the phantom / odd false decoded suspect spots

1 post / 0 new
Last seen: 2 weeks 5 days ago
Joined: 2014/09/10 - 10:33
the phantom / odd false decoded suspect spots

As seen on the Stations Participating per Day we can find a constantly high level of about 700 calls heard during the last months but a decreasing number of about 500 reporting stns.

So my question was: is it due to the phantom odd false decoded signals?

Let us have a look on month 2015-07 (July) (for example)

and even more: on a 24h-map (20m band) one can see some odd false (phantom) decoded signals.
These suspect spots are decoded only once by one stn in this month!
In addition: The TX-QRA is 4 characters in the database.
This means: there are no RX-reports from this Station back to the database.
Some more informations on phantom signals could be found in written by Paul, VK4ZBV

If we assume these odd false decoded signals are reported only by one RX-stn, we find
in July 2015 archive-file 3621 TX-stns reported only once - this is about 0,041 % of the spots of the month.

When we take the average per day, it is about 117 stns. So 117 stns are reported as "calls heard" which never had reported to the wspr-database
There could be another reason: Stns only use a TX and no RX

Let us have a focus on the odd false decoded signals

some typical reports (sorted by TX-Call) out of the CSV-archive file


and with the noID-Decoder (both must be the K9AN-decoder):


but also very realistic calls (France) but with wrong TX-QRA


So my questions was: Are there decoders, having high rate of odd false spots?
The 3621 odd false spots of July show the following distribution:

decoder__________% relative _______ abs %
2.0_r1714_______48.68______________ 0.108
2.12_r3617______ 5.52______________ 0.108
1.6.0-deve______18.25______________ 0.064
noID(?K9AN?)____ 9.66______________ 0.051
0.8_r3058_______ 4.86______________ 0.013

the "relative %" is based on the 3621 spots, the "absolute %" are relative phantom signals on the base of spots with these decoders during the month.
As it is stated in the last column: the 0.108 % of 2.0_r1714 and the 2.12_r3617 means that every 925 spots will be one odd false decoded signal.
It is unclear to me, if the last line is statistical significant for the "0.8_r3058".
We can see, that the noID-decoder has nearly half the phantom decoded signals of the mostly used decoders. (every 2000 spots one phantom signal)

I feel, the 0,041 % false positive spots has to be tolerated, even when a 24h-map with some of these odd false /phantom spots looks very surreal. It seems to happen with all momentary used decoders due to the principal of the encode / decode and FEC, as stated earlier.

It may be one day in the future, that the database admin can set up another (daily/monthly) filter to clean up these odd false spots, when they become more than 0.1 %. But we hope for better decoders!