A method based on equal frequency resampling is proposed to suppress laser nonlinear frequency sweeping for the ultimate spatial resolution in optical frequency domain reflectometry. Estimation inaccuracy of the sweeping frequency distribution caused by the finite sampling rate in the auxiliary interferometer can be efficiently compensated by the equal frequency resampling method. With the sweeping range of 130 nm, a 12.1 µm spatial resolution is experimentally obtained. In addition, the sampling limitation of the auxiliary interferometer-based correction is discussed. With a 200 m optical path delay in the auxiliary interferometer, a 21.3 µm spatial resolution is realised at the 191 m fibre end. By employing the proposed resampling and a drawing tower FBG array to enhance the Rayleigh backscattering, a distributed temperature sensing over a 105 m fibre with a sensing resolution of 1 cm is achieved. The measured temperature uncertainty is limited to ±0.15 °C.

Phase-sensitive optical frequency domain reflectometer (Φ-OFDR) is investigated to deliver an accurate distributed measurement with high spatial resolution. It is found that random phase noise and quadrant discrimination during phase calculation are the main reasons for the random hopping in Φ-OFDR. By efficiently eliminating random hopping in the phase unwrap and suppressing the laser-induced nonlinear sweep for the theoretical spatial resolution, the proposed Φ-OFDR is proved to be able to decouple the limitation between resolution and accuracy in coherent OFDR (C-OFDR). Distributed strain measurement with 20 mm spatial resolution in Φ-OFDR is obtained and analysed. Measurement with little deviation and uniform sensitivity between applied strain and phase change both validate the efficient noise suppression for extreme resolution measurement. Then the influence of the initial sweep frequency between two times measurements is studied. With a further reduced 800 µm spatial resolution, the proposed Φ-OFDR is able to retain accurate distributed measurement compared to conventional C-OFDR methods. Besides, the computation time of the Φ-OFDR is only 3.2% of the C-OFDR.