MECHATRONICS LAB

ME 140L

INTRODUCTION TO CONTROL SYSTEMS

I. LECTURE 2: TRANSFER FUNCTIONS, EIGENVALUES
Example: mass-spring-damper

Equation of motion (assuming all terms linear):

m&x& + Bx& + kx = F
Transfer Functions
Let F = Fest , then x = Xest → Note est can take on many forms

Apply to differential equation :

m&x& + Bx& + kx = F → ⎛⎜ ms 2 + Bs + k ⎞⎟ Xest = Fe st
⎝ ⎠
1 1
X(s) 1 m m
Transfer function : = = =
F(s) ms + Bs + k s 2 + s +
2 B k s + 2ξ ω n s + ω 2n
2

m m
ω2n = k and 2ξ ω n = B , where ξ = damping ratio, ω n = natural frequency
M M

Denominator polynomial of the transfer function is the characteristic

B k
equation (CE) when set to zero s 2 + s + = s 2 + 2 ξ ωn s + ω2n = 0
m m

1
 -B ⎛ B ⎞2 k
± ⎜ ⎟ −4
CE yields eigenvalues: s =
m ⎝m⎠ m
= -ξ ω n ± ω n ξ 2 - 1
2
⎛ 2 ⎞
⎜ - ξ ωn ± ωn ξ - 1 ⎟t
Thus the homogeneous solution: x = Ae⎝ ⎠

Note values of the eigenvalues depend upon ξ and ω n , which are functions
of m, B, and k for the mass-spring-damper system (see Lecture 2)

Case A: Undamped ξ = 0 and s = ± jω n

Case B: Underdamped 0 < ξ < 1 and s = - ξ ω n ± jω n 1 - ξ 2 = -ξ ω n ± jω d
Case C: Critically-damped ξ = 1 and s = - ω n ,-ω n
⎧ ⎫
Case D: Overdamped ξ > 1 and s = s1 , s 2 = ω n ⎨ - ξ ± ξ 2 − 1⎬
⎩ ⎭

LESSON: eigenvalues determine response of system.

Objective of control: change the eigenvalues to get the desired response.

2
II. INTRODUCTION TO AUTOMATIC CONTROL
Objective: Change eigenvalues of the closed-loop system to allow the
system to exhibit the desired closed-loop performance.

A. Block Diagram (graphical representation)

Gc, Gp, and H are transfer functions relating output/input descriptions.

Closed Loop Transfer Functions

Y= GcG p E and E=R-HY Combining and solving for Y

G cG p
Y= R Input/Output → Closed Loop TF
1 + G cG p H

Closed loop Characteristic Eq: 1+ G cG p H =0

Roots of closed-loop characteristic equation are closed-loop
eigenvalues which determine the closed loop response of the system.

Notice the eigenvalues from 1+ G cG p H =0 are different from the

eigenvalues of Gc, Gp, and H

Behavior of the closed loop system is different than the behavior of

the plant by itself.

3
B. Standard Controllers (P, I, PD, PI, PID)
Let m(t) be the output signal from the controller and e(t) be the error input
signal to the controller.

One-term controller (P, I)

Proportional: Gc(s)=Kc m(t)= Kc e(t)
1 e(t)
Integral: Gc(s)= m(t) = ∫ dt
Ti s Ti
Two-term controllers (PD, PI)
Proportional /Derivative: Gc(s)=Kc (1+ Td s ) {
m(t)= K ce(t) + K cTd e& (t) }
⎧ K ⎫
1 ⎪ ⎪
Proportional /Integral: Gc(s)= Kc (1+ ) m(t)= ⎨ K c e(t) + c
∫ e(t)dt ⎬
Ti s ⎪⎩ T ⎪⎭
i

Three-term controller (PID)

1
Proportional/Integral/Derivative: Gc(s)= Kc (1+ +T s)
Ti s d
⎧ K ⎫
m(t)= ⎪⎨ K c e(t) + ∫
ce(t) dt + K c
Td e
& (t) ⎪
⎬
⎪⎩ T ⎪⎭
i

Recall: the closed loop system, defined by

G cG p
Y= R Input/Output → Closed Loop TF
1 + G cG p H

will have the parameters of the controller in the closed loop characteristic
equation: 1+ G cG p H =0

Therefore, the closed loop eigenvalues (time response) will depend upon:
• The chosen controller Gc
• The parameters: Kc, Ti, Td of the chosen controller

4
Comparison of Controllers: Illustration

Consider the mass-spring-damper

system with no spring: k=0

m&x& + Bx& = F
1
X(s) 1 1/B 1/B m
G p (s) = = = m = = where τ =
F(s) ms 2 + Bs s 2 + B s s ( m s + 1) s (τ s + 1) B
m B

m
Let F=unit step input, and B=1 and m=1, so τ = =1
B

Open Loop Step Response of Plant only, No controller:

1
G p (s) =
s(s + 1)

5
Case 1: P Control: G c = K c
X G cG p Kc
= =
F 1 + G cG p H s 2 + s + K
c
Second order system with 2 ξω n = 1 and ω n 2 = K c

Note: controller gain Kc has the mathematical function of a spring

If Kc = 1 ⇒ ξ = 0.5 and ω n = 1 ⇒ s = -0.5 ± j0.87 (Underdamped)

X(s) 1
Closed Loop Step Response: =
F(s) s 2 + s + 1

1
Case 2: I Control: Gc(s)= (Unstable—see appendix)
Ts
i

6
Case 3: PD Control: Gc(s)= Kc(1+Tds)

X G cG p K c (1 + Td s)
= =
F 1 + G c G p H s 2 + (1 + K T )s + K
c d c
Second order system with 2 ξω n = (1 + K c Td ) and ω n 2 = K c

Note : K c functions as a spring and Td affects the system damping

If K c = Td = 1 ⇒ ξ = 1 and ω n = 1 ⇒ s = −1,−1 (Critically damped)

X(s) s +1
Closed Loop Step Response: =
F(s) s 2 + 2s + 1

7
 1
Case 4: PI Control: G c (s) = K c (1 + )
T s
i
Kc
(T s + 1)
X G cG p Ti i
= =
F 1 + G cG p H 1 K
s 3 + s 2 + K cs + c
Ti Ti
Note : Third order system
If K c = Ti = 1 ⇒ s = −1, ± j1 Oscillator (no damping)
If K c = 1 and Ti = 2 ⇒ s = −0.65,-0.18 ± j0.86

X(s) s +1
Closed Loop Step Response: =
F(s) s 3 + s 2 + s + 1

8
 1
Case 5: PID Control: G c (s) = K c (1 + + T s)
T s d
i
Kc
G cG p (Td Ti s 2 + Ti s + 1)
X Ti
= =
F 1 + G cG p H Kc
s 3 + (1 + K c Td )s 2 + K cs +
Ti
Note : Third order system

If K c = Ti = Td = 1 ⇒ s = −1.75, - 0.12 ± j0.74

X(s) s2 + s +1
Closed Loop Step Response: =
F(s) s 3 + 2s 2 + s + 1

LESSION FROM THESE PLOTS: Different controllers will yield

different closed loop responses since the characteristic equations are
different.

9
Best controllers appear to be using P
(Case 1) and PD (Case 3)

Recall that we arbitrarily chose

values=1 for the controller
parameters. In general you must
choose these values to get better
responses—called controller tuning.

Tuning Example: Consider using a PD controller but now with different

control parameters:

X(s) s +1
Recall PD with Kc=1, Td=1: = Eigenvalues: -1, -1
F(s) s 2 + 2s + 1

X(s) 5(0.1s + 1)
Now choose Kc=5, Td=0.1 : = Eigenvalues: -0.75 ± j2.11
F(s) s 2 + 1.5s + 5

LESSON FROM THESE TWO

PLOTS:
For the same controller, choosing
different control parameters will
change the closed-loop response of
the system because the eigenvalues
are different.

10
CONTROLLER TUNING: TRIAL AND ERROR

Example of PD Controller Tuning (same as Case 3 above)

PD Control: Gc(s)= Kc(1+Tds)

X G cG p K c (1 + Td s)
= =
F 1 + G c G p H s 2 + (1 + K T )s + K
c d c

Case a: Keep Td=0.5sec and change Kc : X K c (1 + 0.5s)

=
F s 2 + (1 + 0.5K )s + K
c c
Kc Eigenvalues
1 − 0.75 ± j0.66
5 − 1.75 ± j1.39
10 − 3 ± j1
50 − 24,−2

Unit Step Responses with PD Control with Ti=0.5 and Various Kc

1.4

1.2

0.8
x(t)

Kc=1
0.6
Kc=5
Kc=10
0.4 Kc=50

0.2

0
0 1 2 3 4 5 6
Time, sec

11
Case b: Keep Kc =10 and change Td: X 10(1 + Td s)
=
F s 2 + (1 + 10T )s + 10
d
Td Eigenvalues
0.1 − 1± j3
0.5 − 3 ± j1
1 − 1,−10
5 -0.2,-50

Unit Step Response with PD Control with Kc=10 and Various Td

2

1.8 Td=0.25
Td=0.5
1.6
Td=1
1.4 Td=5

1.2
x(t)

0.8

0.6

0.4

0.2

0
0 1 2 3 4 5 6 7 8 9 10
Time, sec

Note: you may wish to tune the

controller to satisfy some criteria
(see Lecture 2 on 2nd order
systems):

12
Appendix:

1
Case 2: I Control: Gc(s)=
Ts
i
1
X G cG p Ti
= =
F 1 + G cG p H s3 + s 2 + 1
Ti
Note : Third order system

If Ti = 1 ⇒ System is unstable ⇒ s = −1.466, 0.23 ± j0.79

Note underdamped, but eigenvalue has positive real part,
therefore is unstable

X(s) 1
Closed Loop Step Response: =
F(s) s 3 + s 2 + 1

13