Modeling and Mitigation of Thermal Drift in Reflective Fiber-Optic Current Sensors With a Spun Quarter-Wave Plate via Eight-State Square-Wave Modulation

In reflective fiber-optic current sensors (FOCS), the polarization transfer of a quarter-wave plate (QWP) plays a decisive role in the interferometric visibility. Its temperature sensitivity can cause visibility to drift, which can lead to demodulation errors in current measurement. In this work, we derive a round-trip Jones-matrix model for a spun QWP fabricated from gradient-spun polarization-maintaining (PM) fiber. We then perform temperature and wavelength sweeps to quantify the resulting parameter drift. The results indicate that such drift can markedly change the return state of polarization, and consequently lead to visibility variations under external perturbations. Because conventional two-state and four-state square-wave schemes do not readily provide online observability of visibility variations, we propose an eight-state square-wave modulation and demodulation method. The proposed method is compatible with closed-loop current-phase feedback while enabling simultaneous estimation and compensation of half-wave voltage drift and interferometric visibility changes. Controlled heating experiments were conducted on the spun QWP from 20 $^\circ$C to 80 $^\circ$C in 10 $^\circ$C increments, with a 10 minutes stabilization at each temperature point. The experiments were repeated under two SLD drive currents to verify robustness against optical power changes. After compensation, the full-scale normalized coefficient of variation (FS-CV) decreases from 6.37% to 0.67% (@165 mA SLD drive current), corresponding to an improvement of approximately one order of magnitude. Moreover, the temperature dependence of the projected return power measured by a polarimeter closely tracks the uncompensated current drift. This agreement indicates that the dominant thermal error originates from spun QWP induced polarization changes that modulate the interferometric visibility. Finally, simulations incorporating a 30 dB extinction ratio reproduce the observed nonlinearity. The simulated response is well described by a sinusoidal fit, which provides practical guidance for component selection and system optimization. Taken together, this work mitigates demodulation errors caused by the thermally sensitive spun QWP through coordinated modeling, algorithm design, and experimental validation. The proposed approach offers a reproducible route toward wide-temperature, high-stability fiber-optic current sensors.