> MDS, S/N and similar measurements with NR on must be peformed
> completely with the signal present, and not by switching the
> signal on and off.
> This means that the only way to do it is by spectral
> analysis, with due care paid to subtle issuses like windowing
> functions, bin size, etc.
I've been poking at NR on and off for a couple of days now on 2.032. Using
an Elecraft N-gen noise generator to a Daven RF attenuator (to control the
fixed output of the N-gen), with an 8640B providing the signal, and both fed
to the radio through a TT 651 hybrid. And watching the whole thing on
Spectrum Lab. AGC off (really very fast), and the RF gain reduced enough to
keep the signal out of the AGC range.
With no signal present, NR drops the noise output to about -30dB (relative).
But the baseline noise level increases with the application of a signal to
about -18dB. NR drops both signal and noise level when turned on relative
to the levels without. The noise component in the passband is rolled off
fairly gently by what appears to be a a very broad filter effect, so it's
most likely that SNNR is increased, but I'm not sure how to measure it from
what appears on the display. The signal peak actually seems to decrease by
2-3 dBu relative to the noise at the baseline right in the vacinity of the
signal, but there is noise rolloff above and I think below the signal
(although the lower side is harder to ascertain). It shows up best if SPOT
is set relatively high, such as around 1 Khz.
There is no indication of any very narrow or steep-sided bandpass filter
being applied, so the filter itself looks relatively simple and quite broad.
Actually, it looks much like what I see from a couple of external
audio-based DSP gizmos.
My recollection of pre-1.372 NR performance, on SSB signals, is that the
high frequency rolloff in particular was possibly more pronounced.
Grant/NQ5T
_______________________________________________
TenTec mailing list
TenTec@contesting.com
http://lists.contesting.com/mailman/listinfo/tentec
|