The transition delay faults (TDF) model has been widely used in industry to model time-related defects. A dynamic bridging fault (DBF) also has similar delay effect. However, the causes of these two types of faults are quite different: a DBF is due to the bridging effects between two circuit nodes, while a TDF is due to a node itself or the logic connected to the node. It is important to distinguish these two types of faults such that the exact sources of defects can be identified during the yield ramping process. In this paper we analyze the relation of TDFs and DBFs and present an efficient diagnosis pattern generation procedure to distinguish them. A novel circuit model is developed which can transform the problem of distinguishing a pair of a DBF and a TDF into the problem of detecting a DBF. The pattern generation process can then be done by using an ATPG tool for DBFs. All fault pairs can be modeled in a single circuit and dealt with in the same ATPG run. Thus the pattern generation process is quite efficient and very compact pattern sets can be obtained. Experimental results on ISCAS89 benchmarks show that 99.99% of random fault pairs can be either distinguished or identified as equivalent-fault pairs.