Testing a Space Clock: Why Stability Needs a Long Record

How do you establish that a spacecraft's clock keeps a very steady rate? Placing a better clock beside it is not an easy solution when mass and power are limited. Comparison with a reference on Earth is possible, but introduces the communication and measurement chain's own errors. NASA's Deep Space Atomic Clock experiment concerns more than flying a compact atomic clock. It also concerns measuring that clock's long-term behavior reliably from space.

DSAC launched aboard General Atomics' Orbital Test Bed spacecraft on 25 June 2019. Its technology demonstration ended on 18 September 2021, after more than two years. Despite its name, the experiment operated in Earth orbit; it was not a completed navigation system steering itself to another planet. This account considers that experimental record within a knowledge cutoff of 13 October 2026. [2]

The atomic reference inside the clock

An author preprint shared by Eric A. Burt and colleagues in December 2020 describes the instrument and its early long-term results. A transition between specified energy levels of mercury ions provides a stable frequency reference. A microwave signal near 40.5 gigahertz, derived from an electronic local oscillator, interrogates the ions. Their response provides feedback used to determine a frequency correction. [1]

An atomic clock consequently does not eliminate electronic oscillation. A control loop links the atomic reference to an electronic output. Ultraviolet light prepares the ions and reads their state; this is a different function from the microwave clock transition. Describing the instrument as an optical clock or an ultracold laser-cooled atom experiment would change the physical method. The ions are maintained near room temperature using neon buffer gas. [1]

A different operational route

The design contained two ion traps. A four-rod load trap prepared ions, while a multipole trap was intended to expose them to a lower average trapping field during interrogation. Electromagnetic confinement helps avoid interactions with cell walls. It does not remove every motion-related shift or environmental sensitivity. [1]

A neon-depletion issue reported in the manuscript made continuing operation in the intended multipole trap unviable. The team instead operated the clock in the load trap. That region was not optimized for the same short-term performance, but its long-term results still met mission requirements. This change matters to understanding the demonstration. Success does not mean every component operated exactly as originally planned; it concerns the performance established in the configuration actually used.

Short and long times are different tests

Fluctuations over seconds and frequency changes over weeks address different questions. A favorable short-term number does not establish long-term stability. The team could estimate short-term behavior using the clock's signal and operating parameters. Assessing long-term performance required comparison with another clock. [1]

A GPS receiver carried with the payload supported that comparison through time transfer to ground references. This does not mean future deep-space spacecraft can use Earth-orbiting GPS signals everywhere. GPS served as a measurement link here, not as the physical frequency reference replacing the mercury ions. The usefulness of an onboard clock and the infrastructure required to evaluate it during a demonstration are separate issues.

Measuring the comparison chain

GPS time transfer is not perfect. Receiver temperature sensitivity, signal processing, and ground reference clocks contribute to the observation. Corrections for the spacecraft's motion and gravitational environment are also necessary. The researchers did not simply assume a good clock and attribute every change to it. They examined the noise expected from the measurement system as well. [1]

This distinction was especially important at short averaging times, where transfer noise limited separation of the clock's own behavior. Longer averaging permitted clearer investigation of clock performance. Nevertheless, the reported observations include both clock and comparison-system contributions. Describing the measurement as an upper bound for the clock alone acknowledges that the entire observed fluctuation was not necessarily generated inside the clock.

The measurement problem does not disappear because the reference is atomic. An energy-level transition can be a stable anchor while the complete comparison remains affected by receivers, orbital information, and environmental corrections. A convincing result identifies those links rather than treating the word atomic as a guarantee that the observation has no other uncertainty.

What twenty-three days establish

The preprint presents long-term results derived from a continuous fifty-two-day record. At an averaging timescale of twenty-three days, it reports fractional frequency stability of approximately three parts in a quadrillion and time deviation below four nanoseconds. Time deviation here is a statistical stability quantity. It is not a guarantee that the clock's absolute error relative to true time remained below four nanoseconds at every moment. [1]

A small frequency drift was also estimated from the record. Drift concerns the tendency of the output frequency to change with time; it is not simply a number stating how many seconds the clock is ahead. Results obtained with the temperature-control loop disabled and a varying baseplate temperature demonstrated that environmental influences could be managed. They do not imply that temperature or radiation had no effect. Magnetic shielding and the control algorithm were also part of the instrument's performance.

Better knowledge is not always better hardware

NASA's June 2021 account described the new stability assessment as approximately five times better than a result reported in 2020. It explicitly explained that this was not a fivefold improvement in the clock itself. Longer operation and additional data allowed a more precise measurement of stability. This distinction can easily disappear in a technology headline, yet substantially changes the interpretation. [3]

Learning about long-term behavior requires a long record. Selecting a few good seconds cannot establish weekly stability. Slow variations must also be distinguished from the contribution of the comparison chain. Extending the observation therefore does more than increase the sample count: it makes previously unresolved timescales accessible to investigation.

One component of navigational autonomy

A stable onboard clock matters for navigation approaches that evaluate one-way radio observations aboard a spacecraft. The clock does not by itself perform every position and trajectory calculation. Information about the signal source, propagation, and the spacecraft's dynamics is required too. Reducing dependence on ground support does not abolish light-travel time. It changes which calculations can be performed where. [1]

DSAC's concrete contribution has two layers: testing an ion clock in orbit, and establishing the measurement limits within which its behavior was known. Future navigation capabilities can build upon that evidence, but are not substitutes for it. Rather than declaring an autonomous deep-space spacecraft complete, the demonstration supplies a substantial record concerning the long-term reliability of a timing reference that such autonomy requires.

Sources

[1] E. A. Burt and colleagues (2020), The Deep Space Atomic Clock: the first demonstration of a trapped ion atomic clock in space, author preprint, DOI 10.21203/rs.3.rs-122117/v1. Open-access research manuscript. Original English explanation adapted with attribution from this version under CC BY 4.0; figures are not reproduced.

[2] NASA/JPL (5 October 2021), Working Overtime: NASA's Deep Space Atomic Clock Completes Mission. Mission completion record.

[3] NASA/JPL (30 June 2021), Deep Space Atomic Clock Moves Toward Increased Spacecraft Autonomy. Stability-measurement explanation.

Source access and factual review: 15 October 2026. Knowledge cutoff: 13 October 2026. The preprint's main text and methods, and both institutional accounts, were directly read. Raw clock data were not independently reanalyzed; the later journal version's full text was not examined. Historical schedules for future missions are not presented as current plans.