Feedback Control Systems

Fall 2020

September 19, 2020

In This Lecture

Disturbances

r e u y
Controller System
−

ym

Measurements

Objective
I Develop the concept of block diagrams
Feedback System Analysis

I Many nonlinear and complex mechatronic systems can be

approximated by LTI models. In this part, we will study linear
controller design methods for LTI systems
Feedback System Analysis

I Many nonlinear and complex mechatronic systems can be

approximated by LTI models. In this part, we will study linear
controller design methods for LTI systems
I Two types of controllers:
I Open-loop

I Closed-loop
Feedback System Analysis

I Many nonlinear and complex mechatronic systems can be

approximated by LTI models. In this part, we will study linear
controller design methods for LTI systems
I Two types of controllers:
I Open-loop

I Closed-loop

I Will focus on stability, tracking, regulation, and sensitivity

Basic Equations of a System
Open-loop

I Series connection

Yol = GDol R + GW
Basic Equations of a System
Open-loop

I Series connection

Yol = GDol R + GW

I Error

Eol = R − Yol
= R − GDol R − GW
= [1 − GDol ] R − GW
Basic Equations of a System
Open-loop

I Series connection

Yol = GDol R + GW

I Error

Eol = R − Yol
= R − GDol R − GW
= [1 − GDol ] R − GW

I Transfer function

Tol (s) = G(s)Dol (s)

Basic Equations of a System
Closed-loop
Basic Equations of a System
Closed-loop

I Inputs:
I R: Reference (the output is desired to track)
I W : Plant disturbance
I V : Sensor noise
Basic Equations of a System
Closed-loop

I Inputs:
I R: Reference (the output is desired to track)
I W : Plant disturbance
I V : Sensor noise
I Relations:
Ycl = G(U + W )
U = Dcl Q, Q = R − (Ycl + V ),
GDcl G GDcl
Ycl = R+ W− V
1 + GDcl 1 + GDcl 1 + GDcl
Basic Equations of a System
Closed-loop

I Inputs:
I R: Reference (the output is desired to track)
I W : Plant disturbance
I V : Sensor noise
I Relations:
Ycl = G(U + W )
U = Dcl Q, Q = R − (Ycl + V ),
GDcl G GDcl
Ycl = R+ W− V
1 + GDcl 1 + GDcl 1 + GDcl
I Error: Reference - Output
Ecl = R − Ycl
1 G GDcl
= R− W+ V
1 + GDcl 1 + GDcl 1 + GDcl
Basic Equations of a System
Closed-loop
Basic Equations of a System
Closed-loop

I Let’s define two new transfer functions:

1 GDcl
S= , T =
1 + GDcl 1 + GDcl
Basic Equations of a System
Closed-loop

I Let’s define two new transfer functions:

1 GDcl
S= , T =
1 + GDcl 1 + GDcl
I Then, the equations become:

Ycl = T R + GSW − T V
U = Dcl SR − T W − Dcl SV
Ecl = SR − GSW + T V
Stability
Open-loop
Stability
Open-loop

I Define the polynomials:

b(s) c(s)
G(s) = , Dol (s) =
a(s) d(s)

cb
Transfer function: Dol G =
da
Stability
Open-loop

I Define the polynomials:

b(s) c(s)
G(s) = , Dol (s) =
a(s) d(s)

cb
Transfer function: Dol G =
da
I Assume that a(s) contains a pole on the RHP. The only way to
eliminate these poles is to cancel out with. . .
Stability
Open-loop

I Define the polynomials:

b(s) c(s)
G(s) = , Dol (s) =
a(s) d(s)

cb
Transfer function: Dol G =
da
I Assume that a(s) contains a pole on the RHP. The only way to
eliminate these poles is to cancel out with. . .
I However, pole-zero cancellation on the RHP is not safe! The
system would go unstable in case of any disturbance
Stability
Open-loop

I Define the polynomials:

b(s) c(s)
G(s) = , Dol (s) =
a(s) d(s)

cb
Transfer function: Dol G =
da
I Assume that a(s) contains a pole on the RHP. The only way to
eliminate these poles is to cancel out with. . .
I However, pole-zero cancellation on the RHP is not safe! The
system would go unstable in case of any disturbance
I So, the zeros of the controller c(s) or the plant b(s) cannot be
used to stabilize an open-loop system
Stability
Closed-loop

Characteristic equation: 1 + GDcl = 0

Stability
Closed-loop

Characteristic equation: 1 + GDcl = 0

I Substitute the polynomials:

c(s)b(s)
1+ = 0 → a(s)d(s) + b(s)c(s) = 0
d(s)a(s)
Stability
Closed-loop

Characteristic equation: 1 + GDcl = 0

I Substitute the polynomials:

c(s)b(s)
1+ = 0 → a(s)d(s) + b(s)c(s) = 0
d(s)a(s)
I Assume as before that a(s) contains a pole on the RHP. If we put
a zero in c(s) to cancel it out, it will still appear in the
characteristic equation. For example,
1
G(s) = 2 → a(s) = (s + 1)(s − 1)
s −1
and try
K(s + z)
Dcl (s) = → (s + 1)(s − 1)(s + p) + K(s + z) = 0
s+p
Stability
Closed-loop

Characteristic equation: 1 + GDcl = 0

I Substitute the polynomials:

c(s)b(s)
1+ = 0 → a(s)d(s) + b(s)c(s) = 0
d(s)a(s)
I Assume as before that a(s) contains a pole on the RHP. If we put
a zero in c(s) to cancel it out, it will still appear in the
characteristic equation. For example,
1
G(s) = 2 → a(s) = (s + 1)(s − 1)
s −1
and try
K(s + z)
Dcl (s) = → (s + 1)(s − 1)(s + p) + K(s + z) = 0
s+p
Stability
Closed-loop

Characteristic equation: 1 + GDcl = 0

I Substitute the polynomials:

c(s)b(s)
1+ = 0 → a(s)d(s) + b(s)c(s) = 0
d(s)a(s)
I Assume as before that a(s) contains a pole on the RHP. If we put
a zero in c(s) to cancel it out, it will still appear in the
characteristic equation. For example,
1
G(s) = 2 → a(s) = (s + 1)(s − 1)
s −1
and try
K(s + z)
Dcl (s) = → (s + 1)(s − 1)(s + p) + K(s + z) = 0
s+p
I Unlike the open-loop case, we have more freedom. Can cancel out
Tracking

Definition: Tracking is defined as controlling the system output so

that it follows a desired trajectory

Characteristic equation: 1 + GDcl = 0

Tracking

Definition: Tracking is defined as controlling the system output so

that it follows a desired trajectory

Characteristic equation: 1 + GDcl = 0

I If the plant has poles only on the LHP, then the controller can be
chosen to cancel out the poles of the plant and obtain the desired
response
Tracking

Definition: Tracking is defined as controlling the system output so

that it follows a desired trajectory

Characteristic equation: 1 + GDcl = 0

I If the plant has poles only on the LHP, then the controller can be
chosen to cancel out the poles of the plant and obtain the desired
response
I Should be careful with:
I Choose the controller proper because an improper transfer
function cannot be realized
I Pole-zero cancellation close to the imaginary axis jω
Regulation

Definition: Regulation is defined as controlling the system output so

that the error is maintained small under the effect of disturbance
Regulation

Definition: Regulation is defined as controlling the system output so

that the error is maintained small under the effect of disturbance

I In an open-loop structure, the controller has no effect on the

disturbances!
Regulation

Definition: Regulation is defined as controlling the system output so

that the error is maintained small under the effect of disturbance

I In an open-loop structure, the controller has no effect on the

disturbances!
I For closed-loop structures, recall:
GDcl G GDcl
Ycl = R+ W− V
1 + GDcl 1 + GDcl 1 + GDcl
G GDcl
Terms of interest: W and V
1 + GDcl 1 + GDcl
Regulation

G GDcl
Terms of interest: W and V
1 + GDcl 1 + GDcl
I To reduce the effect of W , we increase the magnitude of Dcl
Regulation

G GDcl
Terms of interest: W and V
1 + GDcl 1 + GDcl
I To reduce the effect of W , we increase the magnitude of Dcl
I However, making Dcl large will make GDcl /(1 + GDcl ) tend to
unity so sensor noise will appear at the output
Regulation

G GDcl
Terms of interest: W and V
1 + GDcl 1 + GDcl
I To reduce the effect of W , we increase the magnitude of Dcl
I However, making Dcl large will make GDcl /(1 + GDcl ) tend to
unity so sensor noise will appear at the output
I Solution: Determine the frequency ranges of the plant and sensor
disturbances and adjust the frequency of Dcl accordingly
Sensitivity

G GDcl
Terms of interest: W and V
1 + GDcl 1 + GDcl
I To reduce the effect of W , we increase the magnitude of Dcl
Sensitivity

G GDcl
Terms of interest: W and V
1 + GDcl 1 + GDcl
I To reduce the effect of W , we increase the magnitude of Dcl
I However, making Dcl large will make GDcl /(1 + GDcl ) tend to
unity so sensor noise will appear at the output
Sensitivity

G GDcl
Terms of interest: W and V
1 + GDcl 1 + GDcl
I To reduce the effect of W , we increase the magnitude of Dcl
I However, making Dcl large will make GDcl /(1 + GDcl ) tend to
unity so sensor noise will appear at the output
I Solution: Determine the frequency ranges of the plant and sensor
disturbances and adjust the frequency of Dcl accordingly