RCQT Research Center for Quantum Technology · Chiang Mai University

Interactive lab › Read a sensor's noise like a fingerprint

Interactive lab · 12/66 · Noise and stability

Read a sensor's noise like a fingerprint

Every clock, gravimeter and laser in our laboratories drifts and jitters. Averaging longer helps only while white noise dominates; then flicker sets a floor and random walk and drift take over. Build a record from white, flicker, random-walk, drift and periodic noise, and read it the way metrologists do: as a power spectral density and as an overlapping Allan deviation, with the slopes that identify each process.

What you are seeing

  • The main plot is the overlapping Allan deviation σ(τ) of the simulated record (white), with the theoretical contribution of each noise type (coloured). White noise falls as τ^−½, flicker is flat, random walk rises as τ^½ and drift as τ.
  • The minimum of the curve is the best stability the sensor can reach and the averaging time that gives it; past it, averaging longer makes the answer worse.
  • A periodic disturbance (a building's air conditioning, a pump) leaves a bump near half its period and a line in the spectrum: the two views find different things.
  • Switch the preset between an optical clock, a gravimeter and a free-running laser: the same mathematics, very different numbers.

The numbers in the model

Allan varianceσ²(τ) = ½ ⟨(ȳ_{k+1} − ȳ_k)²⟩
Slopes of σ(τ)white τ^−½ · flicker τ⁰ · random walk τ^½ · drift τ¹
Record16 384 samples; overlapping estimator
SpectrumWelch average of Hann-windowed segments

Go further

More labs on this theme