<xentrac>
so I thought it would be interesting to see if I could do the output filtering and summing purely reactively, in order to get a high-efficiency class-D amplifier with a lot less residual ripple
<xentrac>
I put together a crude simulation at https://is.gd/reactivedac, which does seem to more or less work
<xentrac>
it's tuned for audio reproduction, but for the sim, I'm using 80kHz PWM in order to see how it performs under sort of worst-case circumstances
<xentrac>
and the thing I'm not sure about is whether maybe the inversion stuff doesn't actually contribute anything in this case
<xentrac>
I mean ac-coupling through the inverted signal requires a capacitor, and maybe it would work just as well to, say, use a π-topology LC output filter instead of the two-pole reference design I have at the top of the schematic there
<xentrac>
it does work better than just the two-pole design
<xentrac>
I'm curious what people who really know stuff about audio amplifiers think
<xentrac>
wpwrak had recommended that I talk to someone here, but I can't remember who it was :)
<xentrac>
Woodward's original design used a couple of layers of 74HC04 powered from a precision voltage reference, and had a bunch of other really interesting design features (including the 16-bit resolution feature touted in the title)