Abstract: To detect weak partial discharges and evaluate the severity of partial discharge, a novel corona discharge detection method based on acoustic-sensitive diaphragm nanoscale amplitude demodulation for an optical fiber extrinsic Fabry-Perot (EFPI) ultrasonic sensor was proposed. Firstly, an EFPI probe with a resonance frequency of 55.5 kHz was designed based on theoretical analysis. Then, the phase generated carrier arc-tangent algorithm(PGC-Atan) was investigated to demodulate the weak amplitude of the diaphragm. The EFPI ultrasonic sensor was constructed based on the EFPI and PGC-Atan algorithm, and its performances were verified by a single-point sound source and ignition pulse generator. The resolution of the EFPI ultrasonic sensor to the amplitude of the diaphragm can reach 0.43 nm, and it has good linearity of response and ultrasonic signal detection ability. Finally, the EFPI was installed into the experimental device to detect the air corona discharge, and 13.86 pC partial discharge magnitude was achieved. The relationship between the diaphragm amplitude and the corona discharge development was established by Bootstrap small sample analysis. The experimental results indicate that the amplitude of the diaphragm increases exponentially with the discharge voltage, which is similar to the variation trend of the partial discharge magnitude with the voltage. The amplitude of the diaphragm is related to the severity of corona discharge.