If you want to know what the inventor has to say about "class-I", you can easily find out by looking up patent 5657219 at the "USPTO" ( US Patent and Trademark Office).
it is a variation in circuit, but not really as much a variation in concept as it might seem. The idea of overlapping pulses with the difference constituting the signal is fairly widely used in SMPS, but not in that topology. The Crown topology uses fewer devices, and allows a single-ended output. I agree it is still class-D, just done a bit differently.
The patent mentions some things as "required" in the prior art, but some I have rarely or never seen in actual units..... such as the series inductors for the switch devices (not the reconstruction filter). Nobody I know of uses those at the moment, they are no longer needed.
I also would not describe "other" amplifiers as "vanilla class-D"........ the differences are relatively small, the switching speed advantage is a possible 2:1 difference in losses, but if losses are a few percent now, the absolute reduction is also small.
As for THD differences, I don't know. THD with existing good class-D is quite good, and class-D inherently should be capable of better THD than linear analog amplifiers. THD often comes from changes in the linearity of devices with signal level, and the PWM system has fewer significant changes of that sort than a class-AB amplifier.
Most very good AB amps go to great lengths to keep variations in the operating conditions of outputs to a minimum, generally requiring large numbers of devices, complex tracking regulators, etc. And at the end of the day, the class-AB amp cannot avoid some form of "crossover artifact", however small it may be. Only the class-A amp avoids that.
In some ways, the AB amp improvement process is choosing the brand of lipstick to put on the pig...... but it's still a pig, cosmetic surgery notwithstanding.
My judgement is that there is more room for improvement in class-D, be it alternating, overlapping, etc in terms of how the PWM is accomplished.
If you want to mess around with class-I, the patent is 15 years old in August, so it will be expiring fairly soon.
As for switching loss, as one who has now made BOTH Class-D amps AND VFDs (motor drives) I (and the "Class-I" inventor himself) agree with the contention that losses are generally linear with frequency, at least for mosfets.
For IGBTs, which are at base nothing more than regular bipolar transistors (good ones) set up with a mosfet type input, the losses may get non-linear when you count the "tail current", a slow decay of current which is not present in a mosfet. At a certain switching speed, the IGBT may essentially never turn off.
I don't know that the overlapping "class-I" technique can allow a slow IGBT to work as well as a mosfet for PWM.
