On Wed, 2006-03-22 at 09:30 -0500, Bill Tippett wrote:
> Hi Gerhard,
<SNIP>
>
> >Why does a broader band of noise make the DSP NR more effective?
> Will a roofing filter reduce the effectiveness of DSP noise
> reduction, do wideband I-Q signals improve it?
>
> For the simple reason that DSP NR is
> really a bandwidth-narrowing device.
NOT AT ALL!
> If the
> bandwidth is already narrow, DSP cannot
> improve upon the situation. This is exactly
> what the manual is referring to in the last
> sentence quoted above. If BW is already
> narrowed (either by roofing or DSP filters),
> NR cannot make much, if any, improvement.
All the good NR techniques use the high correlation of a desired signal
and the relative lack of correlation of noise to discriminate between
the two. At least that's what the books about NR that I've read said.
The narrower the preDSP filter, the more correlated the noise. Noise fed
to a 200 Hz crystal filter almost has a tone (more correlated than the
pitch of an operatic soprano or tenor) which to my ear leads to rapid
fatigue copying CW. Because I'm trying to discriminate keyed tones in a
sea of broad tone.
Any bandpass filter effect is secondary to the correlation techniques
that look for a lack of randomness in the incoming signal to
discriminate against noise to enhance the signal.
>
> >Why shouldn't we use a different NR techniques for different modes?
>
> I believe the technique is the
> same but the minimum bandwidths are
> very different depending on the mode.
> Narrowing bandwidth improves S/N because
> you filter noise out of the signal plus
> noise. For a given signal plus noise,
> every time you reduce the bandwidth by
> a factor of two you reduce noise by 3 dB
> and therefore improve S/N by 3 dB. When
> you reach the BW limit determined by the
> fundamental information bandwidth of the
> specific mode, there is no more that DSP
> can do.
But that is a filtering mode, NOT a NR mode.
> I am not sure what lower BW limit
> Orion's NR assumes for each mode, but I
> suspect it is similar to the basic DSP
> BW limit for CW (i.e. 100 Hz). SSB may
> be in the 1700-1800 Hz range but I have
> really not played with it enough to know.
> 1500 Hz is about the limit I find manually
> but it requires some major adjustment of
> PBT which I suspect the DSP NR does not do.
>
> In summary, I've said this before
> but will repeat it here again:
>
> ***NR will not do anything that adjusting DSP
> BW and Gain controls will not do faster and
> more directly.***
>
True when you know where and what the signal is to adjust the BW
controls to find.
> If anyone still questions my statement, please
> again read the last paragraph from the Orion
> manual quoted above. The only reason the
> above might not hold true would be if Orion's
> NR actually creates a filter of less bandwidth
> than can be done by reducing the BW control.
>
> Some of us want to find a way to
> circumvent the basic relationship between
> information, bandwidth and noise described
> by Nyquist, Shannon and Hartley at Bell Labs
> in the late 1940's, but there is simply no free
> lunch, DSP or no DSP, advertising hype (hello
> Icom!) or not. If you want to learn more, there
> are plenty of good references available on the
> Internet by searching "information bandwidth
> noise" and adding one or more of the above
> three names.
>
> 73, Bill W4ZV
>
The benefit of the DSP BW filters is that by using many coefficients,
the bandwidth can be narrow without the ringing inherent in the simpler
analog filters. The rub in practical filters that each added filter
coefficient adds delay and the total number is generally chosen to be a
compromise between delay and ringing and the narrowest bandwidths. By
adding more coefficients (and delay) the narrower the filter can be made
without ringing. The added delay and the narrower the filter makes
tuning increasingly difficult.
--
73, Jerry, K0CQ,
All content copyright Dr. Gerald N. Johnson, electrical engineer
_______________________________________________
TenTec mailing list
TenTec@contesting.com
http://lists.contesting.com/mailman/listinfo/tentec
|