[TowerTalk] Using Stubs to Reject Harmonics

[Jim Lux]

-----Original Message-----
>From: Rick Karlquist <richard at karlquist.com>
>
>
>What Tom says is right on the mark, as long as the last component in
>your amplifier is a capacitor to ground, which is probably the case
>90+% of the time.  In Jim's case, the last component is a series
>inductor, in which case Tom's advice needs to be modified to
>placing the stub at the amp, or a multiple of a halfwave away.
>All of this confusion is yet another one of many reasons, IMHO, to build
>filters out of lumped components instead of coax stubs.  And
>the filters should be elliptic function low pass filters, not bandstop
>filters.  For examples of these filters, study the output filters
>in any solid state transmitter.  For example, see pages 3 and 4 of:
>

I'll go for the lumped design (certainly easier and more predictable), and the recommendation for Elliptic/Cauer (especially if you need some notches, and don't care about ultimate rejection far away). 

I've looked at transmission lines as a replacement for lumped L and C a bunch of times (mostly for making phasing and tuning networks), and at HF, I'm pretty sure that the loss of a TL stub is worse than even cheap Ls and Cs. One advantage of a C made from coax is that it's pretty inexpensive and stands off high voltage, and it's empirically adjustable with a pair of cutters. The main issue with loss is the small conductor size, so the IR losses are there.

In a tuned stub, the low Q might actually be an advantage.  You get broader bandwidth at the expense of ultimate rejection.  If you don't need a gazillion dB of rejection, that might be a good trade.  I haven't thought it through, though.

However, most of the filter designs assume that they're working into a resistive termination on both ends, and a tuned output amplifier or a resonant antenna isn't that.  I suppose you can do the filter synthesis with a suitable model of the termination impedances.  I guess for sufficiently out of band, the Z tends to some reasonable asymptote.  Or, perhaps, in a practical sense, it works "good enough" even if the end sections of the filter don't have quite the terminating impedance.

Hmmm.. maybe you need an adaptive canceller for the interfering signal?

Jim