This paper proposes an online motor-parameter-estimator for a PMSM in an EV-powertrain. The proposed method differs from the conventional approach by using thermal measurements to decouple the resistance estimation from the rest of the estimation. Conventional approaches use the voltage and current measurements to estimate all parameters at once. However the resistance estimation was often found to be unreliable and noisy due to the low-contribution in the voltage-equations. A recursive least-squares filter approach in combination with the discrete-time dynamic voltage-equations was adopted. In this way the estimator was valid in both transient and steady-state operation while providing a robust estimation over the entire operating range. The proposed estimator was validated both using simulations and experimentally. A sensitivity analysis showed the proposed estimation approach is more robust against rotor-position error leading to smaller errors in the estimation. In the experimental validation the proposed estimator showed reliable estimation over the entire operating-range of the PMSM whereas for the conventional method the unreliable resistance, caused estimation error on the other parameters. The proposed method can be adopted for online maximum-torque-per-ampere control and adaptive-torque-control in an EV-powertrain.