5752	2011-01-19 14:23:11	Chris Trask	Interference Cancelling Using Correlation Feedback
Message	Date	From	Subject
I just completed an online monologue on the subject of interference cancellation using correlation feedback and posted the paper on my web page: http://www.home.earthlink.net/~christrask/Paper024.html There are undoubtedly errors in this, but I wanted to get it completed so that I can move on to another project that has been lingering for a few years. Chris Trask N7ZWY / WDX3HLB Senior Member IEEE http://www.home.earthlink.net/~christrask/
5753	2011-01-20 04:47:28	Tim	Re: Interference Cancelling Using Correlation Feedback
Figure 13, the time constant for the loop is circa 1 ms? Does the time constant in figure 15 have to match closely (i.e. the 1% components are really necessary) or is ballpark autocorrelation good enough? I note that many of the commercial synchronous demodulators, have two different time constants, one for "tuning" (short and maybe circa 1ms for 1kHz bandwidth like yours) and one for "listening" (much much longer so that the loop wouldn't wander during deep fades of one sideband), or that the loop only has the long time constant and is simply turned off and set to the center frequency during tuning. I do not have a lot of deep listening and tinkering experience with synchronous demodulators, but I did find using them in the ham bands that some fades and worse, nearby carriers could throw off the short-time-constant loop in audibly disconcerting ways. Probably not a big loss because these sorts of demodulators are often used
5754	2011-01-20 06:11:26	Chris Trask	Re: Interference Cancelling Using Correlation Feedback
> > Figure 13, the time constant for the loop is circa 1 ms? > Yes, but it's just a starting point. > > Does the time constant in figure 15 have to match closely (i.e. the > 1% components are really necessary) or is ballpark autocorrelation > good enough? > Actually, the 1% denotation there is unnecessary, as for the two in Fig. 13. That degree of precision is only needed in the phase shifters. I'll take those out when I revise this later on. > > I note that many of the commercial synchronous demodulators, have two > different time constants, one for "tuning" (short and maybe circa 1ms for > 1kHz bandwidth like yours) and one for "listening" (much much longer so > that the loop wouldn't wander during deep fades of one sideband), or that > the loop only has the long time constant and is simply turned off and set > to the center frequency during tuning. > Yes, I've seen that done using an antiparallel pair of diodes in series with a resistor, those three in parallel with the input resistor. When the voltage from the multiplier deviates sufficiently from zero, it switches the input resistance lower so you get a shorter time constant. I'd be hesitant to use it here as the presence of a strong adjacent interfering signal might trip the time constant and make the loop seemingly unstable. > > I do not have a lot of deep listening and tinkering experience with > synchronous demodulators, but I did find using them in the ham bands that > some fades and worse, nearby carriers could throw off the short-time-constant > loop in audibly disconcerting ways. Probably not a big loss because these > sorts of demodulators are often used on channelized bands, where the next > carrier is going to be 5 or 10kHz away and thus out of the bandwidth of a 1kHz loop. > Yes, that's the problem I was thinking of. It would halp if you had low-pass filters between the demodulators and the loop multiplier, but they would have to be exact so as to eliminate any phase shift between the two. Something like that would be best implemented in DSP. For that matter, most of this could be done in DSP except for the VCXO, IF/LO phase splitters, and mixers. Chris Trask N7ZWY / WDX3HLB Senior Member IEEE http://www.home.earthlink.net/~christrask/