The design and feeding of Driven Elements for VHF/UHF Yagi Antennas

Modeling, observations and some case studies

Graham Daubney F/G8MBI now F5VHX

CONTENTS
1. Introduction-->
2. Why focus on the DE?
3. Return Loss and VSWR
4. DE in Software Models
5. Software Issues in Modelling the DE
6. Folded Dipoles in the Real World 
7. What to do ? ....Modeling
8. Things to watch for in modeling
9. What to do?....Real World
10. Modeling 50 ohm intrinsic Yagis
11. OK, I modeled it... what now?
12. Case studies
13. Conclusion

1. Introduction

What I present here is intended to provoke thought and experimentation. Whilst I do present some practical solutions, to both the modeling and real world of antennas, my intent is to make input to those who have an interest in both designing and building their own antenna systems, share some views and opinions, and perhaps prompt some of those with deeper understanding and/or more sophisticated testing facilities, to contribute in future. What I do not intend is to infer that a huge problem exists, but simply to explore and try to move forward the constant improvement process.

Antenna modeling is now a large part of many peoples interest, it can also however be a 'lone' part of the hobby, it is possible that many people have seen similar results to these and are also wondering. I want to encourage some data/experiences exchange. If you have input or constructive criticism they are welcome at my email address.

I am truly a radio amateur and not a professionally trained engineer or mathematician. I am a kind of 'hands on guy' who likes to play with, and practically use, both modeling software and physical realizations of antennas in my Hobby. My discussion covers a number of areas and poses some questions as well as offering some solutions.

2. Why Focus on the DE ?

The DE can be viewed alternately as both the least important part of a Yagi antenna and yet the most important part of a high performance system.

Its purpose is purely that of coupling your transmitter into the antenna, no more and no less. In purely theoretical terms it has little or no effect on the pattern, gain or other parameters of the antenna. Yet, in practical terms it, can have major effect on the antennas performance. Especially in the area of realized gain and pattern. A good antenna design can be very quickly turned into a mediocre or even poor design by accumulated losses within the DE , especially those coming from any matching system and the cable attached to it, which frequently and unintentionally gets employed as a matching device. This must be viewed very seriously when trying to operate 'on the edge'. As a practical example let me take a recent real-world case.

A 144MHz EME system was being constructed , the antenna chosen was a group of four 5 lambda boom yagis. The particular antenna chosen was picked because it had a rather good gain performance and a group of 4 produced a theoretical gain of close to 21 dBd. Most people spend a lot of time, effort and finally money, to select what they consider to be the very best and will spend hours or weeks consulting papers and other people for opinions and facts, often choosing a difference of 0.2 dB just to have the 'best'.

What happened here was that on checking the DE assembly it became clear that there were 0.2dB losses in the balun and connections associated with the balun and cable connections. Plus a further 0.2dB losses because the balun was in fact not of correct length. Some further potential for additional cable losses because of poorish return loss. The presence of the balun within the yagi was not considered, but one can safely assume that there will be additional degradation because of that also.

What this means is that the overall system performance was only equivalent to a group of 4.5 lambda antennas with DE's that are running more efficiently, in fact when reduced feeder lengths are considered, then the 4.5's are even better. The group of 5 lambda antennas had, in effect, a total of over 4 metres of boom doing nothing. More importantly the 4.5 lambdas would have reduced stacking distance requirements and would also save a total of > 2 metres of stacking frame. The wind loading savings and weight savings are also significant. Why spend weeks choosing or, worse still designing and building an array to have this happen. ? The moral is that in designing or choosing high performance antennas then pay attention to details OTHER than quoted gain numbers.

To design a good yagi is not enough, you also have to get the antenna coupled properly and efficiently into the rest of the system AND you have to be able to actually build a DE in a reliable manner with respect to finding parts to fabricate it with and, for most people, with simple hand tools.

3. Return Loss and VSWR

There is a good deal of focus in the following on return loss and VSWR. That should not be misunderstood, it is worth remembering that VSWR is only an indicator.

Having a poor VSWR does not necessarily indicate that an antenna is bad, but it might give problems and loss opportunities in other system components.

Equally having a very good and rather flat 'in shack' VSWR is not a very good sign either. When one sees this situation it is often time to start looking for losses in the system. In fact the above example is a very good case where the losses in both balun system and construction actually improve VSWR whilst significantly reducing systems efficiency.

Good or bad return loss is not necessarily an indicator of goodness or badness of an antenna or systems efficiency. In most practical systems a return loss of better than 25dB AT the antenna is rather hard to realize except in perhaps one very narrow band spot, equally 16 to 20 dB can probably be considered as good enough, at 144 anyway.

The make up of the match is rather more critical, a good match in narrow bandwidth or with a specific feeder length maybe OK if you know that this is what you have, but if you do not know it, it may lead to a nasty surprise just when you think you have finished building the system.

CONTENTS
1. Introduction
2. Why focus on the DE?
3. Return Loss and VSWR
4. DE in Software Models-->
5. Software Issues in Modelling the DE
6. Folded Dipoles in the Real World 
7. What to do ? ....Modeling
8. Things to watch for in modeling
9. What to do?....Real World
10. Modeling 50 ohm intrinsic Yagis
11. OK, I modeled it... what now?
12. Case studies
13. Conclusion

4. DE in Software Models

Optimization of antennas using computer modeling has progressed a long way in the last 10 years. Over a period of four years I have attempted to play with and understand some of the many software packages now available. I always had the same problem, I could design very nice yagis with apparently good parameters but "what to do about modeling the DE" ? and, even more importantly, "can I really model the matching system" ? and why do my models not correlate well with the real world.

The solution adopted by VE7BQH during his comparison work [1] was to 'normalize' all the antennas by making no attempt to model the respective DE's and their matching systems but to insert a straight split dipole into each model and tweak the length appropriately to reduce any reactance so as not to influence the gain figure unduly. Whilst it can be argued that this process does not truly represent the real world antenna, it is the ONLY current sensible way to make antenna comparisons in modeling.

For the rest of the antenna we have a lot of modeled to real world correlation effort that has provided what appear to be reliable corrections for boom diameters and mounting methods and element length corrections for material size changes or tapering. One can therefore design in software and simply build the antenna with confidence. But we have little in the area of modeling to reality correlation in the area of Driven Elements.

5. Software Issues in Modeling the DE

It seems that modeling software has a number of issues in this area, centred round the ability to model properly the junctions of dissimilar wire sizes and end effects that are not well understood, even in the real world, contributing to a 'fuzzy' comparison between modeling and real world results.

I made many attempts to correlate the performance of well known DE systems (for example F9FT hairpin) by modeling, using the supplied DE matching options in YO [2] The results did not seem to be quite as expected. I have recently been told that in fact on HF the results given by YO are quite close. I suspect (but do not know) that results here are being affected by additional capacitance and inductance that even short connections and proximity have on VHF/UHF. Perhaps the matching menu items have been calibrated against real-world HF results.

In any event the majority of these options utilize components in the matching process, these imply either difficult to find components or difficult fabrication for high power use and, at VHF and above, are unnecessarily lossy. YO is a super program for a quick evaluation or very quickly knocking a new antenna broadly into shape. I strongly recommend however that DE's only be modeled as straight split (and resonant) dipoles, where results appear to be at least in the ballpark.

I quickly moved from YO to AO [3] , which being a full wires based modeler in x,y,z geometry allows a potentially more accurate model to be constructed. It also claims to handle junctions of dissimilar wire diameters. Once again attempts to model well known real world and well used designs gave some strange results. I conclude that modeling anything that uses junctions of dissimilar wire diameters as part of the matching process should not be attempted.

Probably the most popular design for a DE on 144 and 432 is the folded dipole it is a 'logical' choice and clearly the work of DL6WU and later DJ9BV have had large influence here. It also appeared that this should in fact be modeled more accurately than many others such as hairpin, 'T' match or gamma match, because those generally involve, necessarily so to work, dissimilar diameter junctions, whereas a folded can be modeled in uniform diameters.

The folded, by having common material size throughout , was also able to be run in NEC for Wires 2.05 [4] as well as AO. Although not the main theme of this article, I should also note that tables and models presented here have been run at 20 segments in NEC for Wires 2.05, it can be argued that higher segmentation density is necessary (or even perhaps less), but the choice of 20 was arrived at as a reasonable compromise between accuracy and processing time, from extensive prior study (by myself and others) it also appears well into the flattening part of the curve that can be plotted of accuracy/change versus segmentation density [5]. Polar plots were run quickly in AO at 10 segments.

Modeling of folded dipoles in software has proven to be an extremely thorny issue. Most of the sets of antennas available from real world designs that target 50 ohms intrinsic (in order to be fed by a folded with 4:1 impedance transformation and then 4:1 balun back to 50 ohm feeder) seem to display very poor modeling results, often with rapidly swinging R component and high J values. Thus making an apparently highly reactive match and somewhat surprisingly, given that prior wisdom was that folded dipoles gave a better bandwidth than straight dipoles, often displaying a very narrow match.

When compared with results from a number of stations actually building/using these designs, it was clear that the modeling software was showing the correct trend, but the real world results, whilst displaying reactance and some overall mismatch were not as extreme as models predicted. They were/are however still apparently wrong. A number of people have also thought up a variety of cunning 'cheats' to correct this after the event, such as bent/re-positioned D1's, re-drilled DE position holes and carefully chosen cable lengths to feed them with. On most, simply 'fiddling' with D1 does not fix it completely.

Great effort was expended by a number of keen amateur modelers in an attempt to construct new antennas in software that utilized folded dipoles, myself included, working mostly in AO and then comparison/verification in NEC for wires 2.05. I could almost never arrive at a satisfactory solution to get from 50ohms to 200ohms, I could however make solutions that looked OK from lower starting values (such as 16 transformed to 75), these lower starting value solutions clearly also worked in the real world, and on later generation antennas, as SM2CEW had used this technique in his very successful array. My results here using this method on 432 also indicate success, but only after some linear length correction of the folded dipole.

The reason was still unclear, but appeared to be that there was a major issue in most modeling packages available. More or less I concluded not to use folded dipoles in models. (or real world if possible, but more on that later)

In 1997 some new software became available called NEC WIN PRO. Lionel VE7BQH then re-visited this subject and has perhaps shone some new light on the area. The arrival of this software prompted Lionel to do further work and resulted in the model of a folded dipole with apparent good results in a NEW design. The complex impedance behaved very nicely with good bandwidth and the antenna itself was rather competitive in gain and G/T terms.

This antenna when run in both AO and NEC for wires 2.05, with a folded dipole modeled, now also shows much more expected and acceptable results, with close correlation to those achieved with a straight split modeled, so maybe the issue was/is not all in software ?

What has been done here is two things, firstly the linear length of the dipole has been adjusted carefully, the actual overall dimensions with respect to length versus depth are not that critical (this was already well known from long ago and recent modeling, at least within sensible ratios) but more critically this antenna has a much broader 50 ohm bandwidth, than those previously examined. The match is much closer to 50 ohms non reactive.

This work appears to have pointed in another direction, namely that to get a folded to work well in either modeling or real world it is necessary to have a more stable and much closer to 50 ohms impedance within the ANTENNA itself than many currently available designs show. This antenna has been built using a straight split dipole and showed good real world to model correlation like that, what has not yet been tested is a folded dipole.

It is also necessary in the real world to fabricate both the folded itself and the Balun VERY carefully, as pointed out in prior construction notes that have accompanied these type of designs. Length of tails on the balun and low inductance paths are essential.

My current conclusion is that earlier work with folded dipoles utilized them in relatively 'tame' designs with respect to bandwidth, based mostly on the work of DL6WU. Whilst these have not necessarily presented an excellent return loss, they have shown a relatively flat return loss. It seems that the majority of people are also only checking return loss or VSWR 'in shack' and not necessarily close enough to the antenna. So big attention has not been drawn to the issue.

As more and more software has become available and an increased number of people have been able to create new designs then the 'race' to find highly optimized designs is well and truly on. That has highlighted the return loss issue as it is no longer flat in large bandwidths but displays a reasonably rapid rate of change. It has become much more difficult to get an acceptable part of the curve (high return loss part) to match where you want to operate when realizing a real world design from computer constructed models. THAT is the issue, not necessarily one of bad or poor designs but how to get good correlation from models to real world.

In fact the process of "optimization" in yagi antennas is not really that at all, it is perhaps more correctly described as a process of "compromisation". No matter how good the software or the skill and knowledge of the operator (designer) then you can never get something for nothing. The pressure to constantly strive for better gain and better G/T in high performance yagis introduces the need to compromise in some other area, the easiest compromise to accept is Bandwidth. At least on your computer screen! This may be responsible for highlighting the issue but most likely not its fundamental cause.

Although one possibly cannot trust entirely the software outputs that NEC for Wires 2.05 and NEC WIN PRO arrive at, they do appear to at least show a general trend. They also show some trend of this issue as each generation of antennas is arriving.

The earlier DL6WU 10 element, first generation, and BV 4 lambda, second generation, show at least partial correlation between the folded and straights models with the transformation being close to 4:1 although the modeled resulting VSWR is not that wonderful it is relatively flat. By the time the third generation of BVO-4wl arrives then a faster swinging j component is arriving. Coming completely up to date we arrive at the fourth generation with BVO-3wl, by now we have a very broadly swinging j component. Please remember I am not claiming these modeling numbers are exactly correct, but only that they show the (inevitable?) trend.

Modeling results of some of these popular antennas are shown in Table 1. BEFORE you go and look at that I would like to clearly state that it is NOT my intention to imply or state that DL6WU or DJ9BV antennas are bad, have a major problem, or should all be taken down and thrown away immediately. It is indeed a compliment that the majority of antennas available to study, both in real world and modeling, emanate from the work of these people, both have been prolific in output and responsible for a continuous improvement in, and understanding of, Yagis. 

It is however an ongoing and emerging fact that there are some issues arising here which we should all address in order to continue the improvement process. With the recent release of the new BVO series of Yagis, Rainer had once again produced competitive designs that are excellent performers in Gain and G/T [6]. However the issue of how to centre the usable match portion of the antenna/DE to the frequency you want to use it on, remains.

You must also be careful exactly what you model and how. The folded dipoles modeled in this table have been software constructed with 'flat ends' and the linear lengths have therefore been corrected to compensate. I have attempted to model more accurately constructed folded dipoles with curved ends, the effort needed to do that in x,y,z geometry is high and results show little deviation from those obtained here. Which indeed may be indicative of part of both the modeling and correlation to real world issues.

CONTENTS
1. Introduction
2. Why focus on the DE?
3. Return Loss and VSWR
4. DE in Software Models
5. Software Issues in Modelling the DE
6. Folded Dipoles in the Real World -->
7. What to do ? ....Modeling
8. Things to watch for in modeling
9. What to do?....Real World
10. Modeling 50 ohm intrinsic Yagis
11. OK, I modeled it... what now?
12. Case studies
13. Conclusion

6. Folded Dipoles in the Real World

My first reaction to the modeling results was "OK, but it cannot really be true" . Many people have made antennas with folded dipoles and nobody seemed to raise this issue, there are perhaps thousands of people out there using them.

My real world testing ability on this issue is somewhat limited. I have extremely infrequent and short term access to a network analyser. On a regular basis I can only perform rather simple return loss tests and check for reactance by insertion of random lengths of coax.

Local tests indicated that it was in fact true that a folded was frequently pretty reactive and correlation from 'on screen' dimensions to real world dimensions is poor. Although not quantified, it appears that the result is not as extreme as models may indicate, but nevertheless it does exist.

During this time I got involved with quite a number of European EME operators looking at many aspects of their antenna system design in models. Not entirely surprisingly to me I started to receive real-world return loss data from some of these guys that indicated both reactance and a funny resonance within folded dipoles, on several different antennas.

There also seems to be a consistent factor of best return loss being low in frequency. Implying that the most popular folded designs may be on the long side.

I am also of the opinion that having the balun within the antenna is having some confusing effect on correlation attempts. Although difficult to model, my attempts at modeling pieces of cable in close proximity to the other antenna parts does indicate some serious interference potential in this area.

This is now a very confused and difficult area to explore accurately, it appears that one can summarize the issues into 5 areas.

  1. Perhaps poor modeling within software, not yet proven either way
  2. Existence of significant and un-quantified 'end' and other effects in the realization of a folded dipole.
  3. Effects of balun construction.
  4. Effect of the physical presence of the balun within close proximity to critical and sensitive parts of the antenna.
  5. More highly 'optimised' (effectively bandwidth compromised) designs highlighting items 1 to 4.

7. What to do ? ....Modeling

For ALL forms of matches such as 'T', Gamma, Hairpin etc. the modeling results are unreliable due to issues such as junctions between dissimilar wires, and may also be distorted further by the anomalies seen in folded dipoles.

For folded dipole results to date, it seems that results point the right trend, but again are not really close enough to model something and then build it with any great confidence.

For straight split dipoles the results from the better modeling packages look OK, and when compared with real-world results they seem rather close.

Given that a straight split dipole also has the easiest construction method, the least loss and the simplest form of adjustment (just cut it or slide it) then I recommend to use straight split simple dipoles in all modeling.

If you adopt that as a modeling philosophy then it also seems logical to adopt it in the real world, if possible.

CONTENTS
1. Introduction
2. Why focus on the DE?
3. Return Loss and VSWR
4. DE in Software Models
5. Software Issues in Modelling the DE
6. Folded Dipoles in the Real World 
7. What to do ? ....Modeling
8. Things to watch for in modeling-->
9. What to do?....Real World
10. Modeling 50 ohm intrinsic Yagis
11. OK, I modeled it... what now?
12. Case studies
13. Conclusion

8. Things to Watch For in Modeling

Over time I have developed a list in my head of indications that may point to trouble coming with bandwidth reduction and highly gain optimized antennas. There are a number of characteristics/trends that highly gain pushed designs display, if you start to see these then it is time to look closely at your likelihood of getting a broad and reliable modeling to real world match. None of these mean the antenna is necessarily a bad or poor design, just that you may be getting close to the edge and that you are going to have to pay closer attention to matching, element corrections and construction accuracies. 

  • The gain peak of the antenna moves down, most often the gain peak will be slightly higher than the centre design/match frequency, if you start to see convergence of the gain peak and the centre design frequency then look out. I have personally deliberately exploited this in a number of my own, as yet unpublished, designs as this enables a very competitive g/t number to be generated.
  • The Front/Rear ratio of the antenna starts to degrade VERY rapidly on the high side of the design frequency. It generally will consist of one large tail, rather than a number of lobes with nulls between.
  • The impedance and reactance (and most often = return loss or VSWR) swings rapidly on one side of the design frequency, the high side, but maintains at least some reasonable bandwidth on the other.
  • To create a combination of reasonable match and get back some of the Front/Rear ratio you have to allow the last director to "go where it wants" during optimization, this frequently leads to the last director being longer than the prior one and often longer than several prior ones. This in particular is very indicative.
  • In order to maintain any match at all, the length of the driven element, even a straight simple split dipole, starts to deviate from what one would normally call a common length, mostly significantly shorter. As a general rule of thumb if you start to see DE's outside the range of 980mm to 990mm on 144MHz, this is why.
  • You may also start to see VSWR changes in poor weather.

Some of these will show up in many designs, but a combination of perhaps two or more of these is a sure sign that the design is being 'pushed'.

As good examples, the BVO-3wl shows ALL of these symptoms as do the SM5BSZ designs I have looked at. That is because they are very competitive designs, one could even say "leading published" designs.

9. What to do?....Real World

I make no claim for originality here, only offer an opinion and strategy that works.

Apart from the difficulties of modeling, all forms of DE matching employing additional tubes, movable parts, or components, have either difficult to fabricate parts and/or potential for loss.

The traditional and well accepted folded driven has both advantages and disadvantages. It is undoubtedly the best theoretical choice: it offers, with the 4:1 balun, a good transition from balanced to unbalanced, it puts the centre line of the driven element in close plane match with the parasitic elements, and it is not too hard to fabricate.

On the negative side it is very easy to accumulate losses in the many soldered and exposed junctions, tails and the balun itself. What to do with the coax balun? It can and does cause interference to the antenna match and pattern by being in close proximity to other elements; this may in fact account for at least part of the real-world results in the more recent highly gain pushed designs. It is difficult and maybe not possible to model accurately and then simply produce to the dimensions 'apparently' required. Finally, the folded dipole is completely impossible to adjust - hence all the bent D1's and clipped D1's you can find in the world.

The straight split dipole also has some negatives, the potential for current on the braid affecting the antenna pattern and match. The 'offset' in geometry necessary by having the DE either above or below the boom could cause both gain drop and pattern asymmetry in the H plane.

On the positive side it is exceptionally easy to build, it is 'lossless' , very easy to adjust in situ by both trimming and small repositioning, and correlates rather well with modeling results so can be built with confidence.

Although not the focus of this piece it is also worth mentioning that in crossed yagis the coupling and interactions between straight splits appears to be less than that of folded dipoles - the coax balun of the folded is extremely difficult to locate in a crossed antenna and the problem of passing coax around or through a folded diople is also not easy to address [7].

CONTENTS
1. Introduction
2. Why focus on the DE?
3. Return Loss and VSWR
4. DE in Software Models
5. Software Issues in Modelling the DE
6. Folded Dipoles in the Real World 
7. What to do ? ....Modeling
8. Things to watch for in modeling
9. What to do?....Real World
10. Modeling 50 ohm intrinsic Yagis-->
11. OK, I modeled it... what now?
12. Case studies
13. Conclusion

10. Modeling 50 ohm Intrinsic Yagis

Both the folded and straight split DE's require that the 'natural' (or intrinsic) impedance of the yagi structure be 50 ohms . Whatever final real-world solution you choose, folded or straight, the basic modeling requirement is the same.

No matter how you go about optimizing your model whether by using optimizers (such as YO and AO) or by trial and error, or by manually noting and following a trend, the result will always be that highly gain optimized antennas will trend to require a low feedpoint impedance. Do not allow this to fall too low, even if you plan to use another matching system - allowing the impedance to fall too low will lead to higher losses.

You can create close to 50 ohm antennas in models by following the process outlined below. I have assumed, and therefore not described in detail here, good modeling practice. For example, reasonable segmentation, a sensible spread of optimization frequencies, etc. etc.

Optimize your antenna exactly as you wish, whilst maintaining a 'reasonable' feed impedance requirement. When you think you are finished, insert a new D1 in relatively close proximity to the DE (thus the old D1 becomes D2, old D2 becomes D3 etc.) fix/freeze positions and lengths of all other directors and focus on 'forcing' the impedance up to 50 ohms. Initially change only D1, both length and position.

It seems that if you have optimized pushed the antenna then this procedure will probably not be enough to get either a 50 ohm match or a match with any decent bandwidth and a low j component. To achieve that, it will be necessary to move the DE back towards the reflector, and to move D1 in tandem with it. It may also be necessary to have a slight tweak of the reflector. Changing the length of the DE should be avoided if at all possible.

Some minor positional or length adjustment of D2 and D3 will also help to restore any minor pattern degradations that may have taken place during the matching process.

Once you practice this process it can come rather quickly and you will also be able to visually identify a likely 50 ohm design in future. The rear of the boom holding REF, DE and D1 gets compressed and of course D1 is rather close to DE.

In a really heavily gain pushed design you may also find it necessary to allow the last director to lengthen a bit (remember the points made above). This sometimes allows a final matching tweak, and may get back any lost front/rear ratio.

This whole process should in general have little effect on any prior pattern you had achieved. It will however lower the gain a little, as the adding of the extra director increases the internal losses. It almost certainly will not lower the effective real-world gain, because if left with a low impedance feedpoint then you will be forced to use a lossy matching device anyway, which will likely cost you more. By adopting this strategy of 50 ohms intrinsic and using D1 to achieve it, you will probably save much loss and a lot of the real-world constructional difficulty that is necessary with other matching systems.

11. OK, I Modeled It... What Now?

Now you have a 50 ohm antenna.

I would strongly recommend that you now do some build sensitivity testing before taking hacksaw to aluminium. In your model try changing the length of a number of directors by 1 or 2mm randomly, maybe even in a one sided manner. Also try moving one, or several, elements by 3 or 4mm back and forth. This will simulate any possible construction errors and deviations as you translate your model to a physical antenna. If you see sudden and severely degraded parameters then you should really stop and think if this antenna is truly realizable in the physical world.

Check especially the reflector and early directors near the DE, and also the last director. If very small deviations cause sudden changes in the complex impedance or gain then you may be heading for trouble.

If you plan to use a straight split DE then you have not yet finished. Now you have to take a look at the antenna performance with the DE offset from the plane of the parasitic elements. Adjust your model in steps to check this by moving the DE up, or down, out of perfect plane alignment with the other elements. If you have already decided the boom size and driven element mounting technique then you know the exact amount to move.

Each antenna will perform slightly differently in this respect. The first thing to check is gain drop, an example of this can be seen in Table 2. The antennas modeled here are the BVO-5wl, BVO-3wl and BVO-70 13wl. 

On 144MHz the gain drop amounts to little more than 'modeling noise'. The table shows a result on 432MHz that is probably good enough for tropo use - but can you afford to lose this much on EME? You will have to decide that, based on your construction capabilities and other factors; for example, if building crossed yagis then maybe getting the boom 'clutter' of baluns and folded dipoles out of the way is the best way. In any case a gain drop of 0.15 to 0.2 dB by using a simple DE is likely to only be equal to, or even less, than those losses in a more complex one.

The next thing to check is asymmetry caused by the offset. To do this you MUST switch symmetry off in plotting (for example in AO options menu, or by 'runtime switches' in other packages). The H plane plot of a BVO-5wl with DE in perfect alignment can be seen in Plot 1. and the comparative plot of the same antenna with a DE centre line 40 mm below other elements in Plot 2.

Plot 1

Plot 2

The level of sidelobe change can be seen in these plots. It can also be seen that this is relatively small, so it will not degrade much. But in fact the asymmetry could be used to good advantage if the DE is located on the underside of the boom, as the first few (major) ground lobes actually reduce leading to a possible G/T improvement. You will also have to remember that symmetry may be returned in groups of yagis during the stacking process, so consider and utilize that also.

The results on 432MHz with 30mm offset are shown in Plots 3 & 4. The offset could possibly be reduced to 25mm. To get less than that, even with small size booms, is surprisingly difficult.

Plot 3

Plot 4

I have tried physical and modeling experiments with 'saddle' style DEs where alignment is improved and also 'gull wing' shaped DEs for the same purpose. Results are mixed with advantages and disadvantages both ways. In the end I think for EME the folded dipole remains the optimum choice on 432MHz where our required bandwidth in percentage terms is lower and major side lobe disruption could cause more significant G/T issues. It would be worthwhile to construct folded dipoles with movable ends or at least a prototype one with some kind of trombone adjustment until you are sure that you have the dimensions required correct. By the way, it seems that the correlation from modeling to the real-world is even worse on 432!

Although not as fully explored here as the 144MHz results, it is now quite clear that highly optimized and later generation 432MHz yagis also have potential for matching issues, and likely more potential for balun interference within the antenna.

I recommend making your own versions of these plots and tables for the chosen or designed antenna. When making small compromise choices it is best to understand them all thoroughly first.

After construction a straight split dipole is easy to adjust with minor length corrections or a small movement either back or forward on the boom. When doing this please check with three or even more short but random lengths of cable inserted to check that the match is consistent. As more coax is added to the feeder the return loss or VSWR should only ever improve, not worsen (losses are being added). In most practical systems this will not be precisely the case, small changes are acceptable, but wild swings indicate trouble.

During final adjustment be wary of sliding the DE too close to D1, because this will affect the yagi pattern and close proximity software modeling results of this situation are unreliable and may not show the whole picture. Wherever you end up with the DE in the physical world, it is a good idea to go back and model the solution you arrive at to see what you may have done to the pattern and other parameters.

If you really are nervous about having current on braid and subsequent pointing squint without a balun, then you could add a simple sleeve balun to comfort yourself. However with the straight split DE not boom grounded, the coax taped off frequently and tightly to the boom, those of us without baluns have not seen any real trouble.

Another suggestion could be the use of ferrite rings threaded onto the feeder, I have not experimented with this and it needs some thought and calculation.

CONTENTS
1. Introduction
2. Why focus on the DE?
3. Return Loss and VSWR
4. DE in Software Models
5. Software Issues in Modelling the DE
6. Folded Dipoles in the Real World 
7. What to do ? ....Modeling
8. Things to watch for in modeling
9. What to do?....Real World
10. Modeling 50 ohm intrinsic Yagis
11. OK, I modeled it... what now?
12. Case studies-->
13. Conclusion

12. Case Studies

I am indebted to those mentioned here for providing much real world data and help. I have tried to document simply and accurately the experiences and hope they will assist others. In most cases they did the field work, I did the modeling. 9H1PA has done much modeling of his own also.

BVO-5wl

G4SWX: with a folded dipole of 980mm and 66mm height the return loss from 144.0 to 144.3MHz was acceptable but the best return loss was at 143.5MHz. Moving the balun within the antenna gave variable results. To help with my modeling and real world correlation, John did a lot of testing (often in the snow!). Folded dipoles of 930, 940, 960mm were also tried, but none correlated with the models. They all showed reactance and rather rapid swing. 

A straight split dipole of length 980mm overall, with around 10mm centre gap in 9.5mm tubing produced the best result. In the end the simple split dipole was adopted, with no balun, and tested extensively on tropo and EME. A centre frequency (within the DX portion of 2 metres) return loss of >25dB was achieved. A hosing down test to simulate poor weather moved the best match frequency LF, but maintained very acceptable return loss in-band.

G4YTL tested some T matches on the same design with poor results. He has adopted the split dipole solution as G4SWX.

The team at 9A1CCY have also built this antenna with the straight split dipole solution and the return loss result from the prototype (they plan 4) is also good. By the way they have implemented a crossed version and will use simple polarity switching. 9A1CCY has also incorporated the same straight split dipole into a 10 element DL6WU array with good results.

Thanks to Juergen, OZ1HNE for supplying various measurements, data and input on his crossed version of this antenna also.

BVO-4wl

9H1PA built his with the standard 990mm by 66mm folded dipole and 4:1 balun. Moving the balun around in the antenna makes quite some difference. Again best return loss was in the 143.5 to 143.7MHz area. Attempts to fold the balun up and fix it on top of the feed box out of the way led to even worse results. VSWR was 1.3 to 1.4 in the 144.1 to 144.5MHz range, and the random coax insertion test also showed big changes. 

Phil did much modeling and real-world moving around, and finally he changed the DE from 365 to 395mm and D3 from 1355 to 1340mm. Now he has VSWR of 1.15 and only small changes over the DX portion of the band, or with random coax insertion tests.

BVO-3wl

9H1PA: the conventional folded of 990mm length was rather unhappy, with best VSWR at 143.70MHz; by 144.30MHz it roe to 1.4. A folded dipole of length 968mm was tried which improved the VSWR in the 144.10 to 144.30MHz portion. However the final solution was to revert to the conventional folded dipole of 990mm overall, and to move the DE position to 323mm (from 391) and D1 to 637mm (from 669). This has had the effect of moving best return loss/VSWR into the range 144.05 to 144.30MHz, at 1.1:1. 

Modeling of these changes shows no significant change of either gain or F/R. Neither does it show a very good match either(!) but it clearly works. I would suggest adopting these dimensions on a provisional basis before drilling final holes and testing with various cable lengths. This antenna is a real winner (small boom but good gain, nice G/T and little gain drop with non insulated stacking frame) so I expect many people to be using small groups of this as an entry level to EME. Further feedback from anyone trying these changes would be very much appreciated.

F9HS: on my recommendation Claude has gone for a straight split driven element with no balun. He has a good return loss from 144.00 to 144.50MHz. The final tweaked length was 953mm overall with around 10mm centre gap in 10mm diameter material. This seems rather short but has shown good results on EME, with 13 stations worked in ARRL first leg on a single, polarity switchable, 3 lambda yagi! Return loss from 144.00 to 144.50MHz ranges from worst of 17dB to best of 27dB. More importantly the return loss moves very little in wet weather and cable length deviation by 3/4 lambda shows only 15% increase in returned power.

Again, modeling of this solution does not show fully the good results that are possible in the real world. By the way, this antenna shows excellent results in crossed format, even with aluminium poles running in the elements (although you have to position the poles correctly). The DE for vertical set of elements was adjusted in situ and best results were yielded with a slightly longer DE; this is due to some matching interference from the aluminium support tube, so you do need to adjust on test.

CONTENTS
1. Introduction
2. Why focus on the DE?
3. Return Loss and VSWR
4. DE in Software Models
5. Software Issues in Modelling the DE
6. Folded Dipoles in the Real World 
7. What to do ? ....Modeling
8. Things to watch for in modeling
9. What to do?....Real World
10. Modeling 50 ohm intrinsic Yagis
11. OK, I modeled it... what now?
12. Case studies
13. Conclusion-->

13. Conclusion

Food for thought and a wish list.

It remains unclear, to me anyway, whether there is a genuine problem with modeling folded dipoles in software. Or does the issue lie in the 50 ohm bandwidth of the antenna designs - and then, in turn, in getting a folded dipole to go in the right spot without being able to model it or having any reliable end-effect data?

If the issue is simply one of needing correction factors for end effects in folded dipoles, then why does it not seem possible to create at least a modeled folded that is happy in these antennas in software (even if the dimensions would need correction for actual construction)?

Without some very strict controlled measurements being made I believe we cannot discount the construction of the balun - or indeed its very presence.

I am not concluding that folded dipoles should be abandoned, but with unsure and unsafe modeling results and unquantified real-world effects I do conclude that in highly optimized designs with swinging R and jX components in the complex impedance, getting an acceptable match with a simple dipole will be easier. It will also not cause any significant detriment to the yagi performance and indeed will surely contain less loss and more reliability.

I wish to see an adjustable folded design that can be easily built/adjusted with simple hand tools and with the balun made in rigid cable at no more than boom width dimensions to keep it out of the antenna.

 

Notes & References

1. G/T simulations by VE7BQH

2. YO is Yagi Optimiser written/supplied by Brian Beezley, K6STI . Commercially available, very useful as a fast input tool as well as an optimiser as it will create x,y,z geometry files for onward use in AO and NEC/Wires.

3. AO is a MININEC- based x,y,z modeler from K6STI.

4. NEC/Wires 2.05 is an x,y,z model evaluator from K6STI.

5. SM5BSZ web site with much very useful information on antenna designing and other EME related subjects.

6. Full details in DUBUS Technik V

7. For more on crossed yagis thoughts and other antenna subjects, such as rotatable polarity arrays for EME, stacking information and principles... see this section of the site


Updated 30 June 2003
Page © 1998-2000 Graham Daubney and
IFWtech Limited