On Thu, 2006-05-04 at 14:34 -0400, Gary Hoffman wrote:
> Well....you are asking for an opinion...always a dangerous thing ! <grin>
>
> But....
>
> In my opinion, if one could choose, in the case of SSB (which is
> comparatively wide band, but highly correlated) the better choice would be
> the correlation based filter. This removes noise Inside the pass band.
> Since SSB needs a fairly wide pass band, then band pass based filters tend
> to be limited in effectiveness.
>
> For CW the choice is less clear. CW is narrow band, and highly correlated.
> But it is also very rapidly choppy. So, a narrow pass band filter can do a
> very good job. So can a correlation based filter, but it might have trouble
> tracking the very high levels of chop. Might make a mess of the signal,
> depending on how well it is implemented. So, I might tend to want to choose
> a very narrow band pass filter for that case.
>
> I guess I just said I want both....didn't I ? <grin>
>
> If I could never the less have only one....I'd have to choose the
> correlation filter, because I do a lot of SSB work, as well as some wider
> digital stuff like RTTY and PSK (yes, I know those are kinda narrow
> compared to SSB, but wider than CW too).
>
> In any case, any radio I would buy would likely have narrow band pass
> filters available anyway, so the correlation filter would add the greatest
> Incremental value, for me.
>
> YMMV.
>
> Regards,
>
> Gary, AA2IZ
>
Filtering is easy for ONE CW signal, but how about three or 17 in the
passband. Sure for working one you would want to filter out the rest,
but say you were at an antipodal DX pedition location and could copy
three or four of those 17 at a time. You wouldn't want the NR to only
recover the one you had already worked (or rejected as not worthy of
being worked) so you need effectively several independent narrow
bandpass filters at the same time with parallel inputs and outputs. Then
it starts to get messier with multitone digital signals where you DO
want to copy them all at the same time and the external computer
software does that. But its unacceptable to have the NR software in the
radio leave only 4 of the 17 tones...
Time is the biggest problem. Given a long enough time, error correcting
codes in the transmitted data, and enough repeats DSP can do phenomenal
things copying signals from probes out beyond Uranus with received
signals at -250 dBm from Jodrel Bank sized receiving antennas. But more
than 100 milliseconds delay in a manually tuned receiver and we
overshoot when tuning by ear and that gets to be so annoying to the user
that the DSP gets turned off.
Latency is partly a function of DSP speed, but also depends on sample
rate (more processing needs more samples) and the repetition rate of the
signal. Once a sine wave has been sampled often enough (at least twice
per cycle) to be reproduced it don't necessarily help the S/N
improvement process by sample it 48 times per cycle.
Noise reduction techniques that work well for Johnson noise from
circuits may not work so well for noise from spherics because the noise
from spherics that hit the antenna as a dirac delta function gets
converted to a ringing envelope by the bandpass filters of the radio,
and in some classical filters gets converted to a continuous almost
unmodulated envelope of noise that is more than a little correlated by
the ringing filter. Hash from switching circuits (whether power or
computation) tends to have even better correlation making noise
reduction by autocorrelation more difficult. And in the REAL world we
will see ALL of these noise sources plus power line noise (like spherics
but more repetitive) and a mix of desired and undesired real signals at
the same time.
--
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
|