5G Network Precision Timing

Advanced 5G use cases require more stringent latency requirements across the network from the central office to the edge. More precise clock synchronization can help solve this challenge. The Intel® Ethernet 700 Series Network Adapter with Hardware-Enhanced Precision Time Protocol (PTP) brings higher accuracy IEEE 1588 PTP synchronization signaling to the Ethernet and onto the Edge platform with lower TCO than existing hardware appliance-based approaches. This video provides a demonstration of Intel Ethernet 700 Series Hardware-Enhanced PTP capabilities for timing-critical applications.


Hello, I'm here to talk with you today about the Intel XXV710-DA2T network adapter that Intel is announcing. This is very similar to the existing hardware channel, XXV710-DA2 network adapter, but it adds some additional connectors on the front. This card, which we call Edgewater Channel, is specifically designed to provide additional timing capabilities to critical industries like telecom, financial markets, industrial, and energy distribution markets.

This adapter has two extra coaxial connectors on the front of it. And these coaxial connectors are designed to connect up to external timing devices, such as GPS devices or other timing-sync solutions. And you can use the second adapter to either audit the output from an individual NIC. Or you can actually use this to cascade solutions to other devices, if you want.

This network adapter has specific capabilities. We have an oscillator onboard this NIC that is about 5,000 times greater stability than the typical oscillator that is on a regular NIC. And together, this allows us to provide a timing solution that can give you the capability to connect up a GPS, or other timing solution, and provide timing across the network.

In this example demo configuration we have, we have two servers that are set up, each with an Edgewater Channel XXV710-DA2T card. And we are connecting them via an optical cable at 25 gigabits between these two. We also have SMA cables hooked up to the SMA connectors on here, so we can get one-pulse-per-second signals. And we are synchronizing these two servers across the network and then auditing the results through the SMA cable to see how well-aligned they are. When we look at these outputs on an oscilloscope, you can see that the output is very well-aligned.

In this demo, we are showing the output of the Linux standard PPP 4L commands that are outputting their synchronization accuracy at one-second intervals. In this configuration, we are doing 16 updates per second, so this is putting out the root mean square, the RMS value, the maximum error that it saw over those 16 values, and then some additional information on frequency and the delay, or timing, between the two servers.

And we are doing these synchronizations on an ongoing basis every second. And you can see on screen here, where we are showing that we have an error in this particular configuration that is in the low-nanosecond range. And if I were to go and look at the oscilloscope, you can see that the one-pulse-per-second output on the oscilloscope is very tightly-aligned as well.

Now if I go into this system, and if I stop the system that is aligning with the other, it will no longer-- it will no longer be providing updates over here. And if you look at the oscilloscope, you can see that the oscilloscope is drifting because we are no longer synchronized. And on that screen, the bottom portion is the zoom-in of the upper portion. You can see that it is beginning to drift slowly and slowly out of there.

But even now, it's only about 50 nanoseconds out, when we have no synchronization going. If I go ahead and restart the synchronization and reset the statistics, you'll see that the synchronization will immediately come down to a few nanoseconds of time. And you can see onscreen how we have synchronized a one-pulse-per-second output.

As long as we are able to maintain our network synchronization, we are tightly-synchronized. And even when we lose synchronization, it very slowly drifts apart. It may look like a lot on the oscilloscope, but compared to a standard NIC, where this synchronization will go immediately into the microsecond range, this provides much greater stability.

Again, to stop the synchronization-- I will stop it. And then I will restart it. And you'll see onscreen that we're getting errors, and we'll see coalesce again on the scope.

As you've seen in this demo, the Intel Ethernet Network Adapter XXV710-DA2T provides very good synchronization capability. Not only does it have an oscillator onboard that's 5,000 times greater accuracy, it also provides additional connectors where each provide synchronization inputs or outputs that can be hooked up to a GPS or other timing-synchronization device or used to audit the performance of each node in the system. This is useful for customers like telco operators who want the ability to be able to get close to 100 nanoseconds' phase accuracy across their network in order to provide tighter locality for emergency services and other navigation-based applications using the capabilities of their advanced 5G networks.