Hi Joe, > Very interesting! The model and received spectra are remarkably close > to identical. In fact, it's hard to see any differences as large as 5 > dB, such as the spikes around 750 Hz in the top right plot.
I zoomed in on the region between 500 and 1000 Hz to better show what’s causing the spikes: In the left-hand plot, the red curve is the spectrum of the received signal and the blue curve is the model. Notice that the red curve is below the blue curve (due to RX/TX filter rolloff) except in the spectral “dip” around 750 Hz. I attribute this to the fact that the model does not include the noise floor whereas the received spectrum does. I can try to use the refspec.dat data and our SNR estimate to come up with a modeled noise floor that would be incorporated into the model spectrum. Alternatively, if we’re going to do fitting (either polynomials, or vector-fitting) I think that we could just ignore data points that correspond to low-snr dips in the spectrum. > I guess the fact that phase differences are mostly in the range +/- 45 > deg (at least over the frequency range +/- 750 Hz) must mean that we're > not losing very much potential sensitivity. This is with a TS-2000 at > both ends; I wonder how much it will differ when other rigs are used? Yes - I wonder the same thing. I’ll try this on some other signals as well. Steve
------------------------------------------------------------------------------
_______________________________________________ wsjt-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/wsjt-devel
