Raghu :"That is a very controversial requirement. And can be a reality only for very slow responding systems with a very heavy inertia. Like for instance the temperature regulation of a large mass. Because of its large thermal inertia, the system response will be slow and you can get away with slow computations."

Dear Raghu:
thank you very much for reply, you are completely right. the slower the system needs more complex PID.
my system is actually slow, the idea was to use just proportional PID and to add a feed back of previous output effect on Error, you may say why not using dError/dt or integral(Error dt), the Idea was you in that way one take into account effect of time on Error without feedback of output(I'm sorry my english is very bad)in that way you have no idea that really time that output was high helped to reduce error or time that output was low or what caused to increase Error. I will appreciate your further comments.


thank you very much