Feedforward is a kind of compensation that uses inputs to predict controller dynamic requirements before the uncompensated result actually occurs on the output.
For example, in a pwm controller that modulates power switch turn-off events, input voltage feedforward can be achieved by modulating the amplitude of a controllers modulation ramp or current control signal slope to increase the slope as the voltage increases. This often is simply a resistor tied to the input rail feeding the capacitor or node where the modulator's controlling ramp or slope is presented.
The effect is to reduce pulsewidth without requiring immediate error amplifier influence - compensation is crudely automated, allowing the controller free to react to smaller errors, rather than gross ones, with reduced dynamic disturbance/recovery effort or delay.
In current mode control circuit, the current ramp is directly related to input voltage (v = L di /dt ) so that input voltage feedforward is pre-existing, and loosely adjustable by magnetic component selection or design ( primary inductance lmag or secondary output choke inductance - both effects being present in the primary signal) .
If slope compensation is also used in current mode control, modulating it with an input voltage dependant signal allows increased feedforward effects without adding many extra parts.
RL