Time Synchronization in Measurement Technology From NTP to PTP and the Growing Role of Time-Based Measurements

Classic, star-shaped measurement systems reach their temporal limits with distributed topologies. Protocols like IEEE 1588 or "White Rabbit" allow distributed measurement data to be synchronized down to the picosecond range today. A precision that is essential for building digital twins and training AI models.

Classical test and measurement systems were typically centralized and star-shaped in the past. A central control computer directed the measurement process, managed the measuring instruments as well as the test object (Device Under Test, DUT), and executed the individual test steps. A real temporal connection between the spatially distributed measurement points was rarely present in these architectures. To sequence individual test steps, simple trigger lines were used, which, however, did not have to meet strict time-related requirements (Figure 1).

This fundamental structure proved itself over decades. However, as early as 1990, the T&M industry recognized the growing need for decentralized systems with precise temporal references. While GPS-based systems offering high accuracy were already available at that time, they were simply too expensive for widespread T&M applications. The goal was therefore to achieve the required time synchronization using the existing network infrastructure.

From the NTP Standard to the Precision Time Protocol (IEEE 1588)

The initial spark occurred in the early 1990s when Hewlett-Packard formed a working group led by John C. Eidson. The goal was to develop a concept for significantly improving the previously used Network Time Protocol (NTP). The results, published in 2000, generated such great interest in the measurement technology industry that it was quickly decided to establish a dedicated IEEE standard for this new type of hardware-based time synchronization.

In 2002, the Precision Time Protocol (PTP) was finally adopted as IEEE 1588-2002. For the first time, this standard enabled synchronization accuracies in the sub-microsecond range over standard networks. As technical requirements rapidly increased, the protocol was extensively updated in 2008 (IEEE 1588-2008). The most recent revision, IEEE 1588-2019 (PTP v2.1), introduced significant improvements for modern industrial and measurement networks:

• Significant increase in time synchronization and accuracy.
• Scalability: From localized subsystems to distributed large-scale facilities.
• Largely administration-free operation (plug-and-play capability).
• Flexible use in both high-performance high-end and resource-constrained low-end devices.
• Extended management for redundant and fail-safe systems.

With the implementation of the IEEE-1588-2019 standard, the foundation was laid to reliably determine the exact timing of a measurement in distributed systems. In a standard Ethernet network, PTP enables typical accuracies of ±20 nanoseconds. Modern, hardware-optimized systems even achieve performance in the single-digit nanosecond range (Image 2).

Initial Use Cases: From Power Plants to Particle Accelerators

The first practical use cases for this highly precise time synchronization were found in large industries, such as nuclear power plants. Here, all decentrally collected measurement data was tagged with PTP timestamps, centrally stored in a database, and then correlated and analyzed offline based on their timestamps.

Even more extreme synchronization requirements were posed by CERN for the Large Hadron Collider (LHC). The accelerator ring has a circumference of nearly 27 kilometers (approx. 16.8 miles). Considering physical limits, a signal needs up to 135 µs for this distance with an idealized signal propagation time of 5 ns per meter of cable. The measurement errors resulting from spatial latencies via standard PTP were unacceptable for particle physics.

White Rabbit: PTP in the Picosecond Range

To overcome this physical limit, CERN independently extended the IEEE 1588 standard. The results were published as the so-called White Rabbit standard under the CERN Open Hardware License. Today, White Rabbit is not only used in the control of the LHC but also in other large-scale scientific projects such as the KM3NeT neutrino telescope in the Mediterranean or the FAIR particle accelerator at the GSI Helmholtz Center for Heavy Ion Research in Darmstadt. The key advantage: the synchronization accuracy of the White Rabbit implementation is in the range of ±200 picoseconds. This makes the system approximately 100 times more accurate than the classical IEEE 1588 protocol over standard Ethernet.

Where Nanoseconds are Crucial

The establishment of the PTP standard has created entirely new possibilities for testing and measurement technology today. Fundamentally, four central scenarios can be identified that rely on highly precise, time-based measurement data acquisition:

• Generation of data for digital twins and hardware-in-the-loop simulations (HiL),
• Database for training AI models,
• Distributed measurements over large spatial distances and
• Time-synchronous triggering of actuators and test sequences.

A particularly significant role in modern T&M systems is played by the Digital Twin. This concept, coined in the early 2000s by Michael Grieves and John Vickers, describes the virtual representation of a real process. The key to a functioning Digital Twin is the data and information connection between the physical and virtual worlds.

In measurement technology, this means: The real collected data must be tagged with extremely precise timestamps (correlated) to make it usable. Only when the temporal causality of all sensor signals is absolutely deterministically recorded does the subsequent simulation model behave exactly like its physical counterpart.

Practical Example: Engine Test Benches and Welding Robots

A classic application can be found in modern engine test benches in the automotive sector. Here, thousands of measurements from various sensors (pressure, temperature, torque, or CAN bus signals) are recorded per second. If all these distributed measurements are provided with a synchronous timestamp via PTP, a highly precise, behavior-accurate simulation model (the Digital Twin of the engine) can be generated. Based on this model, developers can drastically reduce the number of time- and cost-intensive tests on the real test bench and instead conduct virtual tests at their desks.

Another concrete example is the optimization of welding processes in automotive production, as realized in collaboration with research institutions. Distributed measurement systems simultaneously record parameters such as voltage, current, electrode distance, wire feed, and temperature during welding. The correlated temporal analysis of this raw data precisely reveals how, for example, a minimal voltage drop with a delay of fractions of a second impacts the welding temperature.

From Simulation to AI Model

If the principle of the deterministic digital twin is consistently pursued, it inevitably leads to the use of artificial intelligence. Machine learning algorithms and AI models require clean, time-correlated datasets for their training. These are the so-called ground truth data.

The welding example mentioned above illustrates this potential: By feeding a AI model with timestamped process data from hundreds of thousands of real welding processes, including their qualitative evaluation, the AI learns the perfect process parameters. These trained models can then be directly applied to the controllers of industrial robots to monitor the welding process in the production line in real time, adaptively regulate it, and thus maximize quality.

The Software is Meant to Master Complexity

As the use cases show, the configuration effort for time-based, distributed measurement systems is significantly higher than for classic, centralized architectures. To operate these systems efficiently, a high level of automation at the software level is essential.

The first hurdle for the test engineer already lies in the fundamental topology: When using a time-based system, the hierarchy of clock distribution must be defined. Who functions as the "Grandmaster Clock"? How is the time signal physically and logically distributed through switches and routers to the "Slave Clocks" of the individual measurement nodes?

Additionally, PTP systems exhibit a physical peculiarity: they require a certain settling time after startup. The software must therefore ensure that all internal clocks are fully synchronized before the actual measurement sequence begins. During operation, it is also essential to ensure that all recorded data is seamlessly correlated with the corresponding timestamps and stored without errors.

Continuous Monitoring of Clock Quality

To generate reliable simulation or AI data, the quality of synchronization must never be blindly assumed. Modern measurement software must continuously monitor and evaluate the quality of time synchronization (sync status, offset, jitter) in the background (Figure 3). This is the only way to ensure that temporary network overloads or synchronization losses do not unnoticedly distort the measurement data. In case of an error, the system must stop the measurement or mark the corresponding data packets as invalid.

Visual Support for the Test Engineer

It is obvious that test engineers must be relieved of handling these profound network and time parameters through software support. Instead of cryptic command lines or complex configuration files, a graphical visualization (GUI) of the entire system topology is indispensable today.

A modern T&M software interface (as shown in Figure 4) enables the user to intuitively define the distribution of clocks, the topology of network nodes, as well as the physical measurement variables and their parameters in one place. In the background, the software handles the heavy lifting: it automatically configures the corresponding components, loads the required calibration data, aligns the parameters, and monitors the settling phase of the clocks. Only when the system confirms the lock status of all clocks does the software give the green light for precise data acquisition.

Latencies and Calibration as Pitfalls of Physics

Even if the network is synchronized to the picosecond range via PTP or White Rabbit, another challenge awaits at the hardware level: analog physics. In time-based measurements, signal propagation times in cables and the internal processing times of sensors (until a measurement value is ready at the output) play a crucial role.

Here too, the test engineer relies on the support of the system software. The signal delay of the sensors can often be obtained from manufacturer datasheets or determined empirically and stored as a fixed offset in the system. It becomes more complex with the wiring: if sensors or external measurement devices are connected via long cables (e.g., coaxial or trigger cables), the exact propagation times of the electrical signals must be determined.

Modern systems largely automate this step. For precise determination of cable propagation times, D-DMTD algorithms (Digital Dual Mixer Time Difference) are used, for example, which detect phase shifts in the sub-nanosecond range. These measured propagation times, combined with the internal latencies of the measurement cards (which must be provided by the system manufacturer), are then automatically calculated by the software as correction factors into the timestamps of the measurement data.

A Conclusion

Time-based test and measurement systems open the door to entirely new application fields. They form the fundamental backbone for creating highly precise digital twins, hardware-related simulations (HiL), and generating reliable data for training AI models.

Although the underlying network architecture and synchronization (via IEEE 1588 or White Rabbit) are technologically far more complex than in traditional centralized measurement setups, modern software frameworks make the handling just as comfortable for the test engineer today. These frameworks graphically visualize and largely automate complex tasks such as clock distribution, network monitoring, and latency calibration.

Currently, the T&M market is still heavily dominated by classic architectures. However, as the industry increasingly demands deterministic, distributed data for AI and simulation purposes, the seamless integration of time-based measurement functions will become a decisive criterion in the selection of future-proof test and measurement systems. 

*Peter Plazotta is the CEO and founder of Technical Software Engineering Plazo