Photoacoustic spectrum ( PAS) gas detection technology based on wavelength modulated spectrum signal second harmonic component detection has been widely used in electric power, chemical and medical industry. Although the second harmonic component detection technology can effectively reduce the related noise in the photoacoustic signal, there is still non-correlation noise in the second harmonic signal, which affects the detection ability of the detection system to trace gas and the accuracy of the test results. In order to study how to weaken the influence of the non-correlated noise of the photoacoustic spectrum detection system on the second harmonic signal, improve the detection system of the limit of detection, improve the measurement accuracy, this paper built a set of with tunable distributed feedback (DFB) semiconductor laser is first longitudinal resonant photoacoustic spectrum of excitation light source gas detection system, first put forward using fast Fourier transform (FFT) to light the second harmonic signal filtering of the new method. Firstly, the sawtooth scanning signal frequency is optimized according to the noise spectrum of the gas photoacoustic spectrum detection system, so as to minimize the influence of background noise of the detection system on the optical acoustic detection signal. Then, the FFT filter is used to extract the fundamental wave component of the optical acoustic second harmonic signal of the same frequency as the scanning sawtooth wave. Although the amplitude of the extracted fundamental wave component is smaller than the maximum of the second harmonic component of the photoacoustic signal, the background noise of the photoacoustic detection system decreases more, so the purpose of improving the minimum detection limit of the photoacoustic detection system is realized. By analyzing and comparing the test results of C2H2/N-2 mixed gas with the concentrations of 102, 75. 1, 50, 30. 3, 15. 3, 7. 7 1, 0. 79 , 0. 57, 0. 35 and 0. 17 mu L . L-1, the test results are very stable after FFT is used to filter the photoacoustic second harmonic signal. When the SNR was 3, the minimum detection limit of the system was reduced from 0. 43 mu L . L(-1 )to 0. 030 6 mu L . L-1. This shows that FFT filtering is very effective in eliminating the non-correlated noise in the second harmonic component of wavelength modulated spectral signal, which can improve the measurement accuracy and improve the minimum detection limit of the photoacoustic spectrum detection system. The research results in this paper can provide some references for the application of wavelength modulated spectrum signal second harmonic component detection technology.