In this paper, we investigate the performance of positioning algorithms in wireless cellular networks based on time difference of arrival (TDoA) measurements provided by the base stations. The localization process of the mobile station results in a non-linear least squares estimation problem which cannot be solved analytically. Therefore, we use iterative algorithms to determine an estimate of the mobile station position. The well-known Gauss-Newton method fails to converge for certain geometric constellations, and thus, it is not suitable for a general solution in cellular networks. Another algorithm is the steepest descent method which has a slow convergence in the final iteration steps. Hence, we apply the Levenberg-Marquardt algorithm as a new approach in the cellular network localization framework. We show that this method meets the best trade-off between accuracy and computational complexity