Automatic Control

Lecture #4 A First Analysis of Feedback

Bo-Chiuan Chen
National Taipei University of Technology

1
 Open-loop Control System

• Open-loop Control is generally simpler, doesn't require a sensor to measure

the output, and doesn't not, of itself, introduce stability problems.

𝑌𝑜𝑙 = 𝐺𝐷𝑜𝑙 𝑅 + 𝐺𝑊
𝐸𝑜𝑙 = 𝑅 − 𝑌𝑜𝑙
= 𝑅 − 𝐺𝐷𝑜𝑙 𝑅 + 𝐺𝑊
= 1 − 𝐺𝐷𝑜𝑙 𝑅 − 𝐺𝑊
= 1 − 𝑇𝑜𝑙 𝑅 − 𝐺𝑊
open-loop transfer function

2
 Feedback Control System

• Feedback (closed-loop) control is more complex and may cause stability

problems but also has the potential to give much better performance than the
open-loop control.

𝑇𝑐𝑙 : closed-loop
transfer function

𝐺𝐷𝑐𝑙 𝐺 𝐺𝐷𝑐𝑙 𝐷𝑐𝑙 𝐺𝐷𝑐𝑙 𝐷𝑐𝑙

𝑌𝑐𝑙 = 𝑅+ 𝑊− 𝑉 𝑈= 𝑅− 𝑊− 𝑉
1 + 𝐺𝐷𝑐𝑙 1 + 𝐺𝐷𝑐𝑙 1 + 𝐺𝐷𝑐𝑙 1 + 𝐺𝐷𝑐𝑙 1 + 𝐺𝐷𝑐𝑙 1 + 𝐺𝐷𝑐𝑙

𝐺𝐷𝑐𝑙 𝐺 𝐺𝐷𝑐𝑙
𝐸𝑐𝑙 = 𝑅 − 𝑌𝑐𝑙 = 𝑅 − 𝑅+ 𝑊− 𝑉
1 + 𝐺𝐷𝑐𝑙 1 + 𝐺𝐷𝑐𝑙 1 + 𝐺𝐷𝑐𝑙
1 𝐺 𝐺𝐷𝑐𝑙
= 𝑅− 𝑊+ 𝑉
1 + 𝐺𝐷𝑐𝑙 1 + 𝐺𝐷𝑐𝑙 1 + 𝐺𝐷𝑐𝑙
3
 Stability

• All poles of the transfer function must be in the LHP.

• For the open-loop system, they are the poles of 𝐺𝐷𝑜𝑙 .
– A naive engineer might believe that if the plant is unstable with a root in the RHP, the
system might be made stable by canceling this pole with a non-minimum phase zero
of the open-loop controller.
• For the closed-loop system, the system poles are the roots of 1 + 𝐺𝐷𝑐𝑙 = 0.
– More design degree of freedom.
– Avoid the pole-zero cancelation in RHP, because perfect pole-zero cancelation
doesn’t exist for reality.
– Need to design the controller to satisfy the desired performance.
• Exercise (for unity feedback system)
1 𝐾 𝑠+𝛾
𝐺 𝑠 = , 𝐷𝑐𝑙 𝑠 = 𝑠+1 𝑠−1 𝑠+𝛿 +𝐾 𝑠+𝛾 =0
𝑠+1 𝑠−1 𝑠+𝛿

𝑠 2 + 2𝜁𝜔𝑛 𝑠 + 𝜔𝑛2 = 0 Solve for 𝐾 and 𝛿 in terms of 𝜁 and 𝜔

(Hint: matching coefficient)

4
 Tracking

• The tracking problem is to cause the output to follow the reference input as
closely as possible.
• For the open-loop system, if the plant is stable and has neither poles nor zeros
in the RHP, the controller can be selected to cancel the transfer function of the
plant and substitute the desired transfer function.
– The controller function must be proper, i.e. no more zeros than poles.
– Cannot request an unrealistically fast design which might cause large control efforts
to saturate the system.
– Non-perfect pole-zero cancelation might cause stability issues.
• Exercise (for unity feedback system)
1 𝑐2 𝑠 2 + 𝑐1 𝑠 + 𝑐0
𝐺 𝑠 = 2 , 𝐷𝑐𝑙 𝑠 =
𝑠 + 3𝑠 + 9 𝑠 𝑠 + 𝑑1
Desired characteristic equation: 𝑠 + 6 𝑠 + 3 𝑠 2 + 3𝑠 + 9 = 0

𝑐2 = 18, 𝑐1 = 54, 𝑐0 = 162, 𝑑1 = 9

For step reference input, the steady-state error is 0.

5
 Regulation

• The problem of regulation is to keep the error small when the reference is at
most a constant set point and disturbances are present.
• For the open-loop system, the controller has no influence at all on the system
response to either of the disturbance. Thus useless for regulation.
• For the closed-loop system, we find a conflict between 𝑤 and 𝑣 in the search for
a good controller.
1 𝐺 𝐺𝐷𝑐𝑙
𝐸𝑐𝑙 = 𝑅− 𝑊+ 𝑉
1 + 𝐺𝐷𝑐𝑙 1 + 𝐺𝐷𝑐𝑙 1 + 𝐺𝐷𝑐𝑙
The resolution of the dilemma is to observe that each of these terms is a
function of frequency so one of them can be large for some frequencies and
small for others.
• Exercise for a constant bias 𝑤
If 𝐷𝑐𝑙 has a pole at 𝑠 = 0, then the error due to this bias will be zero.
If 𝐺 has a pole at 𝑠 = 0, it does not help with a disturbance bias.

6
 Sensitivity

• Suppose a plant is designed with gain 𝐺 at a particular frequency but in

operation it changes to be 𝐺 + 𝛿𝐺.
• For the open-loop system, the nominal overall gain is thus 𝑇𝑜𝑙 = 𝐺𝐷𝑜𝑙 .
𝑇𝑜𝑙 + 𝛿𝑇𝑜𝑙 = 𝐷𝑜𝑙 𝐺 + 𝛿𝐺 = 𝐷𝑜𝑙 𝐺 + 𝐷𝑜𝑙 𝛿𝐺 = 𝑇𝑜𝑙 + 𝐷𝑜𝑙 𝛿𝐺
𝑇
The sensitivity, 𝑆𝐺 𝑜𝑙 , of a transfer function, 𝑇𝑜𝑙 , to a plant gain, 𝐺, is defined as
𝛿𝑇𝑜𝑙
𝑇 𝑇 𝐺 𝛿𝑇𝑜𝑙 𝐺 𝐷𝑜𝑙 𝛿𝐺
𝑆𝐺 𝑜𝑙 = 𝑜𝑙 = = =1
𝛿𝐺 𝑇𝑜𝑙 𝛿𝐺 𝐺𝐷𝑜𝑙 𝛿𝐺
𝐺
𝐷𝑐𝑙 𝐺 𝐷𝑐𝑙 𝐺 + 𝛿𝐺
• For the closed-loop system, 𝑇𝑐𝑙 = 𝑇𝑐𝑙 + 𝛿𝑇𝑐𝑙 =
1 + 𝐷𝑐𝑙 𝐺 1 + 𝐷𝑐𝑙 𝐺 + 𝛿𝐺
𝑑𝑇𝑐𝑙
𝛿𝑇𝑐𝑙 = 𝛿𝐺
𝑑𝐺
𝛿𝑇𝑐𝑙
𝑇 𝑇 𝐺 𝛿𝑇𝑐𝑙 𝐺 𝑑𝑇𝑐𝑙 𝐺 1 + 𝐷𝑐𝑙 𝐺 𝐷𝑐𝑙 1 + 𝐷𝑐𝑙 𝐺 − 𝐷𝑐𝑙 𝐺𝐷𝑐𝑙 1
𝑆𝐺 𝑐𝑙 = 𝑐𝑙 = = = =
𝛿𝐺 𝑇𝑐𝑙 𝛿𝐺 𝑇𝑐𝑙 𝑑𝐺 𝐷𝑐𝑙 𝐺 1 + 𝐷𝑐𝑙 𝐺 2 1 + 𝐷𝑐𝑙 𝐺
𝐺 7
 The Filtered Case

Open-loop system

Closed-loop system

𝐹 can be designed to
improve the tracking
accuracy.

𝐻 can be designed to improve

𝐺𝐷𝑐𝑙 𝐹 𝐺 𝐻𝐺𝐷𝑐𝑙 the response due to the sensor
𝑌𝑐𝑙 = 𝑅+ 𝑊− 𝑉
1 + 𝐺𝐷𝑐𝑙 𝐻 1 + 𝐺𝐷𝑐𝑙 𝐻 1 + 𝐺𝐷𝑐𝑙 𝐻 noise.

𝑇 1 𝐺𝐷𝑐𝑙 𝐻
𝑇
𝑆𝐹 𝑐𝑙 = 1, 𝑆𝐺 𝑐𝑙 = , 𝑆𝐻𝑇𝑐𝑙 =
1 + 𝐺𝐷𝑐𝑙 𝐻 1 + 𝐺𝐷𝑐𝑙 𝐻
8
 Control of Steady-State Error

• Previously
1 𝐺 𝐷𝑐𝑙 𝐺
𝐸𝑐𝑙 = 𝑅− 𝑊+ 𝑉
1 + 𝐷𝑐𝑙 𝐺 1 + 𝐷𝑐𝑙 𝐺 1 + 𝐷𝑐𝑙 𝐺
1 𝐷𝑐𝑙 𝐺
Sensitivity: 𝑆 = Complementary Sensitivity: 𝑇 = 1 − 𝑆 =
1 + 𝐷𝑐𝑙 𝐺 1 + 𝐷𝑐𝑙 𝐺
1
𝐸𝑐𝑙 = 𝑆𝑅 − 𝑆𝐺𝑊 + 𝑇𝑉 𝐸𝑐𝑙 = 𝑅 If 𝑊 = 𝑉 = 0 and 𝐿 = 𝐷𝑐𝑙 𝐺
1+𝐿
loop transfer function
• Consider polynomial inputs, 𝑅 = 1 𝑠 𝑘+1
– 𝑘 = 0, step input or position input
– 𝑘 = 1, ramp input or velocity input
– 𝑘 = 2, parabola input or acceleration input
1 1 1 1 𝑠𝑛 1
𝑒𝑠𝑠 = lim𝑠𝐸 𝑠 = lim𝑠 = lim = lim
𝑠→0 𝑠→0 1 + 𝐿 𝑠 𝑠 𝑘+1 𝑠→0 𝐿𝑜 𝑠 𝑠 𝑘 𝑠→0 𝑠 𝑛 + 𝐾𝑛 𝑠 𝑘
1 + 𝑠𝑛
System type: 𝑛 = 0 type 0
𝑛 = 1 type 1
𝑛 = 2 type 2

9
 Errors as a Function of System Type

𝐾𝑝 = lim𝐿 𝑠 , 𝑛 = 0
𝑠→0
𝐾𝑣 = lim𝑠𝐿 𝑠 , 𝑛 = 1
𝑠→0
𝐾𝑎 = lim𝑠 2 𝐿 𝑠 , 𝑛 = 2
𝑠→0

10
 System Type Examples

• Ex. 4.1 System Type with P control

𝐴 𝑘𝑝 𝐴
𝐷𝑐 = 𝑘𝑝 , 𝐺 = 𝐷𝑐 𝐺 = 𝑛 = 0, 𝐾𝑝 = 𝑘𝑝 𝐴
𝜏𝑠 + 1 𝜏𝑠 + 1
• Ex. 4.2 System Type with PI control
𝑘𝐼 𝐴 𝐴 𝑘𝑝 𝑠 + 𝑘𝐼
𝐷𝑐 = 𝑘𝑝 + , 𝐺 = 𝐿 = 𝐷𝑐 𝐺 = 𝑛 = 1, 𝐾𝑣 = 𝐴𝑘𝐼
𝑠 𝜏𝑠 + 1 𝑠 𝜏𝑠 + 1
• System type is a robust property with respect to parameter changes in the
unity feedback structure. Robustness is the major reason for preferring unity
feedback over other kind of control structure.
• In general, the system is referred to as Type 𝑘 for the corresponding nonzero
steady state error.
1 1
𝑒𝑠𝑠 = lim𝑠𝐸 𝑠 = lim𝑠 1 − 𝑇𝑐𝑙 𝑠 = lim 1 − 𝑇𝑐𝑙 𝑠
𝑠→0 𝑠→0 𝑠 𝑘+1 𝑠→0 𝑠𝑘

11
 System Type Examples

• Ex. 4.3 System Type for a Servo with Tachometer Feedback

1
𝐺 𝑠 = 𝐷 𝑠 = 𝑘𝑝 𝐻 𝑠 = 1 + 𝑘𝑡 𝑠
𝑠 𝜏𝑠 + 1
𝐸 𝑠 =𝑅 𝑠 −𝑌 𝑠
𝐷𝐺
=𝑅 𝑠 − 𝑅 𝑠
1 + 𝐻𝐷𝐺
1 + 𝐻 − 1 𝐷𝐺
= 𝑅 𝑠
1 + 𝐻𝐷𝐺

1 + 𝐻 − 1 𝐷𝐺 1 1 𝑠 𝜏𝑠 + 1 + 1 + 𝑘𝑡 𝑠 − 1 𝑘𝑝
𝑒𝑠𝑠 = lim𝑠 = lim
𝑠→0 1 + 𝐻𝐷𝐺 𝑠 𝑘+1 𝑠→0 𝑠 𝑘 𝑠 𝜏𝑠 + 1 + 1 + 𝑘𝑡 𝑠 𝑘𝑝
= 0, 𝑘=0
1 + 𝑘𝑡 𝑘𝑝 𝑘𝑝
= , 𝑘=1 Type 1, 𝐾𝑣 =
𝑘𝑝 1 + 𝑘𝑡 𝑘𝑝

12
 System Type with Respect to Disturbances

• The transfer function from the disturbance input 𝑊(𝑠) to the error 𝐸(𝑠) is
formulated as
𝐸 𝑠 −𝑌 𝑠
= = 𝑇𝑤 𝑠
𝑊 𝑠 𝑊 𝑠
If the disturbance-to-error transfer function can be written as
1 1
𝑇𝑤 𝑠 = 𝑠 𝑛 𝑇𝑜,𝑤 𝑠 , 𝑇𝑜,𝑤 0 = , 𝑊 𝑠 = 𝑘+1
𝐾𝑛,𝑤 𝑠
1 𝑠𝑛
𝑒𝑠𝑠 = lim𝑠𝑇𝑤 𝑠 = lim𝑇𝑜,𝑤 𝑠 𝑘
𝑠→0 𝑠 𝑘+1 𝑠→0 𝑠

If 𝑛 > 𝑘, then the error is zero.

If 𝑛 < 𝑘, the error is unbounded.
1
If 𝑛 = 𝑘, the system is type 𝑘 and the error is given by
𝐾𝑛,𝑤

13
 System Type Example w.r.t. Disturbance

• Ex. 4.4 System Type for a DC Motor Position Control

If 𝐷 𝑠 = 𝑘𝑃
−𝐵
𝑇𝑤 𝑠 = = 𝑠 0 𝑇𝑜,𝑤
𝑠 𝜏𝑠 + 1 + 𝐴𝑘𝑃
𝐴𝑘𝑃
Type 0, 𝐾0,𝑤 = −
𝐵
𝑘𝐼
If 𝐷 𝑠 = 𝑘𝑃 +
𝑠
−𝐵𝑠
𝑇𝑤 𝑠 = 2 = 𝑠1 𝑇𝑜,𝑤
𝑠 𝜏𝑠 + 1 + 𝑘𝑃 𝑠 + 𝑘𝐼 𝐴

𝐴𝑘𝐼
Type 1, 𝐾𝑛,𝑤 =−
𝐵

14
 PID Control

• Proportional Feedback Control (P control)

𝑈 𝑠
𝑢 = 𝑘𝑃 𝑒 = 𝐷𝑐 𝑠 = 𝑘𝑃
𝐸 𝑠
• Proportional plus Integral Control (PI control)
𝑡
𝑈 𝑠 𝑘𝐼
𝑢 𝑡 = 𝑘𝑃 𝑒 + 𝑘𝐼 𝑒 𝜏 𝑑𝜏 = 𝐷𝑐 𝑠 = 𝑘𝑃 +
𝑡0 𝐸 𝑠 𝑠
• Proportional-Integral-Derivative Feedback Control (PID control)
𝑡
𝑈 𝑠 𝑘𝐼
𝑢 𝑡 = 𝑘𝑃 𝑒 + 𝑘𝐼 𝑒 𝜏 𝑑𝜏 + 𝑘𝐷 𝑒 = 𝐷𝑐 𝑠 = 𝑘𝑃 + + 𝑘𝐷 𝑠
𝑡0 𝐸 𝑠 𝑠
1
or 𝐷𝑐 𝑠 = 𝑘𝑃 1+ + 𝑇𝐷 𝑠
𝑇𝐼 𝑠
Derivative rate
Reset rate

Anticipatory nature of D control

15
 Proportional Control (P)

• If the plant transfer function is written as

𝐴
𝐺 𝑠 = 2
𝑠 + 𝑎1 𝑠 + 𝑎2
The characteristic equation with
proportional control is
1 + 𝑘𝑃 𝐺 𝑠 = 0
𝑠 2 + 𝑎1 𝑠 + 𝑎2 + 𝑘𝑃 = 0

𝑠 2 + 2𝜁𝜔𝑛 𝑠 + 𝜔𝑛2 = 0
• The designer can control the constant term in this equation, which determines
the natural frequency, but cannot control the damping of the equation.
• The system is type 0 and if 𝑘𝑃 is made large to get adequately small steady-
state error. However, the damping may be too low for satisfactory transient
response with p control alone.

16
 Integral Control (I)

• For the same plant 𝐺(𝑠) under integral control

𝑘𝐼
𝐸 𝑠 1 𝑠 𝑈 𝑠 𝑠 𝑘𝐼
= = = =
𝑘 𝑅 𝑆 𝑘
𝑅 𝑆 1 + 𝐼 𝐺(𝑠) 𝑠 + 𝑘𝐼 𝐺(𝑠) 1 + 𝐼 𝐺(𝑠) 𝑠 + 𝑘𝐼 𝐺(𝑠)
𝑠 𝑠
𝑘𝐼
𝑌 𝑠 𝐺(𝑠) 𝑘𝐼 𝐺(𝑠)
=𝑇 𝑠 = 𝑠 =
𝑅 𝑆 𝑘
1 + 𝐼 𝐺(𝑠) 𝑠 + 𝑘𝐼 𝐺(𝑠)
𝑠
• Assume a unit-step reference input 𝑟 𝑡 = 1(𝑡) with 𝑅 𝑠 = 1/𝑠
𝑘𝐼 𝐺 0 0 𝑘𝐼 −1
𝑦 ∞ = = 1, 𝑒 ∞ = = 0, 𝑢 ∞ = =𝐺 0 =1
0 + 𝑘𝐼 𝐺(0) 0 + 𝑘𝐼 𝐺(0) 0 + 𝑘𝐼 𝐺(0)

• Assume a unit-step disturbance input 𝑤 𝑡 = 1(𝑡) with 𝑊 𝑠 = 1/𝑠

𝑌 𝑠 𝑠𝐺(𝑠) 0∙𝐺 0
= 𝑦 ∞ = =0
𝑊 𝑆 𝑠 + 𝑘𝐼 𝐺(𝑠) 0 + 𝑘𝐼 𝐺(0)
𝑈 𝑠 𝑘𝐼 𝐺 𝑠 𝑘𝐼 𝐺 0
=− 𝑢 ∞ =− = −1
𝑊 𝑆 𝑠 + 𝑘𝐼 𝐺 𝑠 0 + 𝑘𝐼 𝐺 0

17
 History of System Error

• Zero steady-state output error in both tracking and disturbance.

• Robust properties
– Plant parameters can be tolerated.
– Regardless of the value of 𝑘𝐼 , the asymptotic tracking and disturbance rejection are
preserved provided that the closed-loop system remains stable.

18
 Derivative Control (D)

• Also called rate feedback.

• The goal
– Improve closed-loop system stability
– Speeding up the transient response
– Reducing overshoot
• A key feature is knowing the slope of the error signal.
– Takes control action based on the trend in the error signal.
– Have an anticipatory behavior.
• Disadvantages
– Tends to amplify noise.

19
 Proportional Plus Integral Control (PI)

• Consider PI control is a speed control example, where the plant is described by

𝐴
𝑌= 𝑈+𝑊
𝜏𝑠 + 1
The characteristic equation with PI control is
𝑘𝐼 𝐴
1 + 𝑘𝑃 + =0
𝑠 𝜏𝑠 + 1

𝜏𝑠 2 + 𝐴𝑘𝑃 + 1 𝑠 + 𝐴𝑘𝐼 = 0 𝑠 2 + 2𝜁𝜔𝑛 𝑠 + 𝜔𝑛2 = 0

𝐴𝑘𝐼 𝐴𝑘 + 1
𝜔𝑛 = , 𝜁= 𝑃
𝜏 2𝜏𝜔𝑛
• If the plant is second order, the characteristic equation with PI control is
𝑘𝐼 𝐴
1 + 𝑘𝑃 + =0
𝑠 𝑠 2 + 𝑎1 𝑠 + 𝑎2
𝑠 3 + 𝑎1 𝑠 2 + (𝑎2 + 𝐴𝑘𝑃 )𝑠 + 𝐴𝐾𝐼 = 0
PI control can only be used to set two coefficients. 20
 PID Control

• An important effect of D control is that it gives a sharp response to suddenly

changing signals. Because of this, the “D” term is sometimes introduced into
the feedback path.

𝑘𝐼 𝐴
1 + 𝑘𝑃 + + 𝑘𝐷 𝑠 2 =0
𝑠 𝑠 + 𝑎1 𝑠 + 𝑎2

𝑠 3 + 𝑎1 + 𝐴𝑘𝐷 𝑠 2 + 𝑎2 + 𝐴𝑘𝑃 𝑠 + 𝐴𝐾𝐼 = 0

Pole placements at desired pole locations.

21
 PID Control of Motor Speed

• Ex. 4.5
Step disturbance input Step reference input

P control usually increases the system response.

D control usually increases the system damping.
I control usually decreases the steady-state error and increase the overshoot.
22
 PI Control for a DC Motor Position Control

• Ex. 4.6
𝐷 𝑠 = 𝑘𝑃
−𝐵
𝑇𝑤 𝑠 = = 𝑠 0 𝑇𝑜,𝑤
𝑠 𝜏𝑠 + 1 + 𝐴𝑘𝑃 ℎ
𝐴𝑘𝑃 ℎ
Type 0, 𝐾0,𝑤 = −
𝐵
𝑘𝐼 −ℎ
If 𝐷 𝑠 = 𝑘𝑃 +
𝑠
−𝐵𝑠 1𝑇
𝑇𝑤 𝑠 = = 𝑠 𝑜,𝑤
𝑠 2 𝜏𝑠 + 1 + 𝑘𝑃 𝑠 + 𝑘𝐼 𝐴ℎ

𝐴𝑘𝐼 ℎ
Type 1, 𝐾𝑛,𝑤 = −
𝐵

23
 Satellite Attitude Control

• Ex. 4.7

If 𝐷 𝑠 = 𝑘𝑃 + 𝑘𝐷 𝑠
1 0
𝑇𝑤 𝑠 = = 𝑠 𝑇𝑜,𝑤
𝐽𝑠 2 + 𝑘𝐷 𝑠 + 𝑘𝑃

Type 0, 𝐾0,𝑤 = 𝑘𝑃
𝑘𝐼
If 𝐷 𝑠 = 𝑘𝑃 + + 𝑘𝐷 𝑠
𝑠
𝑠 1
𝑇𝑤 𝑠 = 3 = 𝑠 𝑇𝑜,𝑤
𝐽𝑠 + 𝑘𝐷 𝑠 2 + 𝑘𝑃 𝑠 + 𝑘𝐼

Type 1, 𝐾𝑛,𝑤 = 𝑘𝐼

24
 Ziegler-Nichols Tuning of PID Regulators

• Closed-loop transient response with Quarter decay ratio.

Process reaction curve

25
 Quarter Decay Ratio

• Ex. 4.8

1
𝑅≅
90
𝐿 ≅ 13

After tuning

1
P control: 𝑘𝑃 = = 6.92
𝑅𝐿
0.9
PI control: 𝑘𝑃 = = 6.22
𝑅𝐿
𝐿 13
𝑇𝐼 = = = 43.3
0.3 0.3
26
 Ziegler-Nichols Tuning of PID Regulators

Impulse response
• Ultimate Sensitivity Method

Neutral stable

27
 Ultimate Sensitivity Method

• Ex. 4.9

𝐾𝑢 = 15.3 After tuning

𝑃𝑢 = 42

P control: 𝑘𝑃 = 0.5𝐾𝑢 = 7.65

PI control: 𝑘𝑃 = 0.45𝐾𝑢 = 6.885

1
𝑇𝐼 = 𝑃 = 35
1.2 𝑢

28
 Feedforward Control

• Plant model inversion can be used to design the feedforward control.

tracking
disturbance
rejection

𝐴
𝐺 𝑠 = 2
𝑠 + 𝑎1 𝑠 + 𝑎2

Possible
problems?
29
 Integrator Anti-Windup

• Feedback system with actuator saturation

with anti-windup

Anti-windup provides local feedback to make the

controller stable alone when the main loop is opened
by signal saturation.
30
 Anti-Windup Compensation for PI controller

• Ex. 9.9

31
 Digital Control

• Block Diagram
Digital controller

• Ex.
𝑈 𝑧 23𝑧 − 21 5 23 21
𝐷𝑑 𝑧 = = 𝑢 𝑘+1 = 𝑢 𝑘 + 𝑒 𝑘+1 − 𝑒 𝑘
𝐸 𝑧 7𝑧 − 5 7 7 7
Difference equation
• Ex.
𝑈 𝑧 1.105𝑧 − 0.895
𝐷𝑑 𝑧 = = 1.4
𝐸 𝑧 𝑧−1
𝑢 𝑘 + 1 = 𝑢 𝑘 + 1.4 1.105𝑒 𝑘 + 1 − 0.895𝑒 𝑘

32
Continuous vs. Discrete Controllers

Output response Control signals

33
 Homework #6

• Problem 4.6 Part (b) has been modified to compute the system 𝐾𝑣 in terms of
parameters 𝐾 and 𝑘𝑡 .
• Problem 4.29 Part (a) has been modified with 𝑒𝑠𝑠 ≤ 0.01 rad/sec.
• Problem 4.30
• Problem 4.32 The expression of 𝑣𝑎 has been modified using non-inverting
𝑡
amplifier, i.e. 𝑣𝑎 = 𝑘𝑃 𝑒 + 𝑘𝐼 0
𝑒𝑑𝑡.

34