Dear Jim & the LF group

Jim wrote:

It would take a little care to implement though; if the drive to only one of the totem poles was varied in phase, changing the output amplitude would also change the output phase, introducing up to 90 degrees unwanted additional phase modulation into the signal. This could be avoided by shifting the phase of both drive signals by equal and opposite amounts.

You are right. I didn't think of that... I did a little trig math and came to 
the same
conclusion. Looking only at the first term of the Fourier series, the 
fundamental
frequency, the voltage difference between the totem pole midpoints is

cos(wt) - cos(wt+p)    (assuming a supply voltage of 1V...)

where w=2*pi*f and p is the phase difference between the totem poles.

Knowing that cos(a) - cos(b) = -2 * sin ((a+b)/2) * sin ((a-b)/2), I inserted
wt and wt+p and got

-2 * sin (wt + p/2) * sin (p/2)

The first sin(...) is the "oscillating part" and the second sin(...) is the 
amplitude.
It is evident from "wt + p/2" that the phase of the resulting signal is shifted
0..90 deg's as the totem pole phase diff' goes from 0..180 deg's (or perhaps
the other way around, I don't care about signs right now, phase is a relative
thing). The amplitude is indeed a function of sin (p/2) as Claudio, IN3OTD,
wrote yesterday. I can't see why an amplitude -> phase lookup table should
be such a nightmare? I think that this could work well if the series resonant
tank is given some more of vitamin Q, less C / more L.

I have the feeling that MOSFETs are probably better than bipolars in this case
since the transistors will (probably) be exposed to "backward currents"
somewhere in the cycle, even in the R+j0 tuned case. MOSFETs will happily
conduct backwards in the "ON state" while bipolar transistors in totem poles
need "antiparallel" diodes with increased losses as result.

(Yes, I am obsessed by the thought of 100% efficiency ;-)

73
Johan SM6LKM