With the continuous development of basic theories and high performance devices, distributed optical fiber acoustic sensing technology has gradually transitioned from the qualitative detection stage to the quantitative detection stage and has been widely used in various applications. Demodulation of disturbance signal is indispensable for quantitative detection, so many demodulation methods intended for optical fiber sensing systems have been derived. The PGC (Phase Generation Carrier) technology has been widely used in optical fiber acoustic sensing systems due to its outstanding advantages of simple structure, high sensitivity, wide dynamic range, good linearity, and strong timeliness. According to the demodulation method type, the PGC technology can be mainly divided into the PGC-DCM (Differential and Cross Multiplying) algorithm and the PGC-Arctan algorithm. In terms of demodulation performance, the PGC-DCM algorithm has relatively high requirements on a circuit and is susceptible to interference from hardware facilities, especially system light intensity disturbances. In contrast, the PGC-Arctan algorithm is less sensitive to light intensity disturbances and has relatively low circuit requirements, but it has high requirements for the modulation depth of the system and faces the problem of phase unwinding. To address this limitation of modulation depth, this study proposes a method that introduces the differential self-division operation in the traditional PGC-Arctan algorithm framework to eliminate the influence of the modulation depth. Theoretical analysis shows that by introducing the differential self-division operation, the first-order and second-order Bessel function values of C can be eliminated, thereby eliminating the influence of the modulation depth drift on demodulation. Then, the original vibration signal is obtained through the square root operation and the arctangent operation. Using the simulation platform to analyze and compare the proposed algorithm and other common algorithms from two aspects: different modulation depths and different SNR. Calculate the amplitude error and the total harmonic distortion of the four algorithms with different modulation depths and signal-to-noise ratios respectively. The amplitude error measures the linear distortion of the demodulated signal, and the total harmonic distortion measures the nonlinear distortion of the demodulated signal. The simulation results show that the amplitude error of the proposed algorithm is less than 0.150% and the total harmonic distortion is less than 0.100%. And with the continuous reduction of the SNR, the linear distortion and nonlinear distortion of the demodulated signal do not fluctuate greatly, indicating that the proposed algorithm has a certain anti-noise performance. An experimental platform is built to verify the simulation results and further study the performance of the proposed algorithm. Firstly, the four algorithms are analyzed and compared under different modulation depths. Comparing the minimum value of the amplitude error and the total harmonic distortion, the minimum amplitude error of the PGC-DSVV (differential self-division) algorithm is 0.105%, which is 0.525%, 0.858%, and 2.900% lower than the other three algorithms, respectively. The minimum total harmonic distortion of the PGC-DSVV algorithm is 0.068%, which is 0.101%, 0.662%, and 0.595% lower than the other three algorithms, respectively. After that, the dynamic range of the proposed algorithm is analyzed. The experimental results show that demodulated signal bandwidth and dynamic range of the PGC-DSVV algorithm are larger than other algorithms. The dynamic range of PGC-DSVV is 62.5 dB when the frequency is 200 Hz, which is 6.1 dB, 11.7 dB, and 14.5 dB higher than the other three algorithms, respectively. And the dynamic range is 31.5 dB when the frequency increases to 5 kHz. The comprehensive analysis results verify that the proposed algorithm can effectively eliminate the modulation depth influence, indicating its good stability and noise resistance. Therefore, the proposed algorithm has a good application prospect in high-performance optical fiber acoustic sensing systems.