We present in this paper two definitions of short‐term frequency stability: (i) time domain, the expected value of the variance of the fractional frequency fluctuations from nominal frequency, 〈σv2(N, T, τ)〉, in which N is the number of samples, T is the averaging time plus the dead time between samples, and τ is the averaging time; and (ii) frequency domain, the power spectral density of the fractional frequency departure from nominal frequency, Sv(f). We discuss the topics of conversion from the frequency domain to the time domain and conversion among time domain measures.
All measurements were made in the time domain, using period counting techniques. An oscillator was offset in frequency by using a specially built quartz crystal unit plated for 10 kHz below the frequency of the other sources. This oscillator was used to obtain the beat frequency required by the period counting approach.
Since the use of 〈σv2(2, T, τ)〉 as a measure of short‐term stability has significant advantages over 〈σv2(N, T, τ)〉, the relationship between the two quantities was investigated. For averaging times of one second or less, 〈v2(2, T, τ)〉 and 〈σv2(N, T, τ)〉 were almost equal.
The short‐term stability of several quartz crystal oscillators and precision frequency generators was measured. The stability over the shorter averaging times was nearly equal for most of the oscillators. At longer times, the stability of each oscillator was unique. The frequency generators demonstrated similar stability over averaging times of 10 and 100 milliseconds, but were unique elsewhere. The accuracy of all measurements was limited by systematic effects from the environment and the measurement equipment itself.