When you say " sample of history " I think you still can do that way - for, the current sample is the " history for the next correction " !

OK in a recent implementation of a tracking set point controller to generate control voltage for a Hydraulic Servo valve - we tried out a " look ahead method ". It works somewhat like this :

I have one register holding the output value that should be for the next sample time.This is from a preprogrammed look up table.(SP)

I have another register that holds the current value of the feedback (PV). Comparing both, I know the required trend - inc / dec / stay put. And the magnitude of the difference ( SP diff PV )gives the quantum of correction. This is plugged into a transfer function and the next command arrived at.

In our case the actuator step response was about 600mSec and the sample interval was 50mSec. Got good results.

In your case the SP ( output volts )is a fixed value instead of a tracking one like ours, and hence should be possible to control as above.

But thats just a hint I am trying to give without knowing the full set up.

Accepted - I may be one of the 6 visually challenged men, feeling the elephant at differnt parts and 'telling' what it looks like !!

Raghu