# How do ultra-stable oscillators for spacecraft work?

Many spacecraft carry "ultra-stable oscillators" for ranging and related measurements, but I can't quickly find any info on how these actually work. Are they just very pure quartz crystals, or do they use some other technology. I'd also be interested in quantitative figures on their stability.

• see this answer to How far from earth have atomic clocks (or ultra-stable oscillators) been placed and monitored? and also this answer to Where would one deploy deep space atomic clocks? – uhoh Aug 16 '19 at 14:50
• Have you checked the wikipedia pages for crystal oscillators (en.wikipedia.org/wiki/Crystal_oscillator) and rubidium clocks (en.wikipedia.org/wiki/Rubidium_standard)? They give a decent overview of the methods, which I'm not sure are quite on-topic for this SE – llama Aug 16 '19 at 21:52
• See also this answer on physics physics.stackexchange.com/a/129311/11645. Basic idea: use light spectroscopy to measure the population of one atomic state, and then use this to tune microwaves to a frequency which drives the maximum number of atoms out of this state and into another hyperfine state (shows as a dip on the light spectroscopy signal). Measure the frequency of these microwaves and divide it down to a useful clock frequency, which is now referenced to the atomic transition. – llama Aug 16 '19 at 21:59

The material depended on the mission but was usually quartz, and the stability is insane.

The development of Ultra-Stable Oscillators (USOs) is ongoing (see here and here).

NASA/the ESA have used things like quartz and rubidium depending on the mission (see the abstract) . Crystal oscillators are preferred because they are "...easier to qualify for deep space than atomic clocks" (same source). Also in that source is the line that "...all ultrastable oscillators in deep space have been quartz-based with the exception of the Huygens probe USO".

In terms of stability, those on the Voyagers had a "Allan Deviation" of $$10^{-12}$$ over 100 seconds. Precisely what the Allan Deviation, or Allan Variance, means is hard to understand, so I'm just going to quote Wikipedia's article on Allan Variance:

An Allan deviation of $$1.3\times10^{−9}$$ at observation time 1 s (i.e. τ = 1 s) should be interpreted as there being an instability in frequency between two observations 1 second apart with a relative root mean square (RMS) value of $$1.3\times10^{−9}$$. For a 10 MHz clock, this would be equivalent to 13 mHz RMS movement.

Newer oscillators, such as those in Mars Global Surveyor, were an order of magnitude better. For reference, a relatively inexpensive gyroscope made by Freescale Semiconductor has an Allan Variance of about $$10^{-2}$$ over the same timescale source, and a cheap crystal oscillator like this one have frequency stability of 100 parts-per-million, which is somewhere around $$10^{-5}$$ or so (although the two measurements aren't exactly equivalent)

• Your cheap reference is particularly bad. I'd mention an oscillator as they are used e.g. in mobile phone which has 10ppm over the whole temperature range, and much better when temperature is stable. They should be better than 1ppm in stability at constant temperature. Maybe also mention that stability != accuracy. – asdfex Aug 16 '19 at 17:17
• @asdfex: I just did a quick Digikey search for one. If you want to find a better one and edit I have no issues – Michael Stachowsky Aug 16 '19 at 17:20
• +1 and slightly related references to Allan Variance : How do I calculate Allan Variance (maybe for pulsars)? and also Why aren't GPS clocks sufficient to synchronize VLBI observations? and possibly (1, 2) – uhoh Aug 16 '19 at 21:29
• Hmm, I wonder if the state of atomic clocks vs quartz oscillators has changed much in the past decade or so (where most of these references seem to date from). It surprises me that atomic clocks aren't more common, since they're a pretty mature technology now and have quite a history of use in earth orbit, and tend to outperform even the best quartz oscillators by quite a large margin, especially on longer timescales – llama Aug 16 '19 at 21:40
• There are much better temperature compensated crystal oscillators sold, this one has a stability of 2.5 ppm and aging per year of 1 ppm. There are crystals oscillators heated to a constant temperature with only 5 ppb (parts per billion). – Uwe Aug 17 '19 at 18:44