Hi Grant,
Just a thought- I assume you made these measurements with a steady state signal.
Well that will be the worst case for a downward expanding type of NR.
However, if the carrier was modulated (like in Sinisa test with continuous
dits), and the RF gain adjusted so that the difference between the signal plus
noise level and the noise level with no signal is maximized, the average SNNR
would be at it's best (by definition). Seems to me, this test is more
applicable; more "real-world".
Wouldn't that be a better test? Let me know what you think.
Almost forgot-I'm assuming a narrow bandwidth, since you have shown quite nicely
that, by itself, the v2.xxxx NR doesn't do much in this regard.
Lin
Grant Youngman wrote:
BW=500Hz: SNNR degrades by approx 2dB with NR=9.
Degrades?...strange result! This
implies a noise bandwidth of ~800 Hz. Simply
turning DSP BW to 100 (which has an actual BW of 150 Hz)
should **improve** S/N by 5.2 dB.
Let me clarify a point here.
What you say is true. However, that isn't exactly what I was looking at.
The absolute value of SNNR does in fact increase, with NR off, and the
bandwidth narrowed as it should But the numbers I gave were the DIFFERENCES
between SNNR with NR off and SNNR with NR=9.
That says that (at the 500HZ bandwidth for example), while the absolute SNNR
increased with narrowed bandwidth and NR off, the "improvement" by turning
on NR was negative. I don't know why that happens, just that it does. It
may be distortion products as Buck suggested.
In any case, I've been arguing all along that the only valid figure of merit
for any NR system is how it changes SNNR, not whether or not the radio
becomes "quiet". At least to my own satisfaction, I don't believe the new
algorithm is doing much relative to that figure of merit. With the signals
set to be under (as far as I can tell anyway) the AGC threshold, it is true
that when NR is turned on the radio gets quiet. Unfortunately, the signal
level decreases right along with the noise level. So I normalized
everything to the value of the signal peak, with and without NR. In that
case, and with a relatively weak signal -- (S+N) set to provide a 3dB
increase over (N) with no NR as measured on a 3400A True RMS meter -- this
very small change up or down is SNNR is what NR seems to be providing.
Grant/NQ5T
_______________________________________________
TenTec mailing list
TenTec@contesting.com
http://lists.contesting.com/mailman/listinfo/tentec
|