Like MMANA, but in a Browser: Antenna Modeling on NEC-2
Almost every radio amateur who has ever computed an antenna knows MMANA. A wire table, w1c instead
of a segment number, an optimizer, a library of hundreds of ready models — more than one generation
learned on it. But it runs only on Windows, its code is closed, and it computes the input impedance
of an antenna over a perfect ground, though the antenna itself hangs over ordinary soil.
The site now has an antenna modeler that keeps the familiar wire table and the files of MMANA but computes with the NEC-2 engine over a real ground. And it runs right in a browser tab — on a computer with any system or on a phone, with nothing to install.
Where MMANA goes wrong
MMANA's engine is MININEC, and its own help says it plainly: it computes the input impedance over a
perfect conducting ground. It takes the real soil into account only in the radiation pattern. While
the antenna hangs a wavelength up or higher, the difference is small. The dipole DP20.MAA from the
library of MMANA-GAL, 20 m up:
| Impedance | SWR | |
|---|---|---|
| MMANA-GAL (over a perfect ground) | 78.6 Ω | 1.61 |
| NEC-2 over the same perfect ground | 78.1 − j9.6 Ω | 1.60 |
| NEC-2 over a real average ground | 75.1 − j6.8 Ω | 1.52 |
The first two rows agree: the programs compute the same model the same way. The third row is what the real soil changes: 3.5 Ω at this height. For a low antenna it is no longer a detail. A 20 m dipole 3 m up, 0.14 of a wavelength:
| Ground | Input impedance | Gain to the zenith |
|---|---|---|
| Perfect, as in MMANA | 39 + j12 Ω | 8.6 dBi |
| Real, average | 58 − j3 Ω | 5.4 dBi |
That is a difference in SWR, in resonance and three decibels of gain — exactly for the antennas strung between two trees at a summer house or off a balcony. They need a real ground most of all.
What the new program does
The model is as in MMANA. A wire table in metres, the radius in millimetres, sources w1c, w2b,
coils, capacitors and traps with a Q. .maa files are opened and written, and after opening one the
program says what of the file could not be taken over. A Yagi is easy to edit by its elements, as in
the Yagi editor of MMANA: the length symmetric about the centre, the position, the diameter, a copy as
a new director.
The engine is NEC-2. It is nec2-rs, our own port of
the public-domain nec2c to Rust, checked against the original report by report and compiled to
WebAssembly. The ground is free space, a perfect one or a real one by the Sommerfeld method. It
computes in a thread of its own, so the page does not freeze even during a long sweep.
Templates that are resonant where they will hang. A dipole, an inverted V, a vertical with radials, a delta loop, a quad of 1–3 loops and a Yagi of 2–6 elements. You give the frequency, the height and the wire diameter — and the engine cuts the driven element to resonance at that very height over the ground, not in free space. The dimensions of the Yagis were found by the program itself: the most gain with a given front/back and an SWR up to 1.5 over ±1.5 % on a 50-ohm cable without matching. A three-element Yagi for 20 m, 12 m up, comes out at 40 Ω, 12.2 dBi and 18.5 dB front/back.
A sweep with a Smith chart. SWR, R and X, gain and front/back against frequency, the working band by SWR — and the impedance curve on a Smith chart, with the circles of SWR 1.5 and 2. You see both the band and where the impedance heads at its edges.
Matching with part values. An L network with a coil and a capacitor for your frequency, a hairpin for a Yagi with its length for tubes of a given diameter, a quarter-wave transformer with a table of common cables.
Comparing variants. Keep a result, change the antenna — and the old numbers stand under the new ones, while the patterns, sweep curves and Smith chart of the kept variant lie dashed on the same charts. What a director, the height or the optimizer gave is plain at once, without a notebook.
And the rest you need at work: an optimizer, as in MMANA; the currents along every element; the Leeson correction for elements of telescoping tubes, the same as in EZNEC; moving an antenna to another frequency; a PDF sheet with a drawing in millimetres to build from; a pop-up hint over every name; a light and a dark theme; Ukrainian and English.
Side by side
| MMANA-GAL | Antenna modeler | |
|---|---|---|
| Runs on | Windows | any modern browser, phones too |
| To install | the program | nothing |
| Engine | MININEC | NEC-2 (nec2-rs) |
| Input impedance over a ground | over a perfect one | over a real one, by the Sommerfeld method |
| Files | .maa | .maa (opens and writes), its own .json, export of .nec cards |
| Code | closed | open, MIT |
| Price | free | free |
A tip for MMANA files
MMANA saves in a file the frequency the model was made at, not the one it was computed at. A colleague sent a six-element Yagi for 2 m saved at 14.15 MHz: computed at that frequency, it gives nonsense of thousands of ohms. At 144–145 MHz the same model gives nearly the numbers of MMANA: the impedance within an ohm or two, the gain within 0.1 dB. The program now notices such files and says which frequency the antenna fits. If a model from MMANA computes oddly, look at the frequency first.
Honestly about the limits
NEC-2 is a model of thin wires. It does not take wire insulation into account, does not model buried radials, and computes a wire touching a real ground poorly. The library of hundreds of MMANA antennas stays in MMANA — but its files open here. A link that carries a model, to share it, is still ahead. And a model is not hardware: masts, houses and insulation move the resonance, so a finished antenna is tuned with an analyzer anyway.
How to try it
Open the program, press “Templates…”, choose an antenna — and “Compute”. What every field means and how to get from a template to a drawing is in the guide with three examples: a dipole for 40 m, a vertical for 20 m and a three-element Yagi. The code of the program is on GitHub; bugs and ideas are welcome there too, in the issues.
