Closed-loop Frequency Response

Bode’s Gain-Phase Relationship

 For any stable minimum-phase system (no poles or zeros in the RHP), the phase of 𝐺 𝑗𝜔
is uniquely related to the magnitude of 𝐺 𝑗𝜔 :
1) Frequency
Response 𝑀 = log 𝐺 𝑗𝜔
1 𝑑𝑀
2) Bode Plot In radians ∠𝐺 𝑗𝜔 = 𝑊 𝑢 𝑑𝑢 ∴ 𝑢 = log 𝜔 𝜔
3) System Stability 𝜋 𝑑𝑢
𝑑𝑀 𝜋 Weighting
4) Closed-loop  If = 𝑐𝑜𝑛𝑠𝑡. 𝑊≅ 𝛿 𝑢
Frequency 𝑑𝑢 2 Function
Response
for a decade after or
5) Design for
Dynamic before 𝜔 :
Compensation 𝜋 𝑑𝑀
6) Case Study ∠𝐺 𝑗𝜔 ≅
2 𝑑𝑢
∠𝐺 𝑗𝜔 ≅ 𝑛 × 90°
𝑛 is the slope of gain plot
(log-log scale).

Example:
𝜔 = 0.1 𝑛 = −2
∠𝐺 𝑗0.1 ≅ −180°

𝜔 = 10 𝑛 = −1
∠𝐺 𝑗10 ≅ −90°
24
 M.R. Emami, 2025
 Closed-loop Frequency Response
Bode’s Gain-Phase Relationship 𝐺 𝑠
Closed-loop System Stability Condition (common case):
1) Frequency
Response
𝐾𝐺 𝑗𝜔 < 1 at ∠𝐺 𝑗𝜔 = −180°
2) Bode Plot  For stability, adjust 𝐾𝐺 𝑗𝜔 so it has a slope of −1 at the “crossover”
3) System Stability frequency 𝜔 (i.e., where 𝐾𝐺 𝑗𝜔 = 1). If the slope is −1 for a decade of
4) Closed-loop
Frequency
frequency range centred at the crossover frequency, then 𝑃𝑀 ≅ 90° .
Response
Rule of Thumb:
5) Design for
Dynamic  Adjust the slope of the magnitude (gain) curve 𝐾𝐺 𝑗𝜔
Compensation
so it crosses over magnitude 1 with a slope of −1 for a
6) Case Study
decade around 𝜔 .
 The above rule typically ensures stability of the closed-
loop system with adequate damping.
 To achieve the desired speed of response, the system
gain is adjusted so that the crossover point is at a
frequency that will yield the desired bandwidth.
NOTE:
 In many systems, the natural frequency 𝜔 , bandwidth
𝜔 , and crossover frequency 𝜔 are close to each
other.
25
 M.R. Emami, 2025
 Closed-loop Frequency Response
Bode’s Gain-Phase Relationship
Example: Spacecraft Attitude Control
1) Frequency
 Find a suitable 𝐾𝐷 𝑠 to provide good damping
Response (𝑀 < 15%) and bandwidth of ~0.2 ⁄.
2) Bode Plot  Determine the frequency where the sensitivity function 𝒮 = 0.7 = −3 𝑑𝑏 .
3) System Stability
4) Closed-loop
PD Control: 𝐾𝐷 𝑠 = 𝐾 𝑇 𝑠 + 1
Frequency  Adjust the gain 𝐾 to obtain the desired bandwidth,
Response and adjust break point 𝜔 = 1⁄𝑇 to have the −1
5) Design for slope well around crossover frequency.
Dynamic
Compensation  Pick a value of 𝐾 to have crossover (≅ bandwidth) at
6) Case Study 0.2 ⁄ , and choose a value 𝜔 about 4 times lower
than 𝜔 to make the slope about −1 around the crossover.

 Plot 𝐺 𝑗𝜔 .
 Modify the plot to include𝐷 𝑗𝜔 with 𝜔 =
1⁄20 𝑟𝑎𝑑⁄𝑠, so the slope will be ≅ −1 well around
𝜔 = 0.2 𝑟𝑎𝑑⁄𝑠.

 Determine 𝐷 𝐺 = 100 where 𝐷 𝐺 curve crosses

the line 𝜔 = 0.2 𝑟𝑎𝑑⁄𝑠, which is where we want
magnitude 1 crossover to be.

 For crossover to be at 𝜔 = 0.2 𝑟𝑎𝑑⁄𝑠:

1
𝐾= = 0.01 𝐾𝐷 = 0.01 20𝑠 + 1
𝐷𝐺 .
26
 M.R. Emami, 2025
 Closed-loop Frequency Response
Bode’s Gain-Phase Relationship
Example: Spacecraft Attitude Control
1) Frequency
 Find a suitable 𝐾𝐷 𝑠 to provide good damping
Response (𝑀 < 15%) and bandwidth of ~0.2 ⁄.
2) Bode Plot  Determine the frequency where the sensitivity function 𝒮 = 0.7 = −3 𝑑𝑏 .
3) System Stability
4) Closed-loop
PD Control: 𝐾𝐷 𝑠 = 0.01 20𝑠 + 1
Frequency 𝐾𝐷 𝐺
Response 𝒯 𝑗𝜔 = 𝜔 = 0.25 𝑟𝑎𝑑⁄𝑠 ≅ 𝜔
1 + 𝐾𝐷 𝐺
5) Design for
Dynamic 1
𝒮 𝑠 =
Compensation 1 + 𝐾𝐷 𝐺
6) Case Study = 0.15 𝑟𝑎𝑑⁄𝑠
𝒮 ≅ 0.7 = −3 𝑑𝑏 at 𝜔

Disturbance Rejection Bandwidth (𝜔 ):

 The frequency at which the sensitivity of the closed-
loop system falls below −3 𝑑𝑏.
 In general, 𝜔 is the maximum frequency at
which the disturbance rejection (the sensitivity
function 𝒮) is below a certain amount (often
− 3 𝑑𝑏 ).

Step Satisfactory
𝑀 ≅ 14%
Response Damping

27
 M.R. Emami, 2025
 Closed-loop Frequency Response
Dynamic Response
 For common systems where the following typical 𝑅 𝑠 + 𝑈 𝑠 𝑌 𝑠
behaviour is observed for their open-loop 𝐾 𝐺 𝑠
_
1) Frequency magnitude (𝜔 is the crossover frequency):
Response
2) Bode Plot 𝐾𝐺 𝑗𝜔 ≫1 𝑓𝑜𝑟 𝜔≪𝜔
3) System Stability 𝐾𝐺 𝑗𝜔 ≪1 𝑓𝑜𝑟 𝜔≫𝜔
4) Closed-loop the closed-loop frequency response 𝐾𝐺 𝑗𝜔 1, 𝜔≪𝜔
Frequency
magnitude is approximated by: 𝒯 𝑗𝜔 = ≅
Response 1 + 𝐾𝐺 𝑗𝜔 𝐾𝐺 , 𝜔≫𝜔
5) Design for  The magnitude 𝒯 𝑗𝜔 around 𝜔
Dynamic
Compensation
heavily depends on the PM.
6) Case Study Example: 𝐾 = 1
𝑃𝑀 = 45° ∠𝐺 𝑗𝜔 = −135°
°
𝐺 𝑗𝜔 =𝑒 = cos −135° + 𝑗 sin −135°
𝐺 𝑗𝜔
𝒯 𝑗𝜔 = = 1.31
1 + 𝐺 𝑗𝜔
 In general, if 𝑃𝑀 = 90° , then 𝜔 = 𝜔 ; and if 𝑃𝑀 < 90° , then : 𝜔 ≤ 𝜔 ≤ 2𝜔
 NOTE: The bandwidth is roughly equal to the natural frequency of
the closed-loop system, with an error typically less than a factor of 2.
°
 NOTE: Bandwidth (𝜔 ) and Phase Margin (PM) are 𝜁 ≅ 𝑃𝑀
𝑃𝑀 < 65°
the two characteristics usually used for defining the 100
desired frequency response of the closed-loop system. 𝜔 ≅ 𝜔
28
 M.R. Emami, 2025
 Closed-loop Frequency Response
Steady State Error
 For Type 0 system, the gain plot of open-loop TF (in a unity
feedback control) starts with a slope 0 at low frequencies.
1) Frequency
Response
𝐾 = lim 𝐾𝐷 𝐺 𝑠 = lim 𝐾𝐷 𝐺 𝑗𝜔
2) Bode Plot → →
L
3) System Stability  𝐾 is the value of open-loop gain at 𝜔 = 0.
4) Closed-loop
Frequency  For Type 1 system, the gain plot of the open-loop TF (in a
Response unity feedback control) starts with a slope -1
5) Design for (−20 𝑑𝑏⁄𝑑𝑒𝑐𝑎𝑑𝑒) at low frequencies. L
Dynamic
Compensation 𝐾 = lim 𝑠𝐾𝐷 𝐺 𝑠 = lim 𝜔 𝐾𝐷 𝐺 𝑗𝜔
→ →
6) Case Study
Thus, at low frequencies:
𝐾
𝐾𝐷 𝐺 𝑗𝜔 ≅ 𝜔
 𝐾 is the value of open-loop gain where the low-frequency
asymptote with slope -1 (−20 𝑑𝑏⁄𝑑𝑒𝑐𝑎𝑑𝑒) crosses 𝜔 = 1.

 For Type 2 system, the gain plot of the open-loop TF (in a

unity feedback control) starts with a slope -2 L
(−40 𝑑𝑏⁄𝑑𝑒𝑐𝑎𝑑𝑒) at low frequencies.
𝐾 = lim 𝑠 𝐾𝐷 𝐺 𝑠 = lim 𝜔 𝐾𝐷 𝐺 𝑗𝜔
→ →

Thus, at low frequencies: 𝐾

𝐾𝐷 𝐺 𝑗𝜔 ≅
𝜔
 𝐾 is the value of open-loop gain where the low-frequency
asymptote with slope -2 (−40 𝑑𝑏⁄𝑑𝑒𝑐𝑎𝑑𝑒) crosses 𝜔 = 1.
29
 M.R. Emami, 2025
 Design for Dynamic Compensation
PD Compensation

 The controller 𝐷 𝑠 is added to the

1) Frequency
Response
feedback control loop to improve stability K

2) Bode Plot
and dynamic response.
3) System Stability 𝐷 𝑠 = 𝑇 𝑠+1
4) Closed-loop
Frequency  Stabilizing effect on the closed-loop system
Response
response by increasing the phase and the
5) Design for
Dynamic corresponding +1 slope at frequencies
Compensation above the break point .
6) Case Study
 Locate the break point so that the
increased phase occurs near the crossover,
i.e., the frequency at which 𝐾𝐷 𝐺 𝑗𝜔 = 1,
thus increasing the PM. D

 With PD controller, compensation continues to

grow with increasing frequency, hence
amplifying the high-frequency noise that is
typically present in any real system.

 Lead compensation is more practical, since it

limits the amplification of high-frequency noise.
D
30
 M.R. Emami, 2025
 Design for Dynamic Compensation
Lead Compensation
 Practically to alleviate the high-frequency K
amplification of PD compensation, a first-order pole is
1) Frequency
added at frequencies much higher (5 times or more)
Response
than the PD break point.
2) Bode Plot Corner Frequencies
𝑇 𝑠+1
3) System Stability 𝐷 𝑠 = 𝛼<1
4) Closed-loop
𝛼𝑇 𝑠 + 1

db
Frequency
Response 𝜙 = ∠𝐷 𝑗𝜔 = tan 𝑇 𝜔 − tan 𝛼𝑇 𝜔
5) Design for 1 1
Dynamic 𝜔| =𝜔 = 𝜙 = tan − tan 𝛼
𝑇 𝛼 𝛼
Compensation
6) Case Study 1−𝛼 1 − sin 𝜙
sin 𝜙 = 𝛼=
1+𝛼 1 + sin 𝜙
1
𝑇 1 1 1
log 𝜔 = log = log + log
𝛼𝑇 2 𝑇 𝛼𝑇
 The lead ratio 1⁄𝛼 is selected to make a trade-off
between a desired PM (good damping) and an
acceptable high-frequency noise amplification. The
lead compensation usually contributes a maximum
70° to the phase.
 For a greater phase lead, double-lead compensation can
be used: 𝑇 𝑠+1 𝑇 𝑠+1
𝐷 𝑠 =
𝛼 𝑇 𝑠+1 𝛼 𝑇 𝑠+1
31
 M.R. Emami, 2025
 Design for Dynamic Compensation
Lead Compensation
Example: DC Motor Lead Compensation (2nd–order position system, servomechanism)
1) Frequency
 Design a lead compensator to obtain response to a unit-ramp input with an overshoot
Response 𝑀 < 25% and a steady-state error of no more than 0.1.
2) Bode Plot 𝑌 𝑠 1
Plant TF: 𝐺 𝑠 = =
3) System Stability 𝑈 𝑠 𝑠 𝑠+1
4) Closed-loop 𝑇 𝑠+1 1
Open-loop TF: 𝐿 𝑠 = 𝐾𝐷 𝑠 𝐺 𝑠 = 𝐾
Frequency 𝛼𝑇 𝑠 + 1 𝑠 𝑠 + 1
Response 1
𝑒 = lim 𝑠 𝑅 𝑠
5) Design for → 1 + 𝐾𝐷 𝑠 𝐺 𝑠
Dynamic 1 1 1
Compensation 𝑅 𝑠 = 𝑒 = lim =
6) Case Study
𝑠 →
𝑠 + 𝐾𝐷 𝑠 1 𝐾𝐷 0
𝑠+1
𝐾 = 𝐾𝐷 0 ≥ 10 𝐾 = 10

𝑀 < 25% 𝑃𝑀 = 45°

 PM for 𝐾𝐺 𝑠 at crossover
is 20° . Add more than 25° ,
since compensator zero will
increase crossover.

𝛼 = 1 − sin 𝜙 ⁄ 1 + sin 𝜙
Add 40° by the lead 1⁄𝛼 = 5
𝑠 +1
𝐷 𝑠 = 10 𝑠 2
Choose 𝑇 to have maximum lead 1⁄𝑇 = 𝜔 𝛼 , trial & error
compensation at desired crossover
𝑇 = 0.5 10 + 1
32
 M.R. Emami, 2025
 Design for Dynamic Compensation
Lead Compensation
Example: DC Motor Lead Compensation (2nd–order position system, servomechanism)
1) Frequency  Design a lead compensator to obtain response to a unit-ramp input with an
Response
overshoot 𝑀 < 25%, and a steady-state error of less than 0.1.
2) Bode Plot
3) System Stability
4) Closed-loop  Root Locus Method
Frequency 𝑠+2
Response 𝐷 𝑠 = 70
𝑠 + 10
5) Design for
Dynamic 𝜁 = 0.56 and 𝜔 = 7.7 𝑟𝑎𝑑⁄𝑠
Compensation
6) Case Study  Rise Time ≅ 0.26
 Overshoot ≅ 22%

 Frequency Response Method

𝑠 +1
𝐷 𝑠 = 10 𝑠 2
10 + 1
 Lower Gain
 Rise Time ≅ 0.33
 Overshoot ≅ 20%

33
 M.R. Emami, 2025
 Design for Dynamic Compensation
Lead Compensation
 In a lead-compensation design, 3 primary
design parameters are usually specified:
1) Frequency
Response  Crossover frequency 𝜔 , which determines the bandwidth
2) Bode Plot 𝜔 (or vice versa), hence speed of response, i.e., rise time 𝑡 and settling time 𝑡 .
3) System Stability  Phase margin 𝑃𝑀, which determines damping ratio 𝜁 and overshoot 𝑀 .
4) Closed-loop  Low-frequency gain 𝐾, which determines steady-state error characteristics.
Frequency
Response  In general, lead compensation increases 𝜔 ⁄ 𝐾𝐷 𝐺 0 . So, if the denominator (numerator)
5) Design for is fixed, the numerator (denominator) is increased (decreased).
Dynamic
Compensation Lead Compensation Design Procedure:
6) Case Study  Determine the gain 𝐾 to satisfy error or bandwidth requirements:
(a) to meet error requirements, pick 𝐾 to satisfy error constants (𝐾 , 𝐾 , 𝑜𝑟𝐾 ), so that steady-state
error specification is met; or alternatively,
(b) to meet bandwidth requirements, pick 𝐾 so that the open-loop crossover frequency is within a
factor of two below the desired closed-loop bandwidth.
 Evaluate the PM of the uncompensated system using the value of 𝐾 obtained from Step 1.
 Allow for extra margin (about 5° or more), and determine the needed phase lead 𝜙 .
 Determine 𝛼 from 𝛼 = 1 − sin 𝜙 ⁄ 1 + sin 𝜙 .
 Pick 𝜔 to be at the (desired) crossover frequency; thus the zero is at =𝜔 𝛼 and the pole is
at =𝜔 ⁄ 𝛼.
 Draw the compensated frequency response and check for the PM.
 Iterate on the design. Adjust compensator parameters (poles, zeros, and gain) until all specifications
are met. Add an additional lead compensator if necessary, i.e., double-lead compensation.
34
 M.R. Emami, 2025
 Design for Dynamic Compensation
Lead Compensation
10
Example: Type 1 Servomechanism System 𝐾𝐺 𝑠 = 𝐾
𝑠 𝑠⁄2.5 + 1 𝑠⁄6 + 1
1) Frequency
Response
 Design a lead compensator to obtain 𝑃𝑀 = 45° and 𝐾 = 10.
1 1 1 1
2) Bode Plot  = = lim 𝑠 𝐾=1
3) System Stability 𝐾 10 → 1 + 𝐾𝐷 𝑠 𝐺 𝑠 𝑠
4) Closed-loop
Frequency
 𝑃𝑀| ≅ −4° at 𝜔 ≅ 4 𝑟𝑎𝑑⁄𝑠
Response
5) Design for  Allow and extra 5° for the 𝑃𝑀:
Dynamic 𝑃𝑀 = 45° + 5° − −4° = 54°
Compensation
6) Case Study
 𝜙 = 54° 𝛼 = 0.1

 Choose a desired 𝜔 of say

4 + 2 𝑟𝑎𝑑 ⁄𝑠: 1 1
𝑇 = ≅
𝜔 𝛼 2
𝑠 +1 1 𝑠+2
𝐷 𝑠 =𝑠 2 =
20 + 1 0.1 𝑠 + 20
 Check the 𝑃𝑀 on Bode plot: 𝑃𝑀 = 23° . More iterations show a single lead
cannot meet the 𝑃𝑀 requirement, due to the high-frequency slope of −3.
 Double the lead compensator (at a higher 𝜔 ):
𝑠 ⁄2 + 1 𝑠 ⁄4 + 1 1 𝑠+2 𝑠+4
𝐷 𝑠 = = 𝑃𝑀 = 46° 35
 M.R. Emami, 2025 𝑠⁄20 + 1 𝑠⁄40 + 1 0.1 𝑠 + 20 𝑠 + 40
 Design for Dynamic Compensation
PI Compensation
 The controller 𝐷 𝑠 is added to the
K
1) Frequency feedback control loop to reduce the steady-
Response state error and keep the bandwidth low.
2) Bode Plot
3) System Stability 1
𝐷 𝑠 = 1+ 10/TI
4) Closed-loop 𝑇𝑠
Frequency
Response  High gain at low frequencies, which reduces
2/TI
5) Design for the steady-state error (increases system type).
Dynamic 1/TI
Compensation  Causes the reduction of phase margin at
6) Case Study frequencies lower than the break point at , 0.2/TI
which degrades the stability conditions. 0.1/TI

 The break point frequency is usually

positioned at a frequency substantially less
than the crossover frequency, so that 𝑃𝑀 of
the closed-loop system is not affected
significantly.
 Susceptible to integrator overflow (windup),
so lag compensation is used as an
approximation to PI control (also avoiding
much phase reduction at low frequencies.)
36
 M.R. Emami, 2025
 Design for Dynamic Compensation
Lag Compensation
 Better stability features than PI control.
1) Frequency 𝑇𝑠+1 K
Response 𝐷 𝑠 =𝛼 𝛼>1
𝛼𝑇 𝑠 + 1
2) Bode Plot
3) System Stability
 Unlike for lead compensation, the pole of lag
compensator has a lower break-point frequency Corner Frequencies
4) Closed-loop
Frequency than its zero.
Response  Gain ≈ 𝛼
 For 𝜔 ≪
5) Design for
Dynamic
 Phase ≈ 0°
Compensation  Gain ≈ 1
 For 𝜔 ≫
6) Case Study
 Phase ≈ 0°
 The low-frequency gain and phase lag (reduction)
depends on 𝛼.
 The effect is increasing the gain at low frequencies
(similar to PI) by choosing 𝛼 properly, but at the cost
of decreasing (lagging) the phase.
 Lag compensator’s pole and zero are selected
relatively close together and well below the
uncompensated system crossover frequency,
i.e., 𝑇 is large, to minimize the effect on the 𝑃𝑀
as well as the system’s dynamic response, while
reducing 𝑒 by a factor of 𝛼.

37
 M.R. Emami, 2025
 Design for Dynamic Compensation
Lag Compensation

1) Frequency K
Response
2) Bode Plot
3) System Stability
4) Closed-loop 𝑇𝑠+1
Frequency 𝐷 𝑠 =𝛼 𝛼>1
Response 𝛼𝑇 𝑠 + 1
5) Design for
Dynamic
Compensation Lag Compensation Design Procedure:
6) Case Study
 Determine gain 𝐾 to satisfy PM requirement without compensation. (Consider addition
of 5°-10° to the desired PM to account for the (small) lag caused by the compensator.)
 Draw the Bode plot of the uncompensated open-loop TF, and evaluate the low-frequency
gain (steady-state error).
 Determine 𝛼 to meet the steady-state error requirement.
 Choose the upper corner frequency (the zero of the lag compensator) to be one
octave to multiple decades below the crossover frequency 𝜔 (of the uncompensated
system) to minimize the added phase lag.
 Obtain the lower corner frequency (the pole location of the lag compensator) from .
 Iterate on the design. Adjust compensator parameters (poles, zeros, and gain) until all
specifications are met.
38
 M.R. Emami, 2025
 Design for Dynamic Compensation
Lag Compensation 𝐾=1
Example: Design a lag compensator for the 𝐺 𝑗𝜔 = 0.38
following plant to obtain a response to step input with
1) Frequency an overshoot less than 15% and steady-state error
Response less than 2%. 115 𝜔 = 2.5 𝑟𝑎𝑑/𝑠
2) Bode Plot 𝐺 𝑠 =
𝑠 + 1 𝑠 + 3 𝑠 + 28
3) System Stability
4) Closed-loop
 𝑀 < 15% 𝑃𝑀 = 56°
𝑃𝑀 = 66°
Frequency -180
Add 10°
Response 𝑃𝑀 = 66°
5) Design for
Dynamic  Draw the Bode plot of
Compensation the uncompensated 𝐾 = 2.63
6) Case Study open-loop TF with 𝐾 = 1.
𝑃𝑀 = 66° 𝜔 = 2.5 𝑟𝑎𝑑/𝑠
1
𝐺 𝑗𝜔 = 0.38 𝐾= = 2.63
0.38 𝜔 = 2.5 𝑟𝑎𝑑/𝑠

 Draw the Bode plot with 𝐾 = 2.63,

𝐾 = 3.6
and determine uncompensated 𝐾 .
1
 𝑒 =
1+𝐾
< 0.02 𝐾 > 49 𝐾 = 50 𝑃𝑀 = 66°

𝐾
𝛼= = 13.9 → 14
𝐾
 Choose one decade below the crossover frequency: 1
𝑇 = 0.25
𝑟𝑎𝑑
𝑠
0.25 4𝑠 + 1
 1
𝛼𝑇 =
14
= 0.018 𝑟𝑎𝑑
𝑠 𝐷 𝑠 = 14
56𝑠 + 1
39
 M.R. Emami, 2025
 Design for Dynamic Compensation
Lag Compensation
Example: Design a lag compensator for the
following plant to obtain a response to step input with
1) Frequency an overshoot less than 15% and steady-state-error
Response less than 2%. 115
2) Bode Plot 𝐺 𝑠 =
𝑠 + 1 𝑠 + 3 𝑠 + 28
3) System Stability
4𝑠 + 1 115
4) Closed-loop 𝐿 𝑠 = 𝐾𝐷 𝑠 𝐺 𝑠 = 2.63 14
Frequency 56𝑠 + 1 𝑠 + 1 𝑠 + 3 𝑠 + 28
Response
5) Design for
1
𝐾 = 50.5 𝑒 = ≅ 0.019 < 0.02
Dynamic
Compensation
 Check the compensated open-loop TF: 1+𝐾
𝑃𝑀 = 60.5° 𝑀 ≅ 10% < 15%
6) Case Study
𝐾 = 50.5

𝑀 = 4%

𝑒 = 1.9%
𝜔 = 2.5 𝑟𝑎𝑑/𝑠

𝑃𝑀 = 60.5°

40
 M.R. Emami, 2025
 Design for Dynamic Compensation
PID Compensation
 To improve bandwidth, 𝑃𝑀 (at 𝜔 ), and low-
frequency gain simultaneously (simultaneous
1) Frequency improvement in transient and steady-state
Response
responses).
2) Bode Plot 1
3) System Stability
𝐷 𝑠 =𝐾 𝑇 𝑠+1 1+
𝑇𝑠
4) Closed-loop
Frequency  Performance of PID control is roughly equivalent
Response to combining lead and lag compensators, i.e., a
5) Design for lead-lag compensator:
Dynamic
Compensation
𝑇 𝑠+1 𝑇𝑠+1
𝐷 𝑠 =𝛾 𝛾>1
6) Case Study 𝑇 𝛾𝑇 𝑠 + 1
𝑠+1
𝛾

41
 M.R. Emami, 2025
 Design for Dynamic Compensation
PID Compensation
Example: Spacecraft Attitude Control
 Sensor lag and disturbing torque are present.
1) Frequency
Response 𝑒 | =0
2) Bode Plot PID controller 𝑃𝑀 = 65°
3) System Stability 𝜔 = 𝑟𝑒𝑎𝑠𝑜𝑛𝑎𝑏𝑙𝑦 ℎ𝑖𝑔ℎ
4) Closed-loop Uncompensated 0.9 2
Frequency 𝐺 𝑠 =𝐺 𝑠 𝐻 𝑠 =
Open-loop TF 𝑠 𝑠+2
Response
5) Design for 𝜔 ≅ 0.95 𝑟𝑎𝑑 𝑠 𝑃𝑀 ≅ −25° unstable
Dynamic
Compensation  The system is unstable for any value of 𝐾. So, ^
derivative term is needed to bring slope to -1 at
6) Case Study
crossover frequency.
𝐷 = 𝑇 𝑠+1
 Adjust 𝑇 to increase 𝑃𝑀:
1 ^
≥ 2 𝑟𝑎𝑑 𝑠 Phase lead only compensates
𝑇 sensor phase lag
1
= 0.1 𝑟𝑎𝑑 𝑠 Largest value that could provide
𝑇 the required 𝑃𝑀 1
^

1 ^
≤ 0.05 𝑟𝑎𝑑 𝑠 No significant increase in the 1

𝑇 crossover (thus bandwidth)

1
0.05 ≤ ≤ 0. 1 𝑟𝑎𝑑 𝑠 ^
𝑇
 A larger (desired) crossover will result in a higher
bandwidth: 1
= 0.1 𝑟𝑎𝑑 𝑠 desired 𝜔 = 0.5 𝑟𝑎𝑑 𝑠 42
 M.R. Emami, 2025 𝑇
 Design for Dynamic Compensation
PID Compensation
Example: Spacecraft Attitude Control
𝑒 | =0
1) Frequency
Response PID controller 𝑃𝑀 = 65°
2) Bode Plot 𝜔 = 𝑟𝑒𝑎𝑠𝑜𝑛𝑎𝑏𝑙𝑦 ℎ𝑖𝑔ℎ
3) System Stability
Θ 𝐺 0.9 𝑠 + 2
4) Closed-loop = =
Frequency 𝑇 1 + 𝐷 𝐻𝐺 𝑠 𝑠 + 2 + 1.8𝐷
Response 1 0.9 𝑠 + 2 1 1
5) Design for 𝑇 = 𝑒 | = lim 𝑠 = lim
𝑠 → 𝑠 𝑠 + 2 + 1.8𝐷 𝑠 → 𝐷
Dynamic
Compensation  To make 𝑒 | = 0 the compensator 𝐷 must contain ^
6) Case Study an integral term: 1
𝐷 𝑠 = 1+
𝑇𝑠
 Choose a factor of 20 lower than ^
1
to have little effect on the phase = 0.005 𝑟𝑎𝑑 𝑠
at crossover (thus not lowering 𝑃𝑀). 𝑇

 To find a proper value for 𝐾, draw the magnitude curve of ^

compensated system with 𝐾 = 1 and find the magnitude
^^
at desired crossover, which is inverse of desired 𝐾:

𝜔 = 0.5 𝑟𝑎𝑑 𝑠 𝐷 𝐻𝐺 𝑠 = 20 𝐾 = 0.05

^^
1
𝐷 𝑠 = 0.05 10𝑠 + 1 1 +
200𝑠
NOTE: The system would be unstable if the gain 𝐾 is
43
 M.R. Emami, 2025 lowered so that 𝜔 ≤ 0.02 ⁄ . (Conditionally Stable System)
 Design for Dynamic Compensation
PID Compensation
Example: Spacecraft Attitude Control
1) Frequency 𝑒 | =0
Response
PID controller 𝑃𝑀 = 65°
2) Bode Plot
𝜔 = 𝑟𝑒𝑎𝑠𝑜𝑛𝑎𝑏𝑙𝑦 ℎ𝑖𝑔ℎ
3) System Stability
4) Closed-loop 1
Frequency 𝐷 𝑠 = 0.05 10𝑠 + 1 1+
Response
200𝑠
5) Design for Θ 𝐷𝐺
𝒯= = step command
Dynamic Θ 1 + 𝐷 𝐻𝐺
Compensation
Θ 𝐺 Θ = 1°
6) Case Study = e.g., 𝑇 = 0.0175 𝑁. 𝑚
𝑇 1 + 𝐷 𝐻𝐺 𝑇 = 0 𝑁. 𝑚

 The relatively slow response to disturbance is due

to the closed-loop pole and zero in the vicinity of
𝑠 = −0.005, since is positioned slow enough to
not impact 𝑃𝑀.
Θ = 0°
NOTE: The orbital period for low-Earth orbit (LEO) 𝑇 = 0.0175 𝑁. 𝑚
satellites is ~6000 𝑠𝑒𝑐 (90-100 minutes).

 If such a response is not accepted, select a lower

value of to obtain some extra 𝑃𝑀, thus
allowing for higher values of without much
compromising the desired 𝑃𝑀 .
44
 M.R. Emami, 2025
 Design for Dynamic Compensation
PID Compensation
Example: Spacecraft Attitude Control
1) Frequency 𝑒 | =0
Response
2) Bode Plot PID controller 𝑃𝑀 = 65°
3) System Stability 𝜔 = 𝑟𝑒𝑎𝑠𝑜𝑛𝑎𝑏𝑙𝑦 ℎ𝑖𝑔ℎ
4) Closed-loop 1
Frequency 𝐷 𝑠 = 0.05 10𝑠 + 1 1+
Response 200𝑠
Θ 𝐷𝐺
5) Design for 𝒯= =
Dynamic Θ 1 + 𝐷 𝐻𝐺
Compensation
1
6) Case Study Sensitivity Function: 𝒮 𝑠 =
1 + 𝐷 𝐻𝐺

Bandwidth (𝜔 ): When 𝒯 𝑗𝜔 crosses

the magnitude 0.707 −3 𝑑𝑏 .
𝜔 = 0.7 𝑟𝑎𝑑/𝑠
Disturbance Rejection Bandwidth (𝜔 ):
When 𝒮 𝑗𝜔 crosses the magnitude
0.707 −3 𝑑𝑏 .
𝜔 = 0.3 𝑟𝑎𝑑/𝑠

NOTE: Most (periodic) disturbances on satellites have a periodicity at the orbital rate of
~0.001 ⁄ , where 𝒮 𝑗𝜔 is very small (~10 ). So, the controller causes a large
attenuation of error around the disturbance frequencies.

45
 M.R. Emami, 2025
 Design for Dynamic Compensation
Design Considerations
 The Bode plot of open-loop TF 𝐿 𝑠 = 𝐾𝐷 𝐻𝐺 𝑠
determines the performance of the closed-
1) Frequency loop system w.r.t. stability and dynamic
Response K
response, steady-state errors, as well as
2) Bode Plot low-frequency disturbance and high-frequency
3) System Stability noise filtering.
4) Closed-loop
Frequency Stability & Dynamic Response:
Response  Identifying proper (high) phase margin 𝑃𝑀 and crossover frequency 𝜔 (and bandwidth 𝜔 ).
5) Design for
Dynamic Steady-state Errors & Low-frequency Disturbances:
Compensation  Placing lower bound on low-frequency gain of open-loop TF (and
6) Case Study identifying proper 𝜔 ).
High-frequency Noise, Dynamic Resonance & Sensitivity:
 Gain Stabilization: Placing upper bound on high-frequency gain
and bandwidth (using lead, lead with double pole, or double-lead.)

 Phase Stabilization: Using notch filters at sensor/system

resonant frequencies if they are known exactly.
𝑊: Low-frequency lower bound on frequency response.
𝑊 : High-frequency upper bound on frequency response.
 Select a crossover near the desired bandwidth within the two
bounds, which should occur at a slope of −1 (or slightly steeper)
for good 𝑃𝑀 and hence stability and damping.
46
 M.R. Emami, 2025
 Design for Dynamic Compensation
Lead-Lag Compensation
𝑇𝑠+1 𝑇 𝑠+1
𝐷 𝑠 =𝛽 𝛼 < 1,𝛽 > 1
1) Frequency
𝛽𝑇 𝑠 + 1 𝛼𝑇 𝑠 + 1
Response
2) Bode Plot
 Determine gain 𝐾 to satisfy bandwidth requirements without compensation; pick 𝐾 so that the open-loop
crossover frequency is within a factor of two below the desired closed-loop bandwidth.
3) System Stability
4) Closed-loop
 Check whether the uncompensated system (proportional controller) achieves the desired phase margin.
(Consider addition of 5°-10° to the desired PM to account for the (small) phase reduction caused by the lag
Frequency
compensator.) If not, a lead compensator needs to be designed for the plant first. If yes, go to Step (8).
Response
5) Design for  Determine the needed phase lead 𝜙 .
Dynamic  Determine 𝛼 from 𝛼 = 1 − sin 𝜙 ⁄ 1 + sin 𝜙 .
Compensation
6) Case Study
 Pick 𝜔 to be at the (desired) crossover frequency first; thus as a first iteration the zero is at =𝜔 𝛼
and the pole is at = 𝜔 ⁄ 𝛼 . Further iterations may be needed for 𝜔 in Step (6).
 Draw the lead-compensated frequency response and check for the PM. Adjust the gain 𝐾 to have desired PM
at the desired 𝜔 .
 Iterate on the design. Adjust compensator parameters (poles, zeros, and gain) until all specifications of
transient response are met. Add an additional lead compensator if necessary, i.e., double-lead compensation.
 From the Bode plot of the lead-compensated (or uncompensated if lead is not required) open-loop TF, evaluate
the low-frequency gain with respect to steady-state error requirements, i.e., error constants (𝐾 , 𝐾 , or 𝐾 ). If
error requirements are not met, a lag compensator is needed.
 Determine 𝛽 to satisfy the steady-state error requirements.
 Choose the upper corner frequency (the zero of the lag compensator) to be one octave to multiple decades
below the crossover frequency 𝜔 (of the lead-compensated system) to minimize the added phase lag.
11 Obtain the lower corner frequency (the pole location of the lag compensator) from .
12 Iterate on the design. Adjust compensator parameters (poles, zeros, and gain) until all specifications are met.
47
 M.R. Emami, 2025
 Case Study
Lead-Lag Compensation

K
1) Frequency
Response
2) Bode Plot
3) System Stability
4) Closed-loop
Frequency  Example: Quadrotor Drone Control
Response
𝜃 𝑠 1
5) Design for =𝐺 𝑠 =
Dynamic 𝑇 𝑠 𝑠 𝑠+2
Compensation 𝑡 ≤1𝑠 𝜔 ≥ 1.8 𝑟𝑎𝑑 ⁄𝑠
6) Case Study
𝑀 % ≤ 40 𝜁 ≥ 0.3 ; 𝑃𝑀 ≥ 30°
Specs:
𝑡 ≤ 10 𝑠 2% 𝜎 = 𝜁𝜔 ≥ 0.46
𝑒 ≤ 10% 𝑒 = 1 𝐾 ≤ 0.1 Lead-Lag is likely needed.

𝜔 =2 𝜔 ≅2 𝜔 = 0.5𝜔 = 1 𝑟𝑎𝑑⁄𝑠
Choose:
𝑃𝑀 = 30° + 10° = 40°

1 1 1
 𝐺 𝑗𝜔 =
𝑗1
=
𝑗1 + 2−𝑗1 − 2
= 0.447 𝐾 =
0.447
= 2.24
−1
 ∠𝐾 𝐺 𝑗𝜔 = ∠2.24 − ∠ −𝑗1 − 2 = − tan = −206.57° ≅ −207° 𝑃𝑀 ≅ −27°
−2
 𝑃𝑀 = 𝜙 = 27° + 40° = 67°
48
 M.R. Emami, 2025
 Case Study
Lead-Lag Compensation
 Example: Quadrotor Drone Control K
1) Frequency
Response 1 − sin 𝜙 1 − sin 67°
2) Bode Plot  𝛼=
1 + sin 𝜙
=
1 + sin 67°
= 0.042
3) System Stability 1 4.88𝑠 + 1
4) Closed-loop  Choose: 𝜔 =𝜔 =1 𝑇 =
𝜔 𝛼
= 4.88 𝐷 𝑠 =
0.21𝑠 + 1
Frequency
Response
5) Design for
 The desired PM can be obtained at the desired crossover frequency 𝜔 = 1.
Dynamic 𝑗4.88 + 1 2.24 𝐾
𝐾 𝐷 𝐺 𝑗𝜔 = = 4.86 𝐾= = 0.46 𝑃𝑀 = 40°
Compensation 𝑗0.21 + 1 𝑗1 𝑗1 + 2 4.86
6) Case Study
𝐾 = 2.24 𝐾 = 2.24 4.86 = 0.46

49
 M.R. Emami, 2025
 Case Study
Lead-Lag Compensation
 Example: Quadrotor Drone Control K
1) Frequency
Response  Check the time response:
2) Bode Plot
3) System Stability
4) Closed-loop
Frequency
Response
5) Design for
Dynamic
Compensation
𝐾 = 2.24 4.86 = 0.46 𝐾 = 2.24 4.5 = 0.50
6) Case Study

4.88𝑠 + 1 1
 𝐾 = lim 𝑠 𝐿 𝑠 = lim 0.50
→ → 0.21𝑠 + 1 𝑠 + 2
= 0.25

1 𝐾
 𝑒 =
𝐾
= 0.1 𝐾 = 10 𝛽=
𝐾
= 40

1
 Choose:
𝑇
= 0.1𝜔 = 0.1 𝑇 = 10 10𝑠 + 1
𝐷 𝑠 = 40
400𝑠 + 1
𝛽𝑇 = 400

50
 M.R. Emami, 2025
 Case Study
Lead-Lag Compensation
 Example: Quadrotor Drone Control K
1) Frequency 10𝑠 + 1 4.88𝑠 + 1
Response 𝐷 𝑠 = 40 ; 𝐾 = 0.50
400𝑠 + 1 0.21𝑠 + 1
2) Bode Plot
3) System Stability
4) Closed-loop
Frequency
Response
5) Design for
Dynamic
Compensation
6) Case Study

1 20𝑠 + 1
Iterate: = 0.05𝜔 = 0.05 𝑇 = 20 𝛽𝑇 = 800 𝐷 𝑠 = 40
𝑇 800𝑠 + 1

51
 M.R. Emami, 2025
 Case Study
Lead-Lag Compensation
 Example: Quadrotor Drone Control K
1) Frequency
Response Iterate on the lead compensator design:
2) Bode Plot
𝜔 = 1.5 𝑟𝑎𝑑 ⁄𝑠
3) System Stability Choose:
4) Closed-loop 𝑃𝑀 = 30° + 10° = 40°
Frequency
Response 1 1 1
5) Design for
 𝐺 𝑗𝜔 =
𝑗1.5 𝑗1.5 + 2
=
1.5 −𝑗1.5 − 2
= 0.178 𝐾 =
0.178
= 5.625
Dynamic −1.5
Compensation  ∠𝐾 𝐺 𝑗𝜔 = ∠5.625 − 2∠ 1.5 − ∠ −𝑗1.5 − 2 = − tan ≅ −217°
−2
6) Case Study
 𝑃𝑀 = 𝜙 = 37° + 40° = 77° 𝑃𝑀 ≅ −37°
1 − sin 𝜙 1 − sin 77°
 𝛼=
1 + sin 𝜙
=
1 + sin 77°
= 0.013
1
 Choose: 𝜔 = 𝜔 = 1.5 𝑇 =
𝜔 𝛼
= 5.81
5.81𝑠 + 1
𝐷 𝑠 =
0.076𝑠 + 1
𝑗8.715 + 1 5.625
 𝐾 𝐷 𝐺 𝑗𝜔 =
𝑗0.114 + 1 𝑗1.5 𝑗1.5 + 2
= 8.71
𝐾
𝐾= = 0.65 𝑃𝑀 = 39°
8.71
 Check the time response:
52
 M.R. Emami, 2025
 Case Study
Lead-Lag Compensation
 Example: Quadrotor Drone Control K
1) Frequency 5.81𝑠 + 1
Response 𝐷 𝑠 = ; 𝐾 = 0.65
0.076𝑠 + 1
2) Bode Plot
5.81𝑠 + 1 1
3) System Stability  𝐾 = lim 𝑠 𝐿 𝑠 = lim 0.65
→ → 0.076𝑠 + 1 𝑠 + 2
= 0.325
4) Closed-loop
Frequency 1 𝐾
Response  𝑒 =
𝐾
= 0.1 𝐾 = 10 𝛽=
𝐾
= 30.77
5) Design for
1
Dynamic
Compensation
 Choose:
𝑇
= 0.1𝜔 = 0.15 𝑇 = 6.67

6) Case Study
𝛽𝑇 = 205.24
6.67𝑠 + 1
𝐷 𝑠 = 30.77
205.24𝑠 + 1

6.67𝑠 + 1 5.81𝑠 + 1
𝐷 𝑠 = 30.77 ; 𝐾 = 0.65
205.24𝑠 + 1 0.076𝑠 + 1
1
Iterate: = 0.025𝜔 = 0.0375 𝑇 = 26.67
𝑇
26.67𝑠 + 1
𝛽𝑇 = 820.63 𝐷 𝑠 = 30.77
820.63𝑠 + 1
26.67𝑠 + 1 5.81𝑠 + 1
𝐷 𝑠 = 30.77 ; 𝐾 = 0.65
820.63𝑠 + 1 0.076𝑠 + 1
53
 M.R. Emami, 2025
 Case Study
Lead-Lag Compensation
 Example: Quadrotor Drone Control K
1) Frequency 𝜃 𝑠 1
Response =𝐺 𝑠 =
𝑇 𝑠 𝑠 𝑠+2
2) Bode Plot
26.67𝑠 + 1 5.81𝑠 + 1
3) System Stability 𝐷 𝑠 = 30.77 ; 𝐾 = 0.65
820.63𝑠 + 1 0.076𝑠 + 1
4) Closed-loop
Frequency
Response
5) Design for 𝜔 ≅ 2.6 𝑟𝑎𝑑 𝑠
Dynamic
Compensation
𝜔 ≅ 1.9 𝑟𝑎𝑑 𝑠
6) Case Study 𝜔 ≅ 0.93 𝑟𝑎𝑑 𝑠

54
 M.R. Emami, 2025
 Case Study
A Summary

1) Frequency
 PD control adds phase lead at all frequencies above the break point. If there is no
change in gain on the low-frequency asymptote, PD compensation will increase the
Response crossover frequency and the speed of response. The increase in magnitude of the
2) Bode Plot frequency response at the higher frequencies will increase the system's sensitivity
3) System Stability to noise.

 Lead
4) Closed-loop
Frequency compensation adds phase lead at a frequency band between the two break
Response points, which are usually selected to bracket the crossover frequency. If there is no
5) Design for change in gain on the low-frequency asymptote, lead compensation will increase
Dynamic both the crossover frequency and the speed of response over the uncompensated
Compensation
system.
6) Case Study

 PIbreak
control increases the frequency-response magnitude at frequencies below the
point, thereby decreasing steady-state errors. It also contributes phase lag
below the break point, which must be kept at a low enough frequency to avoid
degrading the stability excessively.

 Lag compensation increases the frequency-response magnitude at frequencies

below the two break points, thereby decreasing steady-state errors. Alternatively,
with suitable adjustments in K, lag compensation can be used to decrease the
frequency-response magnitude at frequencies above the two break points, so that it
yields an acceptable PM. Lag compensation also contributes phase lag between the
two break points, which must be kept at frequencies low enough to keep the phase
decrease from degrading the PM excessively. This compensation will typically
provide a slower response than using lead compensation.
55
 M.R. Emami, 2025