TimeSyncJapan
2026

GNSS is widely used as a fundamental technology for high-precision time synchronization in telecommunications, broadcasting, finance, power systems, and many other applications. However, GNSS performance can be affected by multipath propagation, ionospheric delay, and, more recently, intentional threats such as jamming and spoofing. This presentation reviews the principles and vulnerabilities of GNSS-based time synchronization and discusses approaches to improving the resilience of timing systems centered on GNSS receivers. In particular, we introduce a multi-source timing architecture that combines GNSS with complementary and alternative technologies such as PTP, LEO-PNT, and OCXO-based holdover. We also examine the functional and performance requirements of GNSS Disciplined Oscillators (GNSS-DOs), which integrate GNSS receivers with OCXOs. Furthermore, we present evaluation results obtained from the annual Jammertest conducted in Norway, demonstrating GNSS receiver behavior under interference conditions and discussing future directions for improving the resilience of time synchronization systems.

This presentation introduces the time-related features of ISDB-T3, a new digital terrestrial television broadcasting system standardized by the Association of Radio Industries and Businesses (ARIB) in 2025. In addition, it briefly provides the technical overview particularly in countries transitioning to ATSC 3.0.

Modern AI infrastructure coordinates thousands of GPUs, CPUs, NICs, switches, storage devices, and sensors, yet each device maintains its own imperfect notion of time. This becomes a fundamental problem when events observed on different machines must be compared: clock offset and uncertainty can be as large as, or larger than, the latency being diagnosed. This talk discusses why precision time is becoming an important infrastructure capability for AI systems. Near-term applications include cross-host trace alignment, one-way and per-hop latency measurement, and correlation of compute and network events for faster root-cause analysis. Emerging applications include time-aware traffic pacing and coordinated operations in large AI clusters, while physical AI systems require accurate capture-time alignment across cameras, LiDAR, and other sensors. A key requirement is not simply a precise timestamp, but a trustworthy bound on its error, together with information about clock health and provenance. The talk will introduce the Open Compute Project Time Appliances Project (OCP-TAP) and its open ecosystem spanning Time Cards, Open Time Server, the Data Center PTP Profile, Precise Time APIs, and timing validation tools. Together, these technologies provide a path toward an open, observable, and resilient clock plane for future AI infrastructure.