------------------------------------------------------------------------
S. Skogestad, 
"Tuning for Smooth PID Control with Acceptable Disturbance Rejection", 
Ind. Eng. Chem. Res. 2006, 45, 7817-7822.
------------------------------------------------------------------------

You may be familiar with my SIMC tuning rules from my 2003-paper in Journal of Process Control. 
For a first-order with delay plant (gain k, delay theta, time constant tau1), we get the PI-settings:
    (1) Kc = (1/k) (tau1/(tauc+theta)),
    (2) tauI=min[tau1,4(tauc+theta)],
For details see http://www.nt.ntnu.no/users/skoge/publications/2003/tuningPID/

There is one tuning parameter - the closed time constant, tauc.

In the 2003-paper I consider the case with  "tight" tuning subject to achieving
reasonable robostness (Gain Margin about 3) and I recommend to use: tauc=theta = delay.
This is actually a minimum value, that is, the recommendation is
      taucmin = theta

However, the choice tauc=taucmin=theta may be a bit aggressive, 
so in practice one often uses a "smoother" tuning, that is, tauc > theta.
But tauc should not be too large, because otherwise the output y will go out of bound 
when there are disturbances d.
This is the topic of the present 2006-paper.
The recoomendation is to use
     Kc > Kcmin = u0/ymax 
     where ymax = max. allowed deviation in the output y
           u0   = required input change to reject the disturbance(s) d 

Subsituting Kcmin into (1) one can the obtain taucmax, and we end with a region for tauc:
   taucmin ("tight") < tauc <  taucmax ("smooth")

Please send me feedback on use of these rules - especially in industrial practice!

- Sigurd Skogestad