A small tool to compute antenna radiation patterns with WSPR or FT8
A longer presentation of this project may be found at: https://docs.google.com/document/d/1xli5nsfunJtP1ATBcLF-FFQXslA9Ovz93nYWgzLMxd0/edit?usp=sharing
The idea is to use two receivers connected to two PCs (or one decent PC with two soundcards, or SDRs) and to collect statistics on the received stations over a defined time span (it depends on propagation and on the intended measurement purpose). Try to avoid very long sessions because the reflecting layers changes height, so the incoming vertical angles will change, a lot. So, if your recordings are too long, your results will be difficult to read because they're the average over several kinds of propagation type.
Data may be collected in two main ways:
- Automatically collected from the WSPR db: http://wsprnet.org/olddb?mode=html&band=all&limit=2000&findcall=&findreporter=is0kyb&sort=date
- FT8 reports produced via Robert Morris AB1HL weakmon: https://github.com/mcogoni/weakmon
Note: WSJT-X doesn't produce useful results: SNR is computed in a non-consistent way that depends even on the window size in pixels. Modifying WSJT-x source code is really more complicated than it should and I ended up adapting Robert' code, which is, on the opposite, exemplar for clarity (given the difficult task anyway) and he wrote a FT8 decoder for dummies that really can help you understand how these communication protocols work: https://github.com/rtmrtmrtmrtm/basicft8
My forked version implements a different SNR computation allowing much more precise data. In this case, you should have two weakmon instances running on the same PC or two different PCs. FT8 decoding is very computationally heavy, so don't use slow computers. The advantage of FT8 is that you have so many more signals to acquire from so many different directions and people usually use quite higher power (sic!) than WSPR. So, in general, recording times may be much shorter to reach the same statistics.
The Jupyter Notebook will process the data in steps and the result will be (you can do anything you like here...):
- Angle / distance distribution of the spots;
- SNR difference (between the two RXs) plot;
- Approximate antenna pattern of unknown antenna if you have a omnidirectional antenna as reference;
Of course you should live in a radio quiet area to obtain "scientific" results. In particular, if you have some known local noise coming from a specific direction, the antennas should be in the same spot.
This example pattern was obtained after about 6 hours and by exploiting the known antenna symmetry, since there are very few stations active from the South (Africa).
As told above, try not to mix long and short propagation since it comes from different vertical angles and you would end up obtaining a horizontal pattern assiciated to several vertical angles... this seems unavoidable without special beam-forming antennas.
Be careful to initially characterize the two rx chains by feeding them the same antenna (i.e. via a hybrid splitter) and check on the WSPR website their relative SNR values on the same spots: take the average value and put the number in the code parameter "rx_offset".
If you're unfamiliar with Jupyter Notebook, you can easily run the code on Google Colab: https://colab.research.google.com/github/mcogoni/APET/blob/master/WSPR_Antenna_Pattern.ipynb and run the commands in the first cell to clone the Github code in Google Drive. To use the notebook in read/write mode, you should save it and it will belong to you.
Another use of the same data is to plot the SNR difference between the antennas over time
to directly verify how propagation evolves, which distances are open, etc
As you can see above, from 8:00 UTC to 22:00 UTC, the highest gain of the directional antenna is over 10dB in the morning, but it degrades over time, until North Europe vanishes ~2 hours after sunset. Then only USA remains and the difference between the antennas grows constantly especially for 9000km paths.
73, marco / IS0KYB