Using the IEEE1588 protocol for precision clock synchronisation.
Motion control and many other distributed systems in automation technology depend on precise clock synchronisation for correct operation. With most automation systems moving over to Ethernet as a communication medium, and the fact that Ethernet is not a deterministic medium, a protocol which defines the mechanism for precise time synchronisation becomes essential. It is also important that any such protocol places minimum strain on the network, as well as the CPUs of any equipment involved.
Traditional solutions used in enterprise networks (Network Time Protocol and the Simple Network Time Protocol) do not meet the stringent demands of control systems (though they offer millisecond range synchronisation), even more so when it comes to processes such as motion control.
The IEEE1588 Precision Time Protocol (PTP), first proposed by Agilent Technologies, is designed to address the requirements for time control, namely:
* Defining an international standard.
* Ease of implementation and administration.
* Synchronisation capabilities of sub-microsecond.
* Ethernet capable.
* Low CPU requirements and low bandwidth requirements.
PTP is very effective in addressing all of the above requirements, and, in Ethernet networks (PTP works on any multicast-capable network), it successfully eliminates the effects that variable network latency may have. It also caters for the effect that switches and routers have on the packet transmission, which can be exacerbated by the use of mechanisms such as QoS and ToS.
PTP does this by introducing the concept of a boundary clock, in addition to having master and slave clocks. A boundary clock is positioned at an Ethernet device (such as a switch, or a router), while the end devices are either a master or a slave clock.
To successfully synchronise clocks, two aspects must be addressed – the offset and the drift. The former ensures that the inaccurate (or less accurate) clock is set to the accurate one, while the latter ensures that the fact that different clocks may run at different speeds is compensated. Thus, PTP works in two phases. In the first phase, the offset is corrected by the master clock (usually connected to a GPS or other high precision source), which sends a time stamped synchronisation signal at regular intervals to the slave clocks. In addition to this, the master also measures the time at which the SYNC message was sent and sends this value to the slaves as well. The slaves measure the reception time and can calculate the offset (or the correction time) and correct themselves. The second phase is the delay measurement, which is measured by the Delay Request and Delay Response messages. Based on these the drift is calculated and corrected accordingly.
The process, however, depends on the latency between master and slave being symmetrical – something which is almost guaranteed when they are directly connected. This is not the case in a typical network though, when there are switches and routers connecting end devices. This is where the boundary clock concept comes in – each network device will have a clock, which is synchronised with a master directly connected to the device, and then acts as a master to all other devices connected to it. This means that the symmetrical latency requirement is always satisfied, and the synchronisation kept accurate. It is still important to adhere to general good practices for Industrial Ethernet network design, since the latency introduced by the networking devices increases proportionately with the network load.
One of the benefits of the PTP is that the selection of the master and slave clocks is handled by the protocol. This is done using the Best Master Clock algorithm, which runs on every member of the PTP. The algorithm compares the properties of the clocks and distributes them amongst the members, who in turn determine their own status. Thus no status negotiation is necessary.
The implementation of the PTP can be done either in software only, or by using relatively simple and inexpensive hardware. The purely software implementation achieves reasonably good accuracy (between 10 and 250 μs), however it is dependent on the CPU cycles and network utilisation.
Hardware implementations, such as the one used in the Hirschmann range of PTP capable switches, ensures that the dedicated Time Stamp Unit is present, which generates the time stamps directly on the transport medium. Such implementations are good for accuracy of under a microsecond.
For more information contact Vladimir Milovanovic, IAC, +27 (0)12 657 3600, [email protected], www.iaconline.co.za
| Tel: | +27 12 657 3600 |
| Email: | [email protected] |
| www: | www.iacontrol.co.za |
| Articles: | More information and articles about Industrial Automation & Control (IAC) |
© Technews Publishing (Pty) Ltd | All Rights Reserved
printer friendly version