Linear Periodically Time-Varying (LPTV) System

The defining characteristic of an LPTV system is its impulse response h(t, τ), which is periodic in the time variable t with period T₀. This means h(t + T₀, τ) = h(t, τ) for all t and τ. When a stationary input passes through this time-varying channel, the output exhibits cyclostationary statistics with cyclic frequencies at integer multiples of 1/T₀. This periodicity arises naturally in communication systems from symbol-rate clocking, frame structures, and switched analog components.

In the frequency domain, an LPTV system is characterized by its bifrequency transfer function H(f, ν), which maps an input at frequency f to an output at frequency f - ν. The system generates spectral correlation between frequency components separated by the cyclic frequency α = k/T₀. This frequency-shift filtering behavior is the mechanism that creates the distinctive spectral lines observed in the Spectral Correlation Function (SCF) of communication signals after passing through time-varying channels.

The periodic time variation of an LPTV system can be decomposed into a Fourier series with coefficients g_k(τ). Each coefficient represents a frequency-shift channel that shifts the input spectrum by k/T₀ before applying a linear time-invariant filter. This decomposition reveals that an LPTV system is equivalent to a parallel bank of LTI filters with frequency shifters, providing the analytical foundation for FRESH filtering and cyclostationary signal processing techniques used in interference suppression and blind equalization.

When a wide-sense stationary (WSS) input x(t) with power spectral density S_x(f) passes through an LPTV system, the output y(t) becomes wide-sense cyclostationary (WSCS). The output cyclic power spectrum S_y^α(f) is directly determined by the system's bifrequency transfer function:

• Spectral correlation appears at cyclic frequencies α = k/T₀
• The degree of correlation depends on the magnitude of the Fourier coefficients g_k(τ)
• This mechanism is exploited in cyclostationary signature embedding for device authentication

Most digital communication transmitters inherently behave as LPTV systems due to periodic operations:

• Pulse shaping at the symbol rate introduces periodicity
• OFDM cyclic prefix insertion creates a time-varying correlation structure
• Switched power amplifiers in burst-mode transmission generate transient periodicity
• DAC clock jitter and I/Q modulator imbalance produce hardware-specific LPTV signatures These imperfections form the basis for RF fingerprinting using cyclostationary feature extraction.

The LPTV system model directly enables the design of cyclic Wiener filters that optimally estimate signals in cyclostationary noise. By exploiting the periodic structure of the system, these filters achieve signal separation even when signals overlap in both time and frequency domains. The filter design requires knowledge of the cyclic power spectra of both the desired signal and interference, which are derived from the LPTV system parameters and input statistics.