Radio Digital Signal Processing program discussion, originally under Linux ()

Rein A. Smit | 3 Mar 21:02

Re: Noise Reduction

Hi Leif,

Thanks for the message. I will have to go back to your documentation again
in order to ask you the right questions. In partcular how to determine 
whether
my front-end  the TS850 is a limiting factor in my environment or not.

My concern is not so much the blanking of pulses but the suppression of 
noise
I mean the blanking of pulses as in lightning, line noise and the like. 
As this is not
much of a problem here, I never paid much attention to that aspect of  
Linrad,
probably the real strong side of linrad.

It is really the rumbling noise what bothers all of us in the end.

I guess there is on the one hand the ability to do weak signal in a
highly polluted RF environment with strong undesired signals, possibly
overloading our receivers and on the other hand, the rumbling  noise
 that seems to be  inherent with the processing of  noise and what
prevents us from hearing the dots and dashes as dots and dashes.

Then there is also of course selectivity for desired sigals and birdies.

More to follow if I may.... .

73 Rein

leif@... wrote:

(Continue reading)

headers

leif-NjylAwyOilXQT0dZR+AlfA@public.gmane.org | 3 Mar 15:39

Re: Noise Reduction

Hi Rein,

> So far I have always used linrad and winrad with output from a 
> ssb receiver ( 1.8 kHZ filter ) not realizing that this might 
> not be the optimum conditions for the noise reducer and noise 
> blanking to perform..

It depends. The Linrad blanker may work perfectly well if the
noise pulses are separated by more than 0.5 ms or so. (In contrast
to simple blankers that do not compensate for the phase response
of the SSB filter.)

The Linrad calibration procedure sets up a filter that corrects
for sub-optimal amplitude/phase response of an analog receiver.
Some pretty old info is available here:
http://www.sm5bsz.com/pcdsp/blnini.htm

With Linrad one can set the sampling speed at just about twice 
the upper audio frequency. For 300-2100 Hz, sample at 5kHz
(if your soundcard can go that low) The calibration routine
will make noise pulses extend one or two 2 samples, typically
5 times shorter than in the original audio.
(see http://www.sm5bsz.com/linuxdsp/pulresp/pulresp.htm )
Energy is conserved so the pulses also become much higher 
and easier to locate. Linrad will subtract the weak 
oscillations that belong to the short pulse. This is a 
linear operation that does not introduce intermodulation 
or loss of information.

The process works ONLY if the radio has adequate dynamic range

(Continue reading)