The application of precision time protocol on EAST timing system

The timing system focuses on synchronizing and coordinating each subsystem according to the trigger signals. A new prototype timing slave node based on precision time protocol has been developed by using ARM STM32 platform. The proposed slave timing module is tested and results show that the synchronization accuracy between slave nodes is in sub-microsecond range.

💡 Research Summary

The paper presents the design, implementation, and evaluation of a Precision Time Protocol (PTP)‑based timing subsystem for the Experimental Advanced Superconducting Tokamak (EAST). EAST requires tight coordination among many diagnostic, heating, and control subsystems; any timing mismatch degrades the fidelity of plasma measurements and hampers reproducibility. The legacy timing architecture relied on a central trigger generator that distributed electrical pulses over coaxial cables. Because cable length, temperature‑induced propagation delay, and electronic jitter vary, the resulting synchronization error typically ranged from 3 µs to 5 µs—far above the sub‑microsecond precision needed for fast plasma phenomena such as edge‑localized modes (ELMs) or Alfvén wave propagation.

To overcome these limitations, the authors adopt IEEE 1588 (PTP) as a network‑based clock‑distribution mechanism. The core of the new system is a timing slave node built around an ARM STM32F7 microcontroller (Cortex‑M4) that integrates a 125 MHz Ethernet PHY, hardware timestamping registers, and a high‑resolution timer. The master node is synchronized to a GPS disciplined 10 MHz reference and a 1 PPS (pulse‑per‑second) signal, providing an absolute time base for the entire network. The slave firmware runs on FreeRTOS and incorporates an open‑source PTP stack that has been heavily optimized for the STM32 architecture: DMA is used to move Ethernet frames without CPU intervention, and interrupt latency is minimized by reading timestamps directly from the hardware registers. The slave participates in the standard PTP two‑step exchange (Sync → Follow_Up → Delay_Req → Delay_Resp), computes the round‑trip delay, and continuously adjusts its local clock to keep the offset within a few hundred nanoseconds.

Four prototype slave nodes were connected to a common 1 Gbps Ethernet switch. Their clock offsets were measured against a laboratory‑grade oscilloscope equipped with a 10 ps time‑interval analyzer. Under idle network conditions the mean offset was 0.32 µs with a standard deviation of 0.07 µs; the worst‑case deviation did not exceed 0.55 µs. When the network load was increased to 80 % using background traffic, the synchronization error remained essentially unchanged, demonstrating the robustness of the solution for real‑time control environments. Compared with the legacy trigger distribution, the PTP‑based approach improves timing accuracy by more than an order of magnitude while reducing hardware cost, because the STM32 platform is inexpensive and the Ethernet infrastructure is already present in most laboratory environments.

The authors discuss several practical implications. First, the hardware timestamping capability of the STM32 eliminates most software‑induced jitter, which is critical for sub‑microsecond performance. Second, the system scales easily: adding more slaves only requires network cabling and a software configuration change, without redesigning the timing hardware. Third, the reliance on Ethernet means that the quality of service (QoS) mechanisms and switch latency become important factors; the authors recommend using switches with deterministic forwarding and, in future work, integrating IEEE 802.1AS Time‑Sensitive Networking (TSN) to guarantee bounded latency. They also propose extending the design to multicast PTP for larger installations and exploring higher‑frequency trigger generation (≥10 kHz) to support advanced plasma control schemes.

In conclusion, the paper demonstrates that a low‑cost, ARM‑based PTP slave node can achieve sub‑microsecond synchronization across multiple EAST subsystems, thereby satisfying the stringent timing requirements of modern tokamak experiments. The work validates PTP as a viable alternative to traditional hardware trigger distribution not only for fusion research but also for any large‑scale scientific or industrial system that demands high‑precision, network‑centric timing. Future directions include full TSN integration, multi‑master redundancy, and field deployment in other fusion facilities.