Ethernet is rapidly becoming the standard for networks on utility and industrial sites, both for monitoring and automation. Previously Ethernet was considered non-deterministic, however with the development of protocols such as traffic prioritization and VLANs, as well as functions within managed network devices such as broadcast throttling it is slowly changing to a more deterministic technology (With correct configuration of the networking switches and routers). PTP (Precision Time Protocol) or IEEE1588 is a relatively new protocol that allows network devices to be synchronised to within 1 µs (1 microsecond = 1 millionth of a second). PTPv1 was first introduced in 2002, and version 2 was released in 2008.
PTPv1 had limitations as it did not cater for switch latency (The time a packet spends within a switches switching fabric, as it transfers from port to port) as well as assuming that a link between two switches has the exact same latency on the upstream as it does on the downstream. This is not necessarily true on a full duplex network. Because of these limitations PTPv1 could not scale well to large networks, as every new switch would increase the inaccuracy in the synchronisation. The actual maximum size of PTPv1 networks whilst keeping true synchronicity would very much depend on the implementation of PTP within the hardware, and so could differ between different vendors. Finally PTPv1 had a shortest sync interval of 1 second (Meaning it would only check it synchronicity every second.) This sounds like a short time, however when the standard calls for an accuracy of at least 1 millionth of a second, this synch interval becomes too infrequent.
PTPv2 was developed to eliminate the inaccuracies inherent within PTPv1. The biggest change in PTPv2 was the inclusion of a “transparent clock” mode of operation for PTP clocks. Transparent mode means that a switch will update the PTP update packets with the switch latency, meaning that synchronicity will be maintained throughout the network no matter its size, as long as all switches are PTPv2 compatible. PTPv2 also changed the assumption that a link’s latency on the upstream is the same as the latency on the downstream, and so would test each individually. Finally PTPv2 allows a shortest synch interval of 125 ms, almost 10 time more frequent than PTPv1, translating into more accurate synchronization, which is extremely important on critical networks.
An important note to keep in mind is the fact that PTPv2 is NOT backwards compatible with PTPv1. This means that any current PTPv1 switches in a network will, for all intents and purposes, act as non-PTP switches for the PTPv2 network. Any non-PTP switch in a PTPv2 network will slowly decrease the synchronicity. The reason for this is that a PTPv2 switch will not be able to determine the path latency between itself and the non-PTP switch. Also, the non-PTP switches will not update any of the PTPv2 switches on the switch latency they are adding to the system. This means that having even one non-PTP switch can potentially cause synchronization to not meet the standard for 1µs synchronicity. This will depend on the network in question, including traffic amounts, configuration etc. however it is recommended to upgrade ALL critical parts of the network that require the refined time stamping to PTPv2.
With the increase of accuracy and scalability brought about by using a combination of VLANs, prioritization and PTPv2, the feasibility of using it to control time synchronization on highly time sensitive networks is becoming greater. In fact there are plans to include time synchronization as part of the IEEE-61850-3 standard for substations. Having this kind of time synchronization is highly valuable when taking readings of synchophasors across the network, and for combining the synchophasor readings into a true representation of the power grid. Highly accurate time synchronization is also important in protection when needing to trip the correct switchgear at the correct time.
In the industrial world time synchronization can also be critical in certain applications. For instance seismic sensors need to have extremely high synchronization in order to provide a proper overview in the case of seismic activity and allowing much more accurate tracking of the epicenters of any seismic events.
Time synchronization of this degree can also help when troubleshooting networks, as all devices logs will be synchronized, making it much easier to accurately trace a problem around a network.
A basic PTP network works on a Master-Slave principle. A master device will be one which is connected to a highly accurate time source, such as a GPS. Between the master and his directly connected slave devices they will calculate the path delay of the links between them (As will all PTP devices in the network). The master then starts sending out PTP packets to his neighbors. These packets will allow the neighbors to calculate their time difference compared to the master and update themselves. They will also then update the PTP packets with their latency and the relevant path latencies and will pass them to the next switch in line and so on. As stated with PTPv2 the interval between these updates can be set to as low as 125ms and so this update process will be happening 8 times every second! This provides the extremely accurate time synchronization called for by the standard.
PTP routers can also be used to act as boundary clocks. These will allow time synchronization to happen between different networks. Compared to PTPv1, which was extremely limited in the size of networks it could handle, PTPv2 can now be implemented on almost any size LAN, provided the LAN is configured correctly and the correct hardware is implemented.
Upgrading to PTP can be quite costly depending on the current network equipment in place and the required parts of the network that require time synchronization (As every switch on a time sensitive portion of the network must be upgraded to handle PTP, as well as every switch in the path to the master clock, otherwise time synchronization will slowly slip due to latencies caused by the non-PTP devices). For this reason planning for the future is crucial, especially for utility networks. PTP can be implemented over a period of time, however true synchronicity across an entire network will only be achieved once all switches are PTPv2 compatible.
PTPv2 is becoming more and more prevalent throughout both the industrial and utility sectors, and can provide many benefits in both environments. However, planning must be done taking into account factors such as budget and deadlines on when synchronicity will be required. Upgrading to PTP would give the desired synchronicity on a network, and will cater for many other applications. Please contact H3iSquared for more details on RuggedCom and their industrial Ethernet range, including PTP.
Tim Craven
H3iSquared Trading
Tel: +27(0)11 454 6025
Email: info@h3isquared.com