UcD - correlation between phase lead network and propagation delay

I am newbie to classD, and some things are unclear to me.

In Bruno Putzeys AES paper it is said that oscillation in ucd amplifier occurs at the point where the propagation delay and phase lead cancel out because output filter already does 180 degrees of phase shift (well, almost 180).
The oscillation frequency is known factor because we choose it when designing the amplifier. Propagation delay is also known for the components we choose.
So the only thing one needs to calculate is phase network values.

Now, do we need the phase lead network to introduce phase shift of 180-t_prop to have full 180 degrees of phase shift that would add to 180 of output filter to initiate oscillation, or we have to take into account that the comparator in basic ucd circuit is inverting, and in that case the phase shift of lead network would have to be identical to that of the lead network, not 180-t_prop ? This confuses me:confused::confused:
 
Thx for the reply.
I saw that graph, but didn't look at it with understanding before. I see now that propagation delay is way up in phase and exactly t_prop + t_output_filter (which is <<t_prop), and in opposite direction. They literally cancel out :)

But I also realised now that tan (fi) = - tan (180-fi), so when calculating resistor and capacitor values it doesn't matter because we will get correct absolute values even if we understand it wrong :rolleyes: