A method based on improved Mahalanobis distance according to the characteristics of analog circuits is proposed for fault diagnosis. Firstly, the circuit testability is analyzed based on the topology of the test circuit and the ambiguity groups are determined by using test matrix. Then the separable potential faulty components under the assumption of single fault are identified. Finally, improved Mahalanobis distance of the feature values of the test points is applied to classify the suspicious components, resulting in fewer classes and faster speed for fault classifying process. Simulation results show that the proposed method can improve the fault diagnosis rate and is effective for analog circuit fault diagnosis and classification.