The paper presents an initial attempt to create a general procedure for synthesising concurrent self-checking microprogrammed control units having a prescribed quality of checking. The authors consider systems in which diagnostic features are provided by the use of a checking strategy based on special check keys to code the flow graph representing the microinstruction sequence. They investigate the conditions under which the diagnostic features allow faulty transitions in the flow graph to be successfully detected, and derive a method for obtaining probabilistic estimates of faulty transition detection. They present a matrix description of the probabilities of the events involved: fault occurrence at flow graph vertices, faulty transitions between vertices, and the detection of faulty transitions. They include a simple example illustrating the methodology, in which a proposed scheme is first analysed to determine its performance, then modified to meet a higher specification.