System Classification

System classification is the systematic characterization of a system based on fundamental properties such as memory, causality, stability, linearity, and time invariance. Every signal processing or communication system encountered in engineering can be described using these five properties, and knowing these properties determines which analysis tools are applicable to the system.

Core Concept Explanation

A system is a rule or mapping that takes an input signal x(t) and produces an output signal y(t). Before analyzing any system, it must be classified according to five fundamental properties. The first property is memory. A system is memoryless (or static) if its output at any time t depends only on the input at that same time t. The ideal resistor (V = IR, i.e., v(t) = R*i(t)) is memoryless. A capacitor (V = (1/C) integral of I dt) has memory because the current voltage depends on all past currents. Any system containing integrators, differentiators, delay elements, or stored state has memory.

The second property is causality. A system is causal if y(t) depends only on x(tau) for tau <= t (past and present inputs only). A causal system cannot use future inputs that have not yet arrived. All real-time physical systems must be causal. However, in offline processing (such as image processing, recorded audio, financial time-series retrospective analysis), non-causal filters are permitted because the entire signal is available. A causal LTI system has an impulse response h(t) = 0 for t < 0.

BIBO stability (bounded input, bounded output stability) is the third property. A system is BIBO stable if every bounded input produces a bounded output. An input x(t) is bounded if there exists a finite M such that |x(t)| <= M for all t. For an LTI system, BIBO stability requires that the impulse response h(t) be absolutely integrable: the integral of |h(t)| from -infinity to +infinity must be finite. In the Laplace domain, this corresponds to all poles of H(s) being in the left half plane. A single pole on the imaginary axis or in the right half plane makes the system unstable.

Mathematical Expression

The linearity test uses the superposition principle. A system T is linear if for all inputs x1(t), x2(t) and all scalars a, b: T{a*x1(t) + b*x2(t)} = a*T{x1(t)} + b*T{x2(t)}. This must hold for all possible input combinations. If it fails for any specific input pair or specific constants, the system is nonlinear. Examples of nonlinear systems: y(t) = x^2(t) (squaring), y(t) = |x(t)| (absolute value), y(t) = cos(x(t)) (sinusoidal nonlinearity). Examples of linear systems: y(t) = 3*x(t) (scaling alone is linear), y(t) = dx/dt (differentiation is linear and time-invariant).

The time invariance test checks if a time shift of the input produces an identical time shift of the output, with no other change. If x(t) produces y(t), then a time-invariant system must give y(t - t0) when the input is x(t - t0), for any t0. A system y(t) = t*x(t) is time-variant because the explicit function of t in the coefficient makes the response different at different times. A system y(t) = x(t - 3) is time-invariant: shifting the input by t0 shifts the output by t0 regardless of t0. A system y(t) = x(2t) is time-variant because time scaling is not a simple shift.

Practical Understanding

In the context of control systems and filters, LTI systems are the foundation of design. All analog and digital filter design (Butterworth, Chebyshev, Bessel, FIR, IIR) is built on the LTI framework. Transfer functions H(s) (continuous-time) or H(z) (discrete-time) exist only for LTI systems. The frequency response H(f) = H(s)|_{s=j2*pi*f} describes the gain and phase shift as a function of frequency. Knowing that a system is LTI immediately allows the use of all Fourier, Laplace, and z-transform tools for complete analysis and design.

In communication systems, a channel that adds white Gaussian noise to the transmitted signal without distorting it is modeled as a linear time-invariant channel with transfer function H(f) = 1, i.e., a pass-through. Real channels introduce frequency-selective fading, which means H(f) varies with frequency. Multipath channels are LTI but have frequency-selective behavior. Systems that change with time (e.g., mobile channels where the fading statistics change with vehicle speed) are modeled as linear time-variant (LTV) systems, requiring more advanced analysis than simple LTI theory.

Given:
System 1: y(t) = 3*x(t) + 5
System 2: y(t) = x(t - 2)
Classify each for linearity and time invariance.

Why this formula applies:
Linearity test: T{a*x1 + b*x2} must equal a*T{x1} + b*T{x2}.
Time invariance test: T{x(t-t0)} must equal y(t-t0).

System 1: y(t) = 3*x(t) + 5
Linearity test:
T{a*x1 + b*x2} = 3*(a*x1 + b*x2) + 5 = 3a*x1 + 3b*x2 + 5
a*T{x1} + b*T{x2} = a*(3x1 + 5) + b*(3x2 + 5) = 3a*x1 + 3b*x2 + 5(a+b)
These are equal only if 5 = 5*(a+b), i.e., a+b=1. Not true in general.
Conclusion: System 1 is NOT linear (constant offset 5 violates superposition).

Time invariance test for System 1:
Input x(t-t0): y = 3*x(t-t0) + 5
y(t-t0) = 3*x(t-t0) + 5
Both equal: System 1 IS time-invariant.

System 2: y(t) = x(t-2)  [pure delay]
Linearity: T{a*x1+b*x2} = a*x1(t-2)+b*x2(t-2) = a*T{x1}+b*T{x2}. Linear.
Time invariance: input x(t-t0) -> y = x(t-t0-2) = y(t-t0). Time-invariant.
Conclusion: System 2 is LINEAR and TIME-INVARIANT. It is an LTI system.

Final Answer:
System 1: NOT linear, time-invariant only.
System 2: Linear and time-invariant (LTI). Impulse response: h(t) = delta(t-2).

Exam Tip: In GATE, any system with a non-zero constant term (y(t) = a*x(t) + C, C not equal to 0) is NOT linear because it fails the zero-input zero-output condition: when x(t) = 0, y(t) = C, not 0. Linearity strictly requires y = 0 when x = 0. This is the fastest way to eliminate linearity for such systems without going through the full superposition test.

Quick Revision

• Memory: Memoryless if y(t) depends only on x(t). Memory if it depends on past (or future) inputs. Capacitors and inductors have memory.
• Causality: Causal if y(t) depends only on x(tau) for tau <= t. Non-causal uses future inputs. All hardware real-time systems are causal.
• BIBO Stability: Stable if bounded input always gives bounded output. LTI stable iff integral of |h(t)| dt is finite. Poles in left half plane = stable.
• Linearity: T{a*x1 + b*x2} = a*T{x1} + b*T{x2} must hold for all inputs and all constants. Non-zero DC offset violates linearity.
• Time Invariance: If x(t) -> y(t), then x(t-t0) -> y(t-t0) for any t0. Explicit time-multiplied coefficients (e.g., t*x(t)) make systems time-variant.
• LTI: Linear + Time-Invariant. Fully characterized by impulse response h(t). y(t) = x(t)*h(t). Transfer function H(s) or H(f) applies.
• GATE trap: y(t) = x(t) + 5 is NOT linear (non-zero at zero input). y(t) = 2*x(t) IS linear and time-invariant. Always check for additive constants and time-varying coefficients first.