In order to improve the phase calculation speed and detection efficiency of Pulse Phase Thermography(PPT), a fast phase algorithm optimized from Fast Fourier Transform(FFT) was proposed, with which the calculation speed could be increased by 1.9-12.4 times. To solve the optimal sampling length of thermal image sequence and frequency ratio, the calculation method for optimal sampling length was proposed based on the heat diffusion formula. Experiment was carried out on aluminium and iron specimens. Results show that the optimal sampling length is 1.1 s and 3.9 s for aluminium and iron specimens respectively and the phase difference at fundamental frequency has the best capability to distinguish defects. The calculation speed is evidently improved by the proposed fast algorithm and the optimal sampling length can be directly calculated by the improved numerical formula. Operation subjectivity and random settings of parameters can be avoided and the detection efficiency of pulse phase thermography can be improved.