The present best method of predicting proton induced SEU rates in the proton radiation belts is given by W.L. Bendel [1,2, and 3]. It is a semi-empirical technique and describes the upset cross section as a function of a single experimental parameter. Bendel presented a two parameter approach in [3] but decided to emphasize the simpler one parameter approach since it could adequately describe the available data. Shimano et al. [4] presented results showing slight improvements in the fits using the two parameter Bendel model, and a slightly different two parameter model has been used by Normand [5]. Recent energy dependent experiments at this and other laboratories [6] on small feature size devices have produced data that is clearly not adequately fit by the single parameter approach. This paper demonstrates that improved fits of both new and old data can be obtained with a two parameter approach. Shimano's work focussed on four devices, three of which are from older technologies with large feature sizes (>1.5 µm) and are no longer representative of the newer small feature sized (<1.5 µm) devices. This work provides summary analyses with the improved Bendel model on much of the existing published proton data [8–13], and also on recent data [6,14, and 15] on newer devices. © 1990 IEEE