Dear Andy, LF Group,
I have just tried a quick test here, which more or less confirms your
findings. My approach was to receive off-air signals as normal with 300Hz
BW, but to also run Spectrum Lab on the decoding PC, configured in " narrow
CW filter" mode, with centre frequency set to 1475 Hz (the audio frequency
G4JNT comes out on) with a BW of 50Hz. The resulting filtered audio was fed
into a second PC running the WSPR decoder. This has the effect of increasing
the reported SNR by about 6dB to 13dB when the filtered audio is decoded -
however, this is smaller than the increase you saw.
Doing this also produces a lot of false decodes - the spectrogram display in
WSPR shows a couple of broad, bright, noise-like bands where the
passband-stopband transitions of the audio filter occur, with assorted
spectral lines that are probably low-level noise from one PC or another. I
have often noticed the number of false decodes goes up when there are
narrow-band spectral "features" present, rather than just uniform noise. It
is curious why the WSPR spectrogram display shows fairly uniform noise
baseline across the whole 200Hz, with brighter bands of noise where the
filter attenuation is increasing, rather than a bright band of noise in the
50Hz passband and darker in the stopband.
Increasing the RX bandwidth to 3kHz makes essentially no difference to the
normal, un-filtered WSPR decoder over the 300Hz BW.
Cheers, Jim Moritz
73 de M0BMU
----- Original Message -----
From: "Andy Talbot" <[email protected]>
To: <[email protected]>
Cc: <[email protected]>
Sent: Sunday, February 07, 2010 6:20 PM
Subject: LF: WSPR S/N Tests
To try to get to grips with the falsely high values of S/N reported by
some
stations recently, I made a set of controlled measurements over a direct
link using a range of Rx bandwidths and input signal levels.
|
