On Oct 31, 2010, at 7:54 PM, Paul Christensen wrote: > Perhaps a quasi anti-LMS algorithm could be used to establish Auto APF. > Does such an algorithm now exi[s]t by name?
(For those who are not familiar with "LMS," it was an algorithm created by Ted Hoff for his dissertation, under the supervision of Bernie Widrow. Hoff went on later to design the instruction set for a chip which became the Intel 4004, the first microprocessor. Many people know of Hoff as the inventor of the microprocessor, but fewer know him as the inventor of LMS.) I really like your idea of experimenting with an auto APF that borrows from auto-notch algorithms. I think you can experiment with auto-notching very low SNR carriers to see if existing LMS algorithms work well enough to use as the core of an APF circuit. I suspect though, that you *might* need a halfway decent SNR to notch well (or fast enough). This is not a problem with existing auto notchers since they are seldom, if ever, called on to notch away carriers that are right at the noise level. If you can consistently and quickly auto-notch low-SNR carriers, then I think you are spot [sic] on Paul -- an LMS tracked auto APF when RIT is engaged! If auto-notch does not work well notching low SNR carriers, then the existing auto-notch algorithm will probably also not work well as an APF peek-seeker. I suspect that bin averaging FFTs might be better for finding the location to peak. You trade off lots of processor cycles, of course. If you look at the APF, I think it is really perfectly matched for aural reception. The typical APF is very sharp (no "flat top" portion at all in the frequency response), but the skirts are quite wide and monotonic. This allows one to hear a signal even when it is far from the peak. And since it is monotonic, you just keep turning the knob until the signal is the loudest, knowing that you are turning the knob in the correct direction as long as the tone is getting louder. Pretty much the only thing you need to be careful about with high-Q digital IIR filter structures is avoiding the "limit cycle" problem (very low amplitude, periodic output when there is no input, caused by numerical truncation errors). 73 Chen, W7AY ______________________________________________________________ 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