What is the Allan deviation and how does it relate to phase noise and frequency stability?
Stability Metrics
Phase noise is the standard metric for RF and microwave oscillator stability because it directly describes the spectral purity at specific offset frequencies. Engineers use phase noise to calculate the impact on receiver sensitivity, EVM, and radar performance at specific offset frequencies relevant to the signal bandwidth.
Allan deviation is the standard metric for precision frequency references (atomic clocks, GPS disciplined oscillators, and frequency standards) because it characterizes stability over averaging times ranging from microseconds to days. The Allan deviation at averaging time τ tells you how much the frequency varies between measurements spaced τ apart.
The mathematical relationship: σy²(τ) = 2 ∫ Sy(fm) × sin⁴(π×fm×τ) / (π×fm×τ)² dfm, where Sy(fm) is the power spectral density of fractional frequency fluctuations. For practical conversion: σy(τ) ≈ √(2 × ln(2)) × fm_offset × 10^(L(fm)/20) / f_carrier, at the offset frequency fm = 1/(2τ).
Frequently Asked Questions
When do I use one vs the other?
Phase noise: for RF system design (receiver sensitivity, EVM, radar Doppler resolution). Allan deviation: for frequency reference characterization (clock stability, time transfer, navigation). Both describe the same underlying oscillator instability in different domains.
What is a good Allan deviation?
TCXO: σy(1s) ≈ 10^-9 to 10^-10. OCXO: σy(1s) ≈ 10^-12 to 10^-13. Rubidium: σy(1s) ≈ 10^-11 to 10^-12. Cesium: σy(1s) ≈ 10^-11 to 10^-12 (but best at longer τ). Hydrogen maser: σy(1s) ≈ 10^-13.
Does Allan deviation measure drift?
Standard Allan deviation does not separate deterministic drift from random instability. Modified Allan deviation (MDEV) and Hadamard deviation are used for characterizing oscillators with significant frequency drift, as they distinguish between noise types more effectively.