Receiving weather images


This blog article is about receiving and decoding the NOAA weather data, it is an old system that still produces data that you can receive with a SDR and a right handed circular polarized (RHCP) antenna called the QFH that I mentioned in the previous blog post.

There are three satellites called NOAA 15, 18 and 19 that routinely transmit scanned images of the cloud deck in infrared and visual. I decided to evaluate the QFH antenna performance described in the previous blog post, the frequencies of the satellites are 137.1 MHz (NOAA 19), 137.620 MHz (NOAA 15) and 137.9125 (NOAA 18) which is below the designed resonance frequency, but the bandwidth is sufficient to cope with this.

At the before mentioned frequencies you set the demodulation to 60k WFM on a rspduo receiver, redirect the audio output to a virtual audio cable and select the VAC as input into the WXtoimg software. And for the rest you wait as the images come by, it is a system that you can leave on all the time. The first results where rather disappointing because the NOAA satellites transmit as right hand circularized (RHCP), and, if you look at the antenna in the previous post then it is really LHCP. But of a blunder but after some soldering, bending and twisting you finally end up with this:


This QFH antenna has the right sense of rotation, which is a difference of nearly 20 dB, in other words the previous QFH was looking at the reflected signal (RHCP becomes LHCP after a reflection) rather than the direct (through the roof) signal of the satellites, a stupid mistake.

Also, there is 30 meter of cable between the QFH and the SDR, together with the connectors and so on it results in loss of a couple of dBs. The intention of the LNA is to boost the signal by 10 dB at cost of some noise which comes from the LNA itself. Also the LNA can use some improvement compared to the previous post, the best advise I can give is to put it in a shielded metal box. This looks less attractive (perhaps) but you don’t want any other signal on the LNA input than the antenna input signal, or stay signals in the neighborhood of the amplifier.

Also, I tweaked around with several metal matches on the PCB of the LNA until all parasitic signals disappear. It is a bit of testing but the LNA noise figure (NF) appears to be about 2 dB. The purpose of the LNA is to raise the signal by 10dB which compensates for the cable losses and maybe the receiver noise floor. Anything that keeps the LNA amplification within the dynamic range of the receiver is fine, actually, you don’t want to amplify too much because the noise floor of the LNA will also go up. In all I hope that 10dB amplification is enough.


The above image shows LNA version 2 in its box, you feed it with 13.8 volt while the MMIC MAR-3 runs at 9.88 Volt. I rerouted some tracks on the PCB compared to version 1, also there was a bit of soldering but the SMA connectors were already there in the metal box before I started the project. The LNA schematic is as follows (it does 10dB amplification with a NF of 2 dB):

And now some results: The first images that I received looked like the one shown below which arrived on 22-feb-2019 at 15:41 UTC. In this case luck played a role because many subsequent images were far worse (left visual and right is infrared I believe):


A few days later I continued with the project, this is what I got, it is a composite image of both channels, the wxtoimg program does all the processing:


The above image is from NOAA 19, it was right overhead on 24-feb 15:20 UTC, the signal to noise ratio (SNR) was >18dB at the highest point in the pass (while this QFH antenna is under the roof). So this is what you want, a RHCP and a 10 dB LNA, not a very good general purpose antenna because that would be one with more gain at low elevation. In this case you need right handed circular polarization (RHCP) and in particular at elevation angles beyond 20 degrees. No problem with the SDRduo to do this, the only thing to do is to reroute the audio via a virtual audio cable (VAC) to the software (wxtoimg):


A bit of luck also helps, the above image should how the PV elements of the neighbors are a bit of a pain sometimes (it is the inverter that makes some noise), but the luck was that the inverter stripes remained out of the reception window.


The above image is the next good pass, it is of NOAA 15 on 17:54 UTC. Here I played around with the filters in SDRplay because they are crucial to get rid of the zebra stripes which are caused by switching power supplies. SDRplay works fine with the following settings: WFM demodulation, 60k width, VOL in WXtoImg should indicated around 60%, adjust the passband in the demodulator window so that it fits all spikes. Please no frequency adjustment because that messes up the decoding (the Doppler effect is not that much, this demodulation window can easily cope with this). Also, put NHCL on and NBW on. Over the Pyrenees in the above image you see what the zebra stripes are when the NBW filter is temporarily off during reception.


WXtoImg also produces various output files, the above one is the same (NOAA 15 24-Feb-2019 at 17:54 UTC) but it accentuates something (still have to read the documentation). Also, whenever the SNR is below 10 dB you only get noise, in that case the test strips in the left and right margin of the raw images will disappear, you need to play around with the elevation margin settings in WXtoImg to skip the passes below your horizon.

Parts of the text on this blog can also be found in the facebook group SDRplay and on twitter. It is a fun project where you learn that the ideal for all circumstances antenna does not exist. The results that I got with the RHCP QFH/LNA combination are beyond anything that I ever got with the Diamond X30 co-linear antenna, which gave me nothing more than zebra striped weather images whatever I tried. This is the main learning experience of this project. What is circular polarization and how do LNAs perform.

About the NOAA weather satellites

For consideration, up to now I was able to track NOAA satellites numbers 15, 18 and 19. These are aging satellites launched in respectively 1998, 2005 and 2009. I do not think that this system will be replaced by additional satellites. The modern solution would be to get a subscription from EUMETsat and receive their broadcast information from a geostationairy satellite. That requires at least a dish and a LNB to downconvert the signal into the RSPduo tuning range. The LNB dish combination is what I like, but I don’t like the subscription idea (not sure how expensive it is). Maybe a project for in the future.

A short youtube demonstration

On how the signal arrives, and how it is processed is here:

Last update: 26-feb-2019 14:04

Weather report

How is the winter progressing, it became warmer, but we had a storm on the evening of the 22nd and the night of the 23rd. The lower right graph shows the pump actions (blue) and derived from the the estimated hourly rainfall.

UL: entrance temperature, UR: outside temperature, LL: pressure, notice the storm on the 22nd and the 23rd, LR: hourly rainfall estimated from pump actions
300hPa wind data, this was the jetstream on the 22nd, thanks Roderiko
Cloudcover, nice wave pattern, thanks Roderiko
Analyzed weather map, thanks colleagues of Geert Jan
Windgust on the 23rd in m/s, so we had roughly 28 m/s in Rotterdam

Last update: 26-feb-2017