Mathematical Modeling of Dynamic Systems

Introduction

A control system is designed, analyzed, and implemented through a mathematical model of the physical process being controlled. Mathematical modeling converts physical laws, such as Kirchhoff’s laws, Newton’s laws, conservation of energy, and conservation of mass, into equations that describe how system variables evolve with time. The resulting model may be represented as a differential equation, transfer function, state-space model, or a combination of these forms.

Modeling is not merely a mathematical exercise. A model determines which dynamic effects are included, which measurements are required, what controller structures are feasible, and how accurately simulation results represent the real plant. In engineering practice, the model is always an approximation. The objective is therefore not to reproduce every physical detail, but to retain the dynamics relevant to the intended analysis and control task.

This chapter develops models for electrical, mechanical, electromechanical, thermal, and fluid systems. It also introduces linearization, Laplace-domain modeling, model reduction, and parameter identification.

Learning Objectives

After completing this chapter, the reader should be able to:

1. Formulate differential-equation models from physical laws.

2. Derive transfer functions under stated initial-condition and linearity assumptions.

3. Construct state-space representations and distinguish states from general variables.

4. Model electrical, mechanical, electromechanical, thermal, and fluid systems.

5. Linearize nonlinear dynamic equations about an operating point.

6. Apply model reduction while recognizing its limitations.

7. Estimate model parameters from experimental data.

8. Evaluate modeling assumptions and select an appropriate model for control design.

1. Dynamic-System Modeling: Concepts and Terminology

A dynamic system possesses memory: its present behavior depends not only on current input but also on past inputs. Energy storage is the physical origin of this memory. Capacitors and inductors store electrical energy; masses and springs store mechanical energy; thermal masses store internal energy; and tanks store fluid mass.

A system variable is a quantity used to describe system behavior. An input or excitation is imposed from outside the system, whereas an output is the variable of interest. A disturbance is an input that affects the output but is not normally manipulated. A parameter is a quantity assumed constant during a particular analysis, such as resistance, mass, capacitance, or heat-transfer coefficient.

The order of a differential-equation model is the highest derivative of the selected output, after simplification. Physically, it often corresponds to the number of independent energy-storage elements, although constraints and ideal couplings can reduce the order.

A model is linear if it satisfies superposition: the response to a sum of inputs is the sum of the individual responses, and scaling an input scales the response by the same factor. Otherwise, it is nonlinear. A model is time invariant if its parameters do not explicitly change with time. A linear time-invariant system is commonly abbreviated as an LTI system.

A lumped-parameter model represents physical quantities as concentrated at discrete elements, producing ordinary differential equations. A distributed-parameter model accounts for spatial variation and generally produces partial differential equations. Control-system analysis commonly begins with lumped models, but distributed effects may become important in flexible structures, transmission lines, and fluid pipelines.

2. General Modeling Procedure

A reliable modeling workflow is:

1. Define the system boundary, input, output, and disturbances.

2. Select reference directions and sign conventions.

3. Identify energy-storage, dissipative, and conversion elements.

4. Apply governing physical laws.

5. Eliminate unwanted variables or retain them as states.

6. Check dimensions, signs, limiting cases, and initial conditions.

7. State assumptions explicitly.

8. Validate the model against measurements or trusted physical behavior.

The choice of system boundary matters. For example, a motor model may include only the armature circuit and rotor, or may also include gear friction, load elasticity, shaft flexibility, and amplifier dynamics. A more detailed model is not automatically better: unnecessary states complicate identification, simulation, and controller design.

3. Differential-Equation Models

Differential equations express the time-domain relationship between input and output. For a general single-input, single-output linear time-invariant system,

an y^(n)(t) + a(n-1)y^(n-1)(t) + ... + a1 y'(t) + a0 y(t)

= bm u^(m)(t) + b(m-1)u^(m-1)(t) + ... + b1 u'(t) + b0 u(t),

where u(t) is the input, y(t) is the output, n is the system order, and the coefficients are constant parameters. Usually n is greater than or equal to m for a proper physical model. The coefficients must have units that make every term dimensionally consistent.

For example, a mass–spring–damper system with applied force F(t) and displacement x(t) obeys

M x''(t) + B x'(t) + K x(t) = F(t),

where M is mass in kilograms, B is viscous damping in N·s/m, and K is stiffness in N/m. Each term has units of newtons.

Differential equations are especially useful when deriving models directly from physical laws and when analyzing nonlinear behavior. However, controller design often benefits from transfer-function or state-space forms.

4. Transfer-Function Models

For an LTI system with zero initial conditions, the Laplace transform converts differentiation into multiplication by s:

L{x'(t)} = sX(s),

provided the initial condition is zero. The transfer function from input U(s) to output Y(s) is

G(s) = Y(s)/U(s),

with all initial conditions set to zero.

For the mass–spring–damper system,

(Ms^2 + Bs + K)X(s) = F(s),

so

G(s) = X(s)/F(s) = 1/(Ms^2 + Bs + K).

The transfer function describes forced response under zero initial conditions; it does not by itself represent independent initial-energy effects. The roots of the denominator are poles, while roots of the numerator are zeros. Poles strongly influence natural modes, stability, and transient response.

A common normalized second-order form is

G(s) = Kp omega_n^2 / [s^2 + 2 zeta omega_n s + omega_n^2],

where Kp is the static gain, omega_n is the undamped natural frequency in rad/s, and zeta is the damping ratio. This form applies when the denominator is second order and has positive natural-frequency parameters.

Transfer functions are convenient for cascade systems and frequency-domain analysis. Their limitations include difficulty representing nonlinearities, initial conditions, multiple inputs, and internal variables. They also conceal the physical meaning of states and can suffer from numerical cancellation when pole-zero cancellation is used carelessly.

5. State-Space Representation

A state is a minimum internal variable set that, together with future inputs, determines future system behavior. For a continuous-time LTI system,

x_dot(t) = A x(t) + B u(t),

y(t) = C x(t) + D u(t),

where x is the n-dimensional state vector, u is the input vector, y is the output vector, and A, B, C, and D are system matrices of compatible dimensions.

The matrix A governs internal dynamics, B describes input coupling, C selects or combines states to form outputs, and D represents direct feedthrough from input to output. State-space models naturally support multi-input, multi-output systems and nonzero initial conditions.

For the mass–spring–damper system, choose

x1 = x, x2 = x_dot.

Then

x1_dot = x2,

x2_dot = -(K/M)x1 -(B/M)x2 +(1/M)F,

y = x1.

Thus,

A = [[0, 1],[-K/M, -B/M]],

B = [[0],[1/M]],

C = [1, 0],

D = 0.

A system may have many valid state-space realizations. Different state coordinates can be related through a nonsingular transformation x = Tz. The transfer function remains unchanged, but the matrices change. Physically meaningful states are often preferred for interpretation and sensor placement; canonical forms may be preferred for analysis.

For an LTI state model, the solution is

x(t) = exp(At)x(0) + integral from 0 to t of exp[A(t-tau)]B u(tau) d(tau).

The first term is the zero-input response and the second is the zero-state response. The matrix exponential therefore provides a compact description of both natural and forced behavior.

6. Models of Important Physical Systems

6.1 Electrical Systems

Kirchhoff’s voltage law states that the algebraic sum of voltages around a loop is zero, while Kirchhoff’s current law states that the algebraic sum of currents at a node is zero. Element relations are

v_R = Ri,

v_L = L di/dt,

i_C = C dv_C/dt.

For a series RLC circuit driven by voltage v_i(t), with capacitor voltage v_C(t) as output, KVL gives

v_i(t) = Ri(t) + L di/dt + v_C(t),

i(t) = C dv_C/dt.

Substitution yields

LC v_C''(t) + RC v_C'(t) + v_C(t) = v_i(t).

Under zero initial conditions,

V_C(s)/V_i(s) = 1/(LCs^2 + RCs + 1).

The inductor current and capacitor voltage are natural state variables because they represent stored magnetic and electric energy.

6.2 Translational Mechanical Systems

Newton’s second law gives the sum of forces equal to mass times acceleration. For a mass attached to a viscous damper and spring,

M x'' + B x' + Kx = F.

For multiple masses, each mass requires a force balance. Coupling springs and dampers create additional states. Coulomb friction, backlash, saturation, and dead zones introduce nonlinearities and should not be represented by viscous damping unless that approximation is justified.

6.3 Rotational Mechanical Systems

The rotational counterpart of Newton’s law is

J theta'' + B theta' + T_L = T,

where J is moment of inertia, B is rotational viscous-friction coefficient, theta is angular position, T is applied torque, and T_L is opposing load torque. If a torsional spring of stiffness K_t is present, the restoring torque K_t theta is included.

Rotational and translational variables can be coupled through gears. For an ideal gear pair, torque and angular velocity transformations depend on the gear ratio, while power conservation requires that torque and speed scale inversely, neglecting losses.

6.4 DC Motor

A separately excited armature-controlled DC motor can be modeled by

v_a = L_a di_a/dt + R_a i_a + K_e omega,

J d omega/dt = K_t i_a - B omega - T_L,

theta_dot = omega.

Here v_a is armature voltage, i_a is armature current, omega is angular speed, theta is shaft position, L_a and R_a are armature inductance and resistance, K_e is back-emf constant, K_t is torque constant, J is total inertia, and B is viscous friction. In SI units, K_e and K_t have numerically equal values when compatible electrical and mechanical units are used.

Taking T_L = 0 and zero initial conditions, the speed transfer function is

Omega(s)/V_a(s) = K_t/[ (L_a s + R_a)(J s + B) + K_e K_t ].

The position transfer function is

Theta(s)/V_a(s) = K_t/[s{(L_a s + R_a)(J s + B) + K_e K_t}].

This model demonstrates electromechanical coupling: electrical current produces torque, while rotor speed produces back emf.

6.5 Thermal Systems

A lumped thermal model follows conservation of energy. For a body with thermal capacitance C_th, temperature T, ambient temperature T_a, heat input q, and thermal resistance R_th to the environment,

C_th dT/dt = q - (T - T_a)/R_th.

If temperature is measured relative to ambient, theta = T - T_a, then

C_th d theta/dt + theta/R_th = q.

The transfer function is

Theta(s)/Q(s) = R_th/(R_th C_th s + 1).

The model assumes a uniform body temperature, constant parameters, and linear heat transfer. Radiation, whose heat flow is proportional to absolute-temperature fourth powers, may require nonlinear modeling. Thermal systems are often slow, and their sensor dynamics may be comparable to the plant dynamics.

6.6 Fluid Systems

For a liquid tank with cross-sectional area A, liquid height h, inflow q_i, and outflow q_o,

A dh/dt = q_i - q_o.

If the outflow is linearized as q_o = h/R_h, where R_h is hydraulic resistance, then

A dh/dt + h/R_h = q_i,

and

H(s)/Q_i(s) = R_h/(A R_h s + 1).

For an orifice, the physical relation is commonly nonlinear, for example q_o = c sqrt(h) for positive h, with coefficient c depending on geometry and gravity. Linearization about a positive operating height h_0 gives

dq_o/dh at h_0 = c/(2 sqrt(h_0)).

The incremental model then uses delta q_o approximately equal to [c/(2 sqrt(h_0))] delta h. The operating point must be specified because the linearized resistance changes with height.

7. Linearization of Nonlinear Models

Many physical systems are nonlinear. Let

x_dot = f(x,u),

y = g(x,u).

Suppose the system operates at (x_0,u_0), satisfying f(x_0,u_0)=0 for an equilibrium. Define perturbations delta x = x - x_0 and delta u = u - u_0. A first-order Taylor expansion gives

delta x_dot = A delta x + B delta u,

delta y = C delta x + D delta u,

where

A = partial f/partial x evaluated at (x_0,u_0),

B = partial f/partial u evaluated at (x_0,u_0),

C = partial g/partial x evaluated at (x_0,u_0),

D = partial g/partial u evaluated at (x_0,u_0).

Terms of second and higher order are neglected. The approximation is reliable only for sufficiently small deviations and away from discontinuities such as ideal switching, backlash, and saturation boundaries.

For a tank with q_o = c sqrt(h), the equilibrium condition is q_i0 = c sqrt(h_0). Since dq_o/dh = c/(2 sqrt(h_0)), the linearized dynamics are

A delta h_dot = delta q_i - [c/(2 sqrt(h_0))] delta h.

Consequently, the local time constant is

tau = 2A sqrt(h_0)/c.

The same nonlinear tank can have different local time constants at different operating levels; this is a fundamental limitation of a single fixed linear model.

8. Model Reduction

High-order models may contain fast modes, weakly coupled modes, parasitic dynamics, or sensor and actuator details unnecessary for a specified bandwidth. Model reduction seeks a lower-order model that preserves relevant input-output behavior.

A basic engineering method is dominant-pole approximation. If a pole is much faster than the closed-loop bandwidth, its factor may be approximated by a constant over the frequency range of interest. For example,

G(s) = K/[(tau_f s + 1)(tau_s s + 1)],

with tau_f much less than tau_s. For frequencies near the slow response, tau_f s is small and the fast factor is approximately one, giving G_red(s) approximately K/(tau_s s + 1).

Other methods include balanced truncation, singular perturbation, modal truncation, and frequency-response fitting. Balanced truncation uses controllability and observability information to identify states that have little influence on input-output behavior. It provides useful error bounds for stable systems, but requires computational effort and careful treatment of unstable or lightly damped modes.

Reduction must preserve the properties important to the application. Removing a fast actuator pole may be acceptable for low-bandwidth control but unsafe for high-gain control, where neglected phase lag can cause instability. A reduced model should therefore be validated over the intended operating frequency and amplitude range.

9. Parameter Identification

Parameter identification estimates unknown model parameters from measured inputs and outputs. It is essential when datasheets are incomplete, parameters vary with operating conditions, or physical construction is complex.

A typical procedure is:

1. Select a model structure from physical knowledge.

2. Design an informative experiment.

3. Measure input, output, and, when possible, internal variables.

4. Preprocess data while preserving relevant dynamics.

5. Estimate parameters using an objective function.

6. Validate the model using independent data.

For a linear regression model y = Phi theta + e, the ordinary least-squares estimate is

theta_hat = (Phi^T Phi)^(-1) Phi^T y,

provided Phi has full column rank. Here theta contains unknown parameters, Phi is the regression matrix, and e is measurement error. In practice, QR or singular-value decomposition is preferred to explicitly forming an inverse because it is numerically more reliable.

For dynamic systems, differentiation of noisy measurements can be problematic. Integral formulations, filtered regressions, prediction-error methods, and subspace identification are often preferable. Experiments should excite the dynamics of interest without violating safety constraints. Validation data must not be the same data used for fitting, otherwise apparent accuracy may result from overfitting.

Identifiability is important: a parameter is identifiable only if the available measurements and excitation contain enough information to distinguish its effect from other parameters. For example, in a motor model, resistance and torque constant may be difficult to estimate independently if only steady-state speed is measured under limited operating conditions.

10. Worked Example: Modeling a DC Motor

Consider a motor with

R_a = 2 ohm, L_a = 0.5 H, K_e = K_t = 0.1, J = 0.02 kg·m^2, B = 0.01 N·m·s/rad.

The speed denominator is

(L_a s + R_a)(J s + B) + K_eK_t.

First,

(L_a s + R_a)(J s + B)

= (0.5s+2)(0.02s+0.01)

= 0.01s^2 + 0.045s + 0.02.

Adding K_eK_t = 0.01 gives

0.01s^2 + 0.045s + 0.03.

Therefore,

Omega(s)/V_a(s) = 0.1/(0.01s^2 + 0.045s + 0.03)

= 10/(s^2 + 4.5s + 3).

The position model is

Theta(s)/V_a(s) = 10/[s(s^2 + 4.5s + 3)].

The speed model has poles obtained from s^2 + 4.5s + 3 = 0. They are approximately -0.814 and -3.686 rad/s, so the isolated motor model is stable. The slower pole dominates long-time speed behavior, but the faster electrical-mechanical mode may matter in a high-bandwidth controller.

At steady state for a constant voltage and zero load torque, the speed gain is G(0) = 10/3 approximately 3.33 rad/s per volt. This value is consistent with the steady-state equations: at zero acceleration, i_a = B omega/K_t and v_a = R_a i_a + K_e omega, giving omega/v_a = 1/[K_e + R_aB/K_t] = 1/(0.1+0.2) = 3.33.

11. Applied Case: Temperature Control of an Industrial Process

Background

Consider an illustrative, fictional case involving a small industrial chamber heated by an electrical element. The objective is to regulate chamber temperature using a feedback controller.

Problem

A detailed finite-element thermal simulation contains many spatial states, but the controller operates at a low sampling rate and only chamber temperature is measured. Engineers require a model suitable for controller tuning and fault monitoring.

Analysis

Initial step-test data show that a change in heater power produces a monotonic temperature response with no evident oscillation. The chamber is therefore initially approximated by a first-order model,

T(s)/Q(s) = K/(tau s + 1),

where K is the steady-state temperature rise per unit heater power and tau is the dominant thermal time constant. Ambient-temperature variation is treated as a disturbance.

A step test is performed within safe temperature limits. The steady-state temperature rise is 40 degrees C for a 200 W power change, giving K = 0.20 degrees C/W. The measured response reaches approximately 63.2% of its final rise after 300 s, so tau is estimated as 300 s, because a first-order step response reaches 1 - e^(-1) of its final value at t = tau.

The identified model is therefore

T(s)/Q(s) = 0.20/(300s + 1) degrees C/W.

Approach and Solution

The model is validated using a second step of different amplitude and a ramp-like disturbance. If the first-order model reproduces the dominant response but misses a short initial delay, a sensor or actuator lag may be added. A controller bandwidth substantially faster than the thermal time constant would be inappropriate if the neglected sensor and heater dynamics introduce significant phase lag.

Outcome

The reduced model is adequate for low-bandwidth temperature regulation and enables rapid simulation. It is not suitable for predicting local heater hotspots, thermal gradients, or behavior near nonlinear radiation-dominated temperatures.

Learning Points

The case illustrates that model selection depends on purpose, data, and operating range. A simple model can be valuable when its assumptions are explicit and its validity is tested. Safety limits, sensor calibration, actuator saturation, and independent validation are part of the engineering modeling process, not optional additions.

12. Challenges, Limitations, and Best Practices

Models are affected by uncertain parameters, unmeasured disturbances, sensor noise, actuator saturation, friction, hysteresis, aging, temperature dependence, and neglected delays. A model fitted at one operating point may fail elsewhere. Structural errors cannot be eliminated by estimating parameters more accurately.

Best practices include:

• Define the intended use before selecting model complexity.
• Preserve physical units and document sign conventions.
• Separate calibration, validation, and test data.
• Include actuator, sensor, delay, and sampling effects when relevant.
• Check conservation laws and limiting cases.
• Compare model predictions with independent measurements.
• Quantify uncertainty rather than reporting unjustified precision.
• Retain nonlinearities that affect safety, saturation, or stability.
• Use physically plausible parameter constraints during identification.
• Record model version, assumptions, data sources, and validation conditions.

In safety-critical systems, model inadequacy must be addressed through verification, validation, fault analysis, and conservative operating margins. AI-assisted identification can help discover model structures or estimate parameters, but it does not replace physical validation. Data-driven models may extrapolate poorly, encode sensor bias, or obscure failure mechanisms. Engineering accountability requires traceable data, interpretable assumptions where possible, and testing under adverse conditions.

Chapter Summary

Mathematical modeling provides the foundation for control-system analysis and design. Differential equations are derived from physical laws and reveal the causal structure of a system. Laplace transforms produce transfer functions for LTI systems under zero initial conditions, while state-space models represent internal dynamics, initial conditions, and multi-variable systems naturally.

Electrical circuits are modeled using Kirchhoff’s laws; mechanical systems using force and torque balances; electromechanical systems by coupling electrical and mechanical equations; thermal systems by energy conservation; and fluid systems by mass conservation and constitutive flow relations. Nonlinear models can be approximated locally by Jacobian-based linearization around an equilibrium. High-order models may be reduced when neglected dynamics are demonstrably irrelevant to the intended bandwidth and operating range. Parameter identification combines experiments, regression or prediction-error methods, and independent validation.

A professionally useful model is not necessarily the most detailed model. It is a model whose assumptions, applicability, uncertainty, and limitations are understood and whose predictions have been tested against physical evidence.

Key Terms

Dynamic system: A system whose behavior depends on time and stored energy.

State: A minimum variable set that determines future behavior with the input known.

Input: An externally applied variable that excites a system.

Output: A system variable selected for observation or control.

Parameter: A quantity describing physical properties, usually treated as constant in a model.

Linear system: A system satisfying scaling and superposition.

Time-invariant system: A system whose parameters do not explicitly vary with time.

Differential-equation model: A time-domain relation involving variables and their derivatives.

Transfer function: The zero-initial-condition ratio of output to input in the Laplace domain.

Pole: A root of a transfer-function denominator.

Zero: A root of a transfer-function numerator.

State-space model: A first-order vector differential-equation representation with output equations.

Linearization: Approximation of a nonlinear model by its first-order Taylor expansion near an operating point.

Operating point: A specified equilibrium or nominal condition.

Lumped-parameter model: A model using discrete elements and ordinary differential equations.

Model reduction: Construction of a lower-order model that preserves important behavior.

Parameter identification: Estimation of unknown model parameters from data.

Identifiability: The ability to determine parameters uniquely from available measurements and excitation.

Validation: Comparison of model predictions with independent physical or experimental evidence.

Review Questions

Short Answer Questions

1. What is meant by a state variable, and why are capacitor voltage and inductor current natural states?

2. State the assumptions required when defining a transfer function as Y(s)/U(s).

3. Distinguish between a lumped-parameter and a distributed-parameter model.

4. Why does a nonlinear tank model have an operating-point-dependent linear approximation?

5. What is the purpose of independent validation data in parameter identification?

Long Answer Questions

1. Derive the transfer function of a series RLC circuit from input voltage to capacitor voltage and identify suitable state variables.

2. Derive the state-space model of a mass–spring–damper system and explain the physical meaning of each matrix.

3. Develop the voltage-to-speed and voltage-to-position models of an armature-controlled DC motor.

4. Explain dominant-pole model reduction and discuss when it can produce an unsafe control design.

5. Compare transfer-function and state-space representations for multi-input, multi-output systems with nonzero initial conditions.

Analytical / Application Questions

1. A thermal plant is modeled by C_th dT/dt = q - (T-T_a)/R_th. Derive its transfer function from heater power to temperature relative to ambient and determine its time constant.

2. A tank has q_o = c sqrt(h). Derive the small-signal model about h_0 and explain how the local time constant changes as h_0 changes.

3. Given a measured high-order plant with one fast pole and two slow poles, propose a reduction strategy, specify the validation tests required, and discuss the risks of neglecting the fast pole.

Multiple Choice Questions with Answers

Question 1

Which condition is required for the standard transfer function G(s)=Y(s)/U(s)?

A. The system must be nonlinear

B. All initial conditions must be zero

C. The output must be a state variable

D. The system must have no energy storage

Correct Answer: B. All initial conditions must be zero

Question 2

For a mass–spring–damper system, which equation is correct?

A. M x + B x' + K x'' = F

B. M x'' + B x' + Kx = F

C. M x' + B x'' + Kx = F

D. M x'' + Bx + Kx' = F

Correct Answer: B. M x'' + B x' + Kx = F

Question 3

In a state-space model, which matrix directly maps the input to the state derivative?

A. A

B. B

C. C

D. D

Correct Answer: B. B

Question 4

The natural state variables of an ideal capacitor and inductor are, respectively:

A. Current and voltage

B. Voltage and current

C. Resistance and current

D. Voltage and resistance

Correct Answer: B. Voltage and current

Question 5

For the thermal model C_th dθ/dt + θ/R_th = q, the time constant is:

A. R_th/C_th

B. C_th/R_th

C. R_th C_th

D. 1/(R_th C_th)

Correct Answer: C. R_th C_th

Question 6

Linearization of x_dot=f(x,u) about an operating point uses which quantities?

A. Second derivatives only

B. The Jacobian matrices of f

C. Only the equilibrium input

D. The transfer-function zeros only

Correct Answer: B. The Jacobian matrices of f

Question 7

For a first-order step response, the output reaches approximately 63.2% of its final change at:

A. One time constant

B. Two time constants

C. Half a time constant

D. Ten time constants

Correct Answer: A. One time constant

Question 8

In a DC motor, back emf is proportional to:

A. Armature resistance

B. Rotor angular speed

C. Load inertia only

D. Shaft position only

Correct Answer: B. Rotor angular speed

Question 9

Which condition is necessary for the ordinary least-squares estimate (Phi^T Phi)^(-1)Phi^T y to be unique?

A. Phi has full column rank

B. The output must be constant

C. Measurement noise must be zero

D. The model must be nonlinear

Correct Answer: A. Phi has full column rank

Question 10

A reduced model is most defensible when the neglected dynamics are:

A. Within the required control bandwidth and strongly unstable

B. Fast relative to the intended bandwidth and validated as insignificant

C. Unknown and untested

D. Responsible for the safety limit

Correct Answer: B. Fast relative to the intended bandwidth and validated as insignificant