Drum-type electrostatic separators have been widely used in mineral processing and are now successfully applied to plastics separation. However, no systematic modeling and optimization work has been done to guide the design. We aim to develop a recovery model and compare a few design options. Particle trajectories were simulated at different conditions, and an empirical trajectory model as a function of various dimensionless groups was developed. A recovery model was then derived through a probability calculation that accounted for particle charge distribution. It was found that the S-shaped partition curve can be fitted with a logistic function, which is widely used for dense-medium separation and other classification methods in mineral processing. Explicit expressions of two key parameters of the curve, Ep (probable error) and q(50) (cut charge, corresponding to a 50% recovery) as functions of design and operating variables, as well as charge distribution parameters have been found. Hence, a connection of this stochastic approach and the traditional partition curve approach has been formalized. Finally, a few design options (1-stage with/without recycle and 2-stage separations) are optimized and compared under various charge distribution conditions.