A new approach to clock modeling and Kalman filter time and frequency prediction

An extension of a previously reported Kalman filter prediction algorithm is presented. In this new algorithm, the time variable is added to the clock model so that both the time and the frequency can be predicted. The long-term limitation of the previous prediction algorithm is removed by calculating the aging time constant using nonlinear fitting with the gradient search technique. The measurement noise covariance can be estimated if the process noise covariance is the null matrix.<
>