UCC28782EVM-030: INPUT FILTER REACTION: Dr. Ridley's work

Hello Robin, 

I would have liked to have attended that Webinar by Dr. Ridley.  It sounds intriguing that filter interaction can be controlled by a signal to the power stage. 

In the UCC28782, there is indeed a link (although indirect) between the CS current source and the input voltage.  The CS current is basically 1/25th of the current pulled out of the VS pin during the lower MOSFET on-time.  The CS current is fed into Ropp to add a small offset to the CS voltage to effectively compensate (cancel) for excess peak current due to turn-off delay of the MOSFET.  The value of Rcs is usually ~3 orders of magnitude lower than Ropp, and is basically ignored in the Ropp value calculation.

In the case of the LMG2610, where the current signal is emulated and scaled down by a factor of  1/1000, Rcs must be increased by x1000.
Since turn-off delays in the lower GaN FET must still be compensated for, but Ropp value is calculated neglecting Rcs, then the actual Ropp value selected must subtract the Rcs value when using the LMG2610.   

That technique works okay for calculated Ropp's that are greater than the LMG2610's Rcs value, but if Ropp_calc is lower than Rcs, then the net effect is an over-compensation of the turn-off delay.  The consequence is that you will get less power out at high line than you get at low-line.  One could reduce Rcs somewhat to regain the required full power at high line, but that would allow excess available power at low line. 

However, to mitigate that case I think one could deliberately introduce just the right amount of low-side turn-off delay (but not turn-on delay) such that calculated Ropp exactly equals Rcs and one could achieve the roughly flat power limit over input line that the Ropp compensation was intended to provide, while actual Ropp value = 0.

Not having seen the Webinar, I can't comment about if/how such line feedforward would/could eliminate line filter interaction, etc.  
If such is the case, I wonder then about the sensitivity to variations in Ropp.  Sometimes Ropp may be specially manipulated to o be larger or smaller than the "optimal" value, to obtain some other benefit at the expense of a flat power-limit curve.  If Ropp must be spot-on to eliminate filter interactions, then it may not be a magic cure-all, but still need some sort of auxiliary damping help.   But if the sensitivity is low, then maybe Ropp can vary widely and help achieve several goals at the same time. 

A load step is often used to assess voltage loop stability.  I suppose it could be used to examine filter interactions, too, but I don't know what criteria (other than obvious oscillatory behavior) to look for.  If a load-step works for that, then I imagine a line-step can also be used.  

Regards,
Ulrich