Understanding Allan Deviation Plots: A Q&A Guide

--- Q: What is an Allan deviation (ADEV) plot, and why does it matter?

A: Allan deviation is the standard tool for characterizing the stability (noise) of oscillators and clocks. An ADEV plot graphs the deviation σ(τ) on a log-log scale against averaging time τ. Each point tells you: "Over this averaging interval, how much does the frequency wander?" Engineers use it to compare oscillators, diagnose noise types, and predict performance in applications like GPS, radar, and communications.

--- Q: How do I read the overall shape of an ADEV curve?

A: The vertical axis (log scale) shows stability—lower is better. The horizontal axis (log scale) shows averaging time τ. A "good" clock has a curve that dips low at the τ of interest. The slope of the curve reveals which noise process dominates at each τ. A straight line with a negative slope on the log-log plot indicates averaging is helping; a flat region means a noise floor has been reached; a positive slope means longer averaging actually hurts—stability is degrading.

--- Q: What does White Frequency Modulation (WFM) look like?

A: WFM appears as a segment with a slope of τ⁻¹/² (i.e., −0.5 on the log-log plot). Each decade of averaging improves stability by one decade. This is the classic white noise region, often dominant at short averaging times for quartz oscillators. It arises from additive white noise on the output signal—random, uncorrelated frequency fluctuations.

--- Q: What does Flicker Frequency Modulation (FFM) look like?

A: FFM produces a segment with a slope of τ⁰ (a flat line, slope = 0). Stability neither improves nor worsens with averaging—it hits a noise floor. This is caused by flicker (1/f) noise in the active device or resonator. On the plot, this is easily spotted as a plateau. It often limits the best achievable stability for mid-range quartz and some atomic standards.

--- Q: What does Random Walk Frequency Modulation (RWFM) look like?

A: RWFM shows a slope of τ⁺¹/² (slope = +0.5). Stability gets worse with longer averaging—the curve turns upward. This is caused by slow, random drift of the oscillator's frequency (e.g., aging, thermal sensitivity, or 1/f³ noise). On the plot, this upturn typically appears at long τ values and signals that averaging has become counterproductive.

--- Q: Quick summary?

Reading an ADEV plot is simply about identifying these slopes, noting where each noise type dominates, and matching the oscillator's performance to your system's averaging-time requirements.