This paper analyzes the ball screw preload loss by acoustic signals through the Hilbert-Huang Transform (HHT) and Short-Term Fourier Transform (STFT). To obtain the robustness of experiment measurement, microphone location for determining the ball screw nut frequency region is conducted first. Second, different preload features are discriminated by Empirical Mode Decomposition (EMD). Finally, the nut stiffness features are compared by using STFT and Marginal Frequency. Experimentally results show the ball screw preload loss can be successfully diagnosed.