You are on page 1of 58

PID controller

From Wikipedia, the free encyclopedia

Jump to: navigation, search


A proportionalintegralderivative controller (PID controller) is a generic control
loop feedback mechanism widely used in industrial control systems. A PID controller
attempts to correct the error between a measured process variable and a desired
setpoint by calculating and then outputting a corrective action that can adjust the
process accordingly.
The PID controller calculation (algorithm) involves three separate parameters; the
Proportional, the Integral and Derivative values. The Proportional value determines
the reaction to the current error, the Integral determines the reaction based on the sum
of recent errors and the Derivative determines the reaction to the rate at which the
error has been changing. The weighted sum of these three actions is used to adjust the
process via a control element such as the position of a control valve or the power
supply of a heating element.
By "tuning" the three constants in the PID controller algorithm the PID can provide
control action designed for specific process requirements. The response of the
controller can be described in terms of the responsiveness of the controller to an error,
the degree to which the controller overshoots the setpoint and the degree of system
oscillation. Note that the use of the PID algorithm for control does not guarantee
optimal control of the system or system stability.
Some applications may require using only one or two modes to provide the
appropriate system control. This is achieved by setting the gain of undesired control
outputs to zero. A PID controller will be called a PI, PD, P or I controller in the
absence of the respective control actions. PI controllers are particularly common,
since derivative action is very sensitive to measurement noise, and the absence of an
integral value may prevent the system from reaching its target value due to the control
action.

A block diagram of a PID controller


Note: Due to the diversity of the field of control theory and application, many naming
conventions for the relevant variables are in common use.

Contents

[hide]

1 Control loop basics


2 PID controller theory
o 2.1 Proportional term
o 2.2 Integral term
o 2.3 Derivative term
o 2.4 Summary
3 Loop tuning
o 3.1 Manual tuning
o 3.2 ZieglerNichols method
o 3.3 PID tuning software
4 Modifications to the PID algorithm
5 Limitations of PID control
6 Physical implementation of PID control
7 Alternative nomenclature and PID forms
o 7.1 Pseudocode
o 7.2 Ideal versus standard PID form
o 7.3 Series / interacting form
8 See also
9 External links
o 9.1 PID tutorials
o 9.2 Simulations
o 9.3 Special topics and PID control applications
10 References

[edit] Control loop basics


A familiar example of a control loop is the action taken to keep one's shower water at
the ideal temperature. The person feels the water to estimate its temperature. Based on
this measurement they perform a control action: use the hot water tap to adjust the
process. The person would repeat this input-output control loop, adjusting the hot
water flow until the process temperature stabilized at the desired value.
Feeling the water temperature is taking a measurement of the process value or process
variable (PV). The desired temperature is called the setpoint (SP). The output from the
controller and input to the process (the tap position) is called the manipulated variable
(MV). The difference between the measurement and the setpoint is the error (e), too
hot or too cold and by how much.
As a controller, one decides roughly how much to change the tap position (MV) after
one determines the temperature (PV), and therefore the error. This first estimate is the
equivalent of the proportional action of a PID controller. The integral action of a PID
controller can be thought of as gradually adjusting the temperature when it is almost
right. Derivative action can be thought of as noticing the water temperature is getting
hotter or colder, and how fast, and taking that into account when deciding how to
adjust the tap.
Making a change that is too large when the error is small is equivalent to a high gain
controller and will lead to overshoot. If the controller were to repeatedly make
changes that were too large and repeatedly overshoot the target, this control loop
would be termed unstable and the output would oscillate around the setpoint in either

a constant, growing, or decaying sinusoid. A human would not do this because we are
adaptive controllers, learning from the process history, but PID controllers do not
have the ability to learn and must be set up correctly. Selecting the correct gains for
effective control is known as tuning the controller.
If a controller starts from a stable state at zero error (PV = SP), then further changes
by the controller will be in response to changes in other measured or unmeasured
inputs to the process that impact on the process, and hence on the PV. Variables that
impact on the process other than the MV are known as disturbances and generally
controllers are used to reject disturbances and/or implement setpoint changes.
Changes in feed water temperature constitute a disturbance to the shower process.
In theory, a controller can be used to control any process which has a measurable
output (PV), a known ideal value for that output (SP) and an input to the process
(MV) that will affect the relevant PV. Controllers are used in industry to regulate
temperature, pressure, flow rate, chemical composition, level in a tank containing
fluid, speed and practically every other variable for which a measurement exists.
Automobile cruise control is an example of a process which utilizes automated
control.
Due to their long history, simplicity, well grounded theory and simple setup and
maintenance requirements, PID controllers are the controllers of choice for many of
these applications.

[edit] PID controller theory


Note: This section describes the ideal parallel or non-interacting form of the PID
controller. For other forms please see the Section "Alternative notation and PID
forms".
The PID control scheme is named after its three correcting terms, whose sum
constitutes the manipulated variable (MV). Hence:

where Pout, Iout, and Dout are the contributions to the output from the PID controller
from each of the three terms, as defined below.

[edit] Proportional term


The proportional term makes a change to the output that is proportional to the current
error value. The proportional response can be adjusted by multiplying the error by a
constant Kp, called the proportional gain.
The proportional term is given by:

Where

Pout: Proportional output


Kp: Proportional Gain, a tuning parameter
e: Error = SP PV
t: Time or instantaneous time (the present)

Change of response for varying Kp


A high proportional gain results in a large change in the output for a given change in
the error. If the proportional gain is too high, the system can become unstable (See the
section on Loop Tuning). In contrast, a small gain results in a small output response to
a large input error, and a less responsive (or sensitive) controller. If the proportional
gain is too low, the control action may be too small when responding to system
disturbances.
In the absence of disturbances, pure proportional control will not settle at its target
value, but will retain a steady state error that is a function of the proportional gain and
the process gain. Despite the steady-state offset, both tuning theory and industrial
practice indicate that it is the proportional term that should contribute the bulk of the
output change.

[edit] Integral term


The contribution from the integral term is proportional to both the magnitude of the
error and the duration of the error. Summing the instantaneous error over time
(integrating the error) gives the accumulated offset that should have been corrected
previously. The accumulated error is then multiplied by the integral gain and added to
the controller output. The magnitude of the contribution of the integral term to the
overall control action is determined by the integral gain, Ki.
The integral term is given by:

Change of response for varying Ki


Where

Iout: Integral output


Ki: Integral Gain, a tuning parameter
e: Error = SP PV
: Time in the past contributing to the integral response

The integral term (when added to the proportional term) accelerates the movement of
the process towards setpoint and eliminates the residual steady-state error that occurs
with a proportional only controller. However, since the integral term is responding to
accumulated errors from the past, it can cause the present value to overshoot the
setpoint value (cross over the setpoint and then create a deviation in the other
direction). For further notes regarding integral gain tuning and controller stability, see
the section on Loop Tuning.

[edit] Derivative term


The rate of change of the process error is calculated by determining the slope of the
error over time (i.e. its first derivative with respect to time) and multiplying this rate
of change by the derivative gain Kd. The magnitude of the contribution of the
derivative term to the overall control action is termed the derivative gain, Kd.
The derivative term is given by:

Change of response for varying Kd


Where

Dout: Derivative output


Kd: Derivative Gain, a tuning parameter
e: Error = SP PV
t: Time or instantaneous time (the present)

The derivative term slows the rate of change of the controller output and this effect is
most noticeable close to the controller setpoint. Hence, derivative control is used to
reduce the magnitude of the overshoot produced by the integral component and
improve the combined controller-process stability. However, differentiation of a signal
amplifies noise and thus this term in the controller is highly sensitive to noise in the
error term, and can cause a process to become unstable if the noise and the derivative
gain are sufficiently large.

[edit] Summary
The output from the three terms, the proportional, the integral and the derivative terms
are summed to calculate the output of the PID controller. Defining u(t) as the
controller output, the final form of the PID algorithm is:

and the tuning parameters are


1. Kp: Proportional Gain - Larger Kp typically means faster response since the
larger the error, the larger the Proportional term compensation. An excessively
large proportional gain will lead to process instability and oscillation.
2. Ki: Integral Gain - Larger Ki implies steady state errors are eliminated
quicker. The trade-off is larger overshoot: any negative error integrated during
transient response must be integrated away by positive error before we reach
steady state.
3. Kd: Derivative Gain - Larger Kd decreases overshoot, but slows down
transient response and may lead to instability due to signal noise amplification
in the differentiation of the error.

[edit] Loop tuning

If the PID controller parameters (the gains of the proportional, integral and derivative
terms) are chosen incorrectly, the controlled process input can be unstable, i.e. its
output diverges, with or without oscillation, and is limited only by saturation or
mechanical breakage. Tuning a control loop is the adjustment of its control
parameters (gain/proportional band, integral gain/reset, derivative gain/rate) to the
optimum values for the desired control response.
The optimum behavior on a process change or setpoint change varies depending on
the application. Some processes must not allow an overshoot of the process variable
beyond the setpoint if, for example, this would be unsafe. Other processes must
minimize the energy expended in reaching a new setpoint. Generally, stability of
response (the reverse of instability) is required and the process must not oscillate for
any combination of process conditions and setpoints. Some processes have a degree
of non-linearity and so parameters that work well at full-load conditions don't work
when the process is starting up from no-load. This section describes some traditional
manual methods for loop tuning.
There are several methods for tuning a PID loop. The most effective methods
generally involve the development of some form of process model, then choosing P, I,
and D based on the dynamic model parameters. Manual tuning methods can be
relatively inefficient.
The choice of method will depend largely on whether or not the loop can be taken
"offline" for tuning, and the response time of the system. If the system can be taken
offline, the best tuning method often involves subjecting the system to a step change
in input, measuring the output as a function of time, and using this response to
determine the control parameters.
Choosing a Tuning Method

Method

Advantages

Disadvantages

Manual
Tuning

No math required. Online method.

Requires experienced
personnel.

Ziegler
Nichols

Proven Method. Online method.

Process upset, some trialand-error, very aggressive


tuning.

Software
Tools

Consistent tuning. Online or offline method.


May include valve and sensor analysis.
Allow simulation before downloading.

Some cost and training


involved.

Good process models.

Some math. Offline method.


Only good for first-order
processes.

CohenCoon

[edit] Manual tuning

If the system must remain online, one tuning method is to first set the I and D values
to zero. Increase the P until the output of the loop oscillates, then the P should be left
set to be approximately half of that value for a "quarter amplitude decay" type
response. Then increase D until any offset is correct in sufficient time for the process.
However, too much D will cause instability. Finally, increase I, if required, until the
loop is acceptably quick to reach its reference after a load disturbance. However, too
much I will cause excessive response and overshoot. A fast PID loop tuning usually
overshoots slightly to reach the setpoint more quickly; however, some systems cannot
accept overshoot, in which case an "over-damped" closed-loop system is required,
which will require a P setting significantly less than half that of the P setting causing
oscillation.
Effects of increasing parameters

Parameter

Rise Time

Kp

Decrease

Increase

Ki

Decrease

Increase

Increase

Eliminate

Small Decrease Decrease

Decrease

None

Kd

Overshoot Settling Time S.S. Error

Small Change Decrease

[edit] ZieglerNichols method


Another tuning method is formally known as the ZieglerNichols method, introduced
by John G. Ziegler and Nathaniel B. Nichols. As in the method above, the I and D
gains are first set to zero. The "P" gain is increased until it reaches the "critical gain"
Kc at which the output of the loop starts to oscillate. Kc and the oscillation period Pc
are used to set the gains as shown:
ZieglerNichols method

Control Type

Kp

Ki

Kd

0.5Kc

PI

0.45Kc 1.2Kp / Pc

PID

0.6Kc

2Kp / Pc KpPc / 8

[edit] PID tuning software

Most modern industrial facilities no longer tune loops using the manual calculation
methods shown above. Instead, PID tuning and loop optimization software are used to
ensure consistent results. These software packages will gather the data, develop
process models, and suggest optimal tuning. Some software packages can even
develop tuning by gathering data from reference changes.
Mathematical PID loop tuning induces an impulse in the system, and then uses the
controlled system's frequency response to design the PID loop values. In loops with
response times of several minutes, mathematical loop tuning is recommended,
because trial and error can literally take days just to find a stable set of loop values.
Optimal values are harder to find. Some digital loop controllers offer a self-tuning
feature in which very small setpoint changes are sent to the process, allowing the
controller itself to calculate optimal tuning values.
Other formulas are available to tune the loop according to different performance
criteria.

[edit] Modifications to the PID algorithm


The basic PID algorithm presents some challenges in control applications that have
been addressed by minor modifications to the PID form.
One common problem resulting from the ideal PID implementations is integral
windup. This can be addressed by:

Initializing the controller integral to a desired value


Disabling the integral function until the PV has entered the controllable region
Limiting the time period over which the integral error is calculated
Preventing the integral term from accumulating above or below predetermined bounds

Many PID loops control a mechanical device (for example, a valve). Mechanical
maintenance can be a major cost and wear leads to control degradation in the form of
either stiction or a deadband in the mechanical response to an input signal. The rate of
mechanical wear is mainly a function of how often a device is activated to make a
change. Where wear is a significant concern, the PID loop may have an output
deadband to reduce the frequency of activation of the output (valve). This is
accomplished by modifying the controller to hold its output steady if the change
would be small (within the defined deadband range). The calculated output must leave
the deadband before the actual output will change.
The proportional and derivative terms can produce excessive movement in the output
when a system is subjected to an instantaneous "step" increase in the error, such as a
large setpoint change. In the case of the derivative term, this is due to taking the
derivative of the error, which is very large in the case of an instantaneous step change.
As a result, some PID algorithms incorporate the following modifications:

derivative of output In this case the PID controller measures the derivative of
the output quantity, rather than the derivative of the error. The output is always
continuous (i.e., never has a step change). For this to be effective, the
derivative of the output must have the same sign as the derivative of the error.
setpoint ramping In this modification, the setpoint is gradually moved from
its old value to a newly specified value using a linear or first order differential
ramp function. This avoids the discontinuity present in a simple step change.

setpoint weighting Setpoint weighting uses different multipliers for the error
depending on which element of the controller it is used in. The error in the
integral term must be the true control error to avoid steady-state control errors.
This affects the controller's setpoint response. These parameters do not affect
the response to load disturbances and measurement noise.

[edit] Limitations of PID control


While PID controllers are applicable to many control problems, they can perform
poorly in some applications.
PID controllers, when used alone, can give poor performance when the PID loop
gains must be reduced so that the control system does not overshoot, oscillate or
"hunt" about the control setpoint value. The control system performance can be
improved by combining the feedback (or closed-loop) control of a PID controller with
feed-forward (or open-loop) control. Knowledge about the system (such as the desired
acceleration and inertia) can be "fed forward" and combined with the PID output to
improve the overall system performance. The feed-forward value alone can often
provide the major portion of the controller output. The PID controller can then be
used primarily to respond to whatever difference or "error" remains between the
setpoint (SP) and the actual value of the process variable (PV). Since the feed-forward
output is not affected by the process feedback, it can never cause the control system to
oscillate, thus improving the system response and stability.
For example, in most motion control systems, in order to accelerate a mechanical load
under control, more force or torque is required from the prime mover, motor, or
actuator. If a velocity loop PID controller is being used to control the speed of the
load and command the force or torque being applied by the prime mover, then it is
beneficial to take the instantaneous acceleration desired for the load, scale that value
appropriately and add it to the output of the PID velocity loop controller. This means
that whenever the load is being accelerated or decelerated, a proportional amount of
force is commanded from the prime mover regardless of the feedback value. The PID
loop in this situation uses the feedback information to effect any increase or decrease
of the combined output in order to reduce the remaining difference between the
process setpoint and the feedback value. Working together, the combined open-loop
feed-forward controller and closed-loop PID controller can provide a more
responsive, stable and reliable control system.
Another problem faced with PID controllers is that they are linear. Thus, performance
of PID controllers in non-linear systems (such as HVAC systems) is variable. Often
PID controllers are enhanced through methods such as PID gain scheduling or fuzzy
logic. Further practical application issues can arise from instrumentation connected to
the controller. A high enough sampling rate, measurement precision, and measurement
accuracy are required to achieve adequate control performance.
A problem with the Derivative term is that small amounts of measurement or process
noise can cause large amounts of change in the output. It is often helpful to filter the
measurements with a low-pass filter in order to remove higher-frequency noise
components. However, low-pass filtering and derivative control can cancel each other
out, so reducing noise by instrumentation means is a much better choice.
Alternatively, the differential band can be turned off in many systems with little loss
of control. This is equivalent to using the PID controller as a PI controller.

[edit] Physical implementation of PID control

In the early history of automatic process control the PID controller was implemented
as a mechanical device. These mechanical controllers used a lever, spring and a mass
and were often energized by compressed air. These pneumatic controllers were once
the industry standard.
Electronic analog controllers can be made from a solid-state or tube amplifier, a
capacitor and a resistance. Electronic analog PID control loops were often found
within more complex electronic systems, for example, the head positioning of a disk
drive, the power conditioning of a power supply, or even the movement-detection
circuit of a modern seismometer. Nowadays, electronic controllers have largely been
replaced by digital controllers implemented with microcontrollers or FPGAs.
Most modern PID controllers in industry are implemented in software in
programmable logic controllers (PLCs) or as a panel-mounted digital controller.
Software implementations have the advantages that they are relatively cheap and are
flexible with respect to the implementation of the PID algorithm.

[edit] Alternative nomenclature and PID forms


[edit] Pseudocode
Here is a simple software loop that implements the PID algorithm:
start:
previous_error = error or 0 if undefined
error = setpoint - actual_position
P = Kp * error
I = I + Ki * error * dt
D = (Kd / dt) * (error - previous_error)
output = P + I + D
wait(dt)
goto start

[edit] Ideal versus standard PID form


The form of the PID controller most often encountered in industry, and the one most
relevant to tuning algorithms is the "standard form". In this form the Kp gain is applied
to the Iout, and Dout terms, yielding:

Where
Ti is the Integral Time
Td is the Derivative Time
In the ideal parallel form, shown in the Controller Theory section

the gain parameters are related to the parameters of the standard form through
and Kd = KpTd. This parallel form, where the parameters are treated as
simple gains, is the most general and flexible form. However, it is also the form where
the parameters have the least physical interpretation and is generally reserved for
theoretical treatment of the PID controller. The "standard" form, despite being slightly
more complex mathematically, is more common in industry.

[edit] Series / interacting form


Another representation of the PID controller is the series, or "interacting" form. This
form essentially consists of a PD and PI controller in series, and it made early
(analog) controllers easier to build. When the controllers later became digital, many
kept using the interacting form.

[edit] See also

Control Theory
Feedback
Instability
Oscillation
Oscillation (mathematics)

[edit] External links


[edit] PID tutorials

What is PID? A Tutorial Overview


PID Tutorial
P.I.D. Without a PhD: a beginner's guide to PID loop theory with sample
programming code
What's All This P-I-D Stuff, Anyhow? Article in Electronic Design
Shows how to build a PID controller with basic electronic components go to
page 22
Tuning loops quickly at start-up

[edit] Simulations

PID controller using MatLab and Simulink


PID controller laboratory, Java applets for PID tuning
Good, basic PID simulation in Excel

[edit] Special topics and PID control applications

Proven Methods and Best Practices for PID Control


Inverted Pendulum Based on Microcontroller, a version of AN964 Microchip
PID Control Primer Article in Embedded Systems Programming

[edit] References

Liptak, Bela (1995). Instrument Engineers' Handbook: Process Control.


Radnor, Pennsylvania: Chilton Book Company, 20-29. ISBN 0-8019-8242-1.

Van, Doren, Vance J. (July 1, 2003). "Loop Tuning Fundamentals". Control


Engineering. Red Business Information.
Sellers, David. An Overview of Proportional plus Integral plus Derivative
Control and Suggestions for Its Successful Application and Implementation
(PDF). Retrieved on 2007-05-05.
Articles, Whitepapers, and tutorials on PID control
Graham, Ron (10/03/2005). FAQ on PID controller tuning. Retrieved on 200705-05.

Retrieved from "http://en.wikipedia.org/wiki/PID_controller"


Categories: Control theory | Control engineering | Control devices | Classical control

How do PID Controllers Work


Common Sense / Practical Approach
Enough of the technical stuff -- let's look at PID control from a common sense /
practical approach. On this page we will build a PID controller using Microsoft Excel
so that you can see, and experiment with, the results that we are talking about.

Let's think about how we would design a PID


controller
We focus on the difference (error) between the process variable (PV) and the setpoint
(SP). There are three ways we can view the error.
1. The absolute error. This means how big is the difference between the PV and
SP. If there is a small difference between the PV and the SP -- then let's make
a small change in the output. If there is a large difference in the PV and SP -then let's make a large change in the output. Absolute error is the
"proportional" (P) component of the PID controller.

2. The sum of errors over time. Give us a minute and we will show why simply
looking at the absolute error (proportional) only is a problem. The sum of
errors over time is important and is called the "integral" (I) component of the
PID controller. Every time we run the PID algorithm we add the latest error to
the sum of errors. In other words Sum of Errors = Error1 + Error2 + Error3 +
Error4 + ...

3. The dead time. Dead Time refers to the delay between making a change in the
output and seeing the change reflected in the PV. The classical example is
getting your oven at the right temperature. When you first turn on the heat, it
takes a while for the oven to "heat up". This is the dead time. If you set an

initial temperature, wait for the oven to reach the initial temperature, and then
you determine that you set the wrong temperature -- then it will take a while
for the oven to reach the new temperature setpoint. This is also referred to as
the "derivative" (D) component of the PID controller. This holds some future
changes back because the changes in the output have been made but are not
reflected in the process variable yet.

Absolute Error / Proportional


One of the first ideas people usually have about designing an automatic process
controller is what we call "proportional". Meaning, if the difference between the PV
and SP is small -- then let's make a small correction to the output. If the difference
between the PV and SP is large -- then let's make a larger correction to the output.
This idea certainly makes sense.
We simulated a proportional only controller in Microsoft Excel (if you click on this
link then you will download the Excel spreadsheet). Here is the chart showing the
results of the first simulation:

In this first example, we assumed that there was no dead time, meaning, that if we
made a change in the output of the controller, the input immediately changed. For
example, zero dead time on our oven means that if we changed the temperature
setpoint on the oven, then the temperature inside the oven instantly changed to the
new setpoint (the oven did not require time to heat up or cool down).
The blue line represents a proportional constant of .1, the magenta lines represents a
proportional constant of .2, the yellow line represents a proportional constant of .4,
and the white line represents the setpoint (SP). From this graph, hopefully two things
jump out at you. First, once the output settles out, the output (blue, magenta, and

yellow lines) are no where near the setpoint (SP) (the white line). Therefore, some
offset has to be added to the output to make the PV reach the SP. Second, the greater
the proportional constant, the less the offset needs to be. For example the yellow line,
with a proportional constant = .4 is closer to the white line than the blue line with a
proportional constant of .1.
If you download the Excel spreadsheet of the PID controller simulator and look at the
effects of increasing dead time you will notice that the outputs settle at the same
output level -- it simply takes longer for the output to reach its final level.
In summary, automatic proportional (only) controllers are not very good because there
is an offset that has to be continually adjusted.

Proportional and Integral Controllers


The integral portion of the PID controller accounts for the offset problem in a
proportional only controller. We have another Excel spreadsheet that simulates a PID
controller with proportional and integral control. Here is a chart of the first
simulation with proportional and integral:

As you can tell, the PI controller is much better than just the P controller. However,
dead time of zero (as shown in the above graph) is not common. So let's take a look
when the dead time equals two.

Now this graph is starting to look more typical of a PID controller. Notice how the
dark blue line quickly goes up to the SP (50) and cycles around 50 a little but quickly
settles down. In contrast, the dark purple line way overshoots the SP of 50, going
above 80, back down to 30, then over 50, and back and forth until it eventually settles
down.
If you download the Excel spreadsheet and look through the different scenarios you
will notice that the P & I parameters that look good for one dead time do not look
optimal for another dead time. In other words, for each process element (valve,
motor, pump, heater, chiller, etc) you are trying to control -- you will have different
process characteristics and will have to determine the optimal P, I, and possibly D
constants. Determining what these constants should be is called "tuning".
Theoretically, you want to minimize the sum of absolute errors, as given in the
spreadsheets.
Let's show one other graph to warn you about a very dangerous condition:

We wanted to show this graph to illustrate what can happen if you choose the wrong
parameters. The green line illustrates an unstable or "out-of-control" controller.
Notice how it continues to get worse and worse. This is not good. This is why you
want to start with very small P, I, and D constants and increase them to improve
performance. If you start with large constants, bad things can happen.

Derivative Control
Derivative control takes into consideration that if you change the output, then it takes
time for that change to be reflected in the input (PV). For example, let's take heating
of the oven. If we start turning up the gas flow, it will take time for the heat to be
produced, the heat to flow around the oven, and for the temperature sensor to detect
the increased heat. Derivative control sort of "holds back" the PID controller because
some increase in temperature will occur without needing to increase the output
further. Setting the derivative constant correctly, allows you to become more
aggressive with the P & I constants.

Actual PID Controllers


PID controllers can be stand-alone controllers (also called single loop controllers),
controllers in PLCs, embedded controllers, or software in Visual Basic or C#
computer programs.

Typically each manufacturer's PID controller acts differently than controllers from
other manufacturers. In other words, do not expect PID controllers from different
manufacturers to act exactly the same.
Many PID controllers today have an "auto tune" feature that will calculate good
values for the P, I, & D constants for a process.
We typically use a product called Expertune which does exactly what it's name
implies.
Remember that as a systems integrator you do not need to be a master of every
technology. You are as good as the resources you manage. Therefore, learn what you
can do, and what you can not do, and then call in the experts listed below to help you
with the difficult applications.

Ziegler Nichols Tuning Method


There is a manual way of calculating PID constants known as the
Ziegler Nichols tuning method. We will give a brief example using
another Excel spreadsheet.
1. Initially set the integral and derivative constants to zero -- proportional control
only

2. Increase the proportional constant until you get a sinusoidal wave with a
constant amplitude (see Excel spreadsheet for examples at different
deadtimes). In our Excel PID simulations, this sinusoidal wave occurs when
the proportional constant = 1.0. See the chart below where the Deadtime = 1,
Proportional constant = 1, and the period = 4. Note that the proportional
constant of 1.0 was determined to give the process a constant amplitude wave.

3. For optimal P & I controller (no derivative), the proportional constant should
be 0.45 times the proportional constant. In this case 0.45 times 1.0 = 0.45.

4. The integral constant is 1.2 / period of the sinusoidal wave. For deadtime = 0,
the period is 2 so the optimal integral constant = 1.2 / 2 = 0.6. For deadtime =
1, the period is 4 so the integral constant = 1.2 / 4 = 0.3. For deadtime = 2, the
period is 6 so the integral constant = 1.2 / 6 = 0.2. The new PI controller
results are shown below.

More PID Information


We have only touched on the "tip of the iceberg" for PID controllers. For small, stand
alone applications, the information you learned from this tutorial may solve many of
the PID applications. For large, complex systems -- what we presented might only get
you in trouble. As with any new ideas or technology -- start working on a small, non
critical application first and as your experience grows then start tackling more
complex projects.

Expertune -- makers of software that automatically determine what your P, I,


& D constants should be. They also have much more elaborate products that
look at the interactions of multiple parameters in control systems.
PID Website by John Shaw -- good tutorial and information on PID control

PID Tutorial
John A. Shaw
Process Control
Solutions
PID Control
Description
Tutorial
Ch. 1
Ch. 2
Ch. 3
Ch. 4
Ch. 5
Ch. 6
Code

Multiple Loop
Cascade
Ratio
Override
Trim

The PID control algorithm is used for the control of almost all loops in the
process industries, and is also the basis for many advanced control
algorithms and strategies. In order for control loops to work properly, the
PID loop must be properly tuned. Standard methods for tuning loops and
criteria for judging the loop tuning have been used for many years, but
should be reevaluated for use on modern digital control systems.
While the basic algorithm has been unchanged for many years and is used
in all distributed control systems, the actual digital implementation of the
algorithm has changed and differs from one system to another and from
commercial equipment to academia.
We will discuss controller tuning methods and criteria. Also discussed will
be the digital PID control algorithm, how it works, the various
implementation methods and options, and how these affect the operation
and tuning of the controller.

Table of Contents
1. The Control Loop
Basic feedback control
Valve Linearity
Valve Linearity:
Fail Open Valves

2. Process responses
Steady state: effect of controller output
Steady state: effect of disturbances
Process Dynamics: Simple lag
Process Dynamics: Dead time
Measurement of dynamics
Disturbances

3. The PID algorithm


Action
Auto/Manual
Key concepts
Proportional
Proportionalunits
ProportionalOutput vs. Measurement
ProportionalOffset
ProportionalReducing offset with manual reset

Adding automatic reset


Reset or integral mode
Derivative
Complete PID response

4. Additional PID concepts


Interactive or Noninteractive algorithm
Converting between interactive and non-interactive
External feedback
Saturation Properties

5. Other controller features


Gain on process rather than error
Derivative on process rather than error

6. Loop tuning
Tuning Criteria
Mathematical criteria
On-line trial tuning
Ziegler Nichols tuning method: open loop reaction rate
Ziegler Nichols tuning method: open loop point of inflection
Ziegler Nichols tuning method: open loop process gain
Ziegler Nichols tuning method: closed loop
Controllability of processes
Flow loops

PID Chap. 1 - The Proce


John A. Shaw
Process Control
Solutions
PID Control
Description
Tutorial
Ch. 1
Ch. 2
Ch. 3
Ch. 4
Ch. 5
Ch. 6
Code

Multiple Loop
Cascade
Ratio
Override
Trim

The Feedback Control Loop

The system measures the process, compares it to a setpoint, and


manipulates the output in the direction which should move the pro
the setpoint.

Valve Linearity

Valves are usually non-linear. That is, the flow through the valve is
same as the valve position. Several types of valves exist:
Linear
Same gain regardless of valve position
Equal Percentage
Low gain when valve is nearly closed
High gain when valve is nearly open
Quick Opening
High gain when valve is nearly closed
Low gain when valve is nearly open

As we will see later, the gain of the process, including the valve, is

important to the tuning of the loop.

If the controller is tuned for one process gain, it may not work for
process gains.

ValveLinearity:
Installedcharacteristics

Theflowvs.percentopencurvechangesduetotheheadlossinthepipin

Atlowflow,theheadlossthroughthepipesisless,leavingalargerdiff
pressureacrossthevalve.

Athighflow,theheadlossthroughthepipeismore,leavingasmallerd
pressureacrossthevalve.

Theeffectistoincreasethenonlinearityofmostvalves.

FailOpenValves
Valvesareusuallyeither:FailClosed,airtoopenor
FailOpen,airtoclose

Regardlessofthewaythevalveoperates,theoperatorisin
theknowingandadjustingthepositionofthevalve,notthe
signal."Upisalwaysopen"

Allcontrollershavesomemeansofindicatingthecontroll
termsofthevalveposition.Whentheoperatorincreasesthe
indicatedonthecontroller,thevalveopens.
IndicationInversion

>
The output indication is inverted.
The controller action takes the valve action into acount.
The flow loop is direct acting.
Most analog controllers work like this.
Signal Inversion

The output signal is inverted. The controller action ignores the val
The flow loop is reverse acting. Some distributed control systems
this.

PID Chap. 2 - The Control


John A. Shaw
Process Control
Solutions
PID Control
Description
Tutorial
Ch. 1
Ch. 2
Ch. 3
Ch. 4
Ch. 5
Ch. 6
Code

Multiple Loop
Cascade
Ratio
Override
Trim

The Process Response to the


Controller
Steady state relationships:
changing load

When the load changes, either the process value changes or the v
position must be changed to compensate for the load change.

Process Dynamics: Simple lag

Process Dynamics: Dead time

Dead Time: A delay in the loop due to the time it takes material to
one point to another
Also called: Distance Velocity Lag
Transportation Lag

Most loop combine dead time and lag.

Measurement of dynamics
The dynamics differ from one loop to another.
However, they usually result in a response curve like this:

L is Lagthe largest lag in the process loop.

D is "Pseudo Deadtime"the sum of the deadtime and all lags oth


the largest lag.

Disturbances
Almost all processes contain disturbances.
Disturbances can enter anywhere in the process.

The effect of the disturbance can depend on where it enters the lo


Most disturbances cannot be measured.

back to index

PID Chap. 3 - PID Algorithm


John A. Shaw
Process Control
Solutions
PID Control
Description
Tutorial
Ch. 1
Ch. 2
Ch. 3
Ch. 4
Ch. 5
Ch. 6
Code

Multiple Loop
Cascade
Ratio
Override
Trim

The PID algorithm


Controller Action
PROCESSACTION
Definestherelationshipbetweenchangesinthevalveandchangesinthe
measurement.
DIRECTIncreaseinvalvepositioncausesanincreaseinthemeasurement.
REVERSEIncreaseinvalvepositioncausesadecreaseinthemeasurement.

CONTROLLERACTION
Definestherelationshipbetweenchangesinthemeasuredvariableandchangesin
thecontrolleroutput.
DIRECTIncreaseinmeasuredvariablecausesanincreaseintheoutput.
REVERSEIncreaseinmeasuredvariablecausesadecreaseintheoutput.

Thecontrolleractionmustbetheoppositeoftheprocessaction.

Auto/Manual
ManualMode:
Theoperatoradjusttheoutputtooperatetheplant.
Duringstartup,thismodeisnormallyused.


AutomaticMode:
Thecontrolalgorithmmanipulatestheoutputtoholdtheprocessmeasurementsat
theirsetpoints.

Thisshouldbethemostcommonmodefornormaloperation.

Keyconcepts
ThePIDcontrolalgorithmdoesnot"know"thecorrectoutputtobringthe
processtothesetpoint.
Itmerelycontinuestomovetheoutputinthedirectionwhichshouldmovethe
processtowardthesetpoint.

Thealgorithmmusthavefeedback(processmeasurement)toperform.

ThePIDalgorithmmustbe"tuned"fortheparticularprocessloop.Without
suchtuning,itwillnotbeabletofunction.
TobeabletotuneaPIDloop,eachofthetermsofthePIDequationmustbe
understood.
Thetuningisbasedonthedynamicsoftheprocessresponse.

ThePIDControlAlgorithm
The PID control algorithm comprises three elements:

Proportional - also known as Gain


Integral - also known as Automatic Reset or
simply Reset

Derivative

- also known as Rate or Pre-Act (TM of


Taylor Instrument Co.)

The algorithm is normally available in several


combinations of these elements:

Proportional only
Proportional and Integral (most common)

Proportional, Integral, and Derivative

Proportional and Derivative

We will examine each of the three elements below:


Proportional
E=MeasurementSetpoint(directaction)

E=SetpointMeasurement(reverseaction)
Output=E*G+k

Theoutputisequaltotheerrortimesthegainplusthemanualreset.
Ifthemanualresetstaysconstant,thereisafixedrelationshipbetweenthe
setpoint,themeasurement,andtheoutput.

Proportionalunits
Theproportionalorgaintermmaybecalibratedintwoways:
GainandProportionalBand
Gain=Output/Input
Increasingthegainwillcausetheoutputtomovemore.
Proportionalbandisthe%changeintheinputwhichwouldresultina100%
changeintheoutput.
ProportionalBand=100/Gain
Wewillusegaininthisdocument.

ProportionalOutputvs.Measurement

(Reverseacting)

Proportionalonlycontrolproducesanoffset.Onlytheadjustmentofthemanual
resetremovestheoffset.

ProportionalOffset
Offsetcanbereducedbyincreasinggain.

Proportionalcontrolwithlowgain

Proportionalcontrolwithhighergain

ProportionalReducingoffsetwithmanualreset
Offsetcanbeeliminatedbychangingmanualreset.

Proportionalcontroldifferentmanualreset

Addingautomaticreset
Withproportionalonlycontrol,theoperator"resets"thecontroller(toremove

offset)byadjustingthemanualreset:

Thismanualresetmaybereplacedbyautomaticresetwhichcontinuestomove
theoutputwheneverthereisanyerror:

Thisiscalled"Reset"orIntegralAction.
Notetheuseofthepositivefeedbacklooptoperformintegration.

Resetorintegralmode
ResetContribution:
Out=gXKrXintegral of error

wheregisgain,Kristheresetsettinginrepeatsperminute.

Unitsusedtosetintegralorreset
Assumeacontrollerwithproportionalandintegralonly.
Calculationofrepeattime:(gainandresettermsusedincontroller)
Withtheerrorsettozero(measurementinput=setpoint),makeachangeinthe
inputandnotetheimmediatechangeinoutput.Theoutputwillcontinuetochange
(itisintegratingtheerror).Notethetimeittakestheoutputto,duetotheintegral
action,repeattheinitialchangemadebythegainaction.

Somecontrolvendorsmeasureresetbyrepeattimeinminutes.Thisisthetimeit
takesthereset(orintegral)elementtorepeattheactionoftheproportional
element.
Othersmeasureresetby"repeatsperminute".

Repeatsperminuteistheinverseofminutesofrepeat
Thisdocumentwilluserepeatsperminute.

Derivative
Firstusedasapartofatemperaturetransmitter("SpeedAct"Taylor
InstrumentCompanies)toovercomelagintransmittermeasurement.
AlsoknownasPreActandRate.
DerivativeContribution:
Out=gxKdxde/dt
wheregisgain,Kdisthederivativesettinginminutes.

Responseofcontrollerwithproportionalandderivative:

Theamountoftimethatthederivativeactionadvancestheoutputisknownasthe
"derivativetime"measuredinminutes.
Allmajorvendorsmeasurederivative(Derivative,Rate)thesame.

CompletePIDresponse
NonInteractive(textbook)form:

Out=G(e+R+D )
Where
G=Gain
R=Reset(repeatsperminute)
D=Derivative(minutes)
Note:SeeInteractivevs.Noninteractive(below)
backtoindex

PID Chap. 4 - PID Concepts


John A. Shaw
Process Control
Solutions
PID Control
Description
Tutorial
Ch. 1
Ch. 2
Ch. 3
Ch. 4
Ch. 5
Ch. 6
Code

Multiple Loop

Additional PID Concepts


Interactive or Noninteractive PID
Algorithm
"Interactive" and "Noninteractive" refer to interaction between the reset and
derivative terms. This is also known as "series" or "parallel" derivative.

Cascade
Ratio
Override
Trim

Almost all analog controllers are interactive.


Many digital controllers are non-interactive, some are interactive
The only difference is in the tuning of controllers with derivative.

Non-Interactive (Parallel):

Out = G(e + R+ D )
Interactive (series):

Out = (RD+1)G(e + R+D )


Converting between interactive and non-interactive
Applies only to 3-mode controllers
To convert from non-interactive to interactive:
Gn = Gi (1 + Ri Di)
Rn = Ri/(1 + Ri Di)
Dn = Di/(1 + Ri Di)
In other words, with a non-interactive controller the gain

should be higher, the reset rate lower, and the derivative lower
than on a commercial interactive controller.

External feedback

The integral function implemented using a positive feedback.

If the input to the positive feedback loop is taken from the signal to the
process, it is called "external feedback" or "reset feedback". At steady state
the controller output is the Gain multiplied by Error added to external
feedback. If the error is zero, the output is equal to the external feedback.

Saturation Properties
Another difference is in the "Saturation Properties"
eg. what happens when output has been at the upper or lower limit.

Standard algorithm
Described on previous page.

Output stays at limit until measurement crosses setpoint.

"Integrated velocity form"


Similar to equation:
Output = Last output + gain x (error - last error + reset x error)
Output pulls away from limit one reset time before measurement crosses
setpoint.

For most applications, there is no difference. For some batch startup


problems, the "integrated velocity form" algorithm works best.
Standard works best for high gain/low reset rate applications.

PID Chap. 5 - Controller Features


John A. Shaw
Process Control
Solutions
PID Control
Description
Tutorial
Ch. 1
Ch. 2
Ch. 3
Ch. 4
Ch. 5
Ch. 6
Code

Multiple Loop
Cascade
Ratio
Override
Trim

Other Controller Features


Gain on process rather than error
In applications with high gain, a step change can result in a sudden, large
movement in the valve.
Not as severe as the derivative effect, but still can upset the process
Solution: let gain act only on process rather than error.

Derivative on process rather than error


A step change in the setpoint results in a step change in the error.
The derivative term acts on the rate of change of the error.
The rate of change of a step change is very large.
An operator step change of the setpoint would cause a very large change

in the output, upsetting the process.


Solution: let derivative act only on process rather than error.
back to index

PID Chap. 6 - Loop Tuning


John A. Shaw
Process Control
Solutions
PID Control
Description
Tutorial
Ch. 1
Ch. 2
Ch. 3
Ch. 4
Ch. 5
Ch. 6
Code

Multiple Loop
Cascade
Ratio
Override
Trim

Loop Tuning
Tuning Criteria
or
"How do we know when its tuned"
Elementary methods
1 The plant didnt blow up.
2 The process measurements stay close enough to the setpoint.
3 They say its OK and you can go home now.
Informal methods
1 Optimum decay ratio (1/4 wave decay).

2 Minimum overshoot.

3 Maximum disturbance rejection.

The choice of methods depends upon the loops place in the process and its
relationship with other loops.

Mathematical criteria
Mathematical methodsminimization of index

IAE - Integral of absolute value of error


ISE - Integral of error squared
ITAE - Integral of time times absolute value of error
ITSE - Integral of time times error squared:
These mathematical methods are used primarily for academic purposes,
together with process simulations, in the study of control algorithms.

On-line trial tuning


or

The "by-guess-and-by-golly" method


1. Enter an initial set of tuning constants from experience. A conservative
setting would be a gain of 1 or less and a reset of less than 0.1.
2. Put loop in automatic with process "lined out".
3. Make step changes (about 5%) in setpoint.
4. Compare response with diagrams and adjust.

Ziegler Nichols tuning method: open loop reaction


rate
Also known as the "reaction curve" method
The process must be "lined out"not changing.
With the controller in manual, the output is changed by a small amount.
The process is monitored.

The following measurements are made from the reaction curve:


X
R

%
%/min.

Change of output
Rate of change at the point of inflection (POI)

D min.
value

Time until the intercept of tangent line and original process

The gain, reset, and Derivative are calculated using:


Gain

Reset

Derivative

X/DR

PI

0.9X/DR

PID

1.2X/DR

0.3/D
0.5/D

0.5D

Ziegler Nichols tuning method: open loop point of


inflection
Another means of determining parameters based on the ZN open loop.
After "bumping" the output, watch for the point of inflection and note:

Ti

min

Time from output change to POI

Process value change at POI

%/min

Rate of change at POI (Same as above method)

Change in output. (Same as above method)

D is calculated using the equation:


D=Ti - P/R
D & X are then used in the equations on the previous page.

Ziegler Nichols tuning method: open loop process


gain
Mathematically derived from the reaction rate method.
Used only on processes that will stabilize after output step change.
The process must be "lined out"not changing.

With the controller in manual, the output is changed by a small amount.


The process is monitored.

Gp is the process gain - the change in measured value (%) divided by the
change in output (%)
The gain, reset, and Derivative are calculated using:
Gain

Reset

Derivative

L/GpD

PI

0.9 L/GpD

PID

1.2 L/GpD

0.3/D
0.5/D

0.5D

Ziegler Nichols tuning method: closed loop


Steps
Place controller into automatic with low gain, no reset or derivative.
Gradually increase gain, making small changes in the setpoint, until
oscillations start.
Adjust gain to make the oscillations continue with a constant amplitude.

Note the gain (Ultimate Gain, Gu,) and Period (Ultimate Period, Pu.)
The Ultimate Gain, Gu, is the gain at which the oscillations continue with a
constant amplitude.

The gain, reset, and Derivative are calculated using:


Gain

Reset

Derivative

0.5 GU

PI

0.45 GU

1.2/Pu

PID

0.6 GU

2/Pu

Pu/8

Controllability of processes
The "controllability" of a process is depends upon the gain which can be
used.
The higher the gain:
the greater rejection of disturbance and
the greater the response to setpoint changes.

The predominate lag L is based on the largest lag in the system.


The subordinate lag D is based on the deadtime and all other lags.

The maximum gain which can be used depends upon the ratio .
From this we can draw two conclusions:
Decreasing the dead time increases the maximum gain and the
controllability.
Increasing the ratio of the longest to the second longest lag also increases
the controllability.

Flow loops
Flow loops are too fast to use the standard methods of analysis and tuning.
Analog vs. Digital control:

Some flow loops using analog controllers are tuned with high gain.
This will not work with digital control.

With an analog controller, the flow loop has a predominate lag (L) of a few
seconds and no subordinate lag.
With a digital controller, the scan rate of the controller can be considered
dead time.
Although this dead time is small, it is large enough when compared to L to
force a low gain.
Typical digital flow loop tuning: Gain= 0.5 to 0.7
Reset=15 to 20 repeats/min..
no derivative.

Loop Tuning Fundamentals


PID loop tuning may not be magic, but its intricacies
do lie somewhere between science and art. The
following proven tuning tips will help you craft
your processes.
Vance J. VanDoren, Control Engineering -- Control Engineering,
7/1/2003

AT A GLANCE
A control loop is a feedback mechanism that
attempts to correct discrepancies between a
measured process variable and the desired setpoint.
The controller applies the necessary corrective
actions via an actuator that can drive the process
variable up or down.

PID explained
Value selection
tips
Open-and closedloop tuning
Software product
listing

A proportional-integral-derivative (PID) controller


tracks the error between the process variable and the
Sidebars:
setpoint, the integral of recent errors, and the rate by
Help is available
which the error has been changing. It computes its
next corrective action from a weighted sum of those
three terms (or modes), then outputs the results to the process and awaits the next
measurement.
PID basics

A PID controller using the ideal or ISA standard form of the PID algorithm computes
its output CO(t) according to the formula shown in Figure 1. PV(t) is the process
variable measured at time t and the error e(t) is the difference between the process
variable and the setpoint. The PID formula weights the proportional term by a factor
of P, the integral term by a factor of P/T I, and
the derivative term by a factor of P. T D where
P is the controller gain, T I is the integral time,
and T D is the derivative time.
This terminology bears some explaining. Gain
refers to the amount by which the error signal CO(t) is the controller's current output
e(t) = SP - PV (t) is the error between the set point
will gain or lose strength as it passes through
(SP) and the process variable PV(t)
the controller en route to becoming part of the P is the controller gain
controller's output. A PID controller with a high T is the integral time T is the derivative time
gain will tend to generate aggressive corrective actions to eliminate errors.
1

The integral time refers to a hypothetical sequence of events where the error starts at
zero then abruptly jumps to a fixed value. Such an error would cause an instantaneous
response from the controller's proportional term and a response from the integral term
that starts at zero and increases steadily. The time required for the integral term to
catch up to the unchanging proportional term is the integral time T I. A PID controller
with a long integral time is more heavily weighted towards proportional action
thanintegral action.
Similarly, the derivative time T D is a measure of the relative influence of the
derivative term in the PID formula. If the error were to start at zero and begin
increasing at a fixed rate, the proportional term would start at zero while the
derivative term assumed a fixed value. The proportional term would then increase
steadily until it caught up to the derivative term at the end of the derivative time. A
PID controller with a long derivative time is more heavily weighted towards
derivative action than proportional action.
Historical note
The very first feedback controllers included just the proportional term. For
mathematical reasons that only became apparent later, a P-only controller tends to
drive the error downward to a small but non-zero value and then quit. Operators
observing this phenomenon would then manually increase the controller's output until

the last vestiges of the error were eliminated. They called this operation resetting the
controller.

An open-loop step test reveals the prcess' time constant T, deadtime d, and
gain k.

When the integral term was


introduced, operators observed
that it tended to perform the reset
operation automatically. That is,
the controller would augment its
proportional action just enough
to eliminate the error entirely.
Hence, integral action was
originally called automatic reset
and remains labeled that way on
some PID controllers to this day.
The derivative term was invented
shortly thereafter and was

described, accurately enough, as rate control.


Tricky business
Tuning is the art of selecting values for the tuning parameters P, T I, and T D so that
the controller will be able to eliminate an error quickly without causing the process
variable to fluctuate excessively. That's easier said than done.
Consider a car's cruise controller, for example. It can accelerate the car to a desired
cruising speed, but not instantaneously. The car's inertia causes a delay between the
time that the controller engages the accelerator and the time that the car's speed
reaches the setpoint. How well a PID controller performs depends in large part on
such lags.
Suppose an overloaded car with an undersized engine suddenly starts up a steep hill.
The ensuing error between the car's actual and desired speeds would cause the
controller's derivative and proportional actions to kick in immediately. The controller
would begin to accelerate the car, but only as fast as the lag allows.
After a while, the integral action would also begin to contribute to the controller's
output and eventually come to dominate it (since the error decreases so slowly when
the lag time is long, and a sustained error is what drives the integral action). But
exactly when that would happen and how dominant the integral action would become
thereafter would depend on the severity of the lag and the relative sizes of the
controller's integral and derivative
times.
This simple example demonstrates a
fundamental principle of PID tuning.
The best choice for each of the tuning
parameters P, T I, and T D depends on
Forcing the closed-loop system into sustained oscillations with a
the values of the other two as well as proportional-only controller reveals the ultimate gain P and the
ultimate period T ..
the behavior of the controlled
process. Furthermore, modifying the tuning of any one term affects the performance
of the others since the modified controller affects the process and the process, in turn,
affects the controller.
u

Ziegler-Nichols tuning

So how can a control engineer designing a PID loop determine the values for P, T I,
and T D that will work best for a particular application? John G. Ziegler and Nathaniel
B. Nichols of Taylor Instruments (now part of ABB Instrumentation in Rochester,
NY) addressed that question in 1942 when they published two loop-tuning techniques
that remain popular to this day.
Their open-loop technique is based on the results of a bump or step test for which the
controller is taken off-line and manually forced to increase its output abruptly. A strip
chart of the process variable's subsequent trajectory is known as the reaction curve
(see Figure 2).
A sloped line drawn tangent to the reaction curve at its steepest point shows how fast
the process reacted to the step change in the controller's output. The inverse of this
line's slope is the process time constant T which measures the severity of the lag.
The reaction curve also shows how long it took for the process to demonstrate its
initial reaction to the step (the dead time d) and how much the process variable
increased relative to the size of the step (the process gain K). By trial-and-error,
Ziegler and Nichols determined that the best settings for the tuning parameters P, T I,
and T D could be computed from T, d, and K as follows:

Once these parameter settings have been loaded into the PID formula and the
controller returned to automatic mode, the controller should be able to eliminate
future errors without causing the process variable to fluctuate excessively.
Ziegler and Nichols also described a closed loop-tuning technique that is conducted
with the controller in automatic mode, but with the integral and derivative actions shut
off. The controller gain is increased until even the slightest error causes a sustained
oscillation in the process variable (see Figure 3).
The smallest controller gain that can cause such an oscillation is called the ultimate
gain P u. The period of those oscillations is called the ultimate period T u. The
appropriate tuning parameters can be computed from these two values according to
the following rules:
P = 0.6 Pu
TI = 0.5 Tu
TD = 0.125 Tu

Caveats

Unfortunately, PID loop tuning isn't really that simple. Different PID controllers use
different versions of the PID formula, and each must be tuned according to the
appropriate set of rules. The rules also change when:

The derivative and/or the integral action are disabled.


The process itself is inherently oscillatory.
The process behaves as if it contains its own integral term (as is the case with
level control).
The deadtime d is very small or significantly larger than the time constant T.

Furthermore, Ziegler and Nichols had a particular closed-loop performance objective


in mind when they settled on their particular tuning rules. They chose to allow some
fluctuations in the process variable so long as each successive peak was no more than
one-fourth the size of its predecessor (so-called quarter wave decay). For applications
that require even less fluctuation, additional tweaking of the tuning parameters is
required.
This is where loop tuning becomes an art. It takes more than a little experience and
sometimes a lot of luck to come up with just the right combination of P, T I, and T D.

What's All This P-I-D Stuff, Anyhow?


Bob Pease | ED Online ID #6131 | June 26, 1995

Recently, I wrote about refrigerators,1 pointing out that there are several ways to control a servo loop, such as
a temperature chamber, or an oven, or a refrigerator using thermoelectric coolers (let's leave bang-bang
controllers and on-off heat-movers out of this). Fuzzy Logic controllers can work pretty well, and so can a P-ID (or PID) controller. That's pronounced "pee-eye-dee", not "pid". Several readers said that they were not very
knowledgeable about PID controllers. They don't teach very much about them in schools these days, I guess.
They asked me, "Please show us a good example of a PID controller." Well, I agree completely that an
example is one of the most powerful tools. I'll show you a couple of examples, so you can see how easy it is to
come up with one yourself. And I will point out that, after a Fuzzy Logic expert showed us his best example of
a nice simple F.L. controller, I had no idea how to make it myself. Do you know how to run a F.L. controller
after seeing an example of one? I don't. I hope that would not be true for my examples.
One example is found in my book on Troubleshooting, 2 where I had to control the temperature of a blast of
heated air. When you apply more watts to the heater, there's a delay before the sensor warms up to its new
temperature. In fact, there are transport delays and thermal lags. This is a fairly interesting kind of system for
closing the loop accurately, but not really difficult. Engineers have known how to do this for many years
about 140 years, I would say. Back in the 1880s, when most servo loops were mechanical or pneumatic, and
the instrumentation was crude, it was a wise person who understood how to close a loop with it good accuracy
and loop stability. But for the last 40 years, when a good operational amplifier costs $22 or even less, it's been
a piece of cake.

Note, a wise old colleague observed that the introduction of the Integral term to Control Theory is credited to
the 1930s. But I found good documentation in my Encyclopedia Brittanica3 that a flyball governor with an
added integral term was invented by Sir W. Siemens in 1853.4 Never bet that the British didn't get there first.
However, I can't say that I've seen the Derivative effect exploited in that 1894 Encyclopedia. So maybe the
PID controller is only about 60 years old...
First of all, let's discuss the nomenclature"PID." That stands for Proportional, and Integral, and Derivative.
You can build some controllers using only P and I, and others using only P and D, but when you need good
performance, using all three terms can provide REAL advantages. Let's see how these terms are made, and
how they are used.
First, these functions are used to operate on the error signal, the difference between the feedback parameter
and the desired parameter. Let's spell out an example. Say that we want to define a precision heater controller
perhaps for an electric frying panwith a sensor for the controlled chamber that puts out 10mV/C. The
input command is -350mV (which corresponds to a desired temperature or "set point" of +35C),the output of
the temperature sensor is +250mV (which coresponds to a temperature of 25C), and the load must be heated
to get the feedback voltage to track (equal but with opposite sign). What's needed, then, is a circuit to operate
on the error, namely (Vout + Vin), or -0.1 V. Op amp A1 does just that (Fig. 1). (Let's keep things simple by
considering primarily linear systems; if the system actually has some nonlinearities, we can address them
later.)
After we generate that error term, you will want to generate a correction signal that's a function of the Error
Signal. As I discussed back in December, you might design your system so that a heater has its watts linearly
Proportional to the temperature error. "If chamber temperature is very cold, turn heat up high" is how the
Fuzzy Logic guys like to say this. This is partly wise, because if the temperature really is too cold, turning on
the heat is one of the good things to do. That is the Proportional term.
Referring to Figure, we follow the error-detector amplifier A1 with a Proportional amplifier A2. We can control
the gain of the Proportional path by adjusting the pot P2 at its output, so as to get the right gain going to the
power amplifier, A5. Eventually, we will figure out what to do with A3 and A4, but right now we can set their trim
pots to ZERO, and then they're out of the picture. Let's keep things simple, one step at a time.
Now, let's say you pour a bucket of very cold water into the electric frying pan, where the controller is set for
+35C. The sensor soon says it's much too cold, so the heater turns on pretty hard. As the temperature of the
sensor gets near the desired temperaturethe set pointthe heater will eventually be turned off. The problem
is that any heater has a delay before its heat gets to the chamber and load and sensor. So, when the error
gets to zero, and you turn off power to the heater, the chamber still keeps on heating for a while and overshoot
occurs. If you turn the gain down low, this overshoot may be minor. But if you decide that you must have very
high accuracy and turn up the gain, overshoot is certain, and oscillation or bad ringing is likely.
Now, how can we avoid this overshoot by foreseeing this situation and recognizing that the power needs to be
shut down a little early? The best solution is to add in A3 to compute the Derivative of the temperature error
the rate of changewhile the proportional amplifier is computing when the error becomes small. When these
two signals are combined properly, the Derivative signal lets the controller decide, "Whoa, we are getting very
close to the set point, and the sensor's temperature is still rising pretty rapidlytime to cut back on the power."
In practice, this works quite well. This is called P-D control, using just A2 and A3 for the Proportional and
Derivative terms. You trim P3 to the setting that gives good resultsnot too much overshoot (not enough
derivative) and not too slow (too much derivative term). The Fuzzy Logic guys achieve this same function by
using the words: "If the temperature error is small (negative) and the rate-of-change is small (positive), heating
power should be small." This works, too.
In theory, you can make a differentiatora rate-of-change computerby taking op amp A3 and just
connecting an input capacitor Cin and a feedback resistor Rf. But in practice, with real op-amps, this will cause
a local oscillation of the amplifier, due to the lag in the feedback loop. The fix is fairly simplefor most cases,
to prevent local oscillations, add a small resistor R111 in series with Cin, and add a small capacitor (Cf) in
parallel with Rf. In practice, if you make R111 about 1/10 or 1/100 of Rf, and Cf about 1/20 or 1/200 of Cin, that
works pretty well. In other cases, it may be a bit more critical which value you choose, either to prevent local
oscillation or avoid degradation of loop stability.
The P-D controller is quite good for many servo control applications. To a large extent, many Fuzzy Logic
ontrollers are quite analogous to a P-D loop, and often they work well. NOW let me invent a case where the PD controller (and the simple F.L. controller begins to work lousy (that's a technical term). Let's take this +35C
controller outside on a cold day. The water stays warm for a while, but the air starts to cool it off. After a while,
the Proportional path turns up the heater. But there's still an erroralways an error. If the fry-pan needs 50 W
to keep the water at about 35C, then the error will be 50 W divided by the gain.
To avoid a large error, it's natural to just turn up the gainwhich is what some people propose to do. But when
you turn up the gain, the loop stability is hurt. How badly? Ahhhm that is hard to predicthard to model. The
reason is that the thermal transfer from the heater to the water and the sensor isn't a simple model. It's not a
simple lag. The transfer might be similar to a cascade of 5 or 10 lags, so that a step of heat causes a slow
change of the sensor temperature. It's possible to do this in a computeror within SPICEbut it's not really
easy. Furthermore, you basically have to measure some real response of the system. You can guess, but it's
not really easy. Anyhow, if you take enough data and generate an accurate-enough model, you can show that
you can turn up the Proportional Gain a certain amount. But, if you go any higher your loop will oscillate, or

ring severely. Let's say that you can only set the Proportional gain as high as 50 W/5C, without oscillation.
Your need is for less than 1C of error. But there remains about 5.
FIRSTLY, you might add as much insulation as you can to cut down the amounts of watts needed. But let's say
there is still more than 2 of error. What can we do?
One alternative (secondly) is to sense the outside temperature. We could then use that to predict that 20 or 40
W of power will be needed. We could add that information into the controllerwhich does work at times,
sometimes very well. This is known as feedforward.
Another thing we could do (thirdly) is add not just extra insulation, but a heated shell around the experiment
perhaps at 30C. That could greatly improve the accuracy, but often this amount of complexity is
unacceptable.
Okay, the fourth optionand a fairly popular and inexpensive oneis to look at that eror signal (the output of
A1), and if there's any dc error, just INTEGRATE that signal using A4. Then feed that output through its
adjustment pot (P4) and sum it with the other signals. This beats the conundrum: to get full accuracy in view of
the rule, "You can only turn on the heat high when the error signal is large." In this case, you can turn on the
heat high even if the error signal is NOT largebut you may have to wait a while for the integrator to do its
job. Why not just turn up the GAIN for the Integrator? You can do that to some extent. If you overdo it, that
makes the loop unstable. So don't overdo it.
The best thing about using the Integrator is that it lets you turn down the gain of the Proportional amplifier. If
you turned up the Proportional gain too high to try to cut down the error, that will cause instability as I
mentioned earlier. When you have the integrator working, you can turn down the Proportional Gain and
improve the loop stability a LOT, yet still have infinite gain at dc. You can have ZERO static error.
But, don't turn the Proportional Gain down TOO FAR. If you did that, the I and D terms would act like an L-C
filter with no damping. So if you turn P2 down to zero, that's sure to cause oscillation, too. Now that you have
the Integral path working, the gain setting for the Proportional term acts as a DAMPING FACTOR to prevent
the loop from ringing. As you can imagine, the optimization of such a loop isn't trivial. Also, in many cases,
these loops may be slow, so it's hard to see if any changes you're making are doing more good than harm.

Hint 1: Use a slow strip-chart recorder so that you can see the shape of the loop's
step response, and if you're making any improvements. Or use a storage scope. Or a
Rustrak meter.

Hint 2: Take a little open-loop data to show the delay from a step of heat to a change
of output temperature. Build up a cascaded R-C network to stimulate that slow lag
(Fig. 2). Then change the lag's response by a factor of 100 by decreasing all of the
capacitors by a factor of 100. Then design your controller to make that loop stable at
a speed that's easy to observe. Then scale that controller 100:1 slower, and you're
fairly close to having a controller that will work. This is one form of Analog Computer.
Hint 3: If the system changesif the amount of water in the fry-pan decreasesyou
will probably need to change the coefficients of your system. You could turn those
pots, or you could use multiplying DACs in place of P2, P3, and P4, to get the
coefficients you want. Not trivialbecause the system won't do it for youyou have
to tell it what you want. But this IS feasible.
Hint 4: This circuit won't work well at all with general-purpose op amps An LM741 at
A3 or A4 would cause HORRIBLE errors, because the resistors in the differentiator
and in the integrator will be quite highperhaps 2 or 5 M forcing you to get op
amps with low bias currents. But, fortunately, good op amps with low imput current
(50pA or 0.05 pA) aren't expensive these days.M
Hint 5: When you have all of this worked out and optimized pretty well, you can do the
whole thing with one op ampyou may not need five amplifiers. In Figure 3, op amp
A6 does the whole thing. Of course, you don't have the flexibility of three independent
controls, but in many cases you don't need that. In this case, the output of A6 is a
summation of the Derivative and the Integral terms, with a Proportional (damping
factor) term also included.

Can Fuzzy Logic likewise take advantage of an Integrator to convert from PD to PID? Yes, and pretty easily, if
you figure out the right trick. Of course, if the system is REALLY nonlinear, or a nonlinear controller is really
needed, then nothing is simple, and you might have to write 125 or 343 rules. Still, a small Integral term could
let you effectively turn down the "gain" of the Proportional path and greatly improve the dc accuracy AND the
loop stability. Of course, this doesn't mean that you can easily get fast settling under difficult conditions, such
as "we have no idea how much water is in the pan." But there's still a definite opportunity for improvement by
adding in the integrator5.
As I mentioned in '93, F.L. does not, by itself, compute a derivative. So if you want a derivative term, you have
to generate a derivative and digitize it, then present it to the F.L. controller: OR, you digitize the proportional

signal and take a DIFFERENCE every few seconds, then present THAT as a derivative signal to the F.L.
controller: So, in exactly the same way, the F.L. controller can't generate an integral. But, you can program a
subroutine to compute the integral of the Error Signal and present it to the F.L. OR, you could compute the
Integral term and just ADD it to the Proportional term, and then process that total without any fanfare. If the
system is fairly linear, nobody will ever know that you cheated, and it will probably work perfectly. You may not
have to write 343 rulesmaybe 25 or 49 will work just fine!
So, if you have a few bucks worth of op amps and a little time, you can make a pretty darned good controller:
Much better than a bang-bang controller. When I came to NSC back in '76, I found several Application Notes
where we had recommended a temperature controller. But either the controller had finite gain (i.e., poor low
gain) or else ran bang-bang, with various kinds of noise and inaccuracyand bad error!! In my App notes, I
recommend that a proportional controller with stability enhanced by the PID terms, can be fairly simple and
effective.
OF COURSE, if the delay from the heat to the sensor is just too slow that makes everything much harder.
Locating the sensor where it gets a prompt response to the heat can help a lot. Also, you may get better
results from having two sensors. The one that drives the Differentiator may be located very close to the heater
as an aid to stability. But the one that drives the integrator may live in the "sweet spot"the exact place where
highest precision is needed. If there's an extra lag there, that will certainly make the loop difficult to engineer.
Then the other tricks mentioned abovethe feedforward path and the oven-within-the-ovenmay be
justifiable. In all of the cases where transport delays occur, system design can be very challenging. But it
needn't be considered impossible or even very difficult. And it's usually not a situation where Fuzzy Logic has
any inherent advantages. In fact, PID usually has advantages over a Fuzzy Logic controller if that controller
tries to do without any Integral term.
Comments invited!/RAP
Robert A. Pease/Engineer
P.S. If you have a heatersuch as a gas furnaceyou do not want to be turning it ON and OFF every few
seconds because you would wear it out. The same is true for an electromechanical refrigerator. But if you
have a thermoelectric cooler, you can turn that ON to any desired extent by driving the number of amperes the
loop calls for. The same thing applies with dc resistive heaters, but BEWAREthe amount of heat is normally
porportional to the SQUARE of the current. If you use the power transistor AND the resistor as heaters, the
total watts is about linearly proportional to the current, but you have to be careful where you locate the
transistor (a small source) compared to the power resistor, which often is made of wire wrapped all around the
temperature chamber. The management of thermal flow is a very important and tricky subject; I can't give any
easy answersyou really have to study it.
Modulating or controlling a high-power heater, such as a kilowatt of 115-V line power, sounds like it would be
much harder, but actually it's easy. You can drive a power Triac using a MOC3030 (zero-crossing firing circuit)
and add a dither circuit to ensure that the average heating of the power resistor is linearly porportional to the
duty cycle. In my book, I showed a 17-Hz dither that turned the Triac ON and OFF about 17 times per second.
Averaged over the internal time constant of the heater, you can hardly see any noise, and the duty cycle is
very well controlled without much heating in the control circuit. (You certainly don't want to control a kilowatt of
resistive heat with a linear amplifier.)
References:
1. "What's All This Refrigerator Stuff, Anyhow?," Electronic Design, Dec. 19, 1994, p. 122.
2. Troubleshooting Analog Circuits, Robert A. Pease, 1991, p.109. Butterworth-Heineman, Available from the
author for $31.95 (inc. tax and mailing).
3. Encyclopedia Brittanica, London, 1894, Vol. XXII, "The Steam Engine," pp. 508-509.
4. "A Differential Governor," Sir W. Siemens, Proceedings of the Royal Institute of Mechanical Engineering,
1853.
5. "The Basics of PID on the NLX220," by Adaptive Logic, Address: 800 Charcot Ave., Suite 112, San Jose,
CA 95131. This arrived recently and shows easy techniques to get full PID control with F.L. Ask for info at
(408) 383-7200.
Originally published in Electronic Design, June 26, 1995
RAP's 1997 comments: THIS is not just about PID. THIS is not just about Analog Computers. THIS is not just
about Op-Amps. THIS is not just about Fuzzy Logic. This is about THE REAL WORLD. The whole world. If you
ask for something and you get it, are you happy? If not, why not? I have seen some F.L. scientists who were
wise enough to agree that they could make do with a LOT LESS than 343 rules, by adding the I term or the D
term to the P term. Of course, this works best when there is not much nonlinearity. And when there is not much
nonlinearity, F.L. does not have much advantage over PID controllers.
Minor correction on the circuit referenced in my book: on page 109, I show a good circuit with a "17 Hz triangle
wave." The main function of the triangle wave is so you can see the LED blinking, and guess when the loop is
pegged, or whether the duty cycle is high or low. I built another copy of this, recently. The LED did not seem to
blink right. I checked, and the frequency was 170 Hz. So please mark up that schematic, to change the
capacitor that is connected to "17 Hz" from 0.01 F to 0.1 F, so you would see the right kind of blinking. This
is the only error I have found in my book. I think we can get the 1998 printing to have the correct C value.
rap.