NEC5 Test Drive

I have a long to-do list for my station. It reflects a lot of ambition. However, my actual pace of work is at a lower rate. Amateur radio is a hobby after all. Some items linger on the list for a long time indeed! NEC5 was one of them. I say 'was' because I've finally purchased a licence for the software and added it to EZNEC.

EZNEC makes it easy to use NEC5 as an alternative calculating engine. Considering the many advantages of NEC5 over NEC2 and NEC4, and that EZNEC is now free, it is very easy to justify US$110 for a licence. I will not bore readers with what can be learned elsewhere about NEC5 and EZNEC, which you can find elsewhere, including following links provided in this article.

These are perhaps the most relevant resources I read before diving into my initial experiments with NEC5. It helps to know the technology you are dealing with before using it and trusting the results.

• EZNEC and the EZNEC 7 manual
• AC6LA's NEC5 notes (access requires a QRZ.com account)
• NEC5 validation manual
  

Next, a brief summary of why to choose NEC5 versus NEC2. Others cover it better and more completely, but a few points are enough to get you started:

• Tapered elements, including loaded elements -- with NEC2 you need the SDC (stepped diameter correction) in EZNEC, or its equivalent in other modelling software, and also stay within its strict constraints
• Antennas with wires meeting at angles, especially acute angles -- it can be done using NEC2 with fairly complex segmentation procedures, although EZNEC and at least one other package will help you with it
• Wire segmentation is in general not at all strict -- NEC2 requires strict segment alignment between parallel conductors when the separation is small
  
• Antenna elements with loads, lumped or distributed, are handled well -- includes capacitance hats, traps and coils inline with tapered elements and more that are often handled poorly in NEC2
  
• Radials can be on the ground or in the ground, and with more accurate results -- no more having to artificially place radials a short distance above ground, and consequent issues
  

All of these have been important to me at various times. Sometimes there are modelling workarounds and other times I had to mathematically calibrate a NEC2 model with measurements of a built antenna. NEC5 promised a far better modelling experience with less opportunity for errors.

There is a cost to gaining these advantages, and I don't mean the license fee. The greatest cost is segments: you need a lot. A peculiarity of NEC5 is that the calculations converge slowly with increasing segments. That is, you need a lot more segments than with NEC2 to get accurate results. The more complex or large the antenna, the more you need.

That slows the run time and can be a bit of a bother at times. Doubling the segments is frequently inadequate so I double again. I've stopped at 4× the amount though some go further. More on this as I go through my initial modelling experiments with NEC5. 

There are other differences that must be understood. For example, NEC5 does not natively support insulated wires. EZNEC includes an algorithm to allow them when using NEC5.

Luckily for most hams NEC2 continues to be perfectly adequate, especially with the enhancements included with EZNEC. Whether to use NEC5 depends on your interests and needs. This article may help you decide.

40 meter reversible Moxon

The initial experimental model of a 40 meter reversible Moxon was the first antenna I modelled with NEC5. I did nothing more than change the calculating engine, just to see how the results would change without the segmentation and other modifications recommended for NEC5 models. Recall that the model uses constant diameter wires (25 mm) throughout; this is not a physically realistic model, only intended for computer experimentation.

The NEC2 result is at the top and those for NEC5 on the bottom. Resonance shifted downward approximately 1% (~75 kHz). Gain and F/B were similar, after adjusting for the frequency change. Doubling the segments for NEC5 had a negligible impact on the results. 

I tentatively concluded that the difference is more likely due to improved accuracy with NEC5 due to the 90° angles between wires. Even for a straight forward antenna like this we can see where NEC2 is challenged to generate accurate results. 

For those practically minded, you may have more luck getting accurate dimensions from one of the online Moxon rectangle generators (typically based on scaling working antennas) than by using NEC2. I can't guarantee that is generally true, but you ought to keep it in mind should you model a Moxon rectangle or an antenna with similar attributes.

I next tried my physical model for the reversible Moxon. This one uses tapered elements for the antenna I am currently building in my workshop. I won't detail the design in this article since it is an ongoing process. I am trying to build it using only the aluminum tubes and pipes that I have in stock, which imposes interesting constraints. I'll write an article about the antenna when it is built and tested.

Unlike the model with constant diameter wires, the initial model was far off. I doubled the segment count which, unlike for the model with constant diameter wires, did reduce model inaccuracy. For the final design I may again double the segments. That would be 950 segments! That's a lot for what seems to be a modest looking antenna, yet it is frequently necessary for accurate NEC5 models

It was necessary to make a few small changes to achieve the expected performance. One was a slight adjustment to the reflector element coil. The reappearance of short stingers on the elements is due to the material I have on hand and not to a change to my previous conclusion that stingers limit performance for this style of antenna. I am trying to keep them short without resorting to buying more aluminum.

The SWR is only slightly higher after aligning the gain and F/B frequencies with that of the earlier model. It may improve further with another doubling of segments. That may not realistically matter since environmental interactions are likely to be greater: the tower, its guys, and other antennas on the tower. 

With NEC2 the model is far from accurate, worse than the relatively small 1% inaccuracy of the model with constant diameter wires. NEC5 is worth the price if only for this one antenna.

3-element 40 meter yagi with capacitance hats

When I began designing this monstrous antenna I knew that NEC2 would be wholly inadequate. I was less concerned with the impedance than with the frequency of operation. NEC2 typically shifts the resonance downward for antennas of this type, but there is no reliable method for predicting by how much. That is why I built an experimental element to calibrate the model. 

After repeated changes to various sections of the element (position of capacitance hats, length and diameter of select tubes, length and diameter of the hats and stinger), and approximate compensation for ground effects, I trusted that the measurements were sufficient to calibrate the model. Calibration means accurately determining the differences between the NEC2 model and the real antenna. It is not practical to repeatedly raise and lower a 300 lb antenna up 150'. It was critical to get it right.

The extraordinary thing was that the antenna seemed to do pretty well once it was installed and the gamma match and driven element were adjusted. However, small doubts lingered since the calibrated model was imperfect and there is no good way to measure gain and F/B with reliable accuracy. With the NEC2 model and manual calibration, the model's behaviour from 6.55 to 6.85 MHz was calculated to match the real performance between 7.0 and 7.3 MHz.

Then I suffered the effects of a poor mechanical design of the clamps that attach the capacitance hats to the element. One arm of one hat on the reflector fractured and fell off a few months after the yagi was raised. The same later happened to the director. At first I thought it would be disastrous, yet I could not discern a performance impact on air. Even the SWR curve was slightly better.

Modelling of the missing arms using my calibrated NEC2 model suggested that the antenna had its operating frequency range increased by about 70 kHz (1% of 7 MHz). That's not good but it also isn't bad. However the F/B at 7.0 MHz should have measurably declined. That didn't seem to happen. 

Although I redesigned and built new capacitance hat clamps, I've only replaced them on the driven element. That was easily done since it is close to the tower and the tips are accessible by rotating the DE on the boom. Replacing the hats on the parasitic elements is more difficult. I haven't rushed since the antenna continues to work well, and no more capacitance hat arms have failed over the following 2-½ years.

With NEC5 installed, I ran the EZNEC model without making any adjustments to accommodate the unique requirements of the calculating engine. At first glance the results were quite good. Performance between 7.0 and 7.3 MHz was about as expected; no calibration required. Then I took a closer look.

Every yagi design is a balancing act between gain, pattern and match. That isn't easy to achieve for a high Q antenna on 40 meters due to its 4.3% bandwidth. The NEC5 calculation for the F/B at 7.3 MHz was poor. Closer inspection indicated that the optimum range of the antenna was at a slightly lower frequency, by about 50 to 75 kHz. Notice something familiar about that number?

The frequency shift is very close to the higher operating range due to those missing capacitance hat arms. That could explain why the antenna is performing so well. I must quickly add that it is only a hypothesis at this point. It will be necessary to increase the segment count and make other adjustments to ensure that the NEC5 calculations are accurate. It is important not to stop the analysis just because the first hint of an insight conforms with one's subjective experience with an antenna. 

I will do that deeper analysis later, and likely with a lot more segments. For this test drive of NEC5 it is enough to achieve these inklings of enlightenment. I look forward to doing a full NEC5 analysis of this important antenna. I may decide to alter the antenna slightly when I install the new capacitance hats so that it works at its best.

5-element 15 meter yagi

The final model I tested was this long boom yagi for 15 meters. I chose it for two reasons. The first is that I've heard that multi-element yagis are challenging to accurately model with NEC5. The second is that several months ago I sent the model file (including a gamma match) to a ham who requested to see how I modelled the gamma match. He then passed it to someone who ran the model with NEC5. The resulting SWR curve was far from what I measured and successfully modelled with NEC2.

This is a complicated antenna simply because it has so many elements and each element is tapered with telescoping tubes. EZNEC with NEC2 and its SDC algorithm produced a model that matched the actual antenna with exceptionally good accuracy. However, when I added a gamma match to the DE it was necessary to replace the DE with a constant diameter wire since the SDC algorithm can't deal with the gamma match. This is not ideal since the current distribution on a stepped diameter element is not the same as on the equivalent constant diameter element calculated by the SDC algorithm, even though both exhibit the same net reactance.

When I modelled the antenna with the gamma match using NEC5 the SWR curve closely resembled what the other ham got with NEC5. I then returned to the version with a tapered DE and no gamma match and, again, the SWR curve was far off the one calculated by NEC2 (including SDC) and as measured on the actual antenna. That's the upper of the two charts above.

When I doubled the segments of every wire the SWR curve (bottom) improved, though still not very accurate. This result implies that the person who ran my model on NEC5 did not increase the segment count in my NEC2 model.

I suspect that I'd have to double the segments again to do better. That's not a pleasant job because there are so many wires in the model. There may be a convenient method to do it that I have not yet discovered.

I might yet do it just to satisfy my curiosity. I also did not revise the model to include the gamma match with the tapered DE since there would be little point until the antenna itself si accurately modelled with NEC5.

I then produced the antenna patterns at several frequencies across the band and found that the gain was similar but the F/B exhibited more variation. The deviation was worst at the bottom of the band. The pattern at 21.0 MHz has a F/B that is about 5 to 6 db lower than the NEC2 model. Since high F/B figures demand accurate and precise calculations of each element's current amplitude and phase it might again be a matter of increasing the segment count.

Other than my curiosity regarding accurate modelling of the gamma match on a tapered element, it would appear that what I've heard about multi-element yagis and NEC5 might indeed be true. I won't pursue this topic further for now since it's not a priority and NEC2 handles these antennas well.

Wrap-up

From what I've seen so far, I am pleased with my purchase of NEC5. It integrates easily with EZNEC and it can far more accurately model a variety of antennas that NEC2 handles poorly. Neither calculation engine is a perfect solution. Every antenna model requires a few moments of thought to decide which engine is most suitable. It is interesting to try both even when you know one of them will do poorly.

Now that I've done my initial experimentation, I will begin applying it to the construction and testing of actual antennas. The obvious first case is the reversible 40 meter Moxon. Unlike for the big 3-element yagi, it appears that I can develop an accurate model using NEC5 without resorting to building a sample element and calibrating the model with field measurements. Once I have more experience with NEC5 so that I can confidently trust the calculations it will save a lot of time and effort, and wondering whether the calibration procedure is sufficiently reliable and accurate.

If you enjoy playing with antennas you should consider purchasing a license for NEC5. It's a tool that I can see becoming indispensable for design and building antennas for my station.