Sei sulla pagina 1di 37

STADIUS - Center for Dynamical Systems,

Signal Processing and Data Analytics

Lead and lag compensators

Design with root locus

Systems and Control Theory 1


STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Definition
Both a lead compensator and a lag compensator have the same
shape:
Lead compensators: Lag compensators:
1 1
+ +
=
1 with 0 < < 1 =
1 with > 1
+ +

1 1 1
So they have a zero at = and a pole at = or

For lead compensators the pole lies more to the left in the
complex plane than the zero and vice versa for lag compensators
Systems and Control Theory 2
STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lead compensators

Systems and Control Theory 3


STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lead compensators: design with root locus


Design based on time-domain quantities that can be
expressed in terms of the dominant pole locations (overshoot,
rise time, settling time, damping ratio, )
This is done based on the root locus method
Previously weve seen how to use the root locus method to
tune the position of the closed loop poles based on one
parameter, but now we have a problem of an entirely
different order: we have to determine the position of a pole
and a zero and tune a parameter ()

Systems and Control Theory 4


STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lead compensators: design with root locus


Our strategy will be the following:
We will look at where we want our dominant poles to lie
(based on the time domain quantities)
We will place our pole and zero (1/ and 1/) such that the
root locus of moves through these desired
positions
Then well find the corresponding from the root locus of

Systems and Control Theory 5


STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lead compensators: design with root locus


A contribution of a lead compensator is that it moves the
closed loop poles to the left
You just have to look at the effect on the centroid (with and
the poles and zeros of and and of ):
+ +
= , with <
#+1# 1 ##
So the centroid is moved to the left, which means the
asymptotic values of the branches will be more negative
This is how you can stabilize a system which the root locus has
no value for which it resides entirely in the left halve plane

Systems and Control Theory 6


STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lead compensators: design with root locus


How can we now find a way to let the root locus go exactly
through a certain point?
We can actually do that quite easily by noticing the following
property of every point on the root locus:
As you know the root locus display the roots of
1+
1 + = 0 for all (with = from
1+
now)
So for a point to be on the root locus there are two
requirements: has to be equal to
180 2 + 1 and has to equal 1
Lets call our desired closed loop position (and its complex
conjugate, ),
then you have to find , and such that
= 180 2 + 1 and = 1
Systems and Control Theory 7
STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lead compensators: design with root locus


The first requirement becomes:
+ + = 180 2 + 1
And the second requirement becomes:
+ 1
=
+
These two requirements determine two of the degrees of
freedom:
For small enough / (= ) and reasonable 180(2 +

Systems and Control Theory 8


STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lead compensators: design with root locus


So in theory we could again put restrictions on the steady-
state error (which leads to a requirement on /)
But we are not going to do that, since that gives us a
complicated system of equations
+ + = (= the angle condition)
+ 1
=
+
= . . .
This system is difficult to solve, a solution is not guaranteed
and we lose control over the position of the new pole (which
might overtake the role of the dominant poles)
So well use our extra degree of freedom to place the zero at
a location such that the angle condition can most likely be
met and without becoming a dominant pole
Systems and Control Theory 9
STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lead compensators: design with root locus


Lets look at the following: we have a system with two poles
( ), which we want to replace to another location ( )
Where should you position the zero of the lead compensator?
It should not be far to the left, since then the risk exists that
there is no valid position of the pole:
+ + = , so the zero adds phase and the pole
subtracts phase; so the added phase thanks to should be large
enough to make a sum of possible
As you can see = 1 2 230 and < 50,
so there exists no such that =

Here you can also see 2 1


graphically that >
for a lead compensator

Systems and Control Theory 10


STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lead compensators: design with root locus


You should also not place it too far to the right, otherwise the
risk exists that the pole will also be placed too far to the right
and interfere too strongly with the dominant poles, strongly
influencing the closed loop system behavior

2 1

Systems and Control Theory 11


STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lead compensators: design with root locus


The last thing we have to know, is how to determine such
that = 180 2 + 1 ?
Well deduce that here:
= + + +
+ = 180 2 + 1 + + +
Now you have to pick and pick which sign has to be
for the right hand side to be the smallest positive
number; well call the resulting right hand side
Now well rewrite +
arctan / + =
/ + = tan
= cot

Systems and Control Theory 12


STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lead compensators: design with root locus


Weve now seen enough to draw up this action plan:
1. Determine the desired locations of the closed loop poles (
and )
from the time domain qualifications; you should take
a safety margin, as the compensators pole and zero will have
an impact (and even more so if it was not a 2nd order system
to start with), if > 60 continue with 2 lead compensators
2. Check whether a proportional controller wont do the trick
3. If not, continue with the design of a lead
compensator by placing its zero somewhere
in the region indicated by the green circle:
4. Find (see previous slide)
5. Find the corresponding -value: = + / +
6. Verify the result and if it does not suffice begin anew
Systems and Control Theory 13
STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lead compensators: design with root locus


We repeat a previous slide here, which links the time-domain
design criteria with the position of the (dominant) poles:
+ , with < 0 and > 0
The damping ratio: = / + (0 1)
The natural frequency: = / = / 1 2
The rise time: 1.8/
The settling time: = 4.6/
The peak time: = / , with = 1 2
/ 1 2
The overshoot: =

Systems and Control Theory 14


STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lead compensators: design with root locus


Well now illustrate our action plan with a simple example:
= 1/ + 2
The overshoot has to be less than 20 %
The rise time has to be less than 0.65 s
1. We first extract desired values for and out of the
requirements:
1 2
= / 0.2
This will lead to 0.4558 X: 0.4558
Y: 0.2000
1.8/ 0.65 s

2.77
1. So with some safety margin we get: = 0.5 and = 3.5, or
= + 1 2 = 1.75 + 3.03
Systems and Control Theory 15
STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lead compensators: design with root locus


2. To check whether a proportional controller isnt a possibility,
well look at the root locus of (with the desired closed
loop pole locations indicated with orange xes)

The root loci do not move through the desired locations, but
to the right of it; hence a lead compensator is a good option
Systems and Control Theory 16
STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lead compensators: design with root locus


3. Now lets place the zero at z = 2.5, which is, hopefully,
far enough to the left in order to have no interference of the
new pole with the dominant poles
4. The value of can now be determined:
= 180 2 + 1 + + +
= 180 2 + 1 + 2 + + 2.5
3.03 3.03
= 180 2 + 1 tan1 tan1 +
1.75 0.25
1 3.03
tan
0.75
= 180 2 + 1 120 85 + 76 = 51
This is very dangerous, tan1 (3.03/1.75) gives 60, but if you
look at the geometry, youll see it has to be 180 + 60 = 120
1. So now we can find = cot = 4.22
Systems and Control Theory 17
STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lead compensators: design with root locus


+
5. Now we can find = = 13.33
+
6. Lets see how the root locus of looks:

1. The root locus moves exactly through and , as expected

Systems and Control Theory 18


STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lead compensators: design with root locus


1. But what we really need to verify is if our time-domain
requirements are met, so lets see how the system reacts to a
step input:
2. step(K*P*C/(1+K*P*C))
3. stepinfo(K*P*C/(1+K*P*C))
RiseTime: 0.4441
SettlingTime: 2.3230
SettlingMin: 0.9017
SettlingMax: 1.1878
Overshoot: 18.7759
Undershoot: 0
Peak: 1.1878
PeakTime: 1.0000
1. Each of our requirements is nicely met
Systems and Control Theory 19
STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lead compensators: design with root locus


An interesting example of how you can use a Matlab tool to
design lead compensators with root locus (check it out!):
http://ctms.engin.umich.edu/CTMS/index.php?example=AircraftPitch&se
ction=ControlRootLocus
It is a more realistic design example:
The design does not aim to get the dominant closed loop
poles at a certain position, but the goal is to get them in a
certain region
They note the current root locus is too far to the right, hence
they resort to a lead compensator to bend it to the left
Both and are placed in a more intuitive manner
We stick to an action plan, but it is meant for you to build up
your intuition; but remember that design is normally carried
out with tools like sisotool
Systems and Control Theory 20
STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lag compensators

Systems and Control Theory 21


STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lag compensators: design with root locus


The goal of lag compensators, when youre designing them
with root locus:
Reduce the steady state error significantly
With a marginal impact on (the relevant part of) the root locus
While maintaining the freedom to adjust your position on
that root locus!
Lets show this graphically:
Lets say we have a system with the
following root locus:
The orange poles indicate the desired
dominant pole locations (which well call
and ), and as you can see they are on the
root locus, for a gain value well call
Systems and Control Theory 22
STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lag compensators: design with root locus


Unfortunately a gain does not suffice, since it corresponds
to a steady state error that is too large
A lag compensator is capable of decreasing the steady state
error (you cant see that on the root locus plot), but while
leaving the root locus practically unaltered:
As the positions of the pole and zero only
determine and ; we still have one degree
of freedom to chose our position on the root
locus:
Now we get to the question of how a lag compensator
succeeds in doing that; as you probably expected, it will be
thanks to the fact that we placed the compensators pole ( )
and zero ( ) close to the imaginary axis

Systems and Control Theory 23


STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lag compensators: design with root locus


As you know, a point in the s-plane is on the root locus of
if = 180 2 + 1 , and if = 1
So if has to (approximately) lie on the root locus of both
1+1/
and (with = 1+1/ in this section; so with the
outside of ) and if this has to be for (approximately) the
same then:


(every root locus that contains automatically also contains ,
its complex conjugate)
This is the case when the lag compensators pole ( ) lies much
closer to its zero ( ) than either lies to ( )

Systems and Control Theory 24


STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lag compensators: design with root locus


It is quite readily visible from the following root locus plot that
the condition for a small impact on (a region of) the root locus
by a lag compensator is that its pole and zero lie much closer
to each other than to that region:
The effect of this pole and
zero on the angle at is
only the angle between the
dotted and the dashed red
line
The effect of this pole and
zero on the magnitude at
is only the ratio between
the length of the dashed
and the dotted red lines
Systems and Control Theory 25
STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lag compensators: design with root locus


We now know how a lag compensator can leave the root
locus almost unaltered
The second thing we need to know is how a lag compensator
with its pole and zero very close to each other can still result
in a significant rise in the DC gain
Thats quite simple in fact, you just have to look at the formula
of the lag compensator and its DC gain:
1+1/
lim = lim =
0 0 1+1/
This shows we can move freely on the root locus by adjusting
and then we can tune the DC gain by adjusting

Systems and Control Theory 26


STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lag compensators: design with root locus


Lets summarize this by looking at our three degrees of
freedom in the equation of the lag compensator ( ):
+1/
= with 1 <
+1/
We use to position ourselves on the root locus
We use to tune the DC gain
We use to push the poles close enough to the imaginary
axis so the pole (at 1/) is close enough to the zero (at 1/)
so they hardly change the root locus in the neighborhood of
and (the desired dominant poles)
The only question that is left open is how to determine the
size of : we recommend to take = 50/

Systems and Control Theory 27


STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lag compensators: design with root locus


Side note: now 1/ lies much closer to the imaginary axis
than , doesnt that make it dominant (which means it
dominates the transient behavior)?
The answer is: yes and no, depending what you mean
with transient behavior
We will try to leave this ambiguity behind us with an example:
1 1
Take = =
+2+3 +23 2 +4+13
+0.04 50
And = , hence, we used = = 25 and
+0.004
= 10 (a typical value)
These poles and zero are located as
follows in the s-plane:

Systems and Control Theory 28


STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lag compensators: design with root locus


Lets first look at how the steady-state is reached in several
situations:

Systems and Control Theory 29


STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lag compensators: design with root locus


It is clear that the pole at 0.004 completely determines
when and how the eventual steady-state is reached, even if
the zero at 0.04 is present
So does this mean that and are no longer the dominant
poles in the system, but 1/ is?
It is, if you find the reaching of the eventual steady-state the
most important characteristic of the transient behavior
It is clear that we do not do that;
to us there is a more important
characteristic; which is when and how
the first significant step towards
equilibrium is made
This changes markedly thanks to the
presence of the zero at 0.04
Systems and Control Theory 30
STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lag compensators: design with root locus


If you zoom in, you can see that the effect of the zero is
indeed that this interpretation of the transient behavior is
unaltered:

Systems and Control Theory 31


STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lag compensators: design with root locus


So due to its location much closer to the imaginary axis, the
pole at 1/ will eventually be the determining factor in
approaching the steady state behavior
But due to the fact that there is a zero that is very close to
that pole, its effect will not be visible as long as there are large
frequencies that dominate the input, which is at the
beginning of the behavior, when the poles at and
dominate the behavior
+0.04
So has two different time-scales:
2 +4+13 +0.004
A fast one, which comes from the original system and which
is to us the most important in the transient behavior
A slow one, which is added by the lag compensator and
which only comes into effect after the transient period
Systems and Control Theory 32
STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lag compensators: design with root locus


This allows us to explain once more how a lag compensator
can succeed in increasing the DC gain without impacting the
time-domain behavior:
1+1/
You add a very slow component ( with and
1+1/
fairly large), which only takes effect after the (fast part of
the) transient behavior is finished
The effect of that component is to increase the gain with
a factor after longer time
A factor

Systems and Control Theory 33


STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lag compensators: design with root locus


Now were ready to provide a recipe to design a lag
compensator with root locus:
1. Determine the desired dominant pole locations, and
2. Determine what gain should be provided with in order
for its closed loop poles to be equal to and this is
You can do this graphically, from the root locus plot
Or analytically from the demand = 1
3. Translate the steady-state requirement into a requirement on
the DC gain of the lag compensator (lim = ) and
0
determine
4. Determine as 50/
5. Verify the behavior

Systems and Control Theory 34


STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lag compensators: design with root locus, example


0.3
Take the system =
+1 +2
We want to increase the DC gain with a factor 30 with respect
to the DC gain of , and we want the dominant poles to be
such that = 0.5 and
= 0.67 rad/s
1. This corresponds to dominant poles at 0.33 .58
2. You can find the required value of for this:
1
= = 3.5

Systems and Control Theory 35


STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lag compensators: design with root locus, example


3. Our steady-state requirement is already translated into a
requirement on the DC gain; it needs to be 30 times the DC
gain of
So lim = 30 lim
0 0
Or lim = = 30
0
we find = 30/ = 8.6
4. Finding is easy: = 50/ = 50/0.33 = 150
5. Lets verify our result now, we will do this by looking at the
old and new positions of the closed loop poles (we will not
do it here, but you should of course also always verify the
steady-state behavior)

Systems and Control Theory 36


STADIUS - Center for Dynamical Systems,
Signal Processing and Data Analytics

Lag compensators: design with root locus, example


3. The closed loop pole and zero locations of and
are shown here below:

4. As you can see there is hardly any distinction (thats why they
werent plotted on the same graph, because they almost
entirely overlap)
5. The reason is that the factor 50 we use in determining is
very large; this gives very small pole-zero values; this also
means the steady-state value is very slowly approached, so
sometimes youll want a smaller factor than 50 and as you
can see in this example, there is quite some margin left
Systems and Control Theory 37

Potrebbero piacerti anche