This paper analyzes how two Kalman Filter (KF) based frequency estimation algorithms react to phase steps. It is demonstrated that phase steps are interpreted as sharp changes in frequency. The paper studies whether the location of the phase step, within the sinusoidal waveform, has any effect on the frequency estimate. Because phase steps are not the product of a permanent change in the underlying frequency, the paper proposes an algorithm to correct frequency estimates deemed erroneous. The algorithm makes use of the residual of the KF to determine when an estimate is incorrect and to trigger a corrective action in which the frequency estimate is replaced by an average of the previous values that were considered accurate. Using synthesized and simulated data with distortions the paper shows the effectiveness of the correction algorithm in fixing frequency estimates.

10aAA14-0061 aWilches-Bernal, Felipe1 aWold, Josh1 aConcepcion, Ricky1 aBudai, Jamie uhttps://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=8974143