This paper describes the basic concepts of error diagnostics and an associated rule system whose application helps to identify potential hardware/software locations of errors which caused a failure observed while executing tests of embedded systems. Further on, an overview over the "Avionics Smoke Detection System" is given, where the main algorithms have been applied. The methods, techniques and tools described here rely on the preceding investigations performed with respect to signal flow and the observation of causal chains in distributed test benches during test suite executions. These results have been elaborated in the KATO project within the German aerospace research programme LUFO III.