Predictive synchronization and restoration of corrupted velocity samples

Abstract Incremental pulse encoders are widely applied to digital speed measuring. The velocity samples obtained by measuring the pulse length contain different types of error. This combined error can be considered as a corrupting noise component that needs to be reduced before the velocity information is utilized, e.g., for feedback control. To ensure smooth and rapid response of the control system, the noisy velocity signal has to be filtered without delaying the primary component. Conventional low-pass filters always cause lag into the primary signal, and thus they cannot be used if fast tracking of the reference velocity is required. To overcome this problem, predictive filters are used in this work to realize a novel signal processing method, capable of synchronizing and restoring the corrupted and unevenly spaced samples. A cascade predictor structure is proposed for good overall noise attenuation. A software implementation of the proposed algorithm has been developed and tested on the TMS320C25 signal processor.