Metrical calibration is very critical for the accuracy of a 3d sensor. Measuring uncertainty of about 1:3000 of the z-range requires a calibration accuracy of about 1:104 . Several model-based approaches for this problem exist. In this article the calibration task for a phase measuring triangulation sensor is described. After a short description of the geometrical distortions that are compensated for a non-model based algorithm for the metrical calibration is presented. The main idea is the splitting into a longitudinal z-calibration and a lateral x-y-calibration. For z-calibration a linear interpolation is used, whereas the x-ycalibration is performed by a polynomial approach. The polynomial coefficients are obtained by a least squares method. The algorithm is tested using a laboratory 3d sensor. The results are discussed and compared with the accuracy of other calibration algorithms.