The design for this transmitter, and several others, is published on the GW3UEP Web site and this particular circuit can be found here.
The results of my testing mirror almost exactly those claimed by the author with only a small variation. Using an IRF540 switching FET in Class-E, I am measuring drain current of 2.33A at 12.77VDC, which calculates to an input power of 30W. Measured output into a 47 ohm load is 98Vp-p or 25.5W, for an overall amplifier efficiency of 85%.
One thing that I noted on this one, as well as an earlier mock up last year was the need for slightly more inductance in coil L2. As can be seen from the picture, I have fine-tuned L2 by inserting a ferrite bar slightly inside the form. Once I begin the final build I will measure the newer inductance and rewind the coil. As well, I'd like to test this at 24VDC to see what changes might occur with higher drain voltage.
The drain and gate waveforms are shown here and from what little I know about class-E circuitry, it looks as though things are still not ideally optimised but are close. Dimitrios, VK1SV, has also published a nice tutorial on his class-E homebrew experience.
Perhaps cleaning up stray capacities and unwanted couplings in a more 'cosmetic' version will help. Running key down for several minutes seems to generate almost no heat in the FET so the larger heat sink is probably not needed.
While testing last night, I used the little transmitter to have a nice cross band QSO with VA7FC on Vancouver Island, about 150km away.
With the small parts count and inexpensive crystal (50 cents), this transmitter could probably be built for $10 or less, making a nice starter rig for 630m work.