AI summary by TwinMind
Lecture 26: Control, Part 3
MIT OpenCourseWare
51m 16s
Control Design Overview and Example Setup
- Presenter introducing control design tradeoffs for general converter example
- Garden variety boost converter example (not extreme in any dimension)
- Circuit components:
- Input voltage U, inductor L, switch (modulating with q(t)), diode, output capacitor C, resistive load R
- Output voltage V across load
- Example parameters chosen:
- L = 10 microhenries
- C = 50 microfarads
- Switching frequency = 100 kHz (pretty low)
- Output reference Vref = 24 volts (fixed)
- Nominal input voltage = 9V, range: 8V < U < 10V (narrow 25% range)
- Load resistance range: 2Ω < R < 10Ω (5:1 power range, heavy to light load)
State Space Modeling Progression
- Started with switched equations of state:
- dIL/dt = U/L * q + (U-V)/L * (1-q)
- dV/dt = -1/(RC) * V + (1/C) * IL * (1-q)
- Applied state space averaging to remove switching ripple:
- Average model eliminates switching function, uses duty ratio D
- dIL_bar/dt = (1/L) * U - (1/L) * V_bar * D'
- dV_bar/dt = -(1/RC) * V_bar + (1/C) * IL_bar * D'
- Where D' = (1-D)
- Average model captures low frequency dynamics, kills ripple
- Nonlinear due to products of duty cycle with state variables
Linearization Process
- Linearized around operating point for small signal analysis
- Perturbation variables: I_tilde, V_tilde, D_tilde (deviations from DC steady state)
- Linear constant coefficient differential equations:
- dI_tilde/dt = (1/L) * U_tilde - (D'/L) * V_tilde + (V/L) * D_tilde
- dV_tilde/dt = -(1/RC) * V_tilde + (D'/C) * I_tilde - (I/C) * D_tilde
- Constants depend on operating point (load current, input voltage, duty ratio)
- Valid for limited range but surprisingly good for switching converters
Transfer Function Derivation
- Laplace transform of linearized equations
- Transfer function H(s) = V_tilde(s)/D_tilde(s) with U_tilde = 0:
- H(s) = [-s * V/(RCD') + VD'/(LC)] / [s² + (1/RC)*s + D'²/LC]
- Second order system with right half plane zero and complex pole pair
Pole-Zero Analysis and Stability Challenges
- Right half plane zero location: s = V/(RC*D')
- Complex pole pair in left half plane (lightly damped)
- Pole locations depend on operating point (D' squared term)
- Right half plane zeros problematic for control:
- Poles move toward zeros as gain increases
- Can easily cause instability
- Physical behavior: output initially moves wrong direction before settling
Physical Explanation of Right Half Plane Zero
- When duty ratio increases to raise output voltage:
- Initially less current feeds output (more time to ground)
- Output voltage droops temporarily
- Eventually inductor current builds up and output rises
- Mathematical model reflects actual converter physics
- Not artifact of modeling - intrinsic system behavior
Control Design Approaches
- Proportional controller (GC(s) = KP):
- Root locus shows poles move toward right half plane zero
- Limited gain before instability
- High steady state error with low gains
- Integral controller (GC(s) = KI/s):
- Zero steady state error capability
- Still limited by right half plane zero
- Can achieve stability with careful gain selection
Operating Point Variation Effects
- Transfer function poles and zeros vary with:
- Load resistance R (affects damping - larger R = less damped)
- Input voltage (changes duty ratio D')
- Design challenge: single compensator must work across all operating conditions
- Poles move closer to jω axis as R increases (less damping)
Closed Loop Analysis with Integral Control
- KI = 5 example across load range
- Heavy load (2Ω): reasonably damped response
- Light load (10Ω): very lightly damped poles
- All operating points stable (poles in left half plane)
- Gain margin = 17 dB, phase margin ≈ 90° at worst case
Load Step Response Analysis
- Light to heavy load step: acceptable settling behavior
- Heavy to light load step: problematic response
- Large voltage overshoot (energy from inductor dumps into capacitor)
- Converter enters discontinuous conduction mode temporarily
- Highly oscillatory response despite good stability margins
- Average model accuracy remains good even for large transients
Control Performance vs Stability Metrics
- Classical stability metrics (gain/phase margins) indicate very stable system
- Actual response shows unacceptable oscillatory behavior
- Lightly damped poles cause persistent voltage ripple (10-30mV on 24V output)
- Application dependent: instrumentation vs resistive loads
- Need to consider pole damping, not just stability
Design Improvement Strategies
- Plant modification approaches:
- Larger output capacitor for energy absorption
- Damping networks (RC snubber across output)
- Component value optimization for better damping
- Control modification approaches:
- Current mode control (full state feedback)
- Monitor both inductor current and output voltage
- Improves pole placement and damping
