On 12/04/2021 03:30, Andy Durbin wrote:
Would someone please point me to an explanation of how the dB report
of a received FT8 signal is derived. My recollection was that it was
independent of the actual receiver bandwidth but simple
experimentation shows that's not true. Simply narrowing the receiver
bandwidth on a signal of interest results in a huge increase in the
reported signal to noise ratio. This seems to imply that the noise
being considered is the noise in the actual receiver passband and not
the total noise in a 2.5 kHz bandwidth.
Did I remember the implementation incorrectly or has the
implementation been changed fairly recently?
Thanks and 73,
Andy, k3wyc
Hi Andy,
noise estimation is mode dependent. Modes that are likely to be used
without other signals in the pass band measure the noise between tones
and extrapolate to the 2500 Hz noise bandwidth we use for SNR reports.
Other modes like FT8 that are often used in very crowded bands, and are
able to decode overlapping signals, need a different approach to noise
estimation, they measure the whole received spectrum and attempt to fit
a baseline curve to estimate the noise. When estimating a baseline it is
essential that the receiver passband is close to the 2500 Hz noise
bandwidth reference or more, otherwise noise will be underestimated and
SNR numbers will be incorrect higher than reality.
In general we do not recommend reducing your Rx bandwidth to less than
2500 Hz, the filtering in the WSJT-X decoders is far more suitable for
these modes than anything your rig might offer. The only exception might
be when there is an exceptionally strong signal in the passband that is
disrupting decoding by requiring greatly reduced RF gain or excessive
AGC action that pushes wanted weak signals below the receiver noise floor.
73
Bill
G4WJS.
_______________________________________________
wsjt-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/wsjt-devel
