Unfamilliar PID Equation

I was looking around online about PID Tuning Software and came across this website:
http://bestune.50megs.com/

On this company's / person's website, they have a document listing the types of PID Equations used in the software (see attached Word document). One of them is "Type C" (an independent PID equation) which uses the PV in the proportional term instead of the error:




Why would you want to control your output based on the process variable, with no reference to the setpoint?

Wouldn't this only be useful when you have I or D control? In that case, you would have a slower loop; you would have to wait for the integral error to add up and the PV to start changing to see contributions from the I and D terms. For sufficiently high PVs, if your integral gain Ki is too small, it would may take forever for integral action to kick in.

Has anyone used this form of he equation before? If so, for what applications?

Type C is often used for processes that have frequent step changes in setpoint which would cause large immediate changes error leading to immediate changes in the control output (i.e., "bumps") due to the proportional and integral terms acting on the error term. It is preferable to avoid bumps on system setpoint changes in some systems.

You are correct about the response. It assumes a fixed setpoint and requires the integrator to ultimately drive the PV to the setpoint.

Their contention is based on the somewhat arbitrary requirement (IMHO) that the PID implementation take the incremental form as opposed to the absolute form. With the incremental form the controller integrates the derivative of the PID equation. There are some perfectly valid reasons to do this. However, it does bring up the case where a step change in the setpoint will cause the proportional term to spike due to the overal derivative. In the incremental form of the PID implementation this spike can have long-lasting effects.

You can handle this on the input side also by profiling the setpoint change. Some controllers have this as an option, others don't.

Keith

Here is an interesting thread

http://groups.google.com/group/sci....gst&q=bestune+Peter+Nachtwey#d5be27255e5504fd

The Bestune guy has a bias towards what I call a I-PD. I-PD controllers do not add zeros to the closed loop transfer function so their bandwidth is relatively low but it also means they don't suffer from the derivative kick. PIDs add zeros which increase bandwidth which is good for motion control but a set point ramp or target generator for proper use. Complex zeros can also cause response problems.

I just happen to have a .pdf file on this topic comparing the response to a sine wave.
ftp://ftp.deltacompsys.com/public/NG/Mathcad - t1p1 pid comparison sine.pdf

The response to step changes should be the same for PID I-PD and PI-D. A PID has a derivative kick only in stupid simple implementations. Anybody with any sense can see that the step causes the derivative to saturate so instead of letting it happen the target history is modified to the new set point or the error history is offset by the difference by the new error. Simple. I don't know why people insist on being lazy or having no imagination in their PID implementations. ( Now I have got to check mine ). We offer both the PID and I-PD. The I-PD is good for those applications where the target is not smooth. This includes applications like following a joystick reference, inner loops of a cascaded loop and many movie types of applications where a master computer keeps downloading new set points at 30 points per second. ( 29.97 ) are the SMPTE rates. These profiles are typically not that smooth so using a stupid simple PID is a bad idea as it results in rough motion.

Thanks for the info Peter (and everyone else).

By the way, that Google thread is hilarious.

I especially love the repeated regurgitation of the Bestune links...