Velocity model-building for depth imaging of PS converted-wave data requires significantly more effort than PP imaging. For PS depth imaging, we must produce a model that flattens events on both PP and PS gathers and images equivalent events at the same depth for both data types. To satisfy these constraints requires a high-quality anisotropic model with accurate estimates of anellipticity and Vp/Vs ratio. We present a ray-trace tomography algorithm with floating event constraints that allows simultaneous update of all velocity model parameters for efficient and accurate velocity model-building with PP and PS data input. Tests on synthetic data with typical North Sea geology illustrate the effectiveness of the joint PP/PS tomography.