Scroll back to top

    Transfer Functions

    What Are Transfer Functions?

    Transfer functions are the Laplace domain representation of signal modification in a system where the input = u(t) or U(s), output = y(t) or Y(s), and the transfer function = h(t) or H(s) where "s" indicates a function in the Laplace domain. The transfer function H(s) is defined as the ratio between the output and the input:

    $$ H(s) = \frac{Y(s)}{U(s)} $$

    Essentially this means that the output is equal to the input multiplied by the transfer function:

    $$ Y(s) = H(s)*U(s) $$
    Example Problem: Finding System Output from a Block Diagram #tf_simple
    Figure 1. A single-block system with an input and transfer function.

    Using the relationship \( Y(s) = H(s) cdot U(s) \):

    $$ Y(s) = \frac{4}{(s+3)(s+4)} $$

    Another important consideration of transfer functions is that they can be combined and simplified, similar to the components of an electrical circuit. Transfer functions in series are multiplied together, while branches that are in parallel are added or subtracted depending on the summing junction signs.

    Example Problem: Single Block Representation of Combined Systems #tf_combined
    Figure 2. Multiple transfer function blocks arranged in series and parallel branches.

    Combining the series cascade and parallel difference gives the overall transfer function:

    $$ H(s) = G_{1}G_{2}(G_{4}G_{5}-G_{3}) $$

    Try practicing with systems that combine these aspects, and use the table of common Laplace transforms (here) to convert your frequency-domain outputs back to the time domain.

    Eigenvalues and Transfer Function Poles

    In linear systems analysis, the roots of the transfer function's denominator polynomial are termed the poles of the system. From the perspective of differential equations and linear algebra, these poles are identical to the eigenvalues of the system's dynamic state matrix. Eigenvalues and poles govern how a system evolves over time along its principal directions, defining the natural modes of the system independent of the applied input.

    State-Space Characteristic Equation and Dynamic Modes

    Consider a continuous-time linear time-invariant (LTI) system expressed in state-space representation with state vector \( \mathbf{x}(t) \in \mathbb{R}^n \), input \( \mathbf{u}(t) \in \mathbb{R}^m \), and output \( \mathbf{y}(t) \in \mathbb{R}^p \):

    $$ \dot{\mathbf{x}}(t) = \mathbf{A}\mathbf{x}(t) + \mathbf{B}\mathbf{u}(t) $$
    $$ \mathbf{y}(t) = \mathbf{C}\mathbf{x}(t) + \mathbf{D}\mathbf{u}(t) $$

    The natural response describes the unforced behavior of the system (\( \mathbf{u}(t) = \mathbf{0} \)):

    $$ \dot{\mathbf{x}}(t) = \mathbf{A}\mathbf{x}(t) $$

    Assuming an exponential solution mode of the form \( \mathbf{x}(t) = \mathbf{v}e^{\lambda t} \), where \( \mathbf{v} \) is a constant non-zero vector and \( \lambda \) is a scalar rate, substituting this into the unforced state equation yields:

    $$ \lambda \mathbf{v} e^{\lambda t} = \mathbf{A}\mathbf{v} e^{\lambda t} \implies (\lambda \mathbf{I} - \mathbf{A})\mathbf{v} = \mathbf{0} $$

    For non-trivial solutions (\( \mathbf{v} \neq \mathbf{0} \)), the characteristic matrix \( \lambda \mathbf{I} - \mathbf{A} \) must have a determinant of zero. This yields the characteristic equation of the system:

    $$ \det(\lambda \mathbf{I} - \mathbf{A}) = 0 $$

    The \( n \) roots \( \lambda_1, \lambda_2, \dots, \lambda_n \) of this polynomial are the eigenvalues of matrix \( \mathbf{A} \). Each eigenvalue \( \lambda_i \) produces a fundamental characteristic time-domain mode of the form \( e^{\lambda_i t} \). These exponential modes serve as the natural basis functions that compose any transient response of the system.

    Equivalence Between Transfer Function Poles and Eigenvalues

    To see how the state-space eigenvalues directly map to the Laplace transfer function, take the Laplace transform of the state-space equations with zero initial conditions (\( \mathbf{x}(0) = \mathbf{0} \)):

    $$ s\mathbf{X}(s) = \mathbf{A}\mathbf{X}(s) + \mathbf{B}\mathbf{U}(s) \implies (s\mathbf{I} - \mathbf{A})\mathbf{X}(s) = \mathbf{B}\mathbf{U}(s) $$

    Solving for \( \mathbf{X}(s) = (s\mathbf{I} - \mathbf{A})^{-1}\mathbf{B}\mathbf{U}(s) \) and substituting into the output equation \( \mathbf{Y}(s) = \mathbf{C}\mathbf{X}(s) + \mathbf{D}\mathbf{U}(s) \) gives the transfer function matrix:

    $$ \mathbf{H}(s) = \frac{\mathbf{Y}(s)}{\mathbf{U}(s)} = \mathbf{C}(s\mathbf{I} - \mathbf{A})^{-1}\mathbf{B} + \mathbf{D} $$

    Recall from matrix algebra that the inverse of \( s\mathbf{I} - \mathbf{A} \) can be written using its adjugate and determinant:

    $$ (s\mathbf{I} - \mathbf{A})^{-1} = \frac{\text{adj}(s\mathbf{I} - \mathbf{A})}{\det(s\mathbf{I} - \mathbf{A})} $$

    Substituting this relation into the transfer function formula reveals:

    $$ \mathbf{H}(s) = \frac{\mathbf{C}\,\text{adj}(s\mathbf{I} - \mathbf{A})\mathbf{B} + \mathbf{D}\det(s\mathbf{I} - \mathbf{A})}{\det(s\mathbf{I} - \mathbf{A})} = \frac{\mathbf{N}(s)}{D(s)} $$

    Notice that the scalar denominator polynomial \( D(s) \) of the transfer function is identically equal to the characteristic polynomial:

    $$ D(s) = \det(s\mathbf{I} - \mathbf{A}) $$

    Therefore, setting the denominator to zero to find the transfer function poles (\( D(s) = 0 \)) is mathematically identical to solving the eigenvalue characteristic equation (\( \det(\lambda \mathbf{I} - \mathbf{A}) = 0 \)):

    $$ D(s) = 0 \implies s = \lambda_i $$

    Under minimal realizations (where no state modes are cancelled between numerator and denominator), the poles of a transfer function are exactly the eigenvalues of the system matrix.

    Real vs. Complex Conjugate Poles and Natural Responses

    When computing the output response via partial fraction expansion, terms appear in the form \( \frac{r_i}{s - p_i} = \frac{r_i}{s - \lambda_i} \). Inverse Laplace transforming each partial fraction term yields an exponential time function \( r_i e^{\lambda_i t} \). Writing the complex pole/eigenvalue in Cartesian form as \( s = \sigma + j\omega \) allows us to classify the natural responses into two primary families:

    Pole / Eigenvalue Type Complex Coordinate Natural Response Mode Physical Behavior
    Purely Real Pole (\( \lambda = \sigma \)) \( \omega = 0 \), lying on the real axis \( y(t) \propto e^{\sigma t} \) Monotonic exponential decay (\( \sigma < 0 \)) with time constant \( \tau = 1/|\sigma| \), or monotonic exponential divergence (\( \sigma > 0 \)). No oscillation occurs.
    Complex Conjugate Pole Pair (\( \lambda = \sigma \pm j\omega \)) \( \omega \neq 0 \), symmetric across the real axis \( y(t) \propto e^{\sigma t} \cos(\omega t + \phi) \) Sinusoidal oscillation at frequency \( \omega \) enveloped by exponential damping (\( \sigma < 0 \)), exponential growth (\( \sigma > 0 \)), or constant amplitude (\( \sigma = 0 \)).

    The physical contrast between these two behaviors is fundamental to systems engineering:

    • Purely Real Poles (\( \lambda = \sigma \)): Represent overdamped or first-order dynamics (such as a simple resistor-capacitor \( RC \) low-pass filter or thermal dissipation). The response changes smoothly without ringing or overshoot. If \( \sigma < 0 \), the signal decays exponentially toward equilibrium at rate \( |\sigma| \). The closer \( \sigma \) is to the imaginary axis (smaller \( |\sigma| \)), the slower the decay, making it a dominant pole that dictates settling time. If \( \sigma > 0 \), the response diverges monotonically to infinity.
    • Complex Conjugate Pairs (\( \lambda = \sigma \pm j\omega \)): Represent underdamped second-order dynamics where two energy-storage elements exchange energy back and forth (such as mass-spring-damper mechanics or inductor-capacitor \( LC \) tank circuits). Because physical differential equations have real coefficients, complex poles always appear in conjugate pairs. Combining their exponential terms via Euler's identity produces:
      $$ y_{\text{pair}}(t) = 2|r| e^{\sigma t} \cos(\omega t + \angle r) $$
      Here, the real part \( \sigma \) serves as the envelope decay/growth rate, while the imaginary part \( \omega \) determines the oscillation frequency (ringing rate in rad/s).

    Connecting Poles to System Stability

    Because the real part \( \sigma = \text{Re}(\lambda) \) appears directly in the exponent \( e^{\sigma t} \), it strictly dictates whether transients decay to zero (stable, \( \sigma < 0 \)), remain bounded (marginally stable, \( \sigma = 0 \)), or blow up to infinity (unstable, \( \sigma > 0 \)).

    Interactive Transfer Function Lab

    Experiment with transfer functions using the interactive simulator below. Choose different system architectures (1st-order lags, 2nd-order resonant systems, series cascades, and parallel branches), test various input waveforms (\( u(t) \) step, \( \delta(t) \) impulse, pulse, or sinusoid), and observe how transfer function poles, damping ratio \( \zeta \), and block algebra directly shape the real-time output signal \( y(t) \).

    Interactive Transfer Function Lab
    Explore how Laplace transfer functions shape time-domain signals and combine in series & parallel.
    1st-Order Lag Exponential Decay
    Presets:
    Block Diagram Signal Flow: Y(s) = H(s) · U(s) 1st-Order Single Block
    Time-Domain Response y(t)
    H(s) = 1.00 / (0.80s + 1) Hover over plot to inspect
    ━ Output y(t) ╌╌ Input u(t) ⋯⋯ Intermediate y1(t)
    1.00
    0.80 s
    0.40
    4.00 rad/s
    1.00
    0.33 s
    0.25 s
    1.00
    1.00
    0.50 s
    1.00
    1.20 s
    Summing Junction Operation:
    DC Gain: 1.00 Settling ts: 3.20 s Rise tr: 1.76 s Overshoot: 0%
    Poles: s = -1.25
    1st-Order Exponential Lag

    Standard 1st-order transfer function. The time constant τ determines response speed: reaches 63.2% of steady-state at t = τ and 98% at t = 4τ. Commonly models 1-compartment drug clearance and RC lowpass filtering.