White Rabbit, Explained: The Sub-Nanosecond Timing Protocol Beyond PTP

What does "sub-nanosecond" actually mean?

Start with a number you can hold in your hand. Light travels through fiber at roughly 5 nanoseconds per meter — so one nanosecond of time equals about 20 centimeters of cable. When White Rabbit promises sub-nanosecond synchronization across a network, it means every clock on that network agrees to within less than 20 centimeters of fiber-length's worth of time. Your network's timekeeping becomes accurate to about the width of a hand.

NTP gives you milliseconds — good enough for logging and Kerberos. PTP (IEEE 1588) gets you into the microsecond class, good enough for most trading timestamps, 5G fronthaul budgets, and power-grid monitoring. But a handful of jobs genuinely need more: locating a grid fault by triangulation, timestamping particle collisions, or proving exactly which of two competing orders reached an exchange first. White Rabbit exists for those jobs.

Why PTP alone can't get there

PTP is a protocol running on top of ordinary packets, and three things cap what packets alone can measure. First, timestamp granularity and jitter: even with hardware timestamping, PTP slaves typically land in the low microseconds, and on congested links worse. Second, the asymmetry assumption — PTP's delay math assumes the path from master to slave takes exactly as long as the path back. It never does: transmit and receive paths through PHYs, SFPs, and fibers differ by tens or hundreds of nanoseconds, and PTP has no way to know by how much. Third, free-running oscillators: a PTP slave's local clock wanders between correction messages, so the servo is always chasing.

Our NTP vs PTP guide covers the middle ground and how to budget a PTP deployment with the PTP Timing Budget Calculator. White Rabbit is what you reach for when PTP's budget cannot close.

The three ingredients

White Rabbit doesn't replace PTP — it surrounds it with two helpers, and adds a measuring instrument.

1. 1. SyncE: one shared heartbeat. Every White Rabbit link runs Synchronous Ethernet at Layer 1, so each node's oscillator is frequency-locked to the same physical signal. This kills the biggest error source first: with no drift between messages, the clocks only ever disagree on phase, never on rate. SyncE alone doesn't synchronize time — it just makes sure everyone is humming the same tune at the same tempo.
2. 2. Enhanced PTP: aligning the phase. With frequency locked, PTP's job shrinks from steering both rate and time to aligning phase only. The White Rabbit profile of PTP uses the standard 1588 machinery — Sync, Follow-Up, Delay_Req, Delay_Resp — but on a network where the hard part (frequency) is already solved, the servo converges tighter and holds.
3. 3. DDMTD: the picosecond ruler. This is the genuinely new piece. Each White Rabbit node has a dual-mixer time-difference circuit that measures the phase difference between its local clock and the clock recovered from the incoming fiber with picosecond-grade precision. PTP's timestamps told you when packets left and arrived; DDMTD tells you, continuously and precisely, how your tick edges line up with your neighbor's. That's the raw measurement that pushes accuracy from microseconds into the sub-nanosecond range.

The real trick: measuring asymmetry instead of assuming it

Here's the insight that matters most. Plain PTP computes path delay as (t2 − t1 + t4 − t3) / 2 — the average of the round trip, which is only correct if the two directions take exactly the same time. They don't. And on fiber, they differ per wavelength: White Rabbit sends each direction on a different wavelength (typically 1310 nm one way, 1490 nm the other over a single fiber), and those two wavelengths travel at slightly different speeds through the same glass.

White Rabbit measures the actual delay of each direction and each wavelength, per link, during calibration — then subtracts the real asymmetry instead of pretending it is zero. That's the step that buys the last two orders of magnitude. The DDMTD gives you a precise phase reading; the asymmetry correction makes sure you're correcting the right phase.

Who actually needs this — and who uses it

White Rabbit was developed at CERN with research partners including GSI Helmholtz Centre for Heavy Ion Research, to synchronize the equipment scattered across CERN's accelerator complex. It stayed in the physics world for years: CERN's LHC is the flagship deployment, and the KM3NeT neutrino telescope uses it to timestamp detectors spread across the Mediterranean seabed. Since then the user list has widened: power-grid operators use sub-nanosecond sync to locate faults by triangulating arrival times at substations; financial institutions use it for trading-system timestamping where MiFID II's clock-synchronization rules (RTS 25) set the regulatory floor; and 5G — with 6G coming — needs tight phase alignment between radios, exactly the job White Rabbit was built for. Deployed networks span more than 1,000 kilometers with thousands of nodes.

For an HFT engineer, the honest framing is this: your matching-engine timestamps probably live in the PTP world (see the HFT latency guide for how timing fits the tick-to-trade budget, and the timing simulator for how PTP error accumulates across a chain). White Rabbit enters the picture where the regulatory or physics argument demands proof at the nanosecond level — timestamping that has to survive an auditor's or a rival's microscope.

The commercial moment: it's not a lab toy anymore

Three things changed White Rabbit from a physics project into something you can buy. First, standardization: IEEE 1588-2019 includes a High Accuracy profile (Annex I.5) that is based on White Rabbit — the protocol's core is now part of the PTP standard itself, not a separate CERN dialect. Second, openness: the hardware designs live in the Open Hardware Repository (ohwr.org), so the ecosystem is inspectable and multi-vendor rather than locked to one supplier. Third, products: timing vendors now ship White Rabbit-capable gear — Oscilloquartz (Adtran's timing subsidiary) offers White Rabbit timing systems, and vendors including Safran and Timebeat ship White Rabbit products (as listed by timing distributor Telnet Networks). That's the moment a protocol becomes a procurement option: standard, open, and buyable.

The honest limits

Sub-nanosecond has a price, and you should know it before you get excited. Every node in the sub-ns path must be White Rabbit-capable — special switches and nodes with the SyncE + PTP + DDMTD hardware. A White Rabbit master talking to a plain PTP slave degrades gracefully to ordinary PTP performance; the magic exists only between White Rabbit endpoints. It's a fiber technology — the phase measurement and wavelength-specific delay correction are designed around optical links. And you budget calibration work: the per-link, per-wavelength delay measurement that kills asymmetry has to be done for every link, and re-done if the fiber plant changes. None of this is a flaw — it's the cost of measuring instead of assuming. If PTP's microsecond-class performance closes your budget, plain PTP with good hardware timestamping is cheaper, simpler, and easier to staff. White Rabbit is for the jobs where it doesn't close.