Holdover NTP: TCXO, OCXO and Rubidium compared

A time server normally gets its precise time from satellites (GPS). But what if that signal drops out for a moment? Then the server has to keep ticking on its own built-in clock without drifting away. That is called holdover. How long it stays good depends on the type of built-in clock, the oscillator.

Three types, from cheap to expensive

• TCXO: holds for minutes to a few hours. Fine for ordinary IT.
• OCXO: holds for a day to several days. The standard for broadcast and industry.
• Rubidium: a miniature atomic clock, holds for weeks. For defence, telecom and critical applications.

Our advice

Many customers buy Rubidium because "it is the best". In 9 out of 10 cases an OCXO is enough, for half the price. More important than an expensive clock is a good GPS antenna: if you have signal almost all the time, holdover is rarely needed. We ask the right questions so you do not overpay.

Holdover is the ability of a time server to stay within tolerance after the external reference (GNSS, IRIG, etc.) has dropped out. The type of oscillator determines how long that lasts, from hours to months.

Which holdover sources exist?

TCXO — Temperature-Compensated Crystal Oscillator

• Stability: ±0.5 ppm typical, drift ~50 μs/hour
• Holdover at UTC spec (±1 ms): 30 minutes to a few hours
• Cost: low
• Application: IT servers, office environments, non-critical applications

A TCXO compensates temperature-induced frequency drift via an internal thermistor. Cheap but relatively inaccurate. For most IT purposes this is more than enough.

OCXO — Oven-Controlled Crystal Oscillator

• Stability: ±0.01–0.05 ppm typical, drift ~1–5 μs/hour
• Holdover at UTC spec: 24 hours to several days
• Cost: medium
• Application: broadcast, financial trading, data centres that need to demonstrate UTC traceability

An OCXO keeps the crystal at a fixed temperature via a small oven (hence the name). That almost entirely eliminates temperature drift. The crystal draws a few watts continuously, be mindful of heat dissipation in dense rack rooms.

Rubidium — Atomic Frequency Standard

• Stability: ±1e-11 typical, drift ~36 ns/hour
• Holdover at UTC spec: several months
• Cost: high
• Application: defence, telecom fronthaul, scientific (radio astronomy), high-frequency trading

A Rubidium standard uses an atomic-physical phenomenon (the hyperfine transition of rubidium-87) to generate its frequency. In effect it is a miniature atomic clock in your rack chassis. Very stable, but the rubidium cell "lives" 10–15 years and has to be replaced after that.

Which oscillator for which situation?

Use case	Recommended	Reason
Corporate servers, log sync	TCXO (NTP100 default)	Cheap, ms accuracy sufficient
Broadcast TV studio	OCXO (NTP100-OSC)	Frame-accurate sync must survive hours
Data centre with SLA	OCXO	Demonstrable holdover for audit
Trading floor MiFID compliance	Rubidium or OCXO + secondary	100 μs over 24 h, no margin for drift
Digital substation IEC 61850	OCXO + PTP	Power Profile requires μs level
Defence mission-critical	Rubidium	Bridging days to months without GNSS

## Why is holdover a chain problem?

The oscillator is one link. Other weak links: antenna-cable shielding, switch jitter, ground loops. A Rubidium with a poor antenna is wasted money. Good GNSS set first, expensive holdover only after that, if you have lock 99% of the time a TCXO or OCXO is already enough.

How does Daylight help you pick the right holdover source?

We see too many customers buying Rubidium because "it is the best". In 9 out of 10 cases an OCXO is enough, at half the price. Call Daylight and we will ask the right questions, expensive is not always necessary.

Sources

• NIST Time and Frequency Division
• Banerjee & Matsakis — Modern Timekeeping (academic reference)