DeltaV PID Forms

I am working with a customer tune a couple of PID controllers on their DeltaV system. And we started talking about all the different PID Forms out there (don't get me started on that one). So, we got a little curious about the DeltaV forms. There are two forms listed in the DeltaV documentation, the Standard PID and the Series PID.. Here is a screen capture of the forms (see below, also attached if imageshack expires).

The standard PID form looks OK, but the series form seems to be a little screwy (that is the technical term I use). The series form seems to have an extra term that doesn't appear in any series form of the PID equation I have ever encountered.

If you look at the series form as written,
Out(s) = Gain *(1+Tds/(a*Tds+1))*((Trs+1)/(Trs))*E(s)+ F(s)
the (1+Tds/...) should be written as (Tds/a..) without the 1+. I figured the brilliant minds of this board would be able to tell if I am an idiot, or if it just a small error in the documentation.

I haven't brought it to the attention to Emerson/DeltaV people yet. Figured I wanted some proof first.

Btw: I have been reading this forum for the past two months, and have learned more about PLC systems then in my 10 years of college (which including using PLCs to build experiments).



Regards,

Bob

It does look odd

What they did is nessary so one still gets a unity gain for both the I and D terms if both the D and I terms are not used.

(1+Tds/(a*Tds+1))*((Trs+1)/(Trs))
Expand this to a parrallel form.

1 + 1/(Tr*s) + Tds/(a*Tds+1) + Td/((a*Tds+1)Tr)

You will see that if Td is 0 the gain of the whole controller is 0 unless the 1 is there. Notice that 4th term is extra and NOT STANDARD.

I don't think it is an error. It is just odd. I see no purpose for the extra confusion caused by that non standard 4th term. BTW, check out the end of the Omron and Siemens thread. The explaination I used for the 2 degrees of freedom came from the PID on the newer Delta V PID controllers.

Peter,

Thanks for the answer. I understand now that their equation works, and makes sense. I guess I just have to add another one to my basic lists of PID forms.

I had figured it was just supposed to be this form:
Out = Kc * ((Tds+1)/(aTds+1))*((Tis+1)/(Tis))
Sometimes, I forget companies do whatever they want with no regard to standards.

On a side note, I work with a university over here, and one of the graduate students there is studying the different PID forms, I will throw this at him and see what his thoughts are.

I wonder if anyone has done an exhuastive comparison of the various PID forms. I am sure that each form has its own benefit and limitations under certain conditions. Maybe if someone came up with a good enough study, the manufactuers could aggree on some sort of standard set of equations. Now if that is idealistic, I don't know what it.

Thanks again for the answer.

Bob

I have......

Bob Rice said:

I wonder if anyone has done an exhuastive comparison of the various PID forms. I am sure that each form has its own benefit and limitations under certain conditions. Maybe if someone came up with a good enough study, the manufactuers could aggree on some sort of standard set of equations. Now if that is idealistic, I don't know what it.

Thanks again for the answer.

Bob

Click to expand...

I have tons of Mathcad worksheets showing the different advantages and disadvantages. Some of these attributes are not apparent until they are implemented. My company makes motion controllers. Just about anything related to motion has been analyzed. I have not studied serial forms of PID because I would just convert them to equivalent parallel forms.

I have analyzed:

1 the position and velocity forms ( I hate the names because they have nothing to do with position or velocity ).
2 most if not all combinations of gains in the forward loop and feedback loop along with higher order derivatives.
3 feed forwards using the SP to calculate the open loop output directly and the other method using SP and the derivatives to form a modified SP around which the loop is closed.
4 equations for calculating the PID and higher order gains and feedforwards given the system model or transfer function. I can do this for critically, overdamped and underdamped responses. Over damped responses work well for motion control.

I am sure there are others that have done this but it is kept secret for competitive advantage reason.
I haven't had the time to write 'Peter Ponders PID' yet.

Originally posted by Peter Nachtwey:

I haven't had the time to write 'Peter Ponders PID' yet.

Click to expand...

...but I'm a-waitin'!!!

Peter's kind of like E.F. Hutton that way.

Keith