PID Loop Tuning Discussion Notes

PID Tuning Discussions

found on Linked In.
Includes a number of links.



Automation Engineers
How to design perfect a P.I.D Control in PLC Program, What Must be consider i.e Kp,Ki,Kd,Sampling time etc


Claude Flandro • Go to and download a free copy of the PID Loop Tuning Pocket Guide.


Michel Ruel • I disagree with Mr Marinescu. The “clock” for you rprocess is the time constant if and only if you tune your loop to respond nicely to SP changes. If your goal is to reduce load changes, the “clock” is the dead time. The dead time is the time needed to observe a change in PV after a controller output change (in Auto or in Manual).
Sampling time should be ~5 to 10 times smaller than dead time.

PV filter time constant should be around one fifth of dead time.
Derivative time (if needed) should be around half the dead time.
The Integral time should be around 2 to 5 times the dead time.

The proportional gain will determine the loop response. Increasing the Kp will reduce the peak error after a load change, will reduce IAE (integral of Absolute Error) but could introduce overshoot and damped oscillations.

A good rule of thumb is to use as a starting point :
Kp~0.2 x Time constant/(Process gain x Dead Time)

All the values above will work with series or ideal (ISA) algo; with parallel algo, values need to be calculated differently.

Do not hesitate to ask for more precisions… Have fun!

See on our web site:


Mike Brown • I am a specialist in loop optimisation, and have found that about 80% of all PID controllers operating in PLC systems, are set up incorrectly so that scientific tuning cannot be used. Main two problems are incorrect triggering of the block so scan rate is incorrect, and scaling when using engineering units on the SP & PV inputs. I agree with Michel that you should use a standard block if possible, as home made ones seldom work properly. I have also found that the manufacturers generally know nothing about the practical application of their controllers to real life practical control and do some very stupid things. Most of their manuals are also pretty useless.


Amit Lalyan • I am a AB PLC programmer. For Perfect PID control use FBD language PID block in place of Ladder language. It gives more accuracy in control & easy for PID tuning also. Auto Tunning can also help you.In RS view32 & SE scada Activ X ready made face plate is available which directly connect to PID Block without making tags data base for the Loop.


Michele Tonon * For Mike Brown: what do you mean “standard block”, and “home made”? I found a lot of bugs in Siemens S7 standard blocks library…are not they made by human beings? We are professionals…of course you have to know what are you developing. For instance PID temperature control auto-tuning provided with S7 library is based on the inflexion point of open loop step response (Ziegler-Nichols, 1940). Not bad, except that it is very sensitive to noise and second, is performed in open loop, hence without any control. I developed an auto-tuning function, based on moments method (Ray,1972) which is better performing and works for most of its time with the pid in closed loop.
For Rizwan Qadeer : there are plenty of rules to tune the pid parameters. I suggest to have a look at this book that collects a huge number of them: HANDBOOK OF PI AND PID CONTROLLER TUNING RULES (3RD EDITION) by O’DWYER AIDAN, IMPERIAL COLLEGE PRESS.


Graham Jefferson • Rizwan, if you have a view to developing your own PID instruction in a PLC then I would suggest you take Mike Brown’s advice and don’t do it. However if you do or if you want to look at the vendors’ PID instruction then it becomes clear that there is a standard of sorts. The reason is that the PID is a fairly simple algorithm however with only a few flavours. Whether it is Velocity or Positional or Series, Ideal or Parallel. Other than that it’s the bells and whistles that really count, things like bumpless transfer, output scaling and clamping, feedforward etc. Let’s return to the standard Vendor PID instructions and as Mike said an awful lot of them are configured incorrectly and that is particuarly so in the Allen Bradley PLC5, ControLogix and the Siemens S7. People often confuse incorrect configuration with bugs. When you configure them correctly they all work and for all intents and purposes they all work the same. The only true bug I’ve encountered is in the Allen Bradley PLC5 PID which is also in the ControLogix Ladder PID instruction and that is when a large error occurs with a high gain the output is clamped at max or min however the calculation is not clamped resulting in serious and deep saturation. The theory of sampling time can indeed be complex however in practical terms settingthe scan rate at 1 second will suffice for just about any industrial process. Remember that most DCS vendors have fixed scan rates (usually 1 second) and they all work fine so why get hung up on the theory. And lastly you mentioned that you have a temperature controller with a Gain of 12000. That clearly is not a sensible gain and suggests that you have a scaling issue between your input and output ranges. Quite simple every PID MUST have its input and output units normalised to be the same. For example if the output is 0 – 100% and the input (PV and SP) are 0 – 1000 then the input must be multiplied by 0.1 to normalise it. If this is not done then the units of Gain become meaningless. Most controllers manage it in their configuration by putting in the relevant ranges and it is taken care of however many controllers like the S7, Quantum and GE Fanuc do not so the user must do it….


Harold Ennulat • Rizwan, I also thought that a gain of 12000 for a temperature loop was very high.

Graham, your comment on the inputs needing to be normalized with the outputs in order to normalize the PID loop gains is a hugh insight…. I would have expected that this normalization would have been done in the PID loop itself… It is typical to scale the inputs in engineering units, such as temperature and scale the output in percent.
It is very disappointing to learn that manufacturers do not always do this.

I’m not sure I’d go so far as to say the gains are meaningless. I’m still thinking I would much prefer to enter the SP and PV in engineering units in spite of the fact that this affects the gains. While it may be possible to predict the intial gain settings at the end of the day the gains are adjusted to match the actual process requirements balanced against the need for long term stability.

I also sent away for the PID tuning pocket guide mentioned by Claude.

Some Great comments.


Roy Kok • Control Station offers an excellent Sister Site called It offers excellent information in the design, use, tuning, tips and tricks for PID control. It is a free resource.


John Kowalcheck, PE, CAP • I just came across a relatively new site called the Control Loop Foundation:

Take a look at the Workshops which give some good overviews on various aspects of process control. The PID section will allow you to experiment with different tuning parameters. Note this is based on the Emerson DeltaV DCS, so the scaling and parameters are geared toward it. As others have mentioned, you may/will need to modify the numbers to match other platforms.


Rizwan Qadeer • …
i am always very interested to simulate a process on computer base application, bit i have not found a single simulator like this ,can you tell me the simulator like these, and where can i get them.


Ralph Koettlitz • Take a simulation environment like Simulink as a part of Matlab.

It’s state of the art and simple systems are really fast simulated. I used this tool already as a student, but you have to describe your system in laplace space.


Hervé de Foucault • Merry Xmas to all,
There’s a free competitor to simulink called “xcos” ( Will it remain free ?
At least for a try …


Claude Flandro • Michael’s suggestion to compare the new PID parameters against the current/old parameters makes a lot of sense, and we recommend it all the time. But it doesn’t make sense to build a simulator from scratch, because of all the traps previously mentioned. Why not use the simulators built for your specific purpose and available through the established PID tuning software? Here is an example of ControlSoft’s simulation.


See the LinkedIn Discussion for omitted details on the pros and cons of Process & PID loop simulators

Claude Flandro • The software tools collect the process response data from a step change sent to the process by the controller (as described previously by Michael). The response data is used by the software to create a dynamic model which is then used in the simulation for the process side. The controller definition and algorithm are defined in the setup of the simulation. Disturbances and setpoint changes can be set for the simulation to compare how the process will respond to different PID parameters.

There is a variety of PID Loop Tuning Software. I represent ControlSoft, who has been supplying this type of software for 25 years. (INTUNE at )

Here is an online link to an article in Control Engineering September 2006, which lists more vendors:


Michael Taube • Alejandro touches on another aspect of loop tuning that is often overlooked: performance monitoring. While there are software applications available from several different vendors that supposedly do this, I haven’t had any personal experience using them; I’ve had to rely strictly on manual monitoring. Nevertheless, this is one of the facets of loop maintenance that is needed to ensure long term operating benefits and proper instrument maintenance.

Loop performance can also be affected by nonlinear behavior resulting from operating in a different region – “Severity”, throughput, product quality, etc. – compared to when the loop was most recently tuned. The classic example is nonlinear valve behavior and is also the most easily addressed. While some (older) control systems do not provide “valve characterization” features for analog outputs, this functionality can be implemented in one of two ways:
1) Insert a calculation or characterization block between the controller and its AO point, or

2) Dynamically updating the controller’s Gain (Kc or PB) based on the current controller output (valve position).

In both cases, the characterization curve is obtained by stepping the valve through its entire range and recording the resulting flow to create a flow v. output curve. If the flow is subject to wide variations in pressure drop, then it may be necessary to generate a family of curves. For the characterization, pairs of values are selected that makes the valves’ flow characteristics appear linear to the controller. For the case of updating the controller gain, the Derivative of the flow v. output curve is used.

I spend a fair amount of time looking for & examining potential nonlinear behavior as this affects both regulatory and advanced control application performance (including model-predictive controllers). It also makes the Operator’s job easier and everything you can do to make that happen generates a great deal of credability with and cooperation from Operations (whom you rely upon to help you in your job). So it is time well spent.


John Kowalcheck, PE, CAP • I agree with Graham in that most control loops are robust enough to tolerate non-linearities and process dynamics changes. Relatively fast acting flow and pressure loops come to mind. But, there inevitably will be a couple of loops which just defy a simple solution. That’s when the input or output characterizations and the dynamic gain schedulers come into play. Of course, there are books written about these advanced topics.


Michael Taube • It comes as no surprise that, lacking any industry or de facto standard definitions, terms like “robust”, “well tuned” and “perfect PID” are subject to each individual’s experience & expectations. As a result, discussions, like this one, can degrade into minature holy wars! Fortunately, this one has not.

Based on what I have gleaned from simiular discussions, I have fairly high expectations and requirements for PID controller performance compared to others in the profession. While the tuning rules I employ produce tuning constants that can tolerate uncertainty and variations in process behavior (and still produce acceptable controller performance), where there is a measurable, predictable and “significant” change in Process Gain (Kp), I take the time to characterize it. Therefore, I find that, while not all, many of the loops I encounter – up to 30%, or more – do require output characterization in order to obtain the same performance over its entire operating range. This is especially true in facilities that experience large & frequent changes in throughtput – primarily refineries – that have control valves operating at >80% output when at high rates and <=50% at nominal rates.

I heartily agree with Graham about the benefits of manual performance monitoring compared to the software-based solutions on the market. There’s no software package that has yet to duplicate the insight & judgement of an experienced engineer! Autotuners fall into the same category.


Mike Brown • I have not joined in this discussion so far, as it is a subject that cannot be dealt with at all adequately in a few short comments, and it also has drifted miles away from the original subject. However a few remarks:

1. There is no such thing as a standard PID control block. Every manufacturer makes them the way they think is right, and they themselves do not generally have internal standards as their controllers often differ widely with each new version of the software. Most of them seem to have little knowledgethemselves of ‘practical’ control, and do very stuipid things a lot of the time. Most controller manuals are rubbish, often with mistakes, and are very little help to the end user. I haven’t come across a single controller manufacturer who offer courses to their users to expalin how their particluar controller and its options work.
2. Tuning: In my 30 years of control, I have tried many tuning methods and packages. Tuning that works properly is ‘pole cancellation’ tuning, and the only software package I have found that really works properly on virtually any industrial process type is the Protuner Loop Analytical software supplied by Techmation Inc. of the USA. It uses a step change on the process, and converts the response from the time to the frequency domain by means of a digital Laplace transform. The tuning is then calculated from the frequency plots, using the proper mathematics, and it is also combined with some expert rules.
3. Most people believe that tuning can solve all problems. However tuning is a waste of time, before thoroughly trouble-shooting and analysing the loop to establish things like the class of process (integating or self-regulating), if any problems exist, the process dynamics, process stability,etc. Only when one is aware of all these things can tuning be performed correctly to give the best tuning response to meet the desired control requirements for that particular control loop.
4. There is almost a complete lack of understanding of ‘practical’ control which is a very different subject to the theory of control as taught in universities, or technical colleges. As a result, our experience gained on working on thousands of loops, in hundreds of plants in many countres, is that in general at least 85% of loops are not working efficiently in automatic, and approximately 50% of loop exhibit problems that usually the users are not even aware of. Many control people often find these statistics unbelievable. However once my colleagues or I work in their plant with them, they realise the truth of these statements.


Catalin Puiu • PID control in a process with wild behavior is quite a stress…

Ralph, Laplace transformation is great, but it works with continuous time, which does not make sense for a digital system. In discrete time you have to switch to z transform. Not a big issue, but a lot more relevant, unless you’re using an analogical controller.

Simulating a process in simulink or anything else can be applicable (and actually pay its time and money) only if it’s a very simple one, as far as I consider. In my experience, taking data out of the process while varying the parameters you believe of relevance (for example mass, in a speed control PID, or air pressure, in a burner control), can be a lot more relevant (and a lot faster).

The same furnace has the P factor of, let’s say, 70, at 100deg. C, and 7000 at 1200 deg. C. How could you simulate such a process? It would take a year just to get the constants out of the system, and by that time the materials are decaying and have changed their properties…

In areas which are hard to model and control, mathematical black box modeling with all the input parameters is a choice. A proper auto tuning might also be of use, but don’t expect it to be available directly on the PLC or SCADA. The parameters automatically determined can be used, in some fortunate cases, but to have a good behavior a lot more is required. I’ve seen small uchip Japanese controllers determine parameters a lot more stable than Siemens or Emerson, but they ain’t sharing the math behind them:).

Mike Brown is right. Canceling poles of a system which seems to behave like a second order or third order process is a good approach to get a reasonable response in delicate systems. But for high order systems, this requires a tailor made PID, made either on top of a good and well tested one, with anti-wind-up, derivative filtering and so on, or designed from scratch. I haven’t had the time to play with the scratch option, though, but the first option can be of use. Anyways, you have to bother with this, as far as I’ve seen, only if you need very fast response times or if the process is very unstable.

Trying to guess the parameters (raise or lower the P, then raise or lower I, then…) works for systems which have time delays of … human scale. But for a process with a time delay of hours and a time constant of another few hours, you would grow a beard next to it before even getting the P decent.

Rizwan, a good starting approach to any thermal process, with big response times, is to get the data out of the PLC, print it and look at it on a large scale. See how it behaves over hours and days while you’re varying the input, try to figure out if there are other parameters which interfere with the input of the system. Probably the PLC or SCADA can give you a print of the data.

Glad to be part of this community, I just joined it a few days ago. I’m now working on some math behind an auto-tuning and this topic was a breath of process control after the holidays. I’m working with PID controllers in industrial furnaces, with a lot less experience than you, only 3 years, but with plenty of enthusiasm.


Tuning 101:
I’m going to go where no one has yet dared to go.  I’m going to give Rizwan the tuning parameters he is asking for…. or at least the starter values.

I am using the info that Rizwan has provided so far.

You will need one other additional tool to do the tuning, you must be able to monitor the process temperature when you make a step change to the setpoint.  You can do this by hand recording the time and SP (setpoint temp) and PV (measured Temperature).  Since this can take hours on a typical temperature loop some kind of trending tool will be most helpful.  The purpose of this is to monitor the response to the step changes to the setpoint you will be making.

Step 1:
For starters I tend to use a gain of 1 or 2  (P = 1 or 2) and an I and D of “0”.  What a gain of 1 or 2 means varies by controller, so I adjust the P value until I see a full output from the PID controller (CV = 100%) when the error (SP-PV) is at a fairly large variation.  In your case adjust the P value until a 50 DegC temperature difference (SP – PV) equals 100% output (CV = 100).
Looking at the response will tell you whether to add more gain or less.  If the temperature stabilizes without overshoot (overshoot is a temperature that is above the final stable temperature.  On a trend graph is readily apparent).

Step 2:
Increase the P until you get some overshoot where the temperature settles down fairly quickly (quickly means within 2 or 3 overshoot cycles).  If the temperature never stops oscillating then 
there is to much gain.

Step 3:
Reduce the gain by 25 – 50% from the value set in step 2 for stability.

Step 4:
Add Integral or “I” to your PID controller to get to setpoint.  (Up until now we have not worried about getting to setpoint, only about the correct response and getting as much gain on the system as possible for stability).
The amount of Integral gain depends on the response observed in your system.  For set it and forget it tuning, most of the time one can just put in an integral term that allows the output to change by 10 – 20% of its total range in the response time of the system. 
The response time of the system is normally when the PV gets to 67% of the final value.  However I tend to use the full response time, so more like getting to 98% of the final value.  For example if it takes 4 hours for the temperature to stabilize after making a setpoint change, then I would adjust the integral term so it adds 20% to the output in 4 hours.

Step 5:
Basic tuning is done.  You’ll have a stable loop about 90% of the time I’m hazarding to guess that eventually gets to setpoint.

Step 6 and beyond:
If this loop will get to setpoint to slowly more integral can be added but this requires additional time which if you have it great.  I would advise to tune these types of loops as early as possible and to have the trend tools necessary to monitor any new tuning and step change responses while you are doing other things.  In this way you’ll have more time for experimenting to “optimize” the tuning before you need to move on.

For temperature loops many in these forums have suggested some “D” to stabilize the loop.  I consider this advanced stuff and suggest to not add any “D” unless you live in the plant and are available to keep an eye on things, especially when something changes in the process operations.

Disclaimers:  There are other ways to tune, some have been mentioned here.  This is just one method.  However it is a common method that many of us use to get close quickly.

Also, If you don’t have some kind of trending tool where you can monitor the temperature, you might want to start by tuning a faster loop like a pressure or flow loop.  With these you can see the response as you watch….

Sorry Rizwan, tuning a PID loop isn’t based on just numbers, it’s based on knowing your process.  Fortunately knowing the process usually just means measuring the response to a step change in the setpoint or some other “upset”, and we typically just measure it on the actual system.
For actual numbers, no one will be able to do any better than P = 12000 and I = 40 as you have told us produces a working system.
Have you tried changing the setpoint and measuring the response?  This will tell you whether you can try adding (or if you need to reduce) the gain.

One final note:  If you change the P gain after you have added an I value, the I value will produce a different response as in the classic PID control the gain is multiplied by each of the 3 terms P, I, and D.


From PID Tuning Method in “Process Control” group


Manmeet Ahuja • Dear friend,

Its true that ZN method is no more used except for basic . it was formulated in 1942 but till now there has been improvements in like Cohen coon, chiu etc. until 1984 Karl Angstrom formulated MPC tuning method. It was used by some companies like ControlSoft and later Expertune to make special software to be connected to hardware with DDE (later OPC) to tune control loops with graphical outputs and better optimised control. There are many companies making such softwares now
This method based on Frequency domain analysis. Later Time domain tuning method developed eliminating bias and other frequency paramters By Picontrol Solutions
All these methods today are software based interfacing with PLC/DCS/SACDA with OPC and eliminating the manual tuning. They also provide fine tuning if so required.


Petr Bartosik • I think the ZN is a bit hard method; I prefer balanced tuning (P. Klan, R. Gorez) because it is friendlier to actuators.
There are a lot of tuning and autotuning methods, offered together with industrial controllers, PLCs etc. Despite this, a lot of professionals use manual methods. The reasons are, I think, 1. lack of time for optimal tuning of controllers, 2. field workers want to stay „in touch“ with process and (auto)tuning methods aren‘t as comprehensible and controllable as manual button turning.
Moreover, manual PID tuning is fit enough for good control of most common industrial processes.


Juni Irawan • I have similar experience.
I just suprised when i just graduated and worked for the first time in the project, when I want to determine PID value for control PCV at PLC Modicon Quantum, the Company Man just said to me to use manual PID tuning instead calculated it or did it in software simulation like MATLAB .

Then he just suggested me to use Kp & Ki. eliminate Kd.
simple and easy method.

then i saw in other projects did the same thing too. manual PID tuning.

sometimes depend on the process, PID was combine with other contol like during start up and normal operation had different value for Kp, Ki ..
we just make in the logic to do auto transfer value of Kp and Ki.

Jason Hise • All,
I’m new to this group and I have used two types of Ziegler-Nichols tunning for chemical application in water/wastewater treatment plants. 1) Ziegler-Nichols with stepped input and measures delay. 2) Ziegler-Nichols closed loop tuning. Personally, I have found that I can tune a loop faster with the closed loop method then I can using step responses. What methods of Ziegler-Nichols is referred to above.
Antonio Panaccione • Jason,
I agree with you. The ZN closed-loop tuning is much faster to tune a system. However, I must admit that I wind up using both methods alternately just to watch the process dynamics at work, and to fine-tune as much as possible….Time permitting of course. I strongly believe that most people who use the “Manual” method of tuning usually never learned how to perform ZN correctly. Actually, the little yellow book from DuPont isn’t a very good read. However, when I watched an ISA Video about PID Tuning, then the lightbulb went off in my head, and I realized how well ZN works. It is so much faster to tune a process once you understand the fundamentals of ZN.


Tom Barker • A simple and quick method that I’ve used for years in chemical plants and refineries and which forms the basis for much of the tuning software discussed above is:
1. Postulate the form of the process transfer function. For most (90+%) loops a first order lag plus deadtime (k*exp(-st)/(1+tau*s) is good enough.
2. Do a step response test. Put the controller in manual and move the output up and down enough to get process response curves.
3. Fit the results of the step test to the postulated transfer function form. This is fun and quick and easy. Knowing your process gives you insight into whether the transfer function fit is real (often you have some intuitive feel for the deadtime and the gain). Getting a transfer function fit matters a lot, because if you cannot get a fit (meaning there isn’t a relationship between the process output and the process variable that can be fit with a linear ordinary differential equation) and you’ve tried a few transfer function forms, the loop may not be tunable as is. Two software packages that do this well are ControlStation (a bit expensive – Dr. Bob Rice is superb) and ControlArts (very affordable, Dr. Hugo is great to work with). Neither Expertune or ControlSoft will expose the transfer function model to you (unless they have changed in the last couple years).

Petr Bartosik •  (training) with PID


Ray Joseph, PE • To Tune or Not to Tune (Part I)

Some of the concerns we may consider about tuning is when, what and why are we going to tune.

So let’s first break this down into to categories of when. Our first opportunity is during commissioning and startup. We will do this to assure the safety of the system; we don’t want processes swinging out of acceptable bounds while running. Loops will typically be damped (low gains) to achieve this. (This will come up again as a topic on simulation.) That is, we want sufficient control to stabilize the operations and we don’t want over manipulation of controls which may produce damped (or undamped!) cycling.

Our next opportunity is when the plant is running and operators are familiar with its characteristics. Then we can begin to tune for other reasons such as safety and economics.


Ray Joseph, PE • To Tune or Not to Tune (Part III)

We see that we can identify the process and use this information to tune the associated loop or loops. We have seen hardware that plugs into the controls and calculates suggested controller parameters. These concepts can be combined to optimally control the process.


Myke King • There’s a new book, “Process Control: A Practical Approach” (ISBN 0470975873) which spends a couple of chapters on the subject. It shows which version of the PID algorithm should be used, why many of the published methods fail and it publishes for the first time a well-proven method. It contains a whole load of other good stuff as well. It’s the first book I know of that gives truly practical guidance on the implementation of all the common process control techniques.

More details at


Jacques Smuts • There are several reasons why tuning rules don’t always work, making tuners give up on using them.

The Ziegler-Nichols open-loop tuning rules have several drawbacks:
1. It tunes the loop for quarter-amplitude damping response, which overshoots and oscillates quite a bit.

1. It leaves the loop with very little robustness, which can quickly lead to instability.

1. The rules give you very poor response if the process is dead-time dominated.

1. The rules are very sensitive to an accurate measurement of dead time, which is difficult on lag-dominated processes with short dead times.

The Ziegler-Nichols closed-loop tuning method does a little better with issues 3 and 4 above, but issues 1 and 2 remain a problem. In addition, this method is very sensitive to control valve problems like dead band (hysteresis) or stiction. More info here:

The Cohen-Coon and many other open loop tuning rules do much better with issue 3, but issues 1, 2 and 4 are still problems.

Lambda tuning rules give you a stable response with no overshoot, and leave the loop with plenty of robustness to accommodate measurement errors. However, these rules result in a very slow response to disturbances on lag-dominated processes. More info here:

You should select the tuning rule to achieve the desired control objective, while considering the constraints above:
* If you need a very stable loop that absorbs disturbances rather than passing them on, use the Lambda tuning rules.

* If you need fast recovery from disturbances, use the Cohen Coon tuning rules, but use only half of the calculated value for controller gain to overcome issues 1 and 2. However, if the process dead-time is very short (issue 4), or the PV is noisy and you can’t measure the dead time accurately, use the Lambda tuning rules.

Note that the PID tuning rules mentioned above have been designed for an interactive / series controller algorithm. The PI tuning rules (no derivative) will work on both interactive and non-interactive algorithms. If you have a controller with a parallel algorithm, you have to convert the calculated settings to work on it.

Also note that the rules calculate controller gain, and not proportional band. And they calculate integral time (as in minutes or seconds), not integral gain (as in repeats per minute or repeats per second). Finally, the rules expect you to make the measurements of dead time and time constant in the same time-base used by your controller’s integral setting, i.e. minutes versus seconds.

A well-designed tuning software package should let you select the control objective and your controller type, and give you appropriate tuning settings.


Mike Ferguson • There are several training books out on the subject. I recommend the book ‘Automated Continuous Process Control’ by Carlos Smith. I trained with Dr. Smith(Universtiy of South Florida-Tampa)and his processes and methods are very sound.


Added January 9, 2011



  1. […] salient posts and comments have been collected and are now available in the PID Tuning Discussion & Notes page on this […]

  2. hi all, i guess this is a quite old post but i am doing my thesis on the same topic and i am looking to develop a PID controller for Siemens S7 as I Have read the comments it seems that there are a lot of experts involved in the discussion as i am just a starter and i am looking for how i can do it ??
    There is a possibility that I will provided with the gains and the sampling time

    This is just a part of my thesis after this I have adjust the process paramets (painting process ) to control the pressure of air which rotating the turbine.

    I can explain in detail also if you want please help me with the idea so that i can work on it… !!
    thanks in advance

    • S7 (as well as most PLC’s) have built in PID controllers. My suggestion is to:
      1 read up on the PID insstructions and select one to try.
      2 program up the PID block in the PLC program attaching some inputs (PV, and Setpoint).
      3 enter a Setpoint and PV that are identical (optionnally write a simulation program and produces an output that feeds into the PV of your PID)block.
      4 Enter some tuning parameters. See the article for suggestions.
      5 Change the PV and/or the setpoint slightly and then in any way you want.
      6 Observe (or Plot or trend) the response, take notes, and you should begin to understand how it works.
      7 Add logic to handle any undesired behaviors. What I call taming the PID loop.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: