Embedded Precision Timing

An introduction to sub-nanosecond network time synchronization

The need for synchronized computer time in network-connected devices is increasing. How can embedded applications benefit from this and what will be the next steps? This is what lies ahead. 

Time is a driving concept in our lives. Day-night cycles, the way the earth circles the sun and how the moon rotates around the earth, all at a very stable and repetitive pace. With current-day atomic clocks we can monitor the passing of days and years with nearly infinite resolution and accuracy. We even adjust our calendars and watches so that we keep most of our systems synchronized with these clocks. Most devices are now connected to a network, allowing automation of the clock synchronization. But what benefits does it bring to have more accurate clocks, and can it be used to solve specific time-related challenges? 

Precision Timing and how it evolved
Nowadays, virtually all devices have a notion of time. Most of them incorporate a so-called real-time clock. In computer networks, individual timekeeping is synchronized regularly and automatically using for instance an internet-connected NTP server (Network Time Protocol). With NTP, milli-seconds accuracy can be achieved. However, an increasing number of systems require more accuracy, expressing the synchronization resolution in pico-second (~10-12 seconds). Achieving this level of accuracy is quite a challenge, especially when computer systems collaborate in a networked environment. This set-up combined with individual clock accuracy requirements that transcend micro-seconds (< 10-6 ) is called Precision Timing.

From railway clocks to GPS 
The problem of synchronizing individual clocks is a century-old problem. The analogue clock from the old days required winding up every week and synchronizing it with the help of a reference clock like the local church clock: the internal clock source was not infinitely accurate. The 3,500 clocks installed at railway stations in the Netherlands are an illustration of how different technologies behind clock synchronization evolved over the years. All clocks are synchronized using a 1 second time pulse, broadcasted by a radio transmitter located in Germany. However, radio signals travel at the speed of light. This means that a clock 600 km away from the transmitter runs approximately 1 msec late compared to a clock just 300 km away. For systems operating with higher accuracies, more recent technologies may use time-synchronized GPS (global positioning system). With GPS, multiple satellites circle the earth at a fixed position far away, all transmitting accurately timestamped messages to the receiver, arriving with a slight time difference due to the receivers position on earth. Synchronizing satellites with sub-nanosecond resolution to pinpoint accurate localization is a characteristic Precision Timing challenge.

Synchronizing issues & how to benefit from Precision Timing 
Traditionally, systems are synchronized by distributing a triggering or sync signal: systems are wired together, and an electrical pulse is distributed to initiate the operation. The trigger ensures a lasting synchronization of the systems. The problem is that cables and PCB traces may vary in length. Also, the trigger signal may be susceptible to noise and, perhaps the largest problem: the propagation delay varies per signal. When surpassing clock rates of 100MHz, this becomes a problem as differences in trace lengths can cause noticeable sampling errors. 

• Increasing accuracy 
  By using Precision Timing via a network connection, the clocks of the system are synchronized within the desired accuracy range. Instead of having to wire a trigger physically, the trigger moment is programmed as a timer event relative to the real-time time clock. When all clocks in the system are in sync, all triggers are issued at the same time. 

• Simplified board design
  In addition to increasing accuracy, simplified board design and wiring interconnections are yet another benefit. Precision Timing is specified by the IEEE 1588 standard. Parts of this standard are better known as PTP (Precision Time Protocol), SyncE and White Rabbit. Instead of implementing these concepts yourself, the standard can be adopted using either Open Source or closed-source protocol stacks for applications. However, Precision Timing requires specific electrical circuitry to achieve the desired clock accuracy. For instance, TOPIC has implemented on its System-on-Modules various circuits enabling synchronized clock operations with an accuracy higher than 1 nano-second. 

Applying Precision Timing 
Applications requiring Precision Timing solutions are abundant. It is mainly applied in larger-scale systems, but also less complicated systems must deal with similar timing restrictions. A famous example is the grand Hadron collider at CERN; the feasibility of experiments depends on time synchronization accuracy of individual applied sensors and actuators within the system. Another example is the square kilometer array (SKA) radio astronomy telescope currently deployed in Australia and South Africa. The Australian setup consists of approximately 130,000 antennas. Each antenna is sampled at a rate higher than 1Gsps and with suitable accuracy time stamped. By correlating and processing the antenna signals adequately, an image of the universe can be created. 

TOPIC and Precision Timing 
Driven by silicon technology developments, the number of customer applications requiring Precision Timing solutions has increased rapidly. Systems are better connected in a collaborating network where physical distances interfere with the critical sampling rates and hit the lightspeed barrier as a limitation. At TOPIC, we observe that customers experience these limitations while developing applications in the domains of photonic and quantum computing, high-performance test-and-measurement equipment and communication networks like 5G. In audio and video applications, you will find “lightweight” implementations of precision timing, such as AVoIP. 

Therefore, TOPIC adopted White Rabbit (WR) as the leading Precision Timing protocol over fiber-optic Ethernet infrastructure. We have also developed a low-cost stand-alone WR module and are studying a method to turn standard Ethernet switches into WR-compliant switches. In addition, TOPIC enhanced their Miami family of system-on-modules with Precision Timing capabilities. This provides embedded Precision Timing at no additional cost but with all the features to achieve sub-nanosecond timing accuracy. 

Our experience with developing applications using embedded precision timing will be published in a technical white paper early March 2024. Would you like to get more insights in this topic and be one of the first to receive the whitepaper? Feel free to contact dirk.van.den.heuvel@topic.nl.