AI summary by TwinMind
Lecture 25: Control, Part 2
MIT OpenCourseWare
50m 52s
Local Average Operator Review
- Definition: x̄(t) = (1/T) ∫[t-T to t] x(τ)dτ
- Properties:
- Linear and time invariant (LTI operator)
- Differentiation property: d/dt x̄(t) = d/dt[x(t)]
- Product approximation: x(t)y(t) ≈ x̄(t)ȳ(t) for small ripple and slow variation
- Purpose: Filters out switching frequency content while preserving low-frequency dynamics
Boost Converter Circuit Analysis
- Two subcircuits based on switch state:
- Q(t) = 1 (switch on): Linear circuit with inductor connected to input
- Q(t) = 0 (switch off): Linear circuit with inductor connected through diode to output
- State variables: Inductor current (iL) and capacitor voltage (V)
- Periodic steady state: V = U/(1-D) where D is duty ratio
State Space Averaging Method
- Unified differential equation approach:
- L(diL/dt) = U·Q(t) + (U-V)·(1-Q(t))
- C(dV/dt) = -V/R·Q(t) + (iL-V/R)·(1-Q(t))
- After local averaging:
- L(dīL/dt) = Ū - V̄·D'
- C(dV̄/dt) = -V̄/RC + īL·D'
- Where D' = 1-D (complement of duty ratio)
- Result: Nonlinear averaged model (no switching, but products of state variables and control input)
Linearization Process
- Small signal expansion: x(t) = X + x̃(t)
- X = DC/steady-state value
- x̃(t) = small variation from steady state
- Taylor expansion around operating point:
- For f(x,r,t) = 0 at operating point
- Linear approximation: dx̃/dt = (∂f/∂x)x̃ + (∂f/∂r)r̃
- Applied to boost converter:
- All DC terms cancel out (steady-state condition)
- Product terms (small×small) neglected as second-order
- Results in linear constant-coefficient differential equations
Transfer Function Derivation
- Small signal model yields:
- L(dĩL/dt) = ũ + V·d̃ - D'·ṽ
- C(dṽ/dt) = -ṽ/RC + D'·ĩL - I·d̃
- Transfer function H(s) = ṽ/d̃:
- H(s) = (-s·IL/C + V·D'/LC) / (s² + s/(RC) + D'²/LC)
- Second-order system (matches two energy storage elements)
- Coefficients depend on operating point (D', IL, V)
Experimental Validation
- Load step response demonstration:
- Current transitions from lower to higher steady-state value
- Voltage shows second-order oscillatory response
- Thousands of switching cycles during transient
- Model accuracy:
- Predicted oscillation period ≈ 1.4ms
- Measured transient period ≈ 1.6-2ms
- Good agreement despite large step (factor of 2 load change)
Control System Implementation
- Feedback control structure:
- Reference voltage (Vref) compared to actual output (V̄)
- Error signal processed through compensator GC(s)
- Compensator output becomes duty ratio command
- Linear model enables standard control theory application
- Operating point dependency:
- Pole locations change with duty ratio
- D' = 0.5 gives different dynamics than D' = 0.25
- Must account for operating point in controller design
Advantages of State Space Averaging
- Compared to circuit averaging: Same results, different implementation approach
- Benefits:
- Captures low-frequency dynamics without switching detail
- Enables linear control theory application
- Provides transfer function for controller design
- Computationally efficient for system analysis
