In the operational version of the Rapid Update Cycle (RUC) at NCEP, objective analysis is performed using the optimum interpolation method. To improve data assimilation and introduce new kinds od data into analysis, a three-dimensional variational analysis (3DVAR) method has been under development for several years. In this presentation, we will describe a new version of RUC 3DVAR which includes a standard filtering approach in the background error model and application of a regression based balancing scheme. Other improvements are optimization of the filter response, an experimental spectral approach, and a new way of handling integrated precipitation water information. Results from both individual experiments and real-time runs will be presented.