In this paper, two approaches for diagnosability analysis of discrete event systems are discussed and experimentally evaluated. The considered approaches are the diagnoser-based approach proposed in [1, 2] and the model-checking reformulation approach proposed in [3, 4]. Experiments are performed on a level crossing benchmark, using the software tools integrating the considered approaches. These two approaches show different features in terms of state-space building and procedure for analyzing diagnosability. Based on the obtained results through the benchmark, a comparative discussion is provided particularly regarding the generated state-spaces and the time consumption for analyzing diagnosability.