Abstract: A method for diagnosing single soft fault in analog circuits with low testability is presented in this paper.The faults of circuit under test are analyzed by fuzzy clustering algorithm and the fault classes are adaptively obtained by using the fuzzy clusters' validity indices.The testable component set is determined by the cluster results.Support vector machine(SVM) is introduced to identify the analog circuit faults.SVM has advantages of simple structure and strong generalization ability.Experimental results on analog and mixed-signal benchmark circuits demonstrated the efficiency of the proposed method for diagnosing analog circuits' single soft fault which based on fuzzy clustering analysis and SVM.