Recently I've been tuning my system by first setting the PID coefficients to 0 and setting the FF* block to make the following error (as seen in halscope) as small as possible. After that I tune the PID to compensate for the vagaries of the system. When I tune this way I think there should be a one stage delay between the command input and the command/feedback adder. The idea being that if the FF* tuning was perfect there would be no following error resulting in a zero out of the PID. Without the delay the PID would always see an error due to acceleration.
Has anyone considered adding this delay stage?
