Transfer functions and block diagrams
A transfer function H(s) = Y(s)/X(s) assumes zero initial conditions. Block diagram reduction uses series (multiply), parallel (add), and feedback (H/(1+GH)) rules. Mason's gain formula provides a systematic alternative for complex diagrams.
Check yourself: Are initial conditions zero, and is the system linear time-invariant?
Time-domain specifications
For a second-order underdamped system, the damping ratio ζ determines overshoot and the natural frequency ωn determines speed. Rise time decreases with increasing ωn. Percent overshoot = 100 exp(−πζ/√(1−ζ²)) for 0 < ζ < 1. Settling time depends on the dominant pole's real part.
Check yourself: Are the dominant poles truly second-order, or do other poles affect the response?
Stability: Routh-Hurwitz criterion
Form the Routh array from the characteristic polynomial. The number of right-half-plane roots equals the number of sign changes in the first column. All first-column entries must be positive for stability. A zero in the first column requires the epsilon method or auxiliary polynomial.
Check yourself: Have you checked every entry in the first column for sign changes?
Frequency response: Bode plots and Nyquist
Bode magnitude plot uses 20 log₁₀|H(jω)| in dB. Each real pole contributes −20 dB/decade above its corner frequency. Gain margin is measured at the phase crossover frequency; phase margin at the gain crossover frequency. The Nyquist criterion counts encirclements of −1 to determine closed-loop stability.
Check yourself: Is the gain margin measured where the phase is −180°, not where the magnitude is 0 dB?
State-space representation
ẋ = Ax + Bu, y = Cx + Du describes a system with state vector x. Eigenvalues of A determine stability. Controllability requires rank [B AB A²B ...] = n. Observability requires rank [C; CA; CA²; ...] = n. Transfer function H(s) = C(sI − A)⁻¹B + D relates the two representations.
Check yourself: Do the eigenvalues of A match the poles of the transfer function?