Please do not listen to Archon's comments. They show little understanding of system dynamics, PID control, and control theory in general. Archon very much shows a "limited knowledge of it." There are plenty of good threads describing PID and its applications on this site, however, this is not one of them. I hate to do what I'm about to do to another member, but I cannot have such an erroneous description of PID control leading members astray.
Archon said:
I don't know your level of experience but even well educated engineers have difficulty with the PID function on instrumentation since it is based in calculus. Most control work is done proportionately using simpler math.
Not true. Well-educated engineers have a thorough understanding of calculus and understand the PID function quite well.
Archon said:
My limited knowledge of it, was that if I was trying to heat up an oven for example I could hit my target temperature OK with proportional control but the temp would overshoot then come back under and after oscillating around the temp it would slowly settle in on the desired setpoint. The valve doesn't know how fast the oven is heating up and can't anticipate hitting its setpoint.
Proportional control in a non-integrating system will not settle in on the desired setpoint.
Archon said:
By using integral control the processor can calculate the instantaneous slope of the line and compensate for the approach to the setpoint thereby giving you a smoother settle-in to the setpoint temperature. Proportional control can do a little of this by using the difference between your setpoint and your current point to control how far the valve is opened and reduce the valve position as it approaches the setpoint to avoid overshoot that occurs with simpler on-off control of a valve.
Calculating the instantaneous slope of the error is the derivative function, not the integral function.
Archon said:
The differential function is even more rarely needed but my understanding is it allows you to program for quick changes in the load i.e. putting a fully loaded pallet onto a conveyor and maintaining the speed of the conveyor as the load is quickly increased (or decreased).
Derivative control will add large changes in control output as the variable it acts on changes rapidly. This may or may not be a good thing, since it can amplify noise or cause rapid jumps during step changes in setpoint. Disturbances in load can generally better be compensated for by feedforward control.
Archon said:
Your application sounds like it would not need the PID control. You really only need 1 analog reference signal coming back to the PLC from the output of the mixing valve. Then you need to do some math in the program to control the analog output to the RO water valve to achieve your desired mmhos. Just subtract the setpoint from the current value and output the difference to the RO valve. As the mmhos goes down your RO water valve should close down proportionately and cutoff entirely at the low point of your scale.
This is simply describing proportional control, and since the process is not an integrating process, will result in a steady-state condition that is not equal to the setpoint.
Archon said:
You may also want to buffer the output by taking samples every few seconds or something to keep the valve from oscillating with any little change.
This is undersampling, and if the system dynamics change rapidly, the valve may operate more smoothly but the system variability may increase and the process variable may exceed control limits. The sampling period should be based on the dynamics of the system, not the requirements of the actuator. Filtering can always be added.
Listen to the posters above who advocated using ratio control. They've given you better advice.