FWIW, just now I had not trouble at all filtering the fifth out of a
square wave, in exactly the way I suggest below. It seems very silly
to me to put nonlinear elements in when you already have plenty of the
harmonic you want. I don't know what logic family you're using, but
if it's a modern one like 74AC, you should have several milliwatts of
fifth harmonic available. A simple series resonant LC from the logic
output to the base of a 2N2219-type transistor should get you at least
a couple mA RMS of fifth harmonic base current, assuming a
grounded-emitter stage with roughly 50 ohms input resistance. Make
the loaded Q of the LC something around 10 to 20, and you won't screw
up the amplifier with other harmonics or the fundamental. Be a bit
careful about the coil you use, because it will be fairly high
inductance for the frequency you're interested in...that is, keep the
unloaded Q and the self-resonant frequency high enough. Use another
moderate-Q tank in the collector circuit; you should be able to get
over 100mW of fifth, with other harmonics down 40dB or more. You can
get more complicated with the filtering if it's necessary, but for
fixed-frequency operation, there's nothing wrong with simple
synchronous tuning of single resonators set to reasonably high Q. If
you insist on using the unnecessary complication of a multi-pole
bandpass filter, be sure the one from the square wave to the amplifier
starts with a series resonator, not a shunt resonator.

Sheesh...all you need to do is selectively amplify the harmonic you
want; it's already there. Don't add complexity trying to generate
something you already have in abundance.


Paul Burridge wrote in message . ..
On Mon, 15 Mar 2004 20:03:24 +0000, John Woodgate
wrote:

No, what has emerged from the discussion is that rather small deviations
from a perfect square waveform can drastically reduce the amount of one
or more harmonics in the spectrum, and this probably is the source of
the problem.

Not this particular problem, it isn't!