A Multi-Variance Analysis in the Time Domain

Recently a new technique for characterizing the noise processes affecting oscillators was introduced 1 2. This technique minimizes the difference between the estimates of several different variances and their values as predicted by the standard power law model of noise. The method outlined in this paper makes two significant advancements it uses exclusively time domain variances so that deterministic parameters such as linear frequency drift may be estimated, and it correctly fits the estimates using the chi-square distribution. These changes permit a more accurate fitting at long time intervals where there is the least information. This technique has been applied to both simulated and real data with excellent results.