Abstract: To diagnose the soft fault of analog circuits with tolerance, an innovative arithmetic based on fuzzy C-means clustering analysis and fuzzy membership function is presented. Because there is the tolerance in analog circuits containing several the tolerance components, such as resistors and capacitors, the parameter of analog circuits with tolerance is permitted to depart the ideal parameter partly. It is very difficult to distinguish the normal tolerance state and the soft fault states. Firstly, the basic theory of fuzzy C-means (FCM) clustering algorithm and membership algorithm of fuzzy control is summarized. Secondly, a soft fault diagnosis example of analog circuit with tolerance is provided to verify the validity of our algorithm: Firstly, the classes of the normal tolerance state and the soft fault states are defined in analog circuits with tolerance. Multisim software is used to simulate the normal state and the soft fault states of analog circuits with tolerance and get the enough samples that needed by future clustering analysis and fault diagnosis. Secondly, samples are analyzed by clustering algorithm. Fuzzy C-means clustering algorithm is applied to classify the normal state and the soft fault states of analog circuits with tolerance; clustering centers are obtained by fuzzy C-means clustering algorithm. Clustering center is the typical parameter of the normal tolerance state and the soft fault states of analog circuit with tolerance. Finally, a random state is simulated in Multisim software. Fuzzy membership algorithm is used to calculate the membership between current state and clustering centers, judge the state of analog circuits with tolerance, diagnose the fault of analog circuits with tolerance. The result shows that this method can classify the normal state and fault states of analog circuits with tolerance accurately, obtain the typical parameter of every state using small samples, diagnose the soft fault of analog circuits with tolerance objectively and effectively.