If I may be so bold as to offer another way in which S+N/N may be improved,
albeit with distortion; that is by using non-linear gain.
The NR, probably still being some form of Doug Smith's "adaptive predictor with
leaky LMS" design, will attenuate periodic waveforms less than random ones,
even
in the narrowest of passbands. This behavior increases the passband attenuation
noise during pauses in speech or between CW elements. When the
periodic/coherent
signal returns, the gain is restored/increased. Thus both the noise and the
signal are heard. In effect, the noise and desired signal are modulated with
the
desired signal envelope. This could very well bring about the S+N/N improvement
Grant has experienced.
This isn't the only thing at work, since the NR filter is frequency savvy, so
it
can attenuate the parts of the passband spectrum that carry random signals, and
pass those that show periodicity; hence, build a filter around the signal, as
others have observed. More than that, the stronger the periodic signal, the
less
it is attenuated. Utterances of speech (Grant, help me out with the correct
nomenclature!) contains many frequency components simultaneously, each at a
different amplitude. Since this type of filter attenuates weaker components
more
that stronger ones, the voice signal becomes distorted, but our brains are able
to recover the info, for the most part.
The thing is, how best do we use the NR to bring the desired effect about? I
found that reducing the rx gain, (by rf gain, adding attenuation, or by AGC
threshold) with no signal present until the noise is now just audible (with NR
on), gives the best improvement to the desired signal. If there is a better way
to set it for best performance, please let us know!
73,
Lin
WB1AIW
Grant Youngman wrote:
>
>
>> Grant, I don't believe the 331A is True RMS.
>> I cannot find anything on Google to confirm that the
>>331A isn't True RMS, but that's what I recall.
>
>
> You're correct. I had to dig around and find my spec sheet. Live and learn
> :-)
>
> It is RMS, calibrated to a sine wave.
>
>
>> Let's just think about your conclusion.
>>8-10 dB S/N improvement over the 100 Hz DSP BW implies a
>>10-15 Hz filter bandwidth
>
>
> I understand that, and appreciate it. The fact remains, there does appear
> to be further improvement in SNNR with NR engaged, even at narrow
> bandwidths.
>
> Perhaps someone with better lab resources will poke around at this stuff,
> too. In the meantime, I'm going to go shopping for a 339A or something
> similar :-)
>
> Grant/NQ5T
>
>
> _______________________________________________
> TenTec mailing list
> TenTec@contesting.com
> http://lists.contesting.com/mailman/listinfo/tentec
>
>
_______________________________________________
TenTec mailing list
TenTec@contesting.com
http://lists.contesting.com/mailman/listinfo/tentec
|