Cesium, Rubidium or Hydrogen Maser: Which Atomic Clock Does Your Network Actually Need?

by | Oct 8, 2026

Every timing network eventually loses GPS. A jammer on a nearby highway, a damaged antenna cable, a contractor who moves a rooftop unit. When that happens, the only thing keeping your network on time is the oscillator inside your grandmaster or time server, and the physics of that oscillator decides whether you have minutes, hours, days or months before your timing drifts out of spec.

Our partner Microchip produced a short video that explains that physics better than most textbooks. It covers how cesium atoms, hydrogen masers and coherent population trapping turn the behavior of individual atoms into a frequency you can trust. We think it is worth a few minutes of your time, and below the video we explain why it matters for the network you actually operate.

Video courtesy of Microchip Technology, a Syncworks partner.

Why the physics matters to an operator

Every clock in your network is a counter attached to something that oscillates. In most equipment that something is a quartz crystal. Quartz is stable over seconds and minutes, but its frequency wanders with temperature and ages over weeks. On its own it cannot hold time for long.

Atoms do not age. A cesium 133 atom switches between two energy states at exactly 9,192,631,770 cycles per second, which is why that number is the official definition of the second. Rubidium and hydrogen have their own fixed transition frequencies. An atomic clock is, in practice, a quartz or similar oscillator that is constantly corrected against one of those atomic transitions.

That is the part of the video that matters most for your network. While GNSS is available, your grandmaster is steered by the satellites. When GNSS goes away, it is steered by whatever is inside the box. A better atomic reference does not make your network more accurate on a good day. It makes your network survive a bad one.

The five references, and where each one belongs

The video walks through the science. Here is how each technology shows up in real deployments, from the most common to the most specialized.

ReferenceWhat it holdsWhere it belongs
OCXO (quartz)Minutes to hours of useful holdoverEdge and access sites with reliable GNSS or a redundant PTP path
RubidiumHours to days. Microchip rates rubidium holdover in the TimeProvider 4100 at within 400 ns over 24 hoursGrandmasters at aggregation sites, substations and data centers that must ride through a GNSS outage
Cesium beamWeeks to months. The 5071A is rated at better than 5E-13 accuracy under all environmental conditionsPrimary references, ePRTC sites and any network that cannot depend on the sky
Hydrogen maserThe best short and medium term stability available. The MHM 2020 is specified at less than 3E-16 per day agingNational timing labs, radio astronomy and GNSS ground segments
Chip scale atomic clock (CPT)Atomic stability in a part that runs on under 300 mWMobile, battery powered and defense equipment where a rack mounted standard will not fit

We covered the first two in depth in our post on the differences between OCXO and rubidium oscillators. The video fills in the other three, and two of them deserve a closer look.

Cesium: the reference that does not need the sky

Cesium is the only technology in the list that is both a primary standard and practical to deploy in a carrier or utility network. That is why it sits at the heart of the enhanced Primary Reference Time Clock (ePRTC). The ITU-T G.8272.1 requirement is demanding: an ePRTC must be accurate to within 30 ns of UTC, and once in holdover its error may only grow to 100 ns over fourteen days. Two weeks without satellites is exactly the scenario cesium is built for.

Cesium also comes with a lifecycle. The beam tube is a consumable, and planning its replacement is part of owning one. If you operate cesium today or are evaluating it, our TimeCesium 4500 page covers the current platform.

Coherent population trapping: atomic timing in your pocket

The most surprising part of the video for most network engineers is coherent population trapping, or CPT. Instead of a microwave cavity, a CPT clock uses a tiny laser to interrogate a vapor of atoms, which is what allows a full atomic clock to shrink to a chip. Microchip's low noise SA65 chip scale atomic clock runs on under 295 mW with frequency drift under 0.9 ppb per month. You will not find one in a central office, but you will find them in mobile radar, field radios and unmanned systems, anywhere an atomic reference has to run on a battery.

Three questions that decide which reference you need

Choosing a reference is not about buying the best clock. It is about matching holdover to risk. When we design a timing architecture with a customer, we start with three questions.

  1. What is your time error limit? A 5G TDD radio needs to stay within about 1.5 microseconds of its neighbors. IEC 61850 protection and sampled values in a digital substation typically need 1 microsecond. A financial trading platform may need far less. The tighter the limit, the faster a weak oscillator runs out of margin.
  2. How long could GNSS realistically be gone? An hour for a cable cut is very different from days of interference near an airport, a port or a border. If you do not know your exposure, our GNSS security resources and the BlueSky GNSS Firewall are a good place to start.
  3. Is this site a reference for other sites? A clock that only feeds itself can tolerate more risk than one that is the source of time for an entire region. Reference sites are where rubidium becomes the floor and cesium becomes the conversation.

Divide your time error limit by your oscillator's drift and you have a rough holdover budget. Compare that to how long GNSS could be gone, and the right reference usually becomes obvious. When it does not, that is a design conversation worth having before the outage, not during it.

What most networks get wrong

In our field work, the most common problem is not choosing the wrong oscillator. It is not knowing which oscillator is installed. Networks grow by acquisition and by project, and a single region can end up with OCXO, rubidium and cesium equipment of different ages, with nobody sure which sites can actually ride through an outage. The first time anyone finds out is during one.

That is exactly what a Syncworks synchronization audit answers. We inventory every timing source, document the oscillator and holdover capability of each one, and map it against the time error limit of the services it feeds. The result is a clear picture of which sites are protected and which are one antenna failure away from an incident.

Frequently asked questions

What is the difference between a rubidium and a cesium atomic clock?

Both lock a quartz oscillator to an atomic transition. Rubidium is smaller, lower cost and holds time for hours to days, which makes it the standard choice inside grandmasters like the TimeProvider 4100. Cesium is a primary frequency standard that can hold time for weeks to months, and is used where a network must operate without GNSS for extended periods.

Do I need a cesium clock if I already have GPS?

GPS is the most accurate source most networks will ever see, but it is also the most exposed. Cesium is not a replacement for GNSS. It is the reference that keeps your network on time while GNSS is unavailable, and it is a core element of an ePRTC architecture.

What is a hydrogen maser used for?

Hydrogen masers offer the best short and medium term frequency stability of any commercial clock. They are used by national metrology labs that contribute to UTC, by radio astronomy and by GNSS ground infrastructure, rather than in carrier or enterprise networks.

What is coherent population trapping in an atomic clock?

Coherent population trapping is a technique that uses a laser instead of a microwave cavity to detect an atomic transition. It allows the physics package to be miniaturized, which is the basis of the chip scale atomic clock used in mobile and battery powered equipment.

Talk to a Syncworks timing engineer

If the video raised questions about the references in your own network, we can help you answer them. Our engineers design, install and support timing architectures for communications, utilities, cable and data center customers across North America.

Request a synchronization audit, ask for a quote or call us at 904-280-1234.

Join our Newsletter

Expert knowledge and company news

CUSTOM SUPPORT

Keep your network running smoothly with SyncCare — expert support, proactive maintenance, and fast issue resolution for your timing infrastructure.