>    error = value - set_value;
>    p = kp * error;
>    i += ki * error;
>    d = kd * (error - last_error);
>    pid = p + i + d;

a couple of (other) common tweaks:

the derivative term is sometimes generated from the input, rather than the 
error. useful if the error term is likely to be noisy, as taking a 
derivative (= highpass filter) will tend to increase noise.

the integral term can be limited to an absolute maximum, in order to 
combat 'integrator wind-up'; a phenomenon which would otherwise cause big 
overshoots in slow moving systems. lots of other inventive solutions may 
be used to acheive the same result.

some stuff to remember / realise:
the phase (sign) of the constants may be either positive or negative, 
depending on how the signal is generated.
the comments will probably not match the code, they will just give you 
some hint about what the original programmer was thinking about at the 
time.
the integral & derivative terms may not be calculated prior to summing, 
but may be inputs to your system (i.e. externally 'calculated')
if the code does not in some way add a proportional term, and integral 
term and a derivative term, then it is not a PID controller.
any changes you make to the 'PID' code, will mean the calibration for the 
system will change. you may even need to change the calibration procedure.

a couple of documents I have found useful in the past (first link on each 
search):
http://www.google.com/search?q=pid+without+a+phd
http://www.google.com/search?q=motion+control+tutorial+newport

hth,
simon
-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist