the bottom by lifting the wire end off the ground. This of course does not permit accurate representation of radials or ground connection for lightning protection.

I suggest ignoring the lightning protection ground protection (impossible in any case with NEC2) as the lesser evil. W8JI provides some graphic examples of how to provide lightning protection for ground-isolated towers if that is of interest to you. Radials can still be connected to the tower bottom when it is above ground level in the model. With EZNEC at least it is easy to extend the tower downward while also keeping radial connections intact for test cases where the radial interactions are more important than accurately modelling tower resonance.

It is not practical to give better guidance since towers have diverse designs, behave differently on different bands and have various cables running down their length (inside and outside the tower), and are often buried where they exit the tower. There may be better ways to attack this problem.

Wire composition

In a comprehensive model of this type there are conductors mode from different metals. These typically include steel (tower, mast and guy wires), aluminum (yagis) and copper (wire antennas). EZNEC, like similar modelling software, often only allows one in a model. We need to choose which one. To do that there are a couple of considerations:
  • Loss: Steel is far lossier than copper, with aluminum intermediate. However zinc-coated steel (galvanized towers and cables) is almost as good a conductor as aluminum at RF. It is also true that in most installations that the steel and aluminum conductors are of large diameter, and therefore better RF conductors.
  • Resonance: As we saw with the tower model it is possible for large conductors and tapered elements to have different effective (RF) lengths from their physical lengths. We usually don't need to worry about this since these conductors are not parts of antennas. In an interaction model these differences can impact results.
For my model I am ignoring the loss implications since these do not directly impact interactions. The resonance and coupling factors are more important than individual antenna performance. This is unlike modelling of antenna performance where the wire composition can be significant. I specify either aluminum or copper in these complex interaction models. It is also acceptable to specify perfect (no loss) wires.

Segmentation

EZNEC tells you, sometimes quite forcefully, when you push the NEC2 engine beyond its capabilities. These warnings tell you when your model is liable to produce unreliable results. Pay attention to these warnings.

In the interaction model you can quite easily elicit warnings because of the wide range of test frequencies, perhaps as low as 1.8 MHz and as high as VHF. Here are the warnings I commonly run across in the interaction model:
  • Segments are too small (or large) for the test frequency. For example, the segment length that is optimum for 20, 15 and 10 meters on the yagi is often too small for 80 meters. The reverse can occur in the opposite case.
  • The tower can violate the maximum diameter-to-length ratio for a wire, especially if you break the tower into multiple wires to simulate the taper of a self-supporting tower. As indicated earlier it is better to model the tower as one wire. Tapering wires (in towers or yagi elements) is in any case a problem with NEC2 so you gain nothing with a manual taper.
  • Wires with short segments that attach to the tower could have one or two segments entirely inside the tower diameter. This is common with wires that form an acute angle with the tower. You can increase segment length (decrease number of segments) of the offending wire or add a short horizontal wire between the tower and the offending wire.
  • Wires that meet at acute angles or are near parallel should have equal segment length. This is especially vital if the wires are part of the same antenna. If one of the wires is only interacting with the other one -- not part of the same antenna -- it may be safe to proceed even if the segment lengths are unequal. Test with different segment length ratios to be sure. But if you can you should make them equal.
The first point requires some further consideration in the model since there is typically no way to avoid segment length violation at all test frequencies without compromising the model integrity of one or more antennas. I suggest paying attention to where the violation occurs and only taking action if there is a high probability of a problem.

For example, with a test frequency of 3.5 MHz on a half-sloper the segments of the yagi elements in my model were too short. After running the test anyway I found that the current on the yagi elements was small. This tells me that any error resulting from the yagi segmentation will also be small. So I ignored those warnings.

Coaxial cables

All transmission lines to my planned antennas are coax. The only time I use open wire lines is to build wire antenna arrays, none of which are in the current plan. The interaction model therefore only needs to model the transmission lines for common mode (direct conduction or coupling) and termination.

For interaction purposes I model coax as wires between the antenna feed point and the first common-mode choke. I am assuming that the chokes are perfect (infinite impedance components), whether coax coils, baluns or ferrite beads/toroids. Even with this restriction there can be a problem for coax running down the tower since if there is substantial current on the tower (e.g. half sloper antenna) it will couple to the coax runs. Also, whether buried or run between tower and shack in the air there can be coupling to horizontal antennas. If that becomes a problem it is perhaps best to add another common-mode choke so that there are no resonant sections of coax (outer conductor) that can couple to nearby antennas. Yes, it's messy but that is the actual situation whether we like it or not.

As already noted parallel conductors must be modelled with equal-length segments. For the purpose of modelling interactions this is typically more important than optimizing the segment length for any particular frequency.

If the coax is not choked at the feed point there should be a directly-connected wire to the feed point representing the outer conductor of the coax. If there is a choke at the feed line the coax can be modelled as a wire with no connection at the feed point. If a choke is placed where the coax leaves the tower (towards the shack) in most cases the coax run from feed point to the coax does not need to be in the model.

With the coax so close to the tower there are modelling issues with NEC2. Expect inaccuracies.

Use your judgment whether the section from the choke towards the shack is close enough to any horizontally-polarized wire antennas that it can have induced current on it. Don't worry about yagis on top of the tower since they will have little radiation in the direction of that section of coax.

Antennas that are not fed in each test (all but the one with an active source) will conduct any induced current towards the shack via their connected transmission lines. The impedance presented will depend on the length of coax and whether the shack end is shorted, open or connected to a rig's receiver or transmitter. My choice is to avoid this complex issue by not modelling the transmission line other than as (above) for common-mode currents. My reason is that if there is significant and unwanted current of that antenna the problem is the induced current and not its precise behaviour due to the transmission line and shack termination. I instead try to reduce or eliminate that current by moving or otherwise modifying one or more antennas.

Concluding notes

Even with its many flaws a comprehensive NEC2 interaction model of a complete antenna installation can prove very useful. It is certainly easier than building antennas and live testing their interactions in various configurations. Without extensive testing it is often impossible to tell if a problem is in an antenna's design or its interactions with other objects. It is therefore beneficial to test interactions before investing time and money, and to select antennas and their placement to optimum effect at the outset.

Be prepared for investing a bit of time and effort to make the model, and then don't be surprised that running the model (SWR or far-field patterns) is slow, even on a fast computer. It's a good thing I have EZNEC+ since with 2 or 3 wire antennas added to the model I have more than the 500 segments permitted by the basic version of EZNEC.

I have used the model I developed to begin testing a variety of low-band antennas. It has been very interesting and enlightening. This is something I'll return to in some future articles so that you can see how these modelling investigations are guiding my thinking about low-band antennas that I'll soon be building and erecting for the fall operating season. I started with an 80 meters half-sloper but have since gone on to inverted vees and loops.

Here are a few concluding notes on where I find that the interaction model is most useful:
  • Wire antennas that utilize the tower and and yagis as active elements.
  • Directional antennas, including yagis, in the vicinity of wire antennas below or beside them.
  • Interaction of cables and transmission lines with adjacent wire antennas.

No comments:

Post a Comment

All comments are moderated, and should appear within one day of submission.