Complete Root Locus
Complete Root Locus
Complete Root Locus
Introduction
The complete root locus (CRL) displays the locations of the closed loop poles of a single
input, single output (SISO) system as one real parameter k (-,) is varied. Figure 1 illustrates
a unity feedback system having the transfer function
s +1
G (s ) =
(1)
s (s + 4 )(s 2 + 2 s + 2 )
in the forward path. Figure 2 shows the loci of the roots as the proportional gain k is varied. The
solid lines are the locations corresponding to positive gains while the dashed line corresponds to
negative gains. The locus for positive gains is called the root locus while the locus for negative
gains is called the complementary root locus. The two loci together are called the complete root
locus. Also shown in the figure is the MATLAB code for generating the root locus.
s +1
s (s + 4 ) s 2 + 2 s + 2
R(s)
Y(s)
Imaginary Axis
4
2
0
-10
-5
Real Axis
10
Although it is easy and fast to generate plots such as Figure 2, it is essential to understand
how to draw the root locus because it provides insight into the influence a compensator
(controller) pole or zero has on the system performance.
Most undergraduate control texts cover the root locus construction for only positive k. The
presentation will cover both signs. The motivation for doing this is completeness, the insight
offered by observing the influence of the other sign, and the opportunity to check the correctness
of the plot construction by observing certain properties.
Contained herein are the details of drawing the root locus and the properties of the open loop
transfer function that provide the key tools that help in drawing the diagram.
+
R(s)
G(s)
Y(s)
H(s)
Figure 3: SISO System
Y (s )
G (s )
=
= T (s ) .
R (s ) 1 + G (s )H (s )
(3)
Once the transfer function is known and the numerator and denominator fractions have been
cleared, the factorization that is shown in Eq. (2) can be performed.
Example 1
A servo position control system with derivative feedback is shown in Figure 4. In Figure 4,
k is a variable gain on the derivative feedback and the constants a, b, and KP are all known.
Determining the closed loop transfer function for the system of Figure 4 and performing the
operations in Eq. (2) shows
a
a
KP
KPa
KPa
s (s + b ) + K P a
Y (s )
s (s + b )
=
=
=
=
.
K a
aK P
R(s )
s (s + b ) + K P a(1 + k ) s (s + b ) + K P a + kK P a
1 + (1 + k ) P
1+ k
s (s + b )
s (s + b ) + K P a
KP
+
R(s)
a
s (s + b )
KP
-
(4)
Y(s)
1+k s
Figure 4: Servo Position Control System
The denominator of the last term of Eq. (4) is now in the correct root locus form.
(6)
N (s )
is referred to as the
D(s )
open loop transfer function. The roots of N(s) are called open loop zeros of which there are m.
The roots of D(s) are called open loop poles of which there are n. It is assumed in the
presentation that there are more open loop poles than there are zeros, i.e. n > m. The quantities n
and m are also the number of finite open loop poles and zeros, respectively. Because n > m, open
loop zeros can also occur as s becomes infinite giving rise to infinite open loop zeros. If the case
of m > n were considered, then there would be open loop poles as s becomes infinite and then
there would be infinite open loop poles. This is the reason behind distinguishing between finite
and infinite open loop poles and zeros. The constraint of n > m will be maintained and more will
be said about open loop zeros occurring as s becomes large when asymptotes are discussed.
Because n > m, the degree of the characteristic equation in Eq. (6) is n. A branch is the path
of a root as the gain k is varied from zero to (or -). For positive k there are n branches on the
root locus and for negative k there are also n branches on the complementary root locus
Each branch of the CRL has a remarkable property. It starts at an open loop pole when k is
zero and ends at an open loop zero when k is either . To see why this occurs, first write the
CE as
D(s ) + kN (s ) = 0 .
(7)
If k is zero, then the roots of the CE occur at the roots of D(s), i.e. the open loop poles. Now
write the CE as
D (s )
k =
.
(8)
N (s )
As s approaches an open loop zero, larger and larger values of k are needed to satisfy the CE.
The branch of the locus then occurs between the open loop pole and the open loop zero. Because
there are more open loop poles than finite open loop zeros, some of the zeros satisfying Eq. (8)
occur at infinite values of s. More will be said about these zeros when asymptotes are covered.
The root locus is drawn in the complex plane. The sketch begins by drawing the complex
plane and entering the finite open loop poles and zeros.
It should be noted that the coefficients of the characteristic equation are all real which means
the roots are either real or complex conjugate pairs. This means that the root locus is symmetric
about the real axis of the complex plane. It should also be appreciated that if there is another line
of symmetry for the open loop poles and zeros, then the root locus is also symmetric about this
other line of symmetry.
Example 2
Consider the root locus plot of Figure 7. The open loop poles and zeros are symmetric with
respect to the vertical line passing through the point -2 on the real axis. The root locus plot is
symmetric about this line as well as the real axis. The MATLAB code necessary to generate
the plot is also shown in Figure 7. It should be noted that the plot in Figure 2 of the CRL of the
system in Figure 1 is only symmetric about the real axis.
Root Locus
8
Imaginary Axis
-2
-4
-6
-8
-10
-8
-6
-4
-2
Real Axis
j
-4
-1
-j
=
.
(12)
odd multiple of if k > 0
N (s o )
Eq. (10), called the magnitude criterion, is useful for determining the gain that locates a CE root
at a particular point on the CRL. Eq. (12), called the angle criterion, is used as a tool for the
construction of the CRL. Any point in the complex plane that satisfies Eq. (12) is a point on the
CRL.
to evaluate the phase of the transfer function as long as the point is anywhere on the positive real
axis to the right of the origin. Thus, the positive real axis to the right of the origin is on the
complementary root locus.
Figure 10: Determining the Phase of the Given Open Loop Transfer Function on the Positive
Real Axis
Going through the same operation in Figure 11(a) as was done in Figure 10, it is seen that the
phase of the open loop transfer function on the interval between origin and the pole to the left of
the origin is -180o. All of the vectors are zero degrees except the one drawn from the origin.
The two complex conjugate poles were eliminated from consideration. The angle of the vector
drawn from the zero less the sum of the angles drawn from the poles is 0 180 o = -180o.
Because this result is true for any point on the interval between the origin and the pole to the left
of the origin, this interval is on the root locus.
On the interval between the zero and the pole to its right, the angle from the zero is 0o as
shown in Figure 11(b). The sum of the angles from the poles to the point on the interval is 360o.
The phase of any point on this interval is 0o 360o or zero which indicates that all points on this
interval are part of the complementary root locus.
In Figure 11(c), any point on the interval between the zero and the repeated pole, it is seen
that the angle from the zero to the point is 180o and the sum of the angles from the poles to this
point is 360o. The difference is 180o 360o or -180o. This result indicates that the entire interval
is on the root locus.
Finally, in Figure 11(d), the point of interest is located to the left of the repeated poles. The
angle from the zero to the point is 180o and the sum of the angles from the poles to this point is
720o. The difference is 180o 720o or -540o or -180o. Thus, the interval to the left of the
repeated poles is on the root locus. It is seen that the immediate intervals to the left and right of
the repeated poles are both root locus.
(a)
j
(b)
j
(c)
j
(d)
Figure 11: Determining the Locus on the Real Axis
The end result is shown in Figure 12. The intervals of the real axis on the root locus are
indicated with solid lines and the intervals that are on the complementary root locus are shown as
dashed lines.
j
-4
-1
-j
N (s )
lim1 + k
= 1 + k lim n = 1 + k lim(s mn ) = 0.
s
s s
s
D (s )
(13)
In taking the limit, it is seen that as sm-n becomes smaller and smaller, the magnitude of k must
grow in order to satisfy the characteristic polynomial. This situation affirms that as k increases
in magnitude, the locus approaches an open loop zero. For a large magnitude of s, Eq. (13) can
be solved for s as
s = ( k )
1 /( n m )
(14)
which has two sets of solutions depending upon the sign of k. If k is positive, then the n m
angles of s satisfying Eq. (14) are
s = (2i 1)
nm
, i = 0,1, 2, ..., m n 1 .
(15)
nm
, i = 0,1, 2, ..., m n 1 .
(16)
The solutions of Eq. (14) appear as lines having slopes given by Eqs. (15) and (16).
Eqs. (15) and (16) provide only slopes and, in order to use these slopes, a point is required.
A short example will demonstrate the validity of Eqs. (15) and (16).
Example 3
for large magnitudes of k having either sign and plot the roots in the complex plane. Sketch in
lines having slopes determined by Eqs. (15) and (16) and compare.
The MATLAB script file of Figure 14 was used in this calculation. The plot is shown in
Figure 15. The asymptotes of the root locus are drawn in solid lines while the asymptotes of the
complementary root locus are drawn as dashed lines. The dots in Figure 15 are the actual roots
of the characteristic equation. In examining Figure 15, the fact that the lines are asymptotes can
be appreciated from the difference between the calculated roots and the asymptote lines. By
choosing larger values of k in the MATLAB script file, the difference between the roots and
asymptotes will diminish.
clear all
N
= [0 0 0 1 4];
D
= [1 9 3 6 12];
%
%
%
%
%
%
%
%
%
Numerator Polynomial
Denominator Polynomial
Same length polynomials are easy to add
Degree of numerator
Degree of denominator
Values of k
Length of the array ks
Use Figure 1
Put hold on so many items can be plotted
m
= 1;
n
= 4;
ks
= logspace(3,5,20);
nl
= length(ks);
figure(1);
hold on;
for i = 1:nl
pk = roots(D+ks(i)*N);
% Roots for positive k
nk = roots(D-ks(i)*N);
% Roots for negative k
plot(pk,'k.');
% Plot roots for positive k as dots
plot(nk,'k.');
% Plot roots for negative k as dots
end
sigmaa = (sum(roots(D))-(-4))/(n-m); % Asymptote intersection point
L
= 50;
% Length of asymptotes
for i=0:n-m-1
angle = (2*i-1)*pi/(n-m);
% Asymptote angle - positive k
X = [sigmaa L*cos(angle)];
% X coordinates
Y = [
0 L*sin(angle)];
% Y coordinates
plot(X,Y);
% Plot line
angle = (2*i)*pi/(n-m);
% Asymptote angle - negative k
X = [sigmaa L*cos(angle)];
% X coordinates
Y = [
0 L*sin(angle)];
% Y coordinates
plot(X,Y,'k--');
% Plot line
end
grid;
title('Example of Asymptotes')
xlabel('Real s');
ylabel('Imaginary s')
D ( s ) = s + a1s
n
n 1
+ a2 s
n2
+ + an 1s + an = (s + pi )
(18)
i =1
and
m
N ( s ) = s m + b1s m1 + b2 s m2 + + bm 1s + bm = (s + zi )
(19)
i =1
and where the ai and bi are the polynomial coefficients while the open loop poles are located at
points pi and the open loop zeros are located at zi. By dividing N(s) into D(s), the quotient
becomes
D (s ) nm
nm
(20)
= s + (a1 b1 ) s n m1 + lower order terms (s a ) = k
N (s )
where a is given by
a +b
a = 1 1
(21)
nm
and where it is seen for large s, Eq.(20) is the same as Eq. (14). By multiplying out the
monomial term on the right of Eq. (20), it is seen that result matches the two highest power terms
of the quotient. The right side of Eq. (20) can be rearranged as
1
1+ k
=0
(22)
(s a )nm
which corresponds to a system consisting of n-m repeated poles at the point a. Eq.(22) can be
solved for s and evaluated for any value of k, even zero. In Example 3, the evaluation of a leads
to a value of -5/3 and n-m is 3. Plotting the values of s that satisfy Eq. (22) for Example 3 in the
complex produces the straight lines of Figure 15, i.e. the asymptotes. The quantity a is called
the asymptote intersection point. This quantity was used in drawing the lines from Example 3
shown in Figure 15.
Example of Asymptotes
50
40
30
Imaginary s
20
10
0
-10
-20
-30
-40
-50
-50
-40
-30
-20
-10
0
Real s
10
20
30
40
50
a1 = pi
(23)
i =1
b1 = zi .
(24)
i =1
Because a1 is the negative of the pole sum, Eq. (23) can be rewritten as
n
(25)
i =1
(26)
i =1
Substituting Eqs. (25) and (26) into Eq. (21) shows that the asymptote intersection point is
a =
i =1
i =1
(27)
nm
0 4 1 + j1 1 j1 (1)
5.
=
nm
4 1
3
From Eq. (15), the angles of the asymptotes on the root locus are
s = -60o, +60o, and +180o.
From Eq. (16), the angles of the asymptotes on the complementary root locus are
s = 0o, +120o, and +240o.
Combining the asymptote lines with the results of Figure 13 shows the sketch of Figure 16.
a =
i =1
i =1
Breakaway Points
Breakaway points occur in the complex plane where the characteristic equation has multiple or
repeated roots. Breakaway points are significant in that they are the only way a locus can either
enter or leave the real axis. The breakaway points offer useful information concerning the locus
and the locus construction.
The way to find breakaway points is to note that at a repeated root of a polynomial, both the
polynomial and the slope of the polynomial are zero. In order to locate the points of multiple
roots, it is sufficient to locate those points where the derivative of the characteristic equation with
respect to s vanishes. Applying this idea to Eq. (7) shows
d
(D(s ) + kN (s )) = 0 = d D(s ) + k = d D(s ) = d 1 + N (s ) = d N (s ) = 0 . (28)
ds
ds N (s )
ds N (s ) ds k D(s ) ds D(s )
From Eq. (28), it can be seen that the points of multiple roots occur at the solutions to either
d D (s )
=0
(29)
ds N (s )
or
d N (s )
= 0.
(30)
ds D (s )
That either Eqs. (29) or (30) provide the same results might seem disconcerting and a closer look
at Eq. (28) reveal the reasoning behind these two equations.
j
a
-4
-1
-j