The International Timing and Sync Forum, aka ITSF, is always one of the event highlights of the year for me. It’s a well-organized event with a great community and a very focused agenda with deep discussions, and it’s always held in great locations. This year the event took place in Seville, Spain in early November. ITSF has a very strong conference element, with a full three-day program with over 80 presenters and more than 30 posters in two poster sessions. The conference covers many aspects of timing and sync, but there are often trends that pop up and reflect the issues of the day. This year coherent network primary reference time clock (cnPRTC) and the progress toward standardization of the concept was one of the recurring themes of the event.
The term cnPRTC can be slightly confusing for those of us from the optical networking industry the first time we hear it, as the term “coherent” in this context is totally different than coherent in coherent optics. Here cnPRTC refers to the concept of connecting, or federating, multiple independent timing domains into a single larger resilient timing domain with very high performance. This concept is still working its way through the standardization process at the ITU, so it is a little way off in terms of commercial deployment, but various aspects of cnPRTC were discussed at ITSF 2024. We’ll come back to cnPRTC a little later in this blog.
Introducing Point-to-Multipoint Optics into the Sync/Timing Network
Over the last three years, I have done a series of presentations at the event that outlined the challenges of transporting sync/timing over DWDM-based optical transport networks and the techniques that we use to overcome them. This year my presentation addressed the new concept of point-to-multipoint XR optics and the impact it has on timing and sync distribution. XR-based networks are very interesting from a sync point of view as they break the 50-year-old one-to-one point-to-point relationship between optics, as we’ve discussed many times here on the Infinera blog. This means that we need to treat them very differently than conventional optics from a sync point of view. A single higher-speed hub optic now transmits data to/from multiple lower-speed optics and may also need to run multiple parallel SyncE frequency sync sessions and multiple parallel 1588v2 phase sync sessions with an embedded T-BC boundary clock function. Earlier in the year, we worked with DZS and Calnex to jointly demonstrate this point-to-multipoint environment with Infinera ICE-X intelligent coherent (in the optical networking sense) pluggables mounted directly in third-party DZS devices. My presentation introduced the concept of point-to-multipoint optics, discussed the sync implications, and gave a high-level overview of the results of the joint testing. It was a lot to squeeze into my allocated time slot.
ITSF draws together the timing/sync community from a broad range of organizations. On the supplier side, there are specialist sync companies, sync component vendors, and sync transport vendors such as Infinera or those from the satellite communications world. The end users attending the conference also show the breath of industries that rely on reliable and secure timing. The event drew attendees from telecommunications network operators, government bodies, research and education networks, power utilities, data center operators, and many more. This reflects the traction we have with our OTC2.0 timing channel solution, where the majority of customers use the solution for robust high-precision timing distribution for 5G, but we are seeing increasing momentum in many of these other industry sectors.
A Deeper Look at cnPRTC
Going back to the cnPRTC topic, my colleague Thomas Hiestand had a poster selected for the conference poster sessions that outlined a proof-of-concept demonstration that the team created using existing Infinera GX Series OTC2.0 hardware for the high-accuracy timing transfer component of a wider cnPRTC implementation. cnPRTC will involve many key components, but one of them will be timing transport/transfer with sub 5-ns, or even possibly sub 1-ns, timing error over hundreds of kilometers over a fiber network. To put this into perspective, the OTC2.0-based timing cloud architecture that Infinera deploys is architected to deliver timing services around a national or regional network with less than 100-ns timing error from the network’s PRTC reference clocks. This target matches the 100-ns PRTC timing error budget allowed within the G.8271.1 standard, which gives the complete timing budget from PRTC to end device, which is typically a cell tower in a 4G/5G network, including all intermediate transport components – switches, routers, 4G/5G hardware, and the optical network. This 100-ns performance for the transport network effectively makes the whole network act like a timing cloud or virtual PTRC. That said, in reality, deployments are typically achieving timing error in the range of 20-30 ns, so substantially better than our initial goal.
Each timing cloud will be tightly synchronized to less than 100 ns of the primary reference time clock (PRTC), and in practice often in the 20-30 ns range, as I mentioned. But each PRTC initially derives its timing from a highly resilient spoofing-/jamming-proof global navigation satellite system (GNSS) receiver such as GPS or Galileo and has a high-performance local clock, such as a Cesium clock, for holdover during GNSS interruptions. These PRTC clocks and timing domains all differ slightly and may only be within 60-70 ns of the true time due to the allowed budget within that part of the timing chain. cnPRTC will connect multiple timing domains, or national/regional timing clouds in the Infinera case, together to create a larger timing domain with improved accuracy and greatly improved resilience. To achieve this goal, getting down from 100 ns, or 20-30 ns, down to the required low single digits or even below 1 ns for timing/sync transfer over optical transport networks is going to be a significant step forward.
Thomas and his colleagues back in our Munich timing lab ran the demo outlined in the poster remotely from our booth in Seville, so we were able to show attendees the demo live. The proof-of-concept demo uses existing GX hardware with some firmware tweaks to connect two Microchip TimeProvider 4500 (TP4500) enhanced PRTC (ePRTC) nodes over 300 km with approximately 1 ns of constant time error (cTE) and 3 ns of dynamic time error (dTE). It was very well received by the audience as it hits that initial sub 5-ns target already. Extrapolation modeling shows that it should be possible to extend this reach to 600-700 km within the 5-ns target with an intermediate TP4500 node. The wider cnPRTC concept will require many innovation steps to come together, but a key element will be the high-accuracy timing transfer that the demonstration showed is possible to achieve.
Wrapping It All Up…
Going back to the blog title, the blink of an eye lasts between 100 and 400 milliseconds, so you’d already miss the sub 100-nanosecond performance levels that we are achieving in today’s networks by a factor of 1,000,000. Future cnPRTC deployments will drive the high-accuracy timing transfer requirements by another factor of 100. When I write a blog about the first deployment of these networks, then I’ll have to think of a better analogy. Even so, ITSF 2024 was another great event, and Thomas and I look forward to updating everyone on this proof of concept in Prague next year for ITSF 2025!
