10 (Control System Design 2)
Uploaded by
Michael10 (Control System Design 2)
Uploaded by
Michael3 IMPROVING STEADY-STATE ERROR WITH A CASCADE COMPENSATOR
ELEC207 - Control Theory [Lecture 10: Control System
Design 2]
Prof Simon Maskell, <[email protected]>
March 2, 2024
1 Lecture Learning Objectives
Design of a Control System via root locus.
2 Compensation
This lecture revisits some of lecture 7’s content, but with the benefit of the techniques learned in 9
related to root-locus and a closer coupling to the definitions of performance described in lecture 6.
There are two types of compensator, a cascade compensator (shown in figure 1(a)) and a feedback
compensator (shown in figure 1(b)), where the plant is shown with transfer function, P (s), the
original controller with transfer function, G(s) and the compensator with transfer function, C(s). In
this lecture the focus is on the design of cascade compensators. The design of cascade compensators,
as described below, is more straightforward than the design of feedback compensators, which is
therefore not covered here.
3 Improving Steady-state Error with a Cascade Compen-
sator
Recall that the steady-state error was defined in lecture 6 in terms of the position, velocity and
acceleration constants.
3.1 PI Control/Ideal Integral Compensator
If we are happy to change the root locus, we can increase the system-type by adding a pole that is
on the origin. See the root locus for an exemplar system in figure 2(a) and the corresponding figure
with an additional pole at the origin in figure 2(b).
We can add an additional zero close to the origin as in figure 2(c). Note that the root locus for
figures 2(a) and 2(c) are almost identical. Such a controller is the PI controller that was discussed
K I
s+ Kp
in lecture 7 (since Kp + KsI = Kp s ). The controller is also sometimes known as an Ideal Integral
Compensator.
ELEC207 Lecture Notes: Lecture 10 (Control System Design 2), Page 1
3.2 Lag Compensator
4 IMPROVING TRANSIENT RESPONSE WITH A CASCADE COMPENSATOR
(a) Cascade Compensator
(b) Feedback Compensator
Figure 1: Two kinds of compensator.
3.2 Lag Compensator
If we are operating with a system type that is just high enough to have finite steady-state error, but
not sufficient to get zero steady-state error, then we can reduce that steady-state error by reducing
the corresponding error constants. An alternative to PI control in such scenarios is to replace the
pole at the origin with one that is near, but not at, the origin. This is called a Lag Compensator
(for reasons we will discover in a future lecture). For example, assume unity negative feedback
system with Open-loop transfer function, G(s). We know (from lecture 6) that steady-state error
in response to a ramp input will be K1v , where Kv = lims→0 sG(s). If we add a zero at s = zc and
s−zc
a pole at s = pc , then the velocity constant will now be lims→0 sG(s) s−p c
= Kv pzcc . So, if we put
the pole nearer to the origin that the zero, we’ll reduce the steady-state error by a factor pzcc .
Lag compensators can be implemented with passive components (ie resistors and capacitors)
whereas PI controllers demand active components (ie an Op Amp).
4 Improving Transient Response with a Cascade Compen-
sator
Recall that the transient response was defined in lecture 6 for first-order and second-order systems.
If one or two of the poles are much (approximately 5 times) closer to the imaginary axis than
the others, then these poles will be the dominant poles. The transient response can then be
approximated as the response of the dominant poles. To reduce the settling time, we need to move
the poles further from the imaginary axis. If we want a certain percentage overshoot, we can then
ELEC207 Lecture Notes: Lecture 10 (Control System Design 2), Page 2
4 IMPROVING TRANSIENT RESPONSE WITH A CASCADE COMPENSATOR
2 2
1.5 1.5
1 1
0.5 0.5
0 0
−0.5 −0.5
−1 −1
−1.5 −1.5
−2 −2
−3 −2.5 −2 −1.5 −1 −0.5 0 0.5 1 −3 −2.5 −2 −1.5 −1 −0.5 0 0.5 1
(a) Uncompensated (b) Additional Pole at s = 0
2 2
1.5 1.5
1 1
0.5 0.5
0 0
−0.5 −0.5
−1 −1
−1.5 −1.5
−2 −2
−3 −2.5 −2 −1.5 −1 −0.5 0 0.5 1 −3 −2.5 −2 −1.5 −1 −0.5 0 0.5 1
(c) Compensated using PI control (d) Compensated using PD control
2
1.5
0.5
−0.5
−1
−1.5
−2
−3 −2.5 −2 −1.5 −1 −0.5 0 0.5 1
(e) Lead Compensated
Figure 2: Root locus diagrams for system with and without controllers of various kinds.
ELEC207 Lecture Notes: Lecture 10 (Control System Design 2), Page 3
4.1 PD Control/Ideal
5 IMPROVING
Derivative Compensation
STEADY-STATE ERROR AND TRANSIENT RESPONSE
choose the value of K (recall from lecture 6 that overshoot is parameterised by damping ratio and
lines of constant damping ratio are straight lines through the origin).
4.1 PD Control/Ideal Derivative Compensation
We can change the root locus by adding a zero to the
controller
using the PD controller discussed
Kp
in lecture 7 (with transfer function Kp + Kd s = Kd s + Kd ). Since this adds a zero to the closed-
loop response, it can move the root locus further to the imaginary axis, speeding up the transient
response that can be achieved with a given percentage overshoot. See figure 2(d) for an example.
Such a controller is also sometimes known as an Ideal Derivative Compensator.
Note that adding PD control will be likely to alter (ie not necessary improve) the steady-state
error.
4.2 Lead Compensator
Instead of adding a single zero, we can add a zero and a pole. We pick the zero arbitrarily and
then choose the pole position to achieve the desired response. Such a controller is called a Lead
Compensator (again, for reasons we will discover in a future lecture). See figure 2(e) for an example
root-locus for the system, with a lead compensator.
An advantage of a lead compensator over a PD controller is that it only requires the use of
passive electronic components. A lead compensator also avoids amplification of high-frequency
noise.
5 Improving Steady-State Error and Transient Response
To simultaneously achieve a desired steady-state error and transient response we can use a com-
bination of a PI and PD controller, ie a PID controller or a combination of lag-compensator and
lead-compensator, a lag-lead-compensator. Such a controller can be designed as follows:
Design the PD controller or lead-compensator to meet the transient response specifications;
this moves the root locus to where it needs to be.
Verify that the transient response is as predicted and iterate the design if necessary (typically
because there poles nearest the imaginary axis are not dominant and other poles are having
a significant effect on the transient response).
Design the PI controller or lag-compensator to achieve the required steady-state error.
Verify that the steady-state error and transient response are as predicted and iterate the
design if necessary (typically because the small changes to the root locus are larger than
anticipated).
