Case study: When Milliseconds Aren't Good Enough

How a European AV research programme replaced a six-year-old timing hack with sub-30ns infrastructure — across a moving car, a field trailer, and a 15km highway corridor.

A Portuguese autonomous vehicle research programme needed to replace a six-year-old custom time server that could no longer meet the nanosecond-level synchronisation demands of a modern AV stack. 

The objective: Achieve sub-30ns timing accuracy across three distinct deployment environments — a field-deployable mobile trailer, an autonomous car in motion, and up to 20 distributed roadside units along a 15km highway corridor — without a cost structure that made corridor-scale deployment impossible. 

Timebeat delivered the OpenTime Appliance Mini deployment covering the vehicle and trailer requirements, confirmed simultaneous gPTP and PTPv2.1 dual-protocol support, eliminating the need for a dedicated translation switch, and is developing a compact roadside reference design that brings per-unit timing infrastructure within research budget constraints.

The Challenge

A research team in Portugal had built something impressive. A 12–15km highway corridor fitted with roadside units, edge computing, V2X communication, and a growing fleet of autonomous cars and robotic platforms. One of the most advanced AV research programmes in the country.

Their timing infrastructure? A Raspberry Pi with a u-blox module, cobbled together six years ago. It was achieving around 30ms accuracy — which was fine when the work was simpler.

It wasn't fine anymore.

Modern sensor fusion doesn't forgive millisecond-level timing. When you're synchronising multiple LiDARs, cameras, radars, and a V2X stack simultaneously — across a vehicle doing highway speeds — the timestamps need to be right to within nanoseconds. A 30ms gap between sensor readings isn't a calibration issue. It's a fundamental problem for anything the system tries to perceive or decide.

So they needed to replace it. And they needed to do it twice — once for a mobile trailer that deploys into the field with no fixed power, and once for an autonomous car where the unit lives inside the vehicle and feeds the sensor fusion pipeline directly.

The problem with off-the-shelf

Most precision timing hardware is designed for a rack in a server room. Fixed power. Stable environment. A known location.

None of that applies here.

The mobile trailer needs a battery-powered unit that can be driven to a field site, lock to GNSS, and distribute PTP to whatever sensors and edge compute are running alongside it. The autonomous car needs something hardened for an in-vehicle environment — barrel jack power, capable of feeding sub-30ns sync to LiDAR, camera, radar, and V2X simultaneously, while the car is moving.

Different environments. Different power. Different connectivity. Same accuracy requirement.

Timebeat deployed the OpenTime Appliance Mini — with battery-powered for the trailer, and  standard for the vehicle. Both deliver GNSS-disciplined PTP at sub-30ns accuracy. No ongoing licensing. No custom firmware. Hardware purchase is sufficient.

The roadside problem nobody had budgeted for

Then came the harder question: the highway corridor has 15–20 roadside units that also need standalone timing sources.

The existing infrastructure cost €20–25k per unit to build — a figure that made scaling precision timing across the full corridor commercially unviable for a research-funded programme.

Timebeat went back to the drawing board on form factor. The answer: OpenTime Card Mini 3.0 on a CM5 carrier board, in a compact custom enclosure — small enough to fit inside existing roadside boxes, and accurate enough to meet the same sub-30ns requirement as the vehicle-side deployment. A reference design that makes corridor-scale timing viable without rebuilding the budget around it.

One more thing the team needed

The autonomous cars run gPTP — IEEE 802.1AS, the automotive standard. The roadside infrastructure runs PTPv2.1. Normally, bridging those two protocols requires a dedicated automotive switch to handle the translation.

Timebeat confirmed simultaneous dual-protocol broadcast on separate interfaces. PTP on one, gPTP on the other. The switch is no longer needed.

For a research programme running both automotive and infrastructure timing domains simultaneously, that's one less integration problem.

Where things stand

The roadside reference design is under engineering evaluation. Full deployment results to follow — this case study will be updated once the hardware is in the field.

If you're working on AV timing infrastructure and running into similar constraints — mobile deployment, mixed protocol environments, or roadside timing at scale — we're happy to talk through the architecture.

Products Used

• OpenTime Appliance Mini — Battery version (mobile/trailer deployment)

• OpenTime Appliance Mini — Standard version (in-vehicle deployment)

• OpenTime Card Mini 3.0 — Roadside unit reference design

Sector: Autonomous Vehicles / Academic Research

Use Case: V2X synchronisation, sensor fusion, distributed roadside PTP infrastructure

Location: Portugal

timebeat.app   |   Precision Timing Infrastructure   |   PTP · GNSS · White Rabbit