WSPR data
|
jcarroll15@...
When WSPR is receiving a signal, does it take the average SNR or the peak SNR during the 2 minute period. There is always Qsb during the transmission. Some times signals can vary many Db’s over a 2 minute period. Can anyone help me understand how it works. I am losing sleep. Lol
John AA6RF |
|
|
|
Jim Lill
It uses what wsprd, the decoder part of wsjt-x , reports On 5/6/22 12:03,
jcarroll15@... wrote:
When WSPR is receiving a signal, does it take the average SNR or the peak SNR during the 2 minute period. There is always Qsb during the transmission. Some times signals can vary many Db’s over a 2 minute period. Can anyone help me understand how it works. I am losing sleep. Lol |
|
|
|
jcarroll15@...
I know that but does it take a peak or average reading over the 2 minute period. Signal levels change over the 2 minute period when conditions are unstable. (QSB)
toggle quoted message
Show quoted text
John On May 6, 2022, at 12:57 PM, Jim Lill <jim@...> wrote:
|
|
|
|
Glenn Elmore
I'm not sure this applies to what you are asking but there is further detail on wsdrdameon.sh noise algorithms at http://wsprdaemon.org/technical.html and in the thread https://forum.kiwisdr.com/index.php?p=/discussion/comment/9228 The charts delivered by wd.sh use two different algorithms: one which looks at the entire interval and the only the quietest 50ms of the quiet time. THeses are sueful to high pkak/average/ such as T-storms v more uniform sources. On 2022-05-06 13:57, Jim Lill wrote:
|
|
|
|
John Carroll <jcarroll15@...>
Glenn, thanks for the reply. I want to explain why I asked this question on this forum. A friend of mine had a 20 meter vertical antenna on his truck. I have a hy-gain AV620 multiband vertical at my house. We were doing A-B testing of both antennas at the same timeframe. On average I was getting more hits then his over a 6 hour timeframe. Average 2 to 3 db better SNR on almost every hit. We live 5 miles apart. Now the next day he was moving and driving to work and suddenly he was getting more hits with better s/n, a couple of db’s better. He parked at work and the readings came back to normal as we tested the day before. I know the AV-620 preforms better then the mobile 20 meter ham stick. If WSPR reads peak signals then this whole experiment makes more sense. It seemed like a simple question to ask but it turns out I was wrong.
toggle quoted message
Show quoted text
Thanks On May 6, 2022, at 1:19 PM, Glenn Elmore <n6gn@...> wrote:
|
|
|
|
Glenn Elmore
If it seems simple, it probaably just means we haven't looked closely eoug! (:>) Understanding absolute noise spectral density, type and coupling mechanisms is a challenging but can be a fruitful career. If you haven't looked at it, perhaps my notes may add some further obfuscation: http://wsprdaemon.org/ewExternalFiles/N6GN_Notes_on_Improving_Station_Noise_Performance03.pdf Glenn On 2022-05-06 15:28, John Carroll wrote:
Glenn, thanks for the reply. I want to explain why I asked this question on this forum. A friend of mine had a 20 meter vertical antenna on his truck. I have a hy-gain AV620 multiband vertical at my house. We were doing A-B testing of both antennas at the same timeframe. On average I was getting more hits then his over a 6 hour timeframe. Average 2 to 3 db better SNR on almost every hit. We live 5 miles apart. Now the next day he was moving and driving to work and suddenly he was getting more hits with better s/n, a couple of db's better. He parked at work and the readings came back to normal as we tested the day before. I know the AV-620 preforms better then the mobile 20 meter ham stick. If WSPR reads peak signals then this whole experiment makes more sense. It seemed like a simple question to ask but it turns out I was wrong. |
|
|
|
John Carroll <jcarroll15@...>
Let me dumb the question down. When you are listening on the radio and over 2 minutes the signal has qsb, from S9 down to S3 and back and forth. SNR and S units are kinda the same. It’s just a reference to signal strength
toggle quoted message
Show quoted text
At this point,I’m throwing in the towel and going with peak reading is the answer. On May 6, 2022, at 2:50 PM, Glenn Elmore <n6gn@...> wrote:
|
|
|
|
Glenn Elmore
Look in Gwyn's detailed description of the two wsprdaemon.sh algorithms. These algorithms are in addition to what 'norml' wsprd from wst-x uses to to report SNR. So yo may be referring to any one of at leat 3 SNR calculations. K1JT describes that wsprd algorithm on his pages but it's something like the lowest 30% of bins over the wspr interval if I remember right. As I said the other two are different with one more similar to Joe's and ont looking olny briefly a a transmit-quiet time. I'm presently in-hospital and a bit low on resources or might try to elaborate but this is generally the situation. It depends which algorithm you are using and how you are calibrated All are of course relative rather than absolute (unless one has carefully calibrated the system to reflect overall antenna factor - power in the real part of the actual radiation resistance converted to the MA connector on the Kiwi. In comparing to contemporanious reports from two systms there are of course any other factors that keep a simple single value answer from telling the whole story. Not only does one need an absolute characterization but polarity and system take-off-angle may enter in. It's a jungle out there! On 2022-05-06 16:09, John Carroll wrote:
Let me dumb the question down. When you are listening on the radio and over 2 minutes the signal has qsb, from S9 down to S3 and back and forth. SNR and S units are kinda the same. It's just a reference to signal strength |
|
|
|
Gwyn Griffiths
Glenn's recollection regarding the lowest 30% is almost the answer. The noise estimation algorithm within wsprd (which is the wspr decoding program within wsjt-x) takes all of the individual FFT frequency binned estimates, sorts them in increasing order of level, and takes the level at the 30th percentile of all frequency bins as the noise level. In an article in Sept/Oct 2020 QEX I wrote, with Glenn and Rob, there's a summary of the algorithm, with example frequency spectrum and sorted levels plots (Figure 3a and 3b) to explain how the noise level is estimated. I've sent you a copy to your email address on QRZ.com.
Within wsprdaemon, as Glenn has said, we use two noise estimation algorithms, the FFT one is nearly the same as in wsprd, but we chose to use the total power in the lowest 30%, scaled, as the noise estimate rather than just the 'spot' value at 30% as in wsprd. As the wspr bands become crowded, the level at 30% may be raised by the weaker signals, one reason why wsprdaemon also has an RMS noise estimate, from the quietest 50 milliseconds between transmissions. Glenn has referred you of a long report we wrote, which has even more detail, but the QEX article is a nice summary. Happy to send to anyone that would like a copy. Gwyn G3ZIL |
|
|