Digital audio recordings can be manipulated by pervasive audio editing software easily. Often forgery would not be naive splicing. Post-processing would be a part of tampering. Post-processing can eliminate the obvious traces of forgery. Noise can cover audible evidence of forgery and destroy traces of other tampering operations. The detection of additive noise in audio signal is a useful tool for audio forensics. In this paper, we investigate the effect of additive noise on audio signal, and propose a feature named "sign change rate" for detecting additive noise. Via theoretical analyze and extensive experiments, it shows the proposed feature is effective in additive noise detection. Also the method can be a potential tool for forgery localization of digital audio.