[Orion] V2.059d Vs. v1.373b5 Noise Reduction Code

[Merle Bone]

Hank (K7HP) said:
"  I think in my case the NR just does very little for me because I operate CW 
> with bandwidths at 100 to 200 cycles 99 percent of the time. And if we take 
> W4ZV's analysis - once you force narrow bandwidths , the NR has little to 
> improve on."
--------------------------------------------------------------------------------------------------

I've look pretty close at the NR function in V1.373b5 and the V2.0XX (Including V2.059d) firmware releases.
Others on this and other lists have as well (Maybe it gets a lot more attention then it deserves based on real functionality). I sent a full set of spectrum analyzer screen shots of both (V2.057 and V1.373b5) to Jack Burchfield as part of a letter exchange he and I were having. 

In V1.372/3 (And I think earlier) it appears there were three different kinds of filters employed in NR. On setting 1 either a low-pass filter or a high-pass filter was constructed - depending on where the greatest energy was in the audio bandwidth. This is the best filter for use in the SSB mode because it is built quickly and "collapses" quickly as the signal appears and disappears. This minimizes the distortion to the audio as the filter is built and eliminated between words. In positions 2 through 9 a bandpass filter is created. The higher the number, the steeper the skirts of the filter. These positions are probably best for use in CW mode. On the highest number the filter built will be 200Hz to 300Hz wide at down 3-6Db. Because of the number of "taps" in these filters they take progressively longer to create - in terms of time. 

In V2.057 and V2.059d a lowpass filter is created in all cases. The higher the number, it appears that more taps are created and the skirt gets steeper. There is also a significant amount of attenuation inserted - particularly at positions 8 and 9. It looks like the down 3DB point on the LP filter on 2.059d is somewhere around 1400Hz. 

It is very difficult to say what people will hear as "a help" or "no help" with a function like NR. Because we mostly use this function with signals that we listen to with our ears, individual "audio perception" will play strongly into the equation of effectiveness. It is very likely true that if you operate CW with a very narrow filter - 100-200Hz - you probably won't find any help from NR. On the other hand, on SSB with a 2.4KHZ filter you may perceive that the noise is reduced and there is "less fatigue" in operating SSB. Gary Barber characterized it as a very "subjective improvement."

I get more help from the V1 NR code but it really is at the "margin" for me.  I really continue to use V1 because of the Notch and Auto-Notch filters that remove the signal from the AGC so there is no AGC pumping (Like in the V2 code) and the significantly better Speech Processing (audio quality and real increase in average talk power over the V2 code). I do think the V2 code for the Orion is getting better and hope there will continue to be improvements.
Merle - W0EWM