Broadband Vibration Signal Measurement Based on Multi-Coset Sampling in Phase-Sensitive OTDR System

The frequency response of phase-sensitive optical time-domain reflectometry (
$\Phi $
-OTDR) is strictly limited by the repetition rate of the injected sampling pulses in conventional sensing scheme. To fully reconstruct an arbitrary vibration signal, the pulse repetition rate is required to be lower bounded by twice the maximum frequency of the vibration signal according to the Nyquist sampling theorem and at the same time upper bounded by the sensing fiber length to avoid generations of mixed Rayleigh backscattered signals from more than one optical pulse within the sensing fiber, making high-frequency vibration sensing over long distance impossible. In this work, we propose a novel method based on the serial multi-coset sampling (MCS) strategy which can recover sparse broadband vibration signals with sub-Nyquist sampling rates, which is thus able to break through the tradeoff between the measurable vibration frequency bandwidth and the sensing range. Pulses with non-uniform intervals are adopted to realize MCS for the Rayleigh backscattered lights along the entire sensing fiber rather than the traditional uniform pulse train. A continuous to finite (CTF) block is used to determine the active spectrum ranges of the sparse signal. By using a serial MCS sequence with an average sampling rate of 4.8 kHz, both a 4 kHz single-frequency sinusoidal signal and a 0.2 kHz bandwidth chirped signal with a center frequency of 4.6 kHz were sampled and recovered successfully in the simulation. In experiments, a single frequency vibration signal of 15 kHz and a chirped vibration signal with frequency ranging from 15 to 16 kHz were successfully identified and reconstructed over a sensing distance of 4 km using an average sampling rate of 21.6kHz, respectively, which however is not possible in conventional
$\Phi $
-OTDR system. Furthermore, accurate identification of a broadband vibration signal with a maximum frequency of 6 kHz over a 10km-long sensing fiber shows that this method is also practically valid in long-distance sensing.