[NLRS] Yagi Design (Was: Club Cheap Yagi Build Day Advice Needed)
James Duffey
jamesduffey at comcast.net
Sun Nov 27 18:34:47 EST 2016
OK Jerry - Thanks for the reply. Glad to see we are now in agreement. I think that MMana-GAL kind of defaults to putting the source in element one; at least it was hard for me to find a way to do it otherwise and to get a file to run.
As a rover, I am very interested in the stacking heights for antennas and pretty much stack things close. It is hard to keep them at good spacing, though, but it all seems to work OK. I try to avoid having antennas with odd harmonic relationships next to each other as there are considerably more currents induced in the odd harmonic relationships than in the even ones. So, starting at 13 ft, my stack looks like: 6M Moxon, 222MHz Yagi, 144MHz Yagi, 902MHz Yagi, 432 MHz Yagi, 1296MHz Yagi. The spacing between 432MHz and 1206 MHz is less than I would like, but it is possible to get better separation isolation wise between 432 and 1296 than it is between 144 and 432. And 432MHz is kind of a crippled band in NM anyway as we are limited to 50W RMS transmitter output power. No, antenna gain is not considered. Yes, that is ridiculous. Of all the considerations, I think that having the 6M antenna as the top one on the stack is critical. I think it is also important to keep each antenna more than 1/2 wavelength above any car metal, and that stacking achieves that.
CSVHF is in Albuquerque next year and I know at least one local ham who has a spare 2M9SSB that I will try to talk him into bringing to the testing. It is a big antenna for that kind of test range though.
I hope you can get to Albuquerque for CSVHF next year. It should be a good time. - Duffey
On Nov 27, 2016, at 2:21 PM, Dr. Gerald N. Johnson <geraldj at netins.net> wrote:
>
>
> On 11/27/2016 2:27 PM, James Duffey wrote:
>> OK Jerry - I finally got some time to spend with MMana-GAL and loaded
>> up your M2 2M9SSB *.maa file. I am used to the elements running left
>> to right from 1 to 9 as the convention for specifying elements, so
>> that is what threw me at first. But you were correct, the source is
>> in the correct element in your file.
>
> I often put the driven element as wire 1 so the segment for the feed is more constant. My model data did show segment 25 as the point, but I changed it to W1C.
>>
>> I ran it and I got the same results you did. So we are on the same
>> page as to modeling with MMana-GAL. I then began to investigate why
>> your model gives results that are different from other modeling of
>> the same antenna as well as the M2 specifications and what one would
>> expect from a well designed antenna of that length. It looks to me
>> like the element radius you specified, 0.8 mm or 1/32 inch, is too
>> small to accurately represent the actual element diameter, which is
>> specified as 3/16” by M2. So, I increased the element radius to
>> 2.38mm, or 3/32”, which is half the 3/16” actual element diameter as
>> specified by M2.
>
> 0.8mm as displayed on the geometry page is the default, think 14 gauge wire for an HF antenna. I entered .09375*i and it displayed 0.002381 mm by doing the math as for meter as the unit which is what it displayed in your file.
>
> The smaller diameter increased resistive losses a bit and the resonant frequency a couple MHz moving the peak gain up to about 149 instead of 145 and that's where most of the gain difference went.
>
> I don't remember that a M2 2mSSB9 has been measured at CSVHF but I will have to take a hard look one of these days.
>>
>> I ran that model and that increased the gain to 11.89dBD, which is
>> about 1.1 dB higher than what your original file calculated and close
>> to the M2 gain specification of 14.1dBi (11.96dBD) as well as close
>> to the VE7BQH calculated value of 12.01dBD. The change in element
>> diameter of three times is pretty significant in moving the gain peak
>> around at those frequencies without any corresponding adjustments to
>> element length. I ran the same dimensions in your file with the
>> corrected radius with YagiCad and modeled a 12 dBD gain, which is
>> also in line with what M2 and VE7BQH list as actual gain. I am not
>> sure if there are additional tweaks or corrections to be found in
>> your file or not, but I suspect that the differences between the
>> calculated gain values with the correct element radius are not
>> significant and may be due to the differences in the MININEC and
>> NEC2 computing engines. I am not sure why this should be though, as
>> the problems that MININEC solve better than the NEC2 engine, the
>> closely spaced elements, are not really present in this antenna.
>> Still, 0.07dB difference is probably not really enough to worry
>> about. I did verify that the boom correction had been subtracted out
>> properly. The model I ran in YagiCad was in free space without the
>> boom, just as your file models. That is a good way tot model the
>> antenna.
>
> I want to see the face of any antenna range operator who claimes to measure gain to .05 dB repeatability. I don't believe that is practical. Easy enough to compute with a VNA but move the antenna one degree or one inch and it will change that much from odd not fully suppressed range reflections.
>>
>> The exercise did satisfy my curiosity as to why you got similar gain
>> results in modeling the 10 ft K1FO and the M2 2M9SSB at 14.5 ft in
>> your paper on stacking - the M2SSB antenna is modeled with the wrong
>> element diameter.
>
> I still like the 10 ft K1FO and plan to put up a pair of them for 2m and 3 for 432 some eon. The 432 versioin that i took to CSVHF twice measured more gain than computed. Stacking will improve the gain without adding to the azimuth directivity and so make for easier pointing. I have a FO-22 on 432 on the roof tripod and aiming it is a real pain, especially when the rotor direction indicator isn't working. So I aim by knowing what direction it starts out at and then watching the sweep second hand on a wall clock, then I go out on the back deck and check for what I see and often correct. I need a lot more height too.
>
> The point of the paper still is fine, that one doesn't need to stack 432 and 2m beams half a boom length apart that there isn't a gain loss until the booms are a wavelength apart which is much closer than the half a boom length intended to keep the capture areas from overlapping. The close spacing is often used by rovers quite successfully even before this paper. Half a boom spacing is about right for maximum stacking gain of identical yagis for the same band.
>
> I was surprised that the effects on the high band pattern from currents in the low band elements was more in the elevation pattern than the azimuth pattern. I expected to see +/- 45 degree azimuth lobes typical of a 3/2 wave dipole.
>
> Well now we have a topic for another CSVHF paper, correcting my 2011 paper on the M2 modeling. No one else has noticed it. Can be a short paper, won't have to redo the stacking parts.
>
> 73, Jerry, K0CQ
>>
>> Here is the *.mma file I used for the MMana-GAL calculations:
>>
>> 2m9ssb * 144.2 ***Wires*** 9 0.25083, -0.49428,
>> 0.0, 0.25083, 0.49428, 0.0,
>> 0.00238, 50 0.0127, -0.51412, 0.0,
>> 0.0127, 0.51412, 0.0, 0.00238,
>> 50 0.57626, -0.46443, 0.0,
>> 0.57626, 0.46443, 0.0, 0.00238,
>> 50 0.87313, -0.46163, 0.0,
>> 0.87313, 0.46163, 0.0, 0.00238,
>> 50 1.41764, -0.46333, 0.0,
>> 1.41764, 0.46333, 0.0, 0.00238,
>> 50 2.12407, -0.45698, 0.0,
>> 2.12407, 0.45698, 0.0, 0.00238,
>> 50 2.90195, -0.45063, 0.0,
>> 2.90195, 0.45063, 0.0, 0.00238,
>> 50 3.76238, -0.44762, 0.0,
>> 3.76375, 0.44762, 0.0, 0.00238,
>> 50 4.4323, -0.4508, 0.0, 4.4323, 0.4508,
>> 0.0, 0.00238, 50 ***Source*** 1, 0
>> w1c, 0.0, 1.0 ***Load*** 0, 1
>> ***Segmentation*** 800, 80, 2.0, 2
>> ***G/H/M/R/AzEl/X*** 0, 0.0, 3, 21.0,
>> 120, 60, 0.0 ###Comment### Element radius increased
>> from 0.8mm to 2.38mm to reflect actual antenna Source changed from 25
>> to w1c as MMana-GAL would not take segment 25 Mod by James R. Duffey
>> KK6MC 11/27/2016 1039 Mod by Gerald N. Johnson, K0CQ 1/23/2011
>> 6:19:23 PM Created by Gerald N. Johnson, K0CQ 1/23/2011 5:08:08 PM
>>
>> In addition to the radius change, I also had to change the location
>> designation of the source in your file from 25 to W1C. My version of
>> MMana-GAL did not recognize 25 as a valid location for the source and
>> wouldn’t run the file. I had to read the MMana-GAL help file to
>> figure that one out. It has been a long time since I ran MMana-GAL,
>> so I am not sure if the way to designate the source has changed or
>> not. I think that the rest of the file is identical to yours. Run it
>> and see what you get. - Duffey KK6MC
>>
>>
>>
>> On Nov 21, 2016, at 3:58 PM, Dr. Gerald N.
>> Johnson<geraldj at netins.net> wrote:
>>
>>> Load the file and do a view. Element 1 is the driven element,
>>> element 2 is the reflector. Its not wrong.
>>>
>>> 73, Jerry, K0CQ
>>>
>>> On 11/21/2016 2:57 PM, James Duffey wrote:
>>>> Jerry - I haven’t had time to run your M2 2M9SSB MMana-GAL file
>>>> yet as I was busy with Sweepstakes all weekend. After a quick
>>>> glance at your *.maa file though, I think I see the reason why
>>>> your modeled gain is lower than what M2, VE7BQH, and myself
>>>> calculate. It looks to me like you placed the source in element
>>>> one, the reflector, rather than in element 2, the driven element.
>>>> Try moving it to the driven element and let us know how it
>>>> models. Should be quick and easy to do. I will also do that when
>>>> I get a chance.
>>>>
>>>> I knew Warren well and we carried on a long discussion on source
>>>> conjugate matching, both through snail mail, and at the
>>>> Iowa/Midwest Division conventions. I was saddened when I got the
>>>> letter with his obituary and a note from his family. We lost a
>>>> great technical ham when he died. Warren taught me the “one
>>>> degree rule”, which he said was passed down to him by Art Collins
>>>> himself. - Duffey KK6MC
>>>>
>>>>
>>>>
>>>>
>>>> On Nov 19, 2016, at 3:20 PM, Dr. Gerald N.
>>>> Johnson<geraldj at netins.net> wrote:
>>>>
>>>>>
>>>>>
>>>>> On 11/19/2016 10:24 AM, James Duffey wrote:
>>>>>> Jerry - We have strayed fairly far from the original topic,
>>>>>> so I have changed the subject accordingly. That way people
>>>>>> who are only interested in the WA5VJB antenna build can skip
>>>>>> over it if not interested.
>>>>>>
>>>>>> I use MMana-Gal as well as YagiCad and 4NEC2 in my modeling
>>>>>> of the WA5VJB antenna. For serious work, not that Bill’s
>>>>>> WA5VJB project isn’t serious, I use a couple of different
>>>>>> tools to confirm modeled results, generating new input files
>>>>>> as I go. Testing without a good antenna range makes it
>>>>>> difficult to confirm results, but if the SWR or directivity
>>>>>> is way off that is usually a clue that there is a problem in
>>>>>> the design or modeling, and if one can wait a year for the
>>>>>> CSVHF conference to roll around, the antenna range there is
>>>>>> useful. A number of years ago when I had access to a licensed
>>>>>> version of CST, a professional (read expensive) 3D
>>>>>> electromagnetic simulator and NEC4 (controlled access only),
>>>>>> I modeled the 3 element WA5VJB in those as well.
>>>>>> Interestingly enough, the electrical properties of the
>>>>>> antenna varied little with which model one used. I was aware
>>>>>> of the closely spaced wires issue from W7EL’s work as well as
>>>>>> conversations with Cebik at Dayton while he was still alive,
>>>>>> so this good agreement was a bit of a surprise to me.
>>>>>> However, when I modeled a dipole, a folded dipole, and Kent’s
>>>>>> J shaped driven element (It is really more accurately
>>>>>> described as a half a folded dipole) with the different
>>>>>> models, they all showed the same gain and pattern. Only the
>>>>>> input impedance and VSWR bandwidth were different, as one
>>>>>> would expect. This makes sense as a dipole is a dipole,
>>>>>> regardless of the physical form it takes. That gave me
>>>>>> confidence that the WA5VJB antenna could be modeled well with
>>>>>> any of those tools. This all makes sense in the context of
>>>>>> how the Yagi works; the driven element serves to launch the
>>>>>> electromagnetic wave in the antenna, and as long as the
>>>>>> driven element geometry is efficient and balanced, then the
>>>>>> particulars of the driven element are really only important
>>>>>> as to matching to the feed line. With this in mind, the
>>>>>> WA5VJB antennas can be modeled with simple dipoles or folded
>>>>>> dipoles as the driven elements as far as things like element
>>>>>> diameter correction, boom correction, and frequency scaling
>>>>>> go. One just needs to account for the driven element
>>>>>> impedance when looking at things like VSWR and matching or
>>>>>> model the VJB driven element exact.
>>>>>
>>>>> I model a yagi like the M2 or K1FO by taking the elements and
>>>>> running the G3SEK basic dos program (element.bas) to remove the
>>>>> boom correction and I model without a boom. I haven't found a
>>>>> method of moments program that handles intersections of
>>>>> different diameter wires or intersections not at 90.0000
>>>>> degrees well. The fundamentals are that the currents are
>>>>> presumed to be at zero diameter at the center of the wire and
>>>>> that the coupling is from the center of the segments nearest
>>>>> the intersection. When at 90 degrees the coupling is figured as
>>>>> zero because the fields are at right angles. Unfortunately the
>>>>> current distribution is not anywhere near like that
>>>>> presumption. Especially in close spaced conductors.
>>>>>
>>>>> I invented an antenna I called a Cat's Cradle involving angled
>>>>> 1.5 wave radiators gathered like a gang of small rhombics.
>>>>> Modeling in EZNEC the gain and patterns varied drastically with
>>>>> slight changes of inputs, as much as over 3 dB change in gain
>>>>> for a tiny change. I whipped one together a day before CSVHF
>>>>> and took it and it displayed the gain of the lowest modeling
>>>>> results, not the peak. It also as a single array worked better
>>>>> with the boom vertical than pointed at the 1296 signal source,
>>>>> much like a diamond array. In my modeling stacking 4 cleaned up
>>>>> the strong elevation lobes. But I only built one.
>>>>>
>>>>> That experience and many claims for great front to side ratios
>>>>> enhanced by flat loop driven elements or angled driven elements
>>>>> that were guaranteed to couple current to the boom caused me to
>>>>> publish
>>>>> http://www.geraldj.networkiowa.com/papers/2013/FinePoints.pdf
>>>>> in both CSVHF and MUD 2013.
>>>>>>
>>>>>> I have never heard the pejorative Crumcraft used when
>>>>>> referring to CushCraft before. I have heard of them referred
>>>>>> to as CushCrap though. They were more crippled by their
>>>>>> marketing department than the performance of their antennas,
>>>>>> although now that MFJ owns them, that may have changed. The
>>>>>> CushCraft antennas based on NBS Tech Notes 688 perform as one
>>>>>> would expect from the NBS data, which is not too surprising
>>>>>> as those antennas were designed to be reproducible if one
>>>>>> followed the directions in the Tech Note for designing and
>>>>>> building the antennas.
>>>>>
>>>>> That is my invention. Their NBS copies did work as the
>>>>> originals but they claimed several more dB claim, like adding
>>>>> ground reflection gain or link gain with the same antennas on
>>>>> both ends of a link. Maybe their reference was a rubber duck
>>>>> antenna inside a fruit juice can. Their outrageous gain claims
>>>>> caused a permanent QST policy that antenna gain claims are not
>>>>> allowed in QST advertisements.
>>>>>
>>>>> Their 11 element 2m and 432 yagis with all but the front
>>>>> director the same length did have some directivity but their
>>>>> poor quality (too wide spaced) gamma match feed took several DB
>>>>> out of the gain. Most yagi gurus don't like the gamma match
>>>>> feed accusing it of killing the symmetry of the yagi pattern
>>>>> and that could probably be applied to the J feed of Kent's
>>>>> antennas. But on a short yagi the pattern is going to be wide
>>>>> enough that any asymmetry is going to be virtually impossible
>>>>> to detect.
>>>>>
>>>>> I took a 432 crumcraft with my own very close spaced gamma
>>>>> match to a CSVHF about 1967 and it didn't give the claimed gain
>>>>> but was more than 3 dB better than the original measured at
>>>>> CSVHF. The gang laughed out loud when they recognized it before
>>>>> it was measured.
>>>>>>
>>>>>> The K1FO designs were optimized not only for gain, but also
>>>>>> for low side lobes and reasonably wide bandwidth, and he was
>>>>>> the first to effectively optimize all three of those
>>>>>> parameters. They were a big improvement over the NBS designs
>>>>>> in that they emphasized important antenna properties in
>>>>>> addition to gain.
>>>>>
>>>>> The were based on DL6WU designs and rarely gained as much as a
>>>>> whole dB in forward gain.
>>>>>>
>>>>>> The NBS antennas lived on through the 90s not as marketable
>>>>>> antennas, but often as starting points for the genetic
>>>>>> algorithms that were becoming available then to optimize
>>>>>> antenna performance. I think the generation of CushCraft
>>>>>> antennas after the NBS clones, such as the 13B2, were done
>>>>>> this way, and they were and are good antennas for the boom
>>>>>> length. There are better antennas available now, but the
>>>>>> 13B2 speaks to the advance in computing power applied to
>>>>>> antenna design from the late 70s to the early 90s.
>>>>>>
>>>>>> YagiCad has no relation to Yagi Optimizer, it is a
>>>>>> completely different program; essentially a Yagi specific
>>>>>> front end interface to a NEC computational engine. Elements
>>>>>> are entered in a spread sheet by position on the boom and
>>>>>> full element lengths are used. It is much improved over
>>>>>> generating a bare NEC card deck. You should try it out if you
>>>>>> have not. It is a very useful piece of software. It has its
>>>>>> limitations, as do most modeling tools, but it gets one
>>>>>> pretty far up the design curve without a lot of problems. It
>>>>>> has an optimizer and does scaling, both for frequency and
>>>>>> element diameter and you can chose several techniques for
>>>>>> doing this. It has a genetic optimizer and a calculator for
>>>>>> various types of matches. If you want to do something outside
>>>>>> of its limitations, it generates .NEC input files which can
>>>>>> be used in any NEC or miniNEC based tool; barebones NEC,
>>>>>> MMana-Gal, EZNec, CocoaNEC, or 4NEC2; and it can generate
>>>>>> 4NEC2 input files so you can use the extended features of
>>>>>> 4NEC2 in your modeling. As with all modeling tools one needs
>>>>>> to study the results of the model carefully and make sure
>>>>>> that they make sense from a physics and physical point. But
>>>>>> that is the case in any modeling. Try it, I think you will
>>>>>> find it useful.
>>>>>>
>>>>>> I think that there is merit to optimizing G/T, even for
>>>>>> terrestrial use, although I realize that is not a universally
>>>>>> held belief. It can make a big difference if your neighbor’s
>>>>>> plasma TV is on a -10dB side lobe or on a -30dB side lobe.
>>>>>> For the same boom length, there is a dB or more in SNR to be
>>>>>> gained with the same length antenna by choosing an antenna
>>>>>> with a clean design. That is a lot to disregard without some
>>>>>> thought. I realize that you live in the country, so that may
>>>>>> be a lesser consideration for you, but I think that relying
>>>>>> on the station on the other end to have an antenna with less
>>>>>> than optimum properties to work me is asking the station on
>>>>>> the other end to give up a lot just for my advantage. I do
>>>>>> agree that having 2 or more short antennas stacked vertically
>>>>>> is a much better setup for contesting than a single long boom
>>>>>> antenna. And much better for the rotor.
>>>>>
>>>>> My contest Q rate to the cities and to Chicago has gone down
>>>>> with the adoption of the better G/T antennas, and then I moved
>>>>> and don't have antennas up on a tower which hurts a lot too.
>>>>> Some contests I've not even worked my own grid but I've won
>>>>> certificates with 2 digit scores like the SOLP class in the
>>>>> whole Midwest Divisioin from home 2015 UHF contest with 48
>>>>> points. Definitely a sign of sparce activity.
>>>>>>
>>>>>> Thanks for pointing to me to your paper on stacking and the
>>>>>> modeled performance of various antennas. It was very
>>>>>> interesting. I don’t see the modeling of the 12 element M2
>>>>>> antenna you referred to in your earlier post in that article
>>>>>> though; did I miss something obvious?
>>>>>
>>>>> I didn't model the 12 element, just the 9 on a boom longer that
>>>>> K1FO with slightly lower gain.
>>>>>
>>>>>> You may want to revisit your modeling of the M2 9 element
>>>>>> antenna. The gain you modeled is over a dB lower than the
>>>>>> gain I modeled for that antenna when I was doing analysis for
>>>>>> a friends stack that contains a 2M9SSB. It is also about that
>>>>>> much lower than the gain for that antenna that VE7BQH quotes
>>>>>> in his table of modeled Yagi data which he generates from
>>>>>> several modeling programs and which is generally considered
>>>>>> an accurate representation of the performance of existing
>>>>>> antennas. I can dig out and send you my .NEC file for that
>>>>>> antenna to compare with your input file and to check your
>>>>>> model output. If you want to send me your input file, I would
>>>>>> like to review that against mine as well to make sure that I
>>>>>> haven’t made any mistakes. The modeling should be consistent,
>>>>>> and when it isn’t it helps one’s future modeling efforts to
>>>>>> find the cause. The modeled gain for the 9 element K1FO
>>>>>> variant looks about where it should be though, so it looks
>>>>>> like your Yagi modeling process with MMana-Gal is fine.
>>>>>>
>>>>>> Keep in touch. - Duffey KK6MC
>>>>>
>>>>> Here is the Mmana-Gal 2m9ssb.maa file that I used for my
>>>>> conclusions. No boom.
>>>>>
>>>>>> 2m9ssb * 144.2 ***Wires*** 9 0.250825, -0.494284, 0.0,
>>>>>> 0.250825, 0.494284, 0.0, 8.000e-04, 50 0.0127, -0.514121,
>>>>>> 0.0, 0.0127, 0.514121, 0.0, 8.000e-04, 50 0.576263,
>>>>>> -0.464434, 0.0, 0.576263, 0.464434, 0.0, 8.000e-04, 50
>>>>>> 0.873125, -0.461632, 0.0, 0.873125, 0.461632, 0.0, 8.000e-04,
>>>>>> 50 1.417637, -0.463329, 0.0, 1.417637, 0.463329, 0.0,
>>>>>> 8.000e-04, 50 2.124075, -0.456979, 0.0, 2.124075, 0.456979,
>>>>>> 0.0, 8.000e-04, 50 2.90195, -0.450629, 0.0, 2.90195,
>>>>>> 0.450629, 0.0, 8.000e-04, 50 3.762375, -0.447624, 0.0,
>>>>>> 3.76375, 0.447624, 0.0, 8.000e-04, 50 4.4323, -0.450799, 0.0,
>>>>>> 4.4323, 0.450799, 0.0, 8.000e-04, 50 ***Source*** 1, 1 25,
>>>>>> 0.0, 1.0 ***Load*** 0, 1 ***Segmentation*** 800, 80, 2.0, 1
>>>>>> ***G/H/M/R/AzEl/X*** 0, 0.0, 3, 21.0, 120, 60, 0.0
>>>>>> ###Comment### Mod by Gerald N. Johnson, K0CQ 1/23/2011
>>>>>> 6:19:23 PM Created by Gerald N. Johnson, K0CQ 1/23/2011
>>>>>> 5:08:08 PM
>>>>>
>>>>> In 1983 at CSVHF in Lenaxa KS, W0WL (silent key according to a
>>>>> recent QST) a long time Collins Radio employee with a ham radio
>>>>> reputation for doing extremely beautiful construction work had
>>>>> a NBS 432 yagi made with a square aluminum boom and aluminum
>>>>> elements welded in place. Looked beautiful, but on the antenna
>>>>> range was down 5 dB from the NBS data. Turned out he ignored
>>>>> boom effects and element diameter effects and used the lengths
>>>>> for wooden boom and 3/16" elements with a 1" square aluminum
>>>>> boom and 3/8" diameter elements (IIRC) from an ARRL Handbook. I
>>>>> had my own version of a NBS standard gain antenna for 432 but
>>>>> it was too heavy to get up on the pole and near the ground it
>>>>> had more than 4 dB ground gain over the theory. Two half waves
>>>>> in front of a 1 wave square reflector or was that a 2 wave
>>>>> reflector? My 7 dBi antenna showed a stronger signal than
>>>>> Warren's where there should have been about 5 dB the other way.
>>>>> That really pointed out the importance of boom effect a
>>> nd element diameters. Its probably also true that element diameter
>>> effects the coupling from element to element but NEC variations
>>> ignore that.
>>>>>>
>>>>>>
>>>>> 73, Jerry, K0CQ
More information about the NLRS
mailing list