Time Synchronization Redundancy in PTP Networks
Korkee, Aleksi (2020)
2020
Tietotekniikan DI-tutkinto-ohjelma - Degree Programme in Information Technology, MSc (Tech)
Informaatioteknologian ja viestinnän tiedekunta - Faculty of Information Technology and Communication Sciences
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2020-05-25
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-202004294309
https://urn.fi/URN:NBN:fi:tuni-202004294309
Tiivistelmä
Accurate time synchronization is important in industrial automation. Precision time protocol (PTP) is often used as it can provide time synchronization accuracy in scale of nanoseconds. In PTP all devices synchronize to a single clock called grandmaster. A failure in the grandmaster affects all the devices that synchronize to it.
Different methods for adding redundancy to PTP networks is presented in this thesis. The aim of this thesis was to add redundancy to an existing time synchronization software supporting IEEE1588 and 802.1AS protocols. A feature was implemented in the product that added redundancy by using multiple grandmasters concurrently in different time domains. Each device in the network calculates a fault tolerant average between synchronization information from all the available grandmasters. Devices adjust their local clock to this average. On system startup only one of the grandmasters is distributing time. Other grandmasters synchronize to this until their offset to it is small enough. After this, all the grandmasters distribute their time through one domain while also synchronizing to the grandmasters in other domains. If grandmaster fails to provide accurate time, the other grandmasters are able to maintain accurate synchronization by using fault tolerant average algorithm.
The feature was tested with a test network. Four devices were configured as grandmasters and three as slaves. Three behaviours were tested: system startup, system's runtime and different fault scenarios. Each grandmaster was able to synchronize to their higher priority grandmasters and start distributing their time through one of the domains. When the system was left running for 24 hours the synchronization was maintained through the whole time. Most of the time each device's calculated offset from average was below 30 ns with some peaks at 50 ns. Offset from one of the master varied more than the others. Offset from that master was mostly below 100 ns but it had some peaking at 200 ns. The reason for this was not found, but one possible cause might have been the air conditioning affecting the device's oscillator's frequency. The fault tolerance test results showed that the system was able to maintain accurate synchronization even when up to two grandmasters out of four became faulty. The system was not able to maintain synchronization when three grandmasters started sending incorrect time at the same time.
The results presented in this thesis were used to improve the implemented feature.
Different methods for adding redundancy to PTP networks is presented in this thesis. The aim of this thesis was to add redundancy to an existing time synchronization software supporting IEEE1588 and 802.1AS protocols. A feature was implemented in the product that added redundancy by using multiple grandmasters concurrently in different time domains. Each device in the network calculates a fault tolerant average between synchronization information from all the available grandmasters. Devices adjust their local clock to this average. On system startup only one of the grandmasters is distributing time. Other grandmasters synchronize to this until their offset to it is small enough. After this, all the grandmasters distribute their time through one domain while also synchronizing to the grandmasters in other domains. If grandmaster fails to provide accurate time, the other grandmasters are able to maintain accurate synchronization by using fault tolerant average algorithm.
The feature was tested with a test network. Four devices were configured as grandmasters and three as slaves. Three behaviours were tested: system startup, system's runtime and different fault scenarios. Each grandmaster was able to synchronize to their higher priority grandmasters and start distributing their time through one of the domains. When the system was left running for 24 hours the synchronization was maintained through the whole time. Most of the time each device's calculated offset from average was below 30 ns with some peaks at 50 ns. Offset from one of the master varied more than the others. Offset from that master was mostly below 100 ns but it had some peaking at 200 ns. The reason for this was not found, but one possible cause might have been the air conditioning affecting the device's oscillator's frequency. The fault tolerance test results showed that the system was able to maintain accurate synchronization even when up to two grandmasters out of four became faulty. The system was not able to maintain synchronization when three grandmasters started sending incorrect time at the same time.
The results presented in this thesis were used to improve the implemented feature.