Monday, March 04, 2013

Classic WSPR vs WSPR-X

Are you a fan of WSPR mode? Have you tried K1JT's new program WSPR-X yet?

Comparing classic WSPR to WSPR-X
I decided to switch to the newer program as the older 'classic' version won't work with VSPE virtual serial ports. But I had a sneaky feeling that WSPR-X was not decoding some of the traces it should. So I decided to run both programs in parallel, using the same sound card, the same radio, the same data source. Sure enough, WSPR-X is missing about 1 decode in 10 compared to WSPR 2.11. There is no apparent common factor between the signals it missed. They are not at the extremes of the frequency range, close to the limit of timing error nor especially faint.

Look at the screenshot above and look at the decodes for 1540. Classic WSPR has decoded two signals for this interval whilst WSPR-X has decoded only one. The signal from W3CSW was missed. Later signals from the same station were decoded. That is just one example. I only needed to wait a few minutes to find another.

I set the older WSPR to save .wav files and when these were processed by WSPR-X using its File Open option the result was the same as when the signals were received off-air. The same transmission was missed in each case.

WSPR-X seems a bit faster to run the decodes than WSPR. It prints them up on the screen before classic WSPR does. There are sometimes slight differences in the dB and DT figures, but not enough to worry about. Has anyone else noticed this?

7 comments:

YO9IRF said...

Maybe this is happening because the undecoded signal was at -23dB ? Is this consistent with the other missed signals ?

I am still on WSPR 2.20 because of the frequency hopping feature, it is priceless.

Julian Moss said...

Nope. I've seen weaker ones get decoded and stronger ones get missed.

I've let Joe K1JT know. It will be interesting to see what he says.

Julian, G4ILO

Paul PC4T said...

I agree with you, I see hardly any difference, I think sometimes I prefer the 'old' WSPR. 73 Paul

PE4BAS, Bas said...

Very interesting Julian. I hate to miss something. Especially weak signals :-) Very curious what Joe has to say about it. There must be a logical explanation as the decoding core would be the same I guess. 73, Bas

Bert, PA1B said...

Hello Julian, I think it is "de-sensing". The signal of SM7BKZ is 30 dB stronger than the station W3CSW that disapared. 7-(-23)= 30 dB.
I see this often in WSPR Propagation Analysis that a a station is suddenly 20 dB weaker in one timeslot, because in that timeslot an other station is received with a much stronger signal, which is De-sensing the receiver.
So please compare difference in SNR of the two stations in the same timeslot. I curious of your findings. Happy hunting. hi
73, Bert

Julian Moss said...

I don't think it is desensing or any other receiver-based phenomenon, Bert. This is the result of using the two programs to decode the same input. you can try it yourself using this test file.

Joe K1JT says WSPR decoding in a crowded 200 Hz slice of spectrum involves some inevitable randomness. When signals or roughly equal strength collide, the outcome may be decoding of station A, or B, or neither, or both. At S/N around -25 dB or below, very small changes in the noise statistics -- including a small change in the exact start time of the synchronized data -- can change the outcome from decode to no decode.

He ran the two programs against each other for 3 hours last night and sent me the all_wspr.txt files. WSPR 2.11 made 204 decodes while WSPR-X made 245. That's a 20% difference in favour of WSPR-X! I don't know what to make of that. If anyone else would like to try the two programs against each other I'd be interested to know what results you get.

Julian, G4ILO

Bert, PA1B said...

Hello Julian, thank you for this excellent comment on signals colliding. 73, Bert