If you use Q90 you only need maybe 13-15db with headroom.
I'll make a comment. BANDPASS the output of Q90, there is way too much stuff coming
out that can even if it meets -43dbc down contribute to IMD and stuff. What I've seen
going into Q90 was already dirty and had poor IMD, carrier leakage. If you don't then
worrying the IMD is pointless.
Hint: for building. Get a small cone shaped Carborundum stone to round the edges of the
FB61/43-2402 balun cores as they tend to have very sharp edges and tends to scrape off
the enamel causes inter-turn shorts. I use a Dremel (motor tool) branded stone with
fingers for that.
IMD, D3 and D5 not fully tested yet. I expect it will be close to many of the 12V systems that
aren't burning current for running class A. Then again if the source and driver stages one
builds are not good your not going to get performance. Grafted to ubitx, why worry what
you get will be what you put in.
The 1db compression is at 13.8V is 14-15W, moves up with voltage to over 20W at 20V.
That was with 125ma idle per-device. Sweet spot on two was at 160ma each.
I have a few more to build and test for comparison and then get to deeper testing.
Its a pleasure working with an amp that is bullet proof stable.
The RD16 does not need 500ma, unless your shooting for class A! more like 200-250 per device and
it also depends on Push-pull (lower) and single ended (higher).
In both cases IMD and idle current are connected to a point.
uBitx is IMD limits are low due to drive and other internal issues before the amp.
My plan is using it with a modded Atlas 210X to replace the 100W amp with
a QRPlabs amp plus driver. The driver for that will be a MPS6514, 2n5109,
2n3545(ALC) and I expect at 10W about the same specs (IMD, carrier,
harmonics) as the original radio. The other mods include replacing the
whole analog VFO with a digital VFO system.