23cm Antenna progress
Some 3 years (god, is it that long? it seems it is!) I started building some antennae for 23cm based on the "usual" DL6WU formulas, JVL calculations etc. Results were not good, the antennas would not resonate where they should, the pattern was sub optimal.
Some of the early failures, with success at the front.
Numerous attempts to fix them failed. The problem was "boom correction" .. I was using 4mm rods in either 15mm or 5/8" (15.875mm) booms, with insulators. Various rules of thumb exist on how much extra length to add to compensate for what is "hidden" by the boom cross section. I specifically did NOT want to use elements directly embedded in the boom with an electrical connection to the boom itself, as while they may work initially, as corrosion between the elements and the boom sets in, performance is variable at best.
The realisation that "just adding 75% of the boom diameter" was not going to work came through building a proven design for a 70cm antenna. [1] I used the correct rods, the insulators as specified by the designer and the result was incorrect, resonant at 440MHz not 432MHz ... after a lot of head scratching, emails with the designer he was very confused ... many had been built, all had been fine .. why was mine wrong? The answer was, I had used 3/4" (19mm) aluminium box instead of 20mm! "take off 0.3mm from each element" was the instruction, which I did and the antenna immediately became resonant on 432 and performed to specification.
The fact that just 0.3mm could have such an effect at 70cm made it obvious the boom correction for 23cm would have to be very carefully controlled. My blundering amateur efforts at adding a few percent were not going to work! I set about trying to understand the physics of the problem, reverse engineer some designs, read all I could find on the subject and built quite a few more test antennas.
The process I finally used used was to go back to basics and model the antenna in NEC, then correct the antenna using some published formulae, finally use a fixed correction figure determined through experiment, as per DG7YBN's article in Dubus 2/2021 (New 1.3GHz Yagis).
I took some "inspiration" from one of Hartmut's 70cm antennas and then scaled it to 23cm to obtain starting points for the boom positions. The element lengths cannot just be scaled because the original design used 4.76mm elements which scale to 1.6mm at 23cm, and I wanted to use 4mm. I applied the formula in Ian White's (GM3SEK) book which seeks to replace one element with another element of different diameter and adjust the length to get equal reactance. This gets you "close" to the optimum element length.
Finally, I wrote some Perl software that read in the NEC wire file and adjusted an element length or position by a small amount, re-ran NEC, parsed the results file and decided if the gain and/or SWR was improving or getting worse, then either made another adjustment in the same direction or abandoned that correction and moved on to the next element.
Several passes were made optimising for either gain, F/B ratio or SWR or a combination of them, and finally, I had an NEC model that used the 4mm wires I had and produced a decent pattern, gain and SWR on 23cm. I got a bit lost in this process, it is iterative and more akin to "art" than "science" I suspect better optimisation tools (such as MMANA-GAL) do a similar or better job, but, I used some hand crafted Perl. The code is in my GitHub[3]
The next step was to apply the initial "boom correction" .. Leif Asbrink (SM5BSZ) published many years ago a wonderful work on boom correction [2] done with models at 144MHz, and produced some FORTRAN code that will correct various antennas at 144 and 432 MHz with proven success. It takes into account the element diameter, the boom size, wall thickness and hole size. Additionally, he found the correction changed as you approach the ends of the boom ... this too is taken into account by BSZ's code.
There are various versions of this around the 'net, but the ones I found suffered from the same "problem", they would (quite sensibly) throw an error if operated outside the limits of the acceptable scales. If an element is too thick, too thin, too short etc to fall within the range modelled by the software it throws an error and quits. I translated the boom correction algorithm to Perl and as a bonus, it reads wires directly from the source NEC data file. My version will print warnings on STDERR and the main output on STDOUT so you get a nice CSV file of the results, with any warnings just displayed on the terminal screen.
Finally, as per Hartmut's article, a fixed amount was determined through experimentation to make up for the effects of low segmentation density in NEC and the bact the BSZ's boom correction algortithm was working well outside the tested limits. To put it into context, when BSZ developed the software it was for typical 144 MHz elements .. when a 4mm element at 23cm is scaled to 144 MHz, it is 36mm diameter, well beyond the limits the formulae were designed to. Hartmut (see Dubus article) had determined that a fixed offset would handle this. He had found 4.52mm worked well for 3/16" elements in a 5/8" boom .. and 2.21mm for 4mm elements in a 15mm boom. I wanted to use 4mm elements in a 5/8" boom and a different insulator.
Several test antennas were built, elements were cut to an accuracy of +- 0.1mm on a small mill. Measurements were taken on a HP 8753 VNA.
Insulators are from "JetPress" type INS014 [4], drilled with a 5.8mm hole they form a nice tight grip on the 4mm element.
Each antenna was tested for resonance, notes made and another set of elements cut with a slightly different fixed offset.
I determined a 4.0mm additional length offset worked well for this bushing in 5/8" box, with a 1/16th wall thickness. The resonance point is centered on 1296.2 and with suitable adjustment of the arms of the "bent element" DE, a nice sharp, deep null is produced at resonance. Build dimensions are here: https://github.com/rszemeti/Antennas/blob/main/23cm/30ele/optimised_30_ele_V2_plus4p0.pdf
Further testing will verify the pattern, but I believe I am on the right track!
Comments
Post a Comment