Re: [Elecraft] [K3] Passband flatness for FSK441
Bill W5WVO wrote: In WSJT's FSK441 screen, there is a useful little window that plots * the amplitude of the noise floor and the decoded signal burst, if any, * against audio frequency. Since FSK441 uses four tones ranging in * frequency from 882 Hz to 2205 Hz, the net passband (after all stages) * should ideally be flat from about 850 to 2250 Hz, given good signal It needs to be substantially flat across each sub-carrier, but could slope across the band. That means that a sharp cutoff at the edge may be worse than having significant passband ripple. The reason for having multiple sub-carriers is to better tolerate frequency selective fading, and group delay variations, so the modulation method is already designed for total channels which do not have a flat frequency response. Incidentally, if the tones are spaced by 441Hz and there is no wasted spectrum, the total passband should be more like 661.5 to 2424.5. * frequency accuracy. However, the WSJT FSK441 software is designed to * find and decode signals as much as +/- 400 Hz from the nominal receive * frequency, which extends the requisite passband from around 450 to 2650 Hz. The sub-carrier frequencies contain an arbitrary offset that has been chosen to put the group in the middle of the passband of typical SSB filters. That's the only reason that the particular frequencies are chosen. You could equally consider them to be --661.5, -220.5, 220.5, and 661.5 Hz. The reason that the decoder can cope with up to 400 Hz off frequency is probably because there are receiver filters around that will allow that, but I don't see that that is a reason for always making the receive filter that wide. The system has been clearly designed to work with filters that are well behaved over a much narrower frequency range. I suspect the 400Hz slop is based on a 2.7 kHz filter, allowing some slop for carrier offset errors and on the basis that the upper and lower channels will be degraded by the filter roll off. Is the K3's passband flat over this range of audio frequencies using * the 2.7/2.8 kHz filter, or if not, can it be made effectively flat by * the use of custom equalization settings? I should think the answer would * be yes, but I'd like to hear from somebody who has perhaps done it and * seen the resultant flat noise profile in the FSK441 display. You also need constant group delay across each sub-carrier, and unless the decoder de-skews the individual sub-carrier streams, across the whole group. If the equaliser users infinite impulse response filters, it may actually exacerbate group delay problems (although it might also cancel some of the group delay variation from the IF filters). If the decoder is good, it will have an adaptive filter that will automatically equalise the signal for good digital decoding. [ Single line paragraphs have been arbitrarily wrapped.] -- David Woolley The Elecraft list is a forum for the discussion of topics related to Elecraft products and more general topics related ham radio List Guidelines http://www.elecraft.com/elecraft_list_guidelines.htm ___ Elecraft mailing list Post to: Elecraft@mailman.qth.net You must be a subscriber to post to the list. Subscriber Info (Addr. Change, sub, unsub etc.): http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/subscribers.htm Elecraft web page: http://www.elecraft.com
Re: [Elecraft] [K3] Passband flatness for FSK441
...If the equaliser users infinite impulse response filters, it may actually exacerbate group delay problems (although it might also cancel some of the group delay variation from the IF filters). It does not. The Rx (and Tx) equalizers are entirely FIR-based. 73, Lyle KK7P ___ Elecraft mailing list Post to: Elecraft@mailman.qth.net You must be a subscriber to post to the list. Subscriber Info (Addr. Change, sub, unsub etc.): http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/subscribers.htm Elecraft web page: http://www.elecraft.com
Re: [Elecraft] [K3] Passband flatness for FSK441
David Woolley (E.L) wrote: The reason that the decoder can cope with up to 400 Hz off frequency is probably because there are receiver filters around that will allow that, but I don't see that that is a reason for always making the receive filter that wide. The system has been clearly designed to work with filters that are well behaved over a much narrower frequency range. Probably so, but it's useful, operationally, to start out with the filter tolerance set to 400 (its max value), because transmitters and receivers are not usually exactly on the same frequency. This TX/RX frequency offset problem is typically worse on the higher-frequency bands (144 MHz, 222 MHz, even 432 MHz). Once a signal has been detected, RIT can be used to bring the received signal close to the transmit frequency, and the WSJT tolerance figure is then decreased to a more appropriate level. This filters out a lot of false decodes on noise, especially when working with a very marginal signal. If the decoder is good, it will have an adaptive filter that will automatically equalise the signal for good digital decoding. All of this is somewhat speculative. :-) Joe Taylor, to the best of my knowledge, isn't on this list. (If you are, Joe, speak up!) And I certainly don't pretend to understand either filter physics or decoding algorithms as well as you obviously do, so I really appreciate your technical explanations. What I am looking forward to, once I get my K3, is experimenting with filter width and equalization settings (versus no equalization, of course) to see if I can improve the decoding of extremely marginal signals, where the worst-case upper channel is at or just above the noise floor. It could be that Joe's algorithms are so powerfully adaptive that nothing one can do at the analog RX end will make things any better. That wouldn't really surprise me. But meteor-scatter is extreme weak-signal work, and every half-dB you can gain on the noise floor is effort well-spent. Of course, the very best thing you can do in that respect is move QTH out to the middle of Kansas somewhere. :-) Bill W5WVO ___ Elecraft mailing list Post to: Elecraft@mailman.qth.net You must be a subscriber to post to the list. Subscriber Info (Addr. Change, sub, unsub etc.): http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/subscribers.htm Elecraft web page: http://www.elecraft.com
Re: [Elecraft] [K3] Passband flatness for FSK441
Of course, the very best thing you can do in that respect is move QTH out to the middle of Kansas somewhere. :-) Jericho, Kansas? ___ Elecraft mailing list Post to: Elecraft@mailman.qth.net You must be a subscriber to post to the list. Subscriber Info (Addr. Change, sub, unsub etc.): http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/subscribers.htm Elecraft web page: http://www.elecraft.com
Re: [Elecraft] [K3] Passband flatness for FSK441
Yeah, that sounds like a good place! :-) (I'm a big fan, too.) Bill W5WVO Lyle Johnson wrote: Of course, the very best thing you can do in that respect is move QTH out to the middle of Kansas somewhere. :-) Jericho, Kansas? ___ Elecraft mailing list Post to: Elecraft@mailman.qth.net You must be a subscriber to post to the list. Subscriber Info (Addr. Change, sub, unsub etc.): http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/subscribers.htm Elecraft web page: http://www.elecraft.com ___ Elecraft mailing list Post to: Elecraft@mailman.qth.net You must be a subscriber to post to the list. Subscriber Info (Addr. Change, sub, unsub etc.): http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/subscribers.htm Elecraft web page: http://www.elecraft.com
[Elecraft] [K3] Passband flatness for FSK441
In WSJT's FSK441 screen, there is a useful little window that plots the amplitude of the noise floor and the decoded signal burst, if any, against audio frequency. Since FSK441 uses four tones ranging in frequency from 882 Hz to 2205 Hz, the net passband (after all stages) should ideally be flat from about 850 to 2250 Hz, given good signal frequency accuracy. However, the WSJT FSK441 software is designed to find and decode signals as much as +/- 400 Hz from the nominal receive frequency, which extends the requisite passband from around 450 to 2650 Hz. With my TS-B2000, I cannot achieve this. Using the widest possible filter selections and the best possible RX equalization and DSP settings, the best that can be achieved is a sort of flat-ish hump-backed curve, with both ends rolled off from the center by a dB or two, the high end being rolled off more than the low end. It still works (on signals of adequate strength), but it is far from ideal. Is the K3's passband flat over this range of audio frequencies using the 2.7/2.8 kHz filter, or if not, can it be made effectively flat by the use of custom equalization settings? I should think the answer would be yes, but I'd like to hear from somebody who has perhaps done it and seen the resultant flat noise profile in the FSK441 display. Bill W5WVO ___ Elecraft mailing list Post to: Elecraft@mailman.qth.net You must be a subscriber to post to the list. Subscriber Info (Addr. Change, sub, unsub etc.): http://mailman.qth.net/mailman/listinfo/elecraft Help: http://mailman.qth.net/subscribers.htm Elecraft web page: http://www.elecraft.com