Some cyclic variables include within each cycle a value that can be determined extrinsically. In such case, the sensed value may differ from the known value by an amount that is a combination of long-term drift of the sensor and random measuring error. The drift component can be evaluated and eliminated by the following method. Once each cycle, for a number of cycles, the sensor measures the variable at a time when its value should equal the extrinsically-known value. The differences are plotted versus time, and a best-fitting straight line is determined, which indicates the drift. Throughout the next cycle as the variable is continuously sensed, the drift determined from the best-fitting straight line is continuously applied to correct the sensed value.