> And if anyone wonders how good NR "could be," load ver 1.371
> and give a listen. That was the last time the NR in the
> Orion worked and it worked really well. It was dramatic,
> almost squelch-like.
The purpose of NR is to increase SNR. It may be nice to have "quiet" copy,
but if SNR doesn't improve, you will have quiet nice quite copy of nothing!
You may recall I measured in some detail the NR performance of the early
(e.g, 1.371, 1.372, 1.373b5 Orion). They were all the same. There may be
other characteristics driving someone to prefer one version over another,
but SNR improvement IS NOT one of them. At 500Hz bandwidth NR added 1.0 dB
(measured, and I'd argue 'heard' = NADA) to SNR. At 100 Hz, it added 0 dB
to SNR. So one may like it for various reasons, but it certainly doesn't
help copy anything at narrow CW bandwidths. At wider bandwidths (3 Khz) ,
it does INCREASE SNR by up to almost 7dB. This is related to my earlier
comments (ignored and tossed off as superfluous because clearly Icom had an
epiphany with the 7800) about noise bandwidth, and what you can expect NR to
do when the IF bandwidth is already very narrow. The early Orion II NR was
neutral or slightly negative (about -2dB, again I'd argue virtually
inaudible) in SNR improvement depending on bandwidth, and was clearly broken
compared with the 1.xx algorithms. 2.037j changed things again, but I have
not had the time (and frankly just haven't had the inclination) to measure
it against the others.
I will say that, on SSB, the Orion II NR is considerably better than the
1.xx Orion NR. In my opinion of course, listening. Far less "muffling"
(high frequency reduction) in the II than the original (any firmware version
you care to compare it with). But that's because the Orion One NR created
bandpass filters to reduce noise bandwidth. Early Orion II NR didn't do
that, but my anecdotal observation in 2.037j was that a bit of that had been
restored.
What I will do, is when my broken II comes back from T-T, is measure the SNR
impact of NR on the latest II firmware -- in a manner consistent with my
earlier measurements -- and publish it here. Maybe I'll do it on my v2
Orion with whatever the latest release is, but that would just become
another point of contention.
And for the record, I'm not defending T-T. But do you really want to pay
for a multi-processor Cell-architecture machine to do perfect 'NE' (Noise
Elimination) so you have a small enough processing delay to still be able to
work QSK? Frankly, the NR button itself, whatever it does or doesn't do, is
a source of contention by just being there and guaranteeing inflated
expectations as a result :-)
Grant/NQ5T
_______________________________________________
TenTec mailing list
TenTec@contesting.com
http://lists.contesting.com/mailman/listinfo/tentec
|