Field Guide · term

Also known as: frequency stability, oscillator stability, frequency accuracy

Frequency stability is how well an oscillator keeps its output at the intended frequency as temperature, supply voltage, load, and time change.1 It is usually specified in parts per million (ppm) — a ±2 ppm source at 450 MHz may wander up to ±900 Hz — and, for the shortest timescales, as Allan deviation. Stability underpins everything a local oscillator is used for, because a receiver that drifts off frequency loses lock on the signals it is trying to decode.

Δf temp ppm tolerance stable drifting
A stable oscillator holds its frequency inside a narrow ppm band across temperature; a poorly compensated one drifts out of tolerance at the extremes.

How it works

The dominant error in a quartz oscillator is temperature: the crystal’s resonant frequency changes as it warms and cools, tracing a characteristic curve. Aging adds a slow, roughly logarithmic drift over months and years as the crystal and its mount settle. Supply-voltage and load changes contribute smaller pulling effects. A datasheet folds these into an overall ppm figure over a stated temperature and time window.

Technologies are layered to fight each error source:

  • A plain crystal oscillator drifts tens of ppm over temperature.
  • A TCXO (temperature-compensated) applies a correction versus temperature, reaching roughly ±0.1–2 ppm.
  • An OCXO (oven-controlled) holds the crystal at a constant elevated temperature, reaching parts per billion.
  • A GPSDO disciplines a local oscillator to the atomic-clock reference in GPS, giving near-perfect long-term accuracy.

Stability is a long-term property and is distinct from phase noise, the short-term random jitter around the carrier. The two are measured differently — ppm and Allan deviation for stability, dBc/Hz for phase noise — and a source can excel at one while being mediocre at the other.

In practice

Where an oscillator is stable but simply offset, the fix is calibration: measure the error against a known reference and apply a ppm frequency correction so software retunes to compensate. Allan deviation is the tool for characterising stability over a chosen averaging interval, revealing where a source is best (often around one second) and where drift or noise takes over. Systems that must interoperate over the air — cellular base stations, trunked simulcast sites — carry disciplined references precisely so every transmitter agrees on frequency and timing.

Relevance to SDR

Frequency stability is a practical daily concern in SDR reception. Inexpensive RTL-SDR dongles ship with basic crystals specified around ±20–30 ppm and drift as they warm up, which at UHF is enough to walk a narrowband signal out of the decoder’s capture range. The standard workflow is to measure the offset against a known transmitter and enter a ppm correction; TCXO-equipped dongles reduce the drift at the source. For trunked systems the tolerance is tight because the decoder must stay locked to a narrow control channel for long periods.

GopherTrunk relies on the front-end’s stability but also compensates for it in software: it applies a configurable frequency correction and its carrier-tracking loops pull in residual offset, so a modest, slowly drifting SDR can still decode reliably. What software cannot cure is a source so unstable that it drifts faster than the loops can follow — which is why a TCXO or better matters for demanding wideband, multi-channel monitoring.

Sources

  1. Frequency drift — Wikipedia, causes of oscillator frequency change over temperature, time, and aging. ↩

See also