False decodes WSPR

[PE3ES / F4VTQ]

Hi all, and Alan,

I am comparing a variety of receivers and software for WSPR decodes. When looking at the map on wspr.rocks I see that the SparkSDR sw for me is creating more false decodes (relatively to the amount of uniques) than my other software like WSJT-x PavelD on Red Pitaya and wsprdaemon. In the order of 2 % of uniques compared to <1 % of uniques. Can that be attributed to a more aggressive approach ?

And what can I do about it as I would like to generate 0 false decodes ;-)

As far as I can tell all my decoding software is using wsprd under the hood. Probably different versions.

Erwin

[ahop...@googlemail.com]

Hi Erwin,

what decode level are you using?  I had improved the callsign verification used for psk reporter uploads but had not applied it to wsprnet uploads, I have just added this so the next release may be better.

73 M0NNB

[PE3ES / F4VTQ]

I use the most aggressive : 3

Results in spots in the middle of the ocean, maybe I need to restrict that a bit. The map is showing uniques for 24 hour period with SparkSDR only (a Hermes Lite V2 9 Rx and a SDRplay RSP1a doing the lower bands including EU 60 and 80)

Erwin

Op donderdag 9 februari 2023 om 12:43:36 UTC+1 schreef ahop...@googlemail.com:

[PE3ES / F4VTQ]

As spots that land in the Oceans have a valid   (but not existing)  callsign  and a wrongly decoded grid, call sign verification might be difficult.

I also see "balloon-like callsigns 0Q... 0V... 0U..., if these are real balloons they are not false decodes.

Op vrijdag 24 februari 2023 om 13:42:11 UTC+1 schreef PE3ES / F4VTQ:

[ahop...@googlemail.com]

Hi, as you are using level 3 it is the wspd decoder detecting false spots (same true for 1&2), so I'm not sure why spark would report more than other software, maybe they have better filtering for bad spots.  I have added the same filter I use for pskreporter to the next release and will check that balloon spots are not dropped.  I shall ponder further.

73 Alan M0NNB