This paper considers the problem of constructing a globally convergent positionand velocity estimator with close-to-optimal noise properties using hydroacoustic long baseline measurements. Three ways of improving the range robustness of the three stage filter for long baseline measurements with unknown wave speed are suggested. One addition is employing depth measurements in addition to pseudo-range measurements, thus increasing range noise robustness and relaxing requirements for transponder placement from not co-planar to not colinear. Furthermore, a Kalman Filter with a linear measurement model is used, instead of a pseudo-linear time-varying measurement model and a step solving an optimization problem is also added. The proposed scheme is validated through simulation and compared to a standard Extended Kalman Filter and a perfect (non-implementable) Linearized Kalman Filter using real states as linearization point. Simulations suggest that the improved three stage filter will have similar stationary performance as the EKF while having significantly better transient performance and stability subjected to inaccurate initial estimates.