Watson A.Name - 'Watt Sun' wrote in message ...
....
How about a helical resonator. They're smaller than a cavity, maybe
not as high Q, but still higher than lumped constant tuned circuit.

I think that's a popular misconception. The resonator Q is
essentially the same as the Q of the same part used as a shielded
inductor, and the shield actually lowers the Q from what it is with an
inductor in free air (so long as it's not large enough to radiate
significantly).

They're tunable, but I'm not sure how much.

They're certainly easily tunable over a few percent, if you need
that...

But the problem as originally stated implies a filter of fairly high
order and low in-band attenuation, which in turn implies resonators of
very high unloaded Q. 20kHz bandwidth at 40MHz in a single tank is a
loaded Q of 2000, and to keep attenuation low, the unloaded resonator
Q should be perhaps 5 times that much. It would be worse for a
multi-pole filter. All this tells me it's silly to even think of an
LC filter. Add to that the extreme difficulty of getting a set of
resonators to tune together. (To get Qu=10000 in a coaxial resonator
at 40MHz would take an air-dielectric line nearly half a meter in
diameter! Just plain silly.)

I'd opt for a front end with very high dynamic range (esp. low
third-order intermod products), into a good IF filter, etc., and a
communications protocol that optimized whatever performance measure I
needed. Talk to the people who build RF communications sytems that go
on aircraft carriers. Or talk with hams who design receivers with
third order intercepts up in the +50dBm region and higher. By the
way, you may do well by putting an ATTENUATOR on the front end, if
interference (distortion products), and not desired-signal-strength,
is the problem. Distortion products will go down faster than the
inserted attenuation.

Cheers,
Tom