Control Systems - Complete Notes

Topics: Time Response | Controllers | Root Locus | State Space | Bode Plot | Routh Stability

1. Time Response (7 April 2026)

1.1 First Order System
Transfer Function: C(S)/R(S) = 1/(ST+1)

1.2 Unit Step Response

R(S) = 1/S
C(S) = 1/(ST+1) × 1/S
Using partial fractions: 1/[S(ST+1)] = A/S + B/(ST+1)
Solving: A = 1, B = -T
C(t) = 1 - e^(-t/T) [Unit step response]
At t = T: C(T) = 0.632 → Time constant T = time at which output magnitude is 63.2% of final steady
state value
At t = 4T: C(t) = 1 - e^(-4T/T) ≈ 0.982 → settling time ≈ 4T

1.3 Unit Ramp Response

R(S) = 1/S²
C(S) = 1/(ST+1) × 1/S²
Partial fractions: 1/[S²(ST+1)] = A/S + B/S² + C/(ST+1)
Solving: A = 1, B = -T, C = T²
C(t) = t - T + Te^(-t/T)

1.4 Settling Time

Settling Time: Time taken for the output magnitude to settle within 2-5% of its final value.
Error: e(t) = r(t) - c(t) = 1 - (1 - e^(-t/T)) = e^(-t/T)
Steady state error: lim(t→∞) e^(-t/T) = 0

2. Second Order System (8 April 2026)

Standard TF: C(S)/R(S) = ωn² / (S² + 2ζωnS + ωn²)
where: ωn = natural frequency, ζ = damping ratio (= ζωn/ωn)
ζωn = actual damping coefficient
ωd = damped natural frequency = ωn√(1 - ζ²)

Unit step input: R(S) = 1/S

C(S) = (1/S) × ωn² / (S² + 2ζωnS + ωn²)
c(t) = 1 - e^(-ζωnt)/√(1-ζ²) × sin(ωdt + φ) [Unit step response]
φ = tan⁻¹(√(1-ζ²)/ζ)

2.1 Damping Cases

i) Underdamped case (0 < ζ < 1): oscillatory, decaying
ii) Undamped case (ζ = 0): sustained oscillations
iii) Critically damped (ζ = 1): fastest non-oscillatory
iv) Overdamped (ζ > 1): sluggish, non-oscillatory

Overdamped: ζ = Actual damping / Critical damping = ζωn / ωn

2.2 Time Domain Specifications

Rise Time (tr): tr = (π - tan⁻¹(√(1-ζ²)/ζ)) / (ωn√(1-ζ²))
Time for response to rise from 10% to 90% of final value (first time after step is applied)

Peak Time (tp): tp = π / (ωn√(1-ζ²))

Settling Time (ts): ts = 4 / (ζωn) [2% criterion]

Overshoot (%Mp): %Mp = e^(-πζ/√(1-ζ²)) × 100

Amount by which system response exceeds final steady state value during transient.

2.3 Numerical Example

Given: ζ = 0.5, ωn = 6 rad/s
tr = (π - tan⁻¹(√(1-0.25)/0.5)) / (6√(1-0.25)) = 0.403 sec
tp = 0.605 sec, ts = 1.33 sec, Mp = 16.3%

3. Steady State Error

E(S)/R(S) = 1 / (1 + Gc(S)H(S))
E(S) = R(S) / (1 + Gc(S)H(S))
Final Value Theorem: ess = lim(t→∞) e(t) = lim(s→⁰) S×E(S)
ess = lim(s→⁰) S × R(S) / (1 + Gc(S)H(S))

3.1 For Unit Step Input [R(S) = 1/S]

ess = 1 / (1 + Kp) where Kp = lim(s→⁰) Gc(S)H(S) [position error coefficient]

3.2 For Unit Ramp Input [R(S) = 1/S²]

ess = 1/Kv where Kv = lim(s→⁰) S×Gc(S)H(S) [velocity error coefficient]

3.3 For Parabolic Input [R(S) = 1/S³]

ess = 1/Ka where Ka = lim(s→⁰) S²×Gc(S)H(S) [acceleration error coefficient]

3.4 System Type vs Steady State Error Table

Input Type 0 Type 1 Type 2
Unit Step 1/(1+Kp) 0 0
Unit Ramp ∞ 1/Kv 0
Unit Parabolic ∞ ∞ 1/Ka

4. Routh Stability Criterion (10 April 2026)

Characteristic equation: q(S) = a₀Sⁿ + a₁Sⁿ⁻¹ + a₂Sⁿ⁻² + ... + a ₙ = 0
4.1 Routh Array Construction
Row Sⁿ: a₀, a₂, a₄, ...
Row Sⁿ⁻¹: a₁, a₃, a₅, ...
Row Sⁿ⁻²: b₁, b₂, ...
Row Sⁿ⁻³: c₁, ...
...
Row S⁰: aₙ

b₁ = (a₁a₂ - a₀a₃) / a₁

4.2 Stability Conditions

① For stable system: each term of 1st column of Routh array is generally +ve.
② Number of sign changes in 1st column = number of roots of characteristic equation in RHS of s-
plane.

4.3 Special Cases

Case 1: Zero row (entire row is zero)
Use auxiliary polynomial from row above. Differentiate, substitute coefficients for that row, continue
array.

Case 2: Zero in first column only

Substitute s = 1/z in characteristics equation, form new array.
Example: s⁴ + s³ + 2s² + 2s + 5 = 0 → substitute s = 1/z → 5z⁴ + 2z³ + 2z² + z + 1 = 0

4.4 Numerical Examples

Example Q7: s⁴ + 2s³ + 3s² + 4s + 5 = 0
b₁ = (3×2 - 4×1)/2 = (6-4)/2 = 1
c₁ = (b₁a₃ - a₁b₂)/b₁ = (1×4 - 3×2)/1 = (4-6)/1 = -2
Two poles in right hand side → unstable

Intersection Point Example: s⁴ + 6s³ + 10s² + 8s + K = 0

Routh Array shows condition: (208/3 - 6K)/3 = 0 → K = 208/18
Auxiliary eq: 26s²/3 + K = 0 → s = ±4/3 = ±1.15i

5. Signal Flow Graph (SFG) (27 March 2026)

5.1 Key Definitions
In output node: all branches are incoming nodes.
In input node: all branches are outgoing nodes.
Transmittance = Laplace transform of O/P / Laplace transform of I/P (when all initial conditions are
zero) [Transfer Function]

5.2 Mason's Gain Formula

T = ΣgkΔk / Δ
gk = gain of kth forward path
Δk = part of Δ not touching kth forward path
Δ = 1 - [Σ individual loop gains] + [Σ two non-touching loop gains] - [Σ three non-touching loop gains]...

5.3 Steps to Draw SFG

i) Find forward paths
ii) Find single loops L₁, L₂, ...
iii) Find two non-touching loops
iv) Find Δ and Δi (Δ₁, Δ₂, ...)

5.4 Steps to Draw SFG from Differential Equation

Step 1: Solve the equation for the highest order derivative.
Step 2: Consider left-hand term as dependent variable, construct branches according to equation.
Step 3: Connect the nodes of highest order derivative to nodes of lower order; transmittance = 1/S.
Step 4: Reverse the sign of branches connecting consecutive higher-order nodes.

6. Controllers (5 May 2026)

6.1 Proportional Controller (Kp)
E(S) ∝ Ea(S) ⇒ Ea(S) = Kp E(S) ⇒ Ea(S)/E(S) = Kp
In proportional controller, the actuating signal is directly proportional to error signal.
For underdamped system: exponentially decaying oscillations.

TF = Kp × Gr(S) / (1 + Kp Gr(S) H(S))

If Gr(S) = ωn² / (S(S + 2ζωn)):
TF = Kpωn² / (S² + 2ζωnS + Kpωn²)

Advantages of Proportional Controller (Kp):

i) It is very simple.
ii) Stability increases.
iii) For higher values of Kp, it gives faster response.

Disadvantages:
i) Steady state error increases.
ii) Max. overshoot.
iii) Cannot handle non-linearity.

6.2 Integral Controller (KI)

In time domain: E(t) → [KI ∫dt] → Ea(t)
Laplace Transform: Ea(S) = (KI/S) E(S) ⇒ Ea(S)/E(S) = KI/S
[For integration: divide by S; For differentiation: multiply by S]

Gr(S) = (KI/S) × (ωn² / S(S + 2ζωn))

Gc(S) = (KI + SKp)ωn² / [S²(S + 2ζωn)]
TF = KIωn² / (S³ + 2ζωnS² + KIωn²) [after multiplying by S]

Adding 1 pole in the system makes it unstable, but eliminates steady state error.

Advantages of Integral Controller:

i) Eliminates steady state error.
ii) Improves system accuracy.
iii) Compensation for constant disturbances.

Disadvantages:
i) Slower response.
ii) Reduced stability.
iii) Takes time to get stabilized output.

6.3 Derivative Controller (KD)

In time domain: E(t) → [KD d/dt] → Ea(t)
Ea(t) = KD × dE(t)/dt
Laplace: Ea(S) = SKD E(S) ⇒ Ea(S)/E(S) = SKD

It can change the order indirectly; it becomes 1st order system.

So it is for improving the stability of the system, but cannot eliminate steady state error.

Advantages:
i) Increases stability.
ii) Faster response.
iii) Max. overshoot decreases.

Disadvantages:
i) It creates high frequency noise.
ii) Steady state error cannot be eliminated.
iii) Tuning is very challenging.

6.4 Proportional Integral (PI) Controller

Gc(S) = (Kp + KI/S) × (ωn² / S(S + 2ζωn))
Gc(S) = (KI + SKp)ωn² / [S²(S + 2ζωn)]

6.5 Proportional Derivative (PD) Controller

Gc(S) = (Kp + SKD) × (ωn² / S(S + 2ζωn))
TF = (Kp + SKD)ωn² / (S² + S(2ζωn + KDωn²) + Kpωn²)

6.6 PID Controller (6 May 2026)

O/P of PID controller: Ea(t) = Kp E(t) + KI ∫ E(t)dt + KD d/dt E(t)
Laplace: Ea(S) = Kp E(S) + (KI/S) E(S) + SKD E(S)
TF: Ea(S)/E(S) = Kp + KI/S + SKD

Gc(S) = (Kp + KI/S + SKD) × ωn²/(S(S + 2ζωn))

Closed loop TF = ωn²(KDS² + KpS + KI) / (S³ + S²(2ζωn + KDωn²) + S(Kpωn²) + KIωn²)

PID Controller Block: Ea(S)/E(S) = Kp + KI/S + SKD

E(t) is error signal, Kp = proportional gain, KI = integral gain, KD = derivative gain

6.7 Effects of Increasing PID Gains

Parameter Increase Rise Time Overshoot ess Stability
 Kp Decrease (↓) Increase (↑) Reduce (↓) Slightly
decrease
KI Decrease (↓) Increase (↑) Eliminates Decreases (↓)
KD Slightly Decrease (↓) No significant Improves (if KD
decrease effect small)

6.8 Benefits of PID Controllers

i) PID controllers are highly valued for minimizing steady state errors.
ii) It reduces overshoot and oscillations.
iii) It lowers the settling time and gives higher stability.
iv) It gives faster transient response.

6.9 PID Tuning Methods

i) Ziegler-Nichols Method:
PID parameters are determined using ultimate gain and oscillation period.

ii) Cohen-Coon Method:

Used for processes with time delay. Provides better tuning than Ziegler-Nichols in some cases.

7. Tuning of Controller
Tuning → trying to predict the values of Kp, KD, KI (units)

7.1 Ziegler-Nichols Rules

1st Method (Process / S-shape method): [Only for 1st order]
Steps: Apply step input to plant, observe S-shaped response curve.
Draw tangent at 90° from inflection point.
L → Delay Time, T → Constant Time

Controller Kp TI TD
Proportional T/L ∞ 0
PI 0.9T/L L/0.3 0
PID 1.2T/L 2L 0.5L
TF = Kp[1 + 1/(TIS) + TDS]
For PID: Kp = 1.2T/L, KI = 2L, KD = 0.5L

2nd Method (Sustained Oscillation Method):

Let TI = ∞ and TD = 0 → Kp → Kcm [Critical proportional gain]
Increase Kp until we get Kcm of sustained oscillation.
Pcm → Time period between two (+ve) peaks at sustained oscillation. The value of Kp is Kcm.

Controller Kp TI TD
P 0.5Kcm ∞ 0
PI 0.45Kcm 1/1.2 Pcm 0
PID 0.6Kcm 0.5Pcm 0.125Pcm
8. PID Tuning Numerical Example
Given: Plant G(S) = 1 / (S(S+1)(S+5)), H(S) = 1
Closed loop TF: C(S)/R(S) = Kp / (S³ + 6S² + 5S + Kp)
Characteristic equation: S³ + 6S² + 5S + Kp = 0

Routh Array:
S³ | 1 5
S² | 6 Kp
S¹ | (30-Kp)/6 0
S⁰ | Kp

Condition: 30 - Kp ≥ 0 ⇒ Kp ≤ 30, and Kp > 0

∴ Kcm = 30 [Above 30: sustained oscillations; below 30: stable; critical count = 30]

Auxiliary eq at Kcm: 6ω² + 30 = 0 ⇒ 6(jω)² = -30 ⇒ (jω)² = -5

ω = √5 (sustained oscillation frequency)
Pcm = 2π/ω = 2π/√5 = 2.81

For PID Controller: Kp = 0.6Kcm = 0.6 × 30 = 18

TI = 0.5Pcm = 0.5 × 2.81 = 1.405
TD = 0.125Pcm = 0.125 × 2.81 = 0.35125

TF of PID controller: Gc(S) = Kp[1 + 1/(TIS) + TDS]

= 18[1 + 1/(1.405S) + 0.35125S]
= 6.3223(S + 1.4235)²/S

C(S)/R(S) = (6.3223S² + 18S + 12.811) / (S⁴ + 6S³ + 11.3223S² + 18S + 12.811)

9. Root Locus (12 May 2026)

Root locus always symmetric about real axis.

9.1 Root Locus - Example 1 (Unity Feedback)

Gc(S) = K / (S(S+2)), H(S) = 1
Step 1: No. Z = 0, No. P = 2, No. of Loci = 2
Step 2: No. of Asymptotes A = 2
Step 3: Angle of Asymptotes: φA = (2k+1)180°/(P-Z)
K=0: φA1 = 90°, K=1: φA2 = 270°
Step 4: Centroid: δA = (ΣP - ΣZ)/(P-Z) = (0-2-0)/2 = -1

9.2 Root Locus - Example 2 (Unity Feedback)

Gc(S) = K(S+2)(S+3) / [(S+1)(S-1)], H(S) = 1
Step 1: No. P = 2, No. Z = 2
Poles: S₁ = -1, S₂ = 1; Zeros: S₃ = -2, S₄ = -3
Step 2: No. of Asymptotes A = 0
Step 5: Breakaway point
K = -(S²-1)/(S²+5S+6)
dK/dS = 5S² + 14S + 5 = 0 ⇒ S = -0.42, -2.37
9.3 Root Locus - Example 3
Gc(S) = K / (S(S+4)(S+5)), H(S) = 1
Step 1: No. of Loci = 3; Poles: S₁ = 0, S₂ = -4, S₃ = -5
Step 2: No. of Asymptotes = 3
Step 3: Angles: φA1 = 60°, φA2 = 180°, φA3 = 300°
Step 4: Centroid δA = -3

Intersection with imaginary axis:

Characteristic eq: S³ + 9S² + 20S + K = 0
Routh: S² row: 180 - K = 0 ⇒ K = 180
Auxiliary eq: 9S² + 180 = 0 ⇒ S² = -20 ⇒ S = ±4.47i

9.4 Root Locus - Example 4

Gc(S) = K / [S(S+4)(S²+2S+2)], Unit feedback
Step 1: No. of Loci = 4
Poles: S₁ = 0, S₂ = -4, S₃ = -1+i, S₄ = -1-i
Step 2: No. of Asymptotes = 4
Step 3: φA1 = 45°, φA2 = 135°, φA3 = 225°, φA4 = 315°
Step 4: Centroid δA = -1.5
Step 5: Breakaway point
K = -[S⁴ + 6S³ + 10S² + 8S]
dK/dS = 4S³ + 18S² + 20S + 8 = 0
S = -3.1, -0.7 (Breakaway points)

Angle of departure:
φB3 = 90°
φ₁ = 180° - tan⁻¹(1/1) = 135°
φ₂ = tan⁻¹(1/3) = 18.4°
φd = 180° - (φB3 + φ₁ + φ₂) = 180° - (90° + 135° + 180.4°) = ±116.5° ±63.4°

Intersection Point: s⁴ + 6s³ + 10s² + 8s + K = 0

Routh: (208/3 - 6K)/3 = 0 ⇒ K = 208/18
Auxiliary eq: 26s²/3 + K = 0 ⇒ s = ±4/3 = ±1.15i

10. State Space Analysis (17 April 2026)

10.1 Introduction
Advantages over Transfer Function model:
i) It is valid for non-linear, time varying systems.
ii) Applicable for MIMO (Multiple I/P Multiple O/P) systems.
iii) It explained internal state of the system.
iv) It gives zero state response.
v) It gives the full past history of the system described by state variables.

Number of state variables = Order of the system

Ẋ = AX + BU [State equation] (A = State Matrix, U = input vector, m×1 = direction)
Y = CX + DU [Output equation]
10.2 State Space Model from Transfer Function
For given TF: Gc(S) = Y(S)/U(S) = K / (S³ + a₃S² + a₂S + a₁)
Y(S) = K U(S) / (S³ + a₃S² + a₂S + a₁)
(S³ + a₃S² + a₂S + a₁)Y(S) = KU(S)
ẏ(S) + a₃ŷ(S) + a₂ẏ(S) + a₁Y(S) = KU(S)
Applying L⁻¹: ỿ(t) = -a₃ẏ(t) - a₂ẏ(t) - a₁y(t) + Ku(t)

Select state variables: let y(t) = x₁, ẏ(t) = ẋ₁ = x₂, ỿ(t) = ẋ₂ = x₃, ỿ̇(t) = ẋ₃
ẋ₃ = -a₃x₃ - a₂x₂ - a₁x₁ + Ku(t)

State Matrix form:

[ẋ₁] [0 1 0] [x₁] [0]
[ẋ₂] = [0 0 1] [x₂] + [0] u(t)
[ẋ₃] [-a₁ -a₂ -a₃][x₃] [K]

y(t) = [1 0 0] x₁(t)

10.3 Example: y(S)/u(S) = 2/(S³+6S²+11S+6)

ỿ(t) = -6ẏ(t) - 11ẏ(t) - 6y(t) + 2u(t)
State variables: x₁ = y, x₂ = ẏ, x₃ = ỿ
[A] = [[0,1,0],[0,0,1],[-6,-11,-6]], [B] = [[0],[0],[2]], y = [1,0,0]x

10.4 Example: Q5 (21 April 2026)

Given TF: Gc(S) = K / (S³ + a₃S² + a₂S + a₁)
[A] = [[0,1,0],[0,0,1],[-a₁,-a₂,-a₃]], [B] = [[0],[0],[K]], [C] = [1,0,0]

10.5 State Transition Matrix (STM) [1 April 2026 - handwritten note: 1/26]
Ẋ = AX + BU; for unforced response: U = 0 ⇒ ẋ(t) = Ax
Apply Laplace: SX(S) - x(0) = AX(S)
(S-A)X(S) = x(0) ⇒ X(S) = (S-A)⁻¹ x(0)
Apply L⁻¹: x(t) = e^(At) x(0)
∴ Φ = e^(AT) → STM (State Transition Matrix)

Φ(t) = L⁻¹[(SI-A)⁻¹] → Resolvent Matrix

TF = C[SI-A]⁻¹ B + D

10.6 Example: Determine TF from State Space

A = [[-3,1],[0,-1]], B = [[1],[1]], C = [1,1], D = [0]
TF = C[SI-A]⁻¹ B + D
[SI-A] = [[S+3,-1],[0,S+1]]

11. Controllability & Observability

11.1 Definitions
Stability of a system: Ability to return to its original state after being subjected to external disturbances
or state condition changes. 'Stability of a system' refers to the ability of the system.
Controllability: A system is said to be controllable if any initial state x(t₀) or x₀ can be transferred to any
fixed final state x(tf) in a finite time interval (tf-t₀), t>0 by some control.
For a system to be completely controllable, the only necessary & sufficient condition is that the
following n×nm matrix Ωc has rank of n:
Ωc = [B : AB : A²B : ... : Aⁿ⁻¹B]

Observability: A system is said to be observable if every state x₀ can be exactly determined from the
measurement of output y over a finite interval of time 0≤t≤tf.

11.2 Kalman's Test

For Controllability: Ωc = [B : AB : A²B ...]
For Observability: Ωo = [CT : AT CT : (AT)² CT ...]

11.3 Need for Controllability Test

The controllability test is necessary to find the usefulness of a state variable. If the state variables are
controllable, then by varying or controlling the state variables, that desired output of the systems are
achieved.

11.4 Need for Observability Test

It is necessary to find whether the state variables are measurable or not. If the state variables are
measurable, then the state of the system can be determined by practical measurement of the state
variables.

11.5 System Observability Condition

For the system to be observable, if the following n×mp matrix Ωo is non-singular, i.e., Ωo has rank (n-
1):
Ωo = [CT: AT CT: (AT)²CT ... (AT)^(n-1) CT]

11.6 Example - Check Controllability & Observability

Ẋ = [[0,1],[-1,-2]] X + [[1],[-1]] U
Y = [1 1] X

AB = [[0,1],[-1,-2]] × [[1],[-1]] = [[0-1],[-1+2]] = [[-1],[1]]

Ωc = [B:AB] = [[1,-1],[-1,1]] → |Qc| = 0 ⇒ Uncontrollable system (rank ≠ order)

CT = [[1],[1]]
AT CT = [[0,-1],[1,-2]] × [[1],[1]] = [[-1],[1-2]] = [[-1],[-1]]
Ωo = [CT: AT CT] → |Qo| = 0 ⇒ Unobservable

12. Bode Plot - Frequency Response Characteristics (26 May 2026)

1 + Gc(S)H(S) = 0; R(t) → [Gc(t)] → C(t)
Sin(ωt) → input; Sin(ωt + β) → output

For Stability Check: ωgc < ωpc ⇒ GM, PM = +ve

ωpc = Phase cross-over frequency
ωgc = Gain cross-over frequency
GM = Gain margin, PM = Phase margin
PM = -ve: below 180°; PM = +ve (about 180°)
Units of GM and PM: dB and °(degrees) respectively

12.1 Steps for Bode Plot

Step 1: Factorize (1+ST)
Step 2: S = jω
Step 3: Magnitude Table
Step 4: Magnitude Plot
Step 5: Phase Table
Step 6: Phase Plot

12.2 Magnitude Table (Cascading Order)

Factor Corner Freq. Slope Resultant Slope Starting Point
P (pole) - -20 dB/dec - -
Z (zero) - +20 dB/dec - -

13. Additional Notes & Formulas

13.1 PID Controller Design for Given Plant
Consider Gc(S) = Y(S)/U(S) = K / (S³ + a₃S² + a₂S + a₁)
Diagonal Canonical Form: Consider the system y + 6ẏ + 11ẏ + 6y = 6u
Applying L⁻¹: ỿ(t) = -6ẏ(t) - 11ẏ(t) - 6y(t) + 6u(t)
State variable matrix:
[ẋ₁] [0 1 0] [x₁] [0]
[ẋ₂] = [0 0 1] [x₂] + [0] u(t)
[ẋ₃] [-6 -11 -6][x₃] [6]
y(t) = [1 0 0] x₁(t)

13.2 PID Controller Numerical (Continued)

Ea(S)/E(S) = Kp + KI/S + SKD
Ea(t) = 2(t) + KI ∫(3+2t)dt + 0.5 d(3+2t)/dt
= 2(3+2t) + (t) + (t) + 0.5 × 2
= 6 + 4t + 3t + 2t²/2 + 0.5×2
∴ Ea(t) = t² + 7t + 7

13.3 Advantages of PID Controller

i) No steady state error.
ii) Reduce overshoot and oscillations.
iii) Lowers settling time and gives higher stability.
iv) Faster transient response.

Disadvantages:
i) Tuning is very complex.
ii) System complexity is more.
13.4 Q: Effects of Increasing PID Gains on System Response (Q14 & Q15)
Rise Time: Time required for the response of a system to rise from 10% to 90% of its final value for the
first time after a step input is applied.
Formula: tr = (π - tan⁻¹(√(1-ζ²)/ζ)) / (ωn√(1-ζ²))

Overshoot (Mp): Overshoot is the amount by which the system response exceeds its final steady-state
value during transient response.
%Mp = e^(-πζ/√(1-ζ²)) × 100

Steady-state error (ess): It's the difference between the desired O/P and actual O/P after the response
settles completely.