Topband: high power filters

[Tom Rauch]

> You normally want an inductor input filter so that the FET
can
> clip if it wants to, but no current will be drawn at
harmonic
> frequencies.  For a given peak to peak voltage, a square
wave
> has 16/( pi squared) = 62% more power in the fundamental
that
> a sine wave does, using Fourier series analysis.  This
gives
> you a lot more headroom.  This seems to be a little known
> fact, since I often see capacitor input filters used.

The reason you see capacitor input filters used is the
overwhelmingly common transient damage mechanism of an FET
is excessive voltage.
FET's can take dozens of times rated current on peaks, but
will fail immediately with excessive voltage.

If you build the FET amplifier board using MRF150's by Helge
Granberg of Motorola you'll find occasional failures of
FET's on higher frequencies because of series inductance
letting the FET's ring on harmonics of the waveform as the
FET's transition between cutoff and conduction. The best
cure for this problem is moving shunt capacitance right up
next to the FET drains.

The last thing I would do with an FET amplifier running near
rated drain-gate breakdown voltage is load the FETs with
inductance on harmonics, or use a high Q filter! Efficiency
would increase, but FET reliability would hit bottom.

You can see how high transits get with a modest Q network
and mismatched load at:
http://www.w8ji.com/demonstation.htm

The correct filter in almost all conventional linear
amplifier applications (other than switching systems) is one
with low operating Q that has a very low input impedance for
harmonics.

73 Tom