Hi James, I'd have to say that your post and the links to what you consider improvement firmly make the case for how subjective this is. Some folks simply cannot stand to listen to the "warbling" effect caused by the NR algorithm moving around the center of the "passband" to match the speech centers and reject the bands with noise only. What you are asking for is what they can't stand... K2 NR has that sound.
The great problem of noise reduction design is the difference in the individual GOALS of using NR. For some, this is to remove the IRRITANT of noise, and for others to remove the UNINTELLIGIBILITY caused by noise. The latter goal produces different methods than the former. The former is doomed from the start because people's hearing range differs with SOME wide hearers intensely annoyed by high frequency noise, others used to listening to 80 meter QRN in the summer may want reduction with ZERO loss of weak signal, and others are very ANNOYED by reduction ARTIFACTS. For me it's just like anything else going on in firmware, I hope Wayne works on the stuff I want first, and that he keeps working on stuff. My list does not have NR improvement near the top. That's because there is a limit to what NR can do that does not mask very weak signals when it's on. NB is another story, and the t3-7 or t2-7 with sharp skirts effect on key clicks has allowed me to copy many signals where copy was not otherwise possible. I will have to agree with you on the volume issues, because I always NOTICE it. BUT, that falls in the IRRITANT category, not the INTELLIGIBILITY category. So I walk on by and wait for Lyle to figure out how to nullify plasma noise. I'm not holding my breath, but there really is a part of me that thinks he might. BTW, I turn NR *OFF* in contests, period. There is not a single NR setting that I cannot demonstrate a very weak signal loss, that occurs where the NR starts to blend with the noise. My ears, and everyone else's, will do a better job than the NR at that point. And particularly, now that we have APF for CW, hearing that tiny peep is all the clue I need to zero in the APF and get it out of the noise. NR on, and I don't even know it is there. I don't think anyone really knows (federal court worthy peer-reviewed scientific proof) how our ears hear discrete signals that are blending in the noise, but I have a suspicion. How many reading this have been driving home after a weekend CW contest and hear CW in the road noise? Our brains have some kind of heuristic anticipatory mechanism. It's the very one that allows me to copy the "QRP" in DL5QQ/QRP and struggle with the rest to my utter annoyance as he sends the /QRP over and over again. (Talk about irritating.) However it works, it is more effective than anything electronic by some orders of magnitude, and NR strips out what makes it work at the very point that the brain/ear makes the greatest difference. Others may disagree in a contest, just because they want to operate in comfort. That's fine by me, just don't apply for a seat in our WFO NY4A efforts, where our getting ALL the very weakest stuff is our score differential over the competion. If you work 2K+ contacts on 40m in the ARRL DX CW, you can be guaranteed that three or four or five HUNDRED of those contacts are too-far-away QRP, radio-trash crap antenna weak, or at the very edge of propagation where they hear your 1.5 kW above their noise, but their 100w and built-in -12 dB disadvantage puts them firmly IN the noise. Turning on NR in a CW contest can cost you hundreds of QSO's at a competitive station. ANYONE can work loud and medium stations. Try hard and most can work moderately weak. The last layer is only solved by the human brain, and the SOUND of it is irritating and tiring. Our in-the-noise hearing is likely honed by millions of years hearing the movement or breathing of a sabre-tooth tiger before it could spring. In that environment the "false positive" of my hearing CW in the road noise, e.g. hearing a sabre-tooth when there really wasn't one, was easily tolerated in favor of having one or two steps moving away before a real tiger charged. What a noise reduction designer is up against is nature honing something absolutely essential for millions of years, vs analysis in a very limited computing environment. And up against a wild variation in customer taste and tolerance for things audio. I'd say that if a box makes something sound like you like it, don't toss it until the new thing demonstrates what you like. There is no natural law that states that superior results for all problems will always be obtained processing at IF in a K3. They're doing it there because it fits in the scheme of all they're trying to do, and the means is commercially limited. On whole they've succeeded brilliantly, but there simply MUST be SOME things they can't do as well that way. 73, Guy. On Sat, Nov 27, 2010 at 7:16 AM, James Sarte (K2QI) <k2qi....@gmail.com>wrote: > Hello Elecrafters, > > Regarding the issue of NR, I know this can be a touchy subject and most > will > say its very subjective. This post isn't intended as a complaint or > criticism. Instead, I hope through constructive discussions, more > improvements to the K3's NR performance will be realized. I'd like to begin > by saying that Lyle and crew have a done a fantastic job to date with the > K3's NR. I've had the privilege of watching the K3's NR function improve > over time. The boys at Aptos should really be commended for listening to > their customers. > > With that said, the recent CQSS has made me realize that there is still > room > for improvement. During the SS, I used NR extensively in combination with > RF gain and filter hi-cut/lo-cut/shift settings. NR when used in > combination with the aforementioned K3 adjustments can and does work well. > The problem that I believe remains however is the algorithm used still > seems too broad in its rejection calculations; what I mean is that SSB > settings (i.e. F5-1 and higher) don't seem to be selective enough to reduce > noise while allowing speech patterns to remain unaffected. In other words, > the DSP sounds like its reducing everything within its passband. To my > ears, the NR doesn't seem to make speech "pop out" quite as effectively has > other NR implementations. This can be seen by a not-so-subtle reduction in > speech volume whenever the NR is turned on. The NR behavior is consistent > regardless of AGC slope or threshold settings. > > Below are several links to audio recordings of a product I used to own. It > is called the BHI ANEM (Mk. II). Sadly, I sold it after getting the K3. I > should have hung on to it as it has become my benchmark for NR performance. > > These are some sample recordings of the ANEM being turned on and off. > These > recordings are found on W4RT's website > > 20 m SSB <http://www.w4rt.com/BHI/20Mband-ssb.wav> > 80 m SSB <http://www.w4rt.com/BHI/80conv-ssb.wav> > 80 m SSB <http://www.w4rt.com/BHI/80Mconv2-ssb.wav> > > I've also compared the K3's NR performance to that of the Icom 756 Pro 3 > which has similar NR reduction properties - albeit not as effective - as > the > ANEM. > > Perhaps I am being subjective or overly critical, but one thing you can > easily notice with the ANEM recordings is that speech volume doesn't get as > affected as the K3s. This is really surprising to me as the NR for the > ANEM > is AF rather than IF like what's used in our radios. This leads me to > believe that the ANEM's NR algorithms are more effective in reducing > background noise while leaving speech unaffected. > > What do you guys think? > > Sorry for the long email. Many thanks in advance for your > ideas/suggestions. > <http://www.w4rt.com/BHI/20Mband-ssb.wav> > -- > 73 de James K2QI > President UNARC/4U1UN > ______________________________________________________________ > Elecraft mailing list > Home: http://mailman.qth.net/mailman/listinfo/elecraft > Help: http://mailman.qth.net/mmfaq.htm > Post: mailto:Elecraft@mailman.qth.net > > This list hosted by: http://www.qsl.net > Please help support this email list: http://www.qsl.net/donate.html > ______________________________________________________________ Elecraft mailing list Home: http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/mmfaq.htm Post: mailto:Elecraft@mailman.qth.net This list hosted by: http://www.qsl.net Please help support this email list: http://www.qsl.net/donate.html
