The error measuring, modeling and compensation techniques for the positioning stage driven by NC linear motors are studied. The error source of the positioning stage is analyzed, the positioning errors are measured by the laser interferometer, and the neural network error model is set up by RBF algorithm. In order to evaluate the accuracy of RBF network prediction method, part of the error samples are used to test. A DSP-core linear motor experimental platform is built up, the error compensation experiments are conducted, the real-time requirement is proved to be met. The simulation and experimental results indicate that the RBF neural network error model trained by samples has a good learning ability and generalization ability, the positioning accuracy is improved significantly, and the effect of random errors on the system is reduced also.