Nisar,

I'm in agreement with Steve about the floating point arithmetic.

Even if the range of setppoint and process feedback exceeds what you can reasonably represent in fixed floating point, consider converting all variables to a percentage of full scale before entering the PID, then convert back, if necessary.

Be sure you attend to scaling the I and D terms to account for your sample frequency. With a sample data system, effective gain goes up as the sample frequency increases.

Bruce