ECE 486: Control Systems

Lecture 20A: Frequency Domain Performance

 Key Takeaways
Most design requirements can be specified in the frequency
domain as bounds:
A) Good reference tracking and disturbance rejection
𝑆 𝑗𝜔 ≪ 1 at low frequencies
B) Good noise rejection
𝑇 𝑗𝜔 ≪ 1 at high frequencies
C) Reasonable control commands
𝐾 𝑗𝜔 𝑆(𝑗𝜔) is bounded
D) Good robustness
𝑆 𝑗𝜔 ≤ 2.5 at all frequencies

2
Requirements: Closed-Loop Stability + Robustness
Fact: Closed-loop is stable if and only if all zeros of
1+G(s)K(s) are in the LHP.
We require:
A) G(s)K(s) has no pole/zero cancellations in the CRHP
B) 𝑆 𝑠 = 1+𝐺 1𝑠 𝐾(𝑠) is stable
We also showed previously that 𝑆 𝑗𝜔 ≤ 2.5 at all
frequencies ensures good disk margins.

3
 Requirements: Reference Tracking
Goal: The output y should track the reference command r.
The transfer function from r to e=r-y is:
𝑆 𝑠 = 1+𝐺 1𝑠 𝐾(𝑠) (Sensitivity)
Consider a sinusoidal reference 𝑟 𝑡 = 𝑅0 cos(𝜔𝑡). Then:
𝑒 𝑡 → 𝑆 𝑗𝜔 𝑅0 cos( 𝜔𝑡 + ∠𝑆 𝑗𝜔 )
We require 𝑆 𝑗𝜔 ≪ 1 for good tracking at 𝜔.
If 𝜔 = 0 then 𝑟 𝑡 = 𝑅0 (step) and 𝑒 𝑡 → 𝑆(𝑗0)𝑅0 .

4
 Requirements: Disturbance Rejection
Goal: The disturbance d should have small effect on output y.
The transfer function from d to y is G(s) in open loop and
G(s)S(s) in closed-loop.
Consider a sinusoidal disturbance 𝑑 𝑡 = 𝐷0 cos(𝜔𝑡). Then:
(OL) 𝑦 𝑡 → 𝐺(𝑗𝜔) 𝐷0 cos( 𝜔𝑡 + ∠𝐺(𝑗𝜔 )
(CL) 𝑦 𝑡 → 𝐺(𝑗𝜔)𝑆 𝑗𝜔 𝐷0 cos( 𝜔𝑡 + ∠𝐺(𝑗𝜔)𝑆 𝑗𝜔 )
We require 𝑆 𝑗𝜔 ≪ 1 for good disturbance rejection at 𝜔.

5
 Requirements: Noise Rejection
Goal: The noise n should have small effect on output y.
The transfer function from n to y is –T(s) where:
𝐺 𝑠 𝐾(𝑠)
𝑇 𝑠 = 1+𝐺 𝑠 𝐾(𝑠)
(Complementary Sensitivity)
Consider a sinusoidal noise 𝑛 𝑡 = 𝑁0 cos(𝜔𝑡). Then:
𝑦 𝑡 → − 𝑇 𝑗𝜔 𝑁0 cos( 𝜔𝑡 + ∠𝑇 𝑗𝜔 )
We require 𝑇 𝑗𝜔 ≪ 1 for good noise rejection at 𝜔.

6
 Requirements: Control Effort
Goal: The control u should remain within allowable limits.
The transfer function from r to u is K(s)S(s).
Consider a sinusoidal reference 𝑟 𝑡 = 𝑅0 cos(𝜔𝑡). Then:
𝑢 𝑡 → 𝐾 𝑗𝜔 𝑆 𝑗𝜔 𝑅0 cos( 𝜔𝑡 + ∠𝐾 𝑗𝜔 𝑆 𝑗𝜔 )
To remain within saturation limits 𝑢 𝑡 ≤ 𝑢𝑚𝑎𝑥 ,
𝐾 𝑗𝜔 𝑆 𝑗𝜔 𝑅0 ≤ 𝑢𝑚𝑎𝑥 ⇒ 𝐾 𝑗𝜔 𝑆 𝑗𝜔 ≤ 𝑢𝑚𝑎𝑥 𝑅0
We also need to ensure that n does not cause large u.

7
 Design Requirements: S(s) vs. T(s)
Reference tracking and disturbance rejection: 𝑆 𝑗𝜔 ≪1
Noise rejection: 𝑇 𝑗𝜔 ≪ 1

However S(s)+T(s)=1 so we can’t have both 𝑺 𝒋𝝎 ≪ 𝟏

and 𝑻 𝒋𝝎 ≪ 𝟏 at the same frequency. This conflict is
resolved by splitting the requirements by frequency:

𝑆 𝑗𝜔 ≪ 1 at low frequencies and 𝑇 𝑗𝜔 ≪ 1 at high

frequencies.

8
 Basic Frequency Domain Trade-offs
Plant:
Controller:
Loop:
Sensitivity:
Complementary Sensitivity:

Bode magnitude
plots for Kp = 1.

9
 Basic Frequency Domain Trade-offs
Low Frequencies: Good reference tracking and disturbance
rejection but poor noise rejection.
High Frequencies: Good noise rejection but poor reference
tracking and disturbance rejection.
Middle Frequencies: Loop bandwidth 𝜔𝐿 is where 𝐿 𝑗𝜔𝐿 = 1.

Loop bandwidth:

Closed-loop time constant:

Higher bandwidths correspond

to faster response.
10
 Basic Frequency Domain Trade-offs
Low Frequencies: Good reference tracking and disturbance
rejection but poor noise rejection.
High Frequencies: Good noise rejection but poor reference
tracking and disturbance rejection.
Middle Frequencies: Loop bandwidth 𝜔𝐿 is where 𝐿 𝑗𝜔𝐿 = 1.

Kp = 1

11
 Basic Frequency Domain Trade-offs
Low Frequencies: Good reference tracking and disturbance
rejection but poor noise rejection.
High Frequencies: Good noise rejection but poor reference
tracking and disturbance rejection.
Middle Frequencies: Loop bandwidth 𝜔𝐿 is where 𝐿 𝑗𝜔𝐿 = 1.

Kp = 10

12
 Control Effort
Plant:
Controller:

𝐾𝑝 𝑠
Closed-loop r to u: 𝐾 𝑠 𝑆 𝑠 = 𝑠+𝐾𝑝

13
 ECE 486: Control Systems

Lecture 20B: Introduction to Loopshaping

 Key Takeaways
Loopshaping is a design method that focuses on the loop L(s).
We build the controller from components targeting low, middle,
and high frequencies.

Low Frequencies: Good reference tracking / disturbance rejection.

𝑆 𝑗𝜔 ≪ 1 ⇔ 𝐿 𝑗𝜔 ≫ 1
High Frequencies: Good noise rejection.
𝑇 𝑗𝜔 ≪ 1 ⇔ 𝐿 𝑗𝜔 ≪ 1

Middle Frequencies (Crossover Region):

Speed of Response: Loop bandwidth 𝜔𝐿 such that 𝐿 𝑗𝜔𝐿 =1
Stability/Robustness: Transition with a shallow slope.
2
 Speed of Response: Bandwidth
For first- and second-order systems we used settling time and/or
rise time as measures of the speed of response.
For higher-order systems, an alternative frequency domain
notion for speed of response is useful: bandwidth.
1. Loop Bandwidth, 𝝎𝑳 : Smallest frequency with 𝐿 𝑗𝜔𝐿 = 1.
2. Sensitivity Bandwidth, 𝝎𝑺 : Highest frequency such that
1
𝑆 𝑗𝜔 ≤ √2 =−3𝑑𝐵 for all 𝜔 ≤ 𝜔𝑆

3. Complementary Sensitivity Bandwidth, 𝝎𝑻 : Lowest

frequency such that
1
𝑇 𝑗𝜔 ≤ √2 =−3𝑑𝐵 for all 𝜔 ≥ 𝜔 𝑇

3
 Speed of Response: Bandwidth
Example: 𝐺 𝑠 = 1𝑠 and 𝐾 𝑠 = 12.5
Bandwidths: 𝜔𝐿 = 𝜔 𝑇 = 𝜔𝑆 = 12.5𝑟𝑎𝑑
𝑠𝑒𝑐
𝑠 1 1
Note that 𝑆 𝑠 = 𝑠+12.5 ⇒ Time Constant 𝜏 = 12.5 𝑠𝑒𝑐 = 𝜔
𝐿

4
 Speed of Response: Bandwidth
−0.5𝑠2 +1250
Example: 𝐺 𝑠 = and 𝐾 𝑠 = 10𝑠+30
𝑠3 +47𝑠2 +850𝑠−3000 𝑠
𝑟𝑎𝑑 𝑟𝑎𝑑
Bandwidths: 𝜔𝑆 = 5 , 𝜔 = 12.5
𝑠𝑒𝑐 𝐿 , 𝜔 𝑇 = 28𝑟𝑎𝑑
𝑠𝑒𝑐 𝑠𝑒𝑐
3
Settling Time is ≈ 0.6𝑠𝑒𝑐 = 𝜔𝑆

5
 Bode Gain-Phase Relation
Loopshaping focuses on |𝐿 𝑗𝜔 | with less emphasis on ∠𝐿 𝑗𝜔 .

Fact: Assume L(s) has all poles and zeros in the LHP. Then:

Comments:
1. The approximation is accurate if the slope is roughly constant
for ω ∈ [ 𝜔100 , 10ω0 ].
2. The approximation arises from an exact formula by Bode.
3. The phase change from 𝜔 = 0 to 𝜔0 is ±90𝑜 for every
𝑑𝐵
± 20𝑑𝑒𝑐𝑎𝑑𝑒 of slope.

6
 Bode Gain-Phase Relation
Loopshaping focuses on |𝐿 𝑗𝜔 | with less emphasis on ∠𝐿 𝑗𝜔 .

Fact: Assume L(s) has all poles and zeros in the LHP. Then:

7
 Requirements on the Loop L(s)
1 𝐿(𝑠)
Recall 𝐿 𝑠 = 𝐺 𝑠 𝐾 𝑠 , 𝑆 𝑠 = 1+𝐿(𝑠), 𝑇 𝑠 = 1+𝐿(𝑠)
Low Frequencies: 𝑆 𝑗𝜔 ≪ 1 ⇔ 𝐿 𝑗𝜔 ≫1
1
Note: 𝐿 𝑗𝜔 ≫ 1 ⇔ 𝐾(𝑗𝜔)𝑆 𝑗𝜔 ≈ |𝐺 𝑗𝜔 |

High Frequencies: 𝑇 𝑗𝜔 ≪ 1 ⇔ 𝐿 𝑗𝜔 ≪1

8
 Requirements on the Loop L(s)
Middle Frequencies (Crossover Region): The slope near 𝜔𝐿
should not be too steep to ensure stability and robustness.
𝑑𝐵
-A slope of ≈−40𝑑𝑒𝑐 means ∠𝐿 𝑗𝜔 ≈ −180𝑜 and closed-loop
will be unstable and/or have poor phase margins.
𝑑𝐵
-Slope should not be steeper than ≈−30𝑑𝑒𝑐 to ensure 45o margin.

9
 ECE 486: Control Systems

Lecture 20C: Controller Components For Loopshaping

 Key Takeaways
Loopshaping builds controllers from the following components:
A) Proportional Gain: A gain (> 1) increases the loop magnitude at all
frequencies. This increases bandwidth and reduces steady state error
but degrades noise rejection.
B) Integral Boost: Increases the low frequency gain but leaves the
high frequencies unchanged. This gives zero steady state error but
has negligible effect on bandwidth and noise sensitivity.
C) High Frequency Roll-off: Decreases the high frequency gain but
leaves the low frequencies unchanged. This improves noise rejection
but has negligible effect on bandwidth and steady-state error.
D) Lead: Makes the slope more shallow near the crossover
frequency. This improves robustness but it slightly degrades both the
low frequency tracking and high frequency noise rejection.

2
 Example
Plant:
Control:

3
 Example
Plant:
Control:

4
 Controller Components

Proportional Integral
Gain Boost

Roll-off Lead

5
 Proportional Gain
Proportional Gain: 𝐾(𝑠) = 𝐾𝑝
Recall the following fact for Bode magnitudes in dB:

Properties:
-If Kp>1 then gain shifts
entire loop mag. up.
-If Kp<1 then gain shifts
entire loop mag. down.
Proportional gain is used
to set the loop bandwidth
(crossover frequency).
6
 Effect of Proportional Gain
Plant:
Control:

7
 Integral Boost
Integral Boost: 𝐾(𝑠) = 𝑠+𝑠𝜔ഥ
Properties:
-Corner frequency 𝜔,
ഥ high frequency gain |𝐾 𝑗𝜔 | = 1, and
𝑑𝐵
low frequency slope of −20𝑑𝑒𝑐 .
-Corresponds to PI control:

Integral boost is used to

increase low frequency
gain and ensure zero
steady-state error.
8
 Effect of Integral Boost
Plant:
Control:

Integral control ensures zero error in steady-state

9
 High Frequency Roll-off
Roll-off: 𝐾(𝑠) = 𝑠+𝜔ഥ𝜔ഥ
Properties:
-Corner frequency 𝜔,
ഥ low frequency gain |𝐾(𝑗𝜔)| = 1, and
𝑑𝐵
high frequency slope of −20𝑑𝑒𝑐 .
-Corresponds to the ODE:

Roll-off is used to decrease

high frequency gain and
attenuate sensor noise.

10
 Effect of Roll-off
Plant:
Control:

11
 Lead
ഥ
Lead: 𝐾(𝑠) = 𝛽𝑠+
𝑠+𝛽𝜔
𝜔
ഥ

Properties:
-Zero at −𝜔𝛽ഥ and pole at −𝛽𝜔,
ഥ
-Low frequency gain 𝛽1 and high frequency gain 𝛽
-Positive slope at 𝜔
ഥ

Lead is used to make the

slope shallower and hence
ensure stability and
robustness.
12
 Effect of Lead
Plant:
Control:

13
 ECE 486: Control Systems

Lecture 20D: Loopshaping Design Process

 Key Takeaways
The basic steps of the loopshaping process are:
1) Use a proportional gain to set the desired crossover frequency.
This sets the bandwidth / speed of response.
2) Use an integral boost to increase |𝐿 𝑗𝜔 | at low frequencies.
This improves the reference tracking and disturbance rejection.
3) Use a roll-off to reduce |𝐿 𝑗𝜔 | at high frequencies. This
improves the noise rejection.
4) Add lead control (if needed) to modify the slope of |𝐿 𝑗𝜔 | near
the crossover. This is used for closed-loop stability and robustness.
This approach can be used on higher-order plants using controllers
that are, in general, more complex than a PID controller.

2
 Basic Design Process
Key design parameter: Desired loop crossover 𝜔𝑐
1
1. Proportional Gain: Select 𝐾𝑝 = ±
|𝐺 𝑗𝜔𝑐 |
Loop 𝐿1 = 𝐺 𝐾𝑝 has the desired crossover, 𝐿 𝑗𝜔𝑐 = 1.
𝑠+𝜔𝑖
2. Integral Boost: Select 𝐾𝑖 𝑠 = with 𝜔𝑖 ≤ 𝜔𝑐
𝑠
Loop 𝐿2 = 𝐺 𝐾𝑝 𝐾𝑖 has improved low frequency tracking.
Good initial choice 𝜔𝑖 = 𝜔𝑐 /3 so that |𝐾𝑖 𝑗𝜔 | ≈ 1 for 𝜔 ≥ 𝜔𝑐 .
𝜔𝑟
3. Roll-off: Select 𝐾𝑟 𝑠 = with 𝜔𝑟 ≥ 𝜔𝑐
𝑠+𝜔𝑟
Loop 𝐿3 = 𝐺 𝐾𝑝 𝐾𝑖 𝐾𝑟 has improved noise rejection / robustness.
Good initial choice 𝜔𝑟 = 3𝜔𝑐 so that |𝐾𝑟 𝑗𝜔 | ≈ 1 for 𝜔 ≤ 𝜔𝑐 .
𝛽𝑠+𝜔𝑐
4. Lead (If needed): Select 𝐾𝑙 𝑠 = with 𝛽 ≈ 3 − 10
𝑠+𝛽𝜔𝑐
Loop 𝐿4 = 𝐺 𝐾𝑝 𝐾𝑖 𝐾𝑟 𝐾𝑙 has improved stability margins
3
 Example 1: First-Order System
0.25
Design a loopshaping controller for 𝐺 𝑠 = −
𝑠+0.5
Desired crossover at 𝜔𝑐 = 1.5 𝑟𝑎𝑑/𝑠𝑒𝑐

4
 Step 1: Proportional Gain
0.25 𝑟𝑎𝑑
Plant 𝐺 𝑠 = − and desired crossover at 𝜔𝑐 = 1.5
𝑠+0.5 𝑠𝑒𝑐
1
𝐾𝑝 = − = −6.32 (Note Kp<0 because G(0)<0).
𝐺 𝑗𝜔𝑐
L1 = G Kp

5
 Step 2: Integral Boost
0.25 𝑟𝑎𝑑
Plant 𝐺 𝑠 = − and desired crossover at 𝜔𝑐 = 1.5
𝑠+0.5 𝑠𝑒𝑐
𝑠+𝜔𝑖 𝜔𝑐 𝑟𝑎𝑑
𝐾𝑖 = with 𝜔𝑖 = = 0.5
𝑠 3 𝑠𝑒𝑐
L2 = G Kp Ki

6
 Step 3: Roll-off
0.25 𝑟𝑎𝑑
Plant 𝐺 𝑠 = − and desired crossover at 𝜔𝑐 = 1.5
𝑠+0.5 𝑠𝑒𝑐
𝜔𝑟 𝑟𝑎𝑑
𝐾𝑟 = with 𝜔𝑟 = 3𝜔𝑐 = 4.5
𝑠+𝜔𝑟 𝑠𝑒𝑐
L3 = G Kp Ki Kr

7
 Step 4: Lead
0.25 𝑟𝑎𝑑
Plant 𝐺 𝑠 = − and desired crossover at 𝜔𝑐 = 1.5
𝑠+0.5 𝑠𝑒𝑐
Loop L3 = G Kp Ki Kr has a “shallow” slope near crossover.

The closed-loop is stable with [0,∞) gain margins and ±72𝑜

phase margins.

No lead control is required.

Final Controller:

8
 Example 1: Matlab Code
>> G = -tf(0.25,[1 0.5]); % Plant
>> wc = 1.5; % Desired crossover, rad/sec

>> Kp = -1/abs(evalfr(G, 1j*wc)); % Proportional Gain

>> wi = wc/3; % Boost frequency, rad/sec
>> Ki = tf([1 wi],[1 0]); % Integral Boost
>> wr = 3*wc; % Roll-off frequency, rad/sec
>> Kr = tf(wr,[1 wr]); % Roll-off
>> K = Kp*Ki*Kr; % Final Controller
>> L3 = G*K; % Final loop

>> S = feedback(1,L3); % Closed-loop sensitivity

>> isstable(S) % Verify closed-loop stability
>> allmargin(L3) % Classical margins
 Example 2: Higher-Order System
Design a loopshaping controller for
4 400 15
𝐺 𝑠 = 2 2
𝑠 𝑠 + 0.08𝑠 + 400 𝑠 + 15
Desired crossover at 𝜔𝑐 = 2.0 𝑟𝑎𝑑/𝑠𝑒𝑐
1
Step 1) Gain: Select 𝐾𝑝 = ≈1
𝐺 𝑗𝜔𝑐

𝑠+𝜔𝑖
Step 2) Boost: 𝐾𝑖 =
𝑠
𝜔𝑐
with 𝜔𝑖 =
3

𝜔𝑟
Step 3) Rolloff: 𝐾𝑟 =
𝑠+𝜔𝑟
with 𝜔𝑟 = 3𝜔𝑐
10
 Example 2: Higher-Order System
Design a loopshaping controller for
4 400 15
𝐺 𝑠 = 2 2
𝑠 𝑠 + 0.08𝑠 + 400 𝑠 + 15
Desired crossover at 𝜔𝑐 = 2.0 𝑟𝑎𝑑/𝑠𝑒𝑐
1
Step 1) Gain: Select 𝐾𝑝 = ≈1
𝐺 𝑗𝜔𝑐

𝑠+𝜔𝑖
Step 2) Boost: 𝐾𝑖 =
𝑠
𝜔𝑐
with 𝜔𝑖 =
5

𝜔𝑟
Step 3) Rolloff: 𝐾𝑟 =
𝑠+𝜔𝑟
with 𝜔𝑟 = 5𝜔𝑐
11
 Step 4: Lead
Loop L3 = G Kp Ki Kr has a “steep” slope near crossover.
Closed-loop is unstable with L3 so lead control is needed.
𝛽𝑠+𝜔𝑐
𝐾𝑙 𝑠 = with 𝛽 = 8
𝑠+𝛽𝜔𝑐
L4 = G Kp Ki Kr Kl → Closed-loop is stable 45o of phase margin.
 Example 2: Matlab Code
>> G1 = tf(1,[1 0 0]);
>> H = 4*tf(400,[1 2*0.02*20 400])*tf(15,[1 15]);
>> G = G1*H; % Plant
>> wc = 2.0; % Desired crossover, rad/sec

>> Kp = 1/abs(evalfr(G, 1j*wc)); % Proportional Gain

>> wi = wc/5; % Boost frequency, rad/sec
>> Ki = tf([1 wi],[1 0]); % Integral Boost
>> wr = 5*wc; % Roll-off frequency, rad/sec
>> Kr = tf(wr,[1 wr]); % Roll-off
>> wl = wc; % Lead frequency, rad/sec
>> beta = 8; % Lead parameter
>> Kl = tf([beta wl],[1 beta*wl]); % Lead
>> K = Kp*Ki*Kr*Kl; % Final Controller
>> L4 = G*K; % Final loop

>> S = feedback(1,L4); % Closed-loop sensitivity

>> isstable(S) % Verify closed-loop stability
>> allmargin(L4) % Classical margins
 PID vs. Loopshaping
PID with approximate derivative:

Loopshaping with proportional, integral boost, and lead:

These are different parameterizations for the same class of controllers.

Loopshaping can be viewed as a generalization of PID that enables
• Additional controller components (rolloff, notches, etc)
• Closer connection to frequency-domain trade-offs
• Extensions to multivariable systems.
 ECE 486: Control Systems

Lecture 20E: Loopshaping Design Theorems

 Key Takeaways
This lecture presents two important “theorems” regarding the
loopshaping design process.

Under mild conditions, the loopshaping design process will yield a

stable closed-loop with good stability margins.

2
 Basic Assumptions on L(s)=G(s)K(s)
1. L(s) has all poles and zeros in the LHP.
2. L(0)>0
3. One crossover 𝜔𝑐

3
 Basic Assumptions on L(s)=G(s)K(s)
1. L(s) has all poles and zeros in the LHP.
2. L(0)>0
3. One crossover 𝜔𝑐
𝑑𝐵
4. Shallow slope (≥ −30𝑑𝑒𝑐 ) for one decade around 𝜔𝑐

4
 Basic Assumptions on L(s)=G(s)K(s)
1. L(s) has all poles and zeros in the LHP.
2. L(0)>0
3. One crossover 𝜔𝑐
𝑑𝐵
4. Shallow slope (≥ −30𝑑𝑒𝑐 ) for one decade around 𝜔𝑐
5. 𝐿 𝑗𝜔 ≥ 2 for 𝜔 ≤ 𝜔1

5
 Basic Assumptions on L(s)=G(s)K(s)
1. L(s) has all poles and zeros in the LHP.
2. L(0)>0
3. One crossover 𝜔𝑐
𝑑𝐵
4. Shallow slope (≥ −30𝑑𝑒𝑐 ) for one decade around 𝜔𝑐
5. 𝐿 𝑗𝜔 ≥ 2 for 𝜔 ≤ 𝜔1
1
6. 𝐿 𝑗𝜔 ≤ for 𝜔 ≥ 𝜔2
2

6
 Loopshaping Design Theorem
1. L(s) has all poles and zeros in the LHP.
2. L(0)>0
3. One crossover 𝜔𝑐
𝑑𝐵
4. Shallow slope (≥ −30𝑑𝑒𝑐 ) for one decade around 𝜔𝑐
5. 𝐿 𝑗𝜔 ≥ 2 for 𝜔 ≤ 𝜔1
1
6. 𝐿 𝑗𝜔 ≤ for 𝜔 ≥ 𝜔2
2

If L(s) satisfies 1-6 then

the closed-loop is stable
with approximate gain,
phase, and disk margins
≥ ±6dB, ≥ ±45o, and
dmin ≥ 0.5
7
Loopshaping Design Theorem With Integrators
1. L(s)=𝑠1𝑘𝐻(𝑠) where H(s) has all poles and zeros in the LHP.
2. H(0)>0
3. One crossover 𝜔𝑐
𝑑𝐵
4. Shallow slope (≥ −30𝑑𝑒𝑐 ) for one decade around 𝜔𝑐
5. 𝐿 𝑗𝜔 ≥ 2 for 𝜔 ≤ 𝜔1
1
6. 𝐿 𝑗𝜔 ≤ for 𝜔 ≥ 𝜔2
2

If L(s) satisfies 1-6 then

the closed-loop is stable
with approximate gain,
phase, and disk margins
≥ ±6dB, ≥ ±45o, and
dmin ≥ 0.5
8