Paul Burridge wrote in message . ..

Exactly how much power do you need?

Only enough to feed another inverter gate.

Egad, Paul! You've been wasting this much net bandwidth just to drive
another HC gate?? All you need is a filter/matching circuit that
steps up the voltage. This is DOG SIMPLE! See below.

Exactly how "clean" (free from
other harmonics) must it be?

Preferably filthy. It's another multiplier (this time only 3X, thank
God!)

Then you need a clean enough input that you'll get the desired output
purity. "Filthy" is likely NOT the right answer and will just get you
into further trouble. But fortunately, "clean" is simple, and "really
clean" isn't at all difficult.

Try this: square wave output -- I don't recall your exact freq; I
used 3.7MHz -- from HC gate, feeds 4.58pF capacitor (make at least
that one tuneable). Other end of cap feeds 20uH inductor, Qu=200.
Other end of that inductor connects to next gate input, and net 18.6pF
of capacitance to ground: say 15pF cap plus 3.6pF of gate input
capacitance. For DC bias, gate input to ground = 22kohms; gate input
to Vcc = 47kohms. That keeps the gate in a valid logic state when
there's no excitation. Assuming the gate's RF input resistance at
18MHz is at least 2.5kohms, you should get a voltage gain at the fifth
harmonic of about 15dB, which will be ample to drive the gate input.
The available current from the filter is low enough that the gate's
input protection diodes should clamp things nicely at the rails. Be
sure to use a gate that has input protection, or else add
low-capacitance, fast diodes externally. Gain at the third and
seventh is down 20dB or so from that. If it needs to be cleaner than
that, you can add a second resonator.

The gate biasing suggested may result in an output duty cycle
significantly different from 50%. If you will always have 3.7MHz
drive, you can bias the input more in the center of its range, or even
rearrange the circuit a bit and use a feedback resistor from output to
input to set the DC bias. The gate's input impedance is then much
lower, but you don't need much voltage to drive it. Don't use that
trick with a Schmitt trigger input, though.

69 turns of 36AWG (0.125mm) wire, spaced 2 wire diameters c-c, on an
0.375" former, should give you about 20uH at Qu=200 and first parallel
SRF about 50MHz, but you should be able to make it more compact using
something like a T-50-2 powdered iron core.