Comparison - SyncE, NTP, GPS, IEEE-1588

Comparison – SyncE, NTP, GPS, IEEE-1588

We, now have enough information to compare various synchronization methods after we went through the clock synchronization basics and the Synchronous Ethernet (SyncE), IEEE-1588 technologies. In this section we would look at various factors through which we can have a comparison – SyncE, NTP, GPS, IEEE-1588 and evaluate these technologies:

  • Network Load: Sync-E works at the physical layer and is independent of congestion or network load. It would work whether traffic is present or the link is idle. Also, except for 10-pps ESMC messages, Sync-E does not load the network in order to achieve synchronization. Even the accuracy achieved is not dependent on the rate of ESMC messages which is only a control protocol for its functioning. IEEE 1588 works at layer-2 and layer-3 and all its messages contribute to the congestion and are affected by it. A higher PDV is the root cause of not being able to achieve the accuracy. A higher accuracy can be achieved by using a higher rate of messages in 1588, but this itself is a cause of congestion on the network and increase in sensitivity to the PDV. NTP (Network Time Protocol) also behaves like 1588 since it uses UDP as its transport protocol and has similar consequences and contribution to Network Congestion. GPS, on the other hand is a time-broadcast service. The GPS signals are always available, any place near the earth that has unobstructed line of sight to GPS satellites.

 

  • Scalability: GPS is highly scalable, and so is Sync-E because an increase in the number of receivers for GPS, does not load the satellite itself. Sync-E’s ESMC protocol is a unidirectional, multicast where clients can join or leave and are fully responsible for their own synchronization. IEEE 1588 (and NTP) involves a lot a work from the master. More clients means more communication and queries that the master has to reply to and are thus scalable to the extent the standards dictate by limiting the number of slaves per master or con-current peer-to-peer connections. Any increase in client requests in 1588 and NTP also contributes to network congestion which is detrimental to their own system.
Related:  Synchronization in Networks

 

  • Support: GPS, 1588 and NTP are supported on all systems that can be connected to the network or that can see the sky without any obstruction. Sync-E can be supported on all varieties of Ethernet; however, there are some complicated issues on 10 Mbps links and 100 Base-T links. 10Base-T transmitter stops sending pulses during idle periods and sends “I’m alive pulse” every 16 ms while in 1000Base-T it is possible to pass-on the synchronization but then you have manually set all alternating ports to Master/Slave.

 

  • Special Equipment: Sync-E requires changes in the timing cards and PLLs as discussed. 1588, for a precise operation, requires time-stamping and classification of PTP messages at the hardware level. NTP is same as 1588 in this regard and works better the more closer the time-stamp is to the hardware or the software driver. GPS requires special GPS receivers.

 

  • Time-of-Day: 1588, NTP and GPS support ToD synchronization with varying accuracies. Sync-E does not deliver ToD.

 

  • Reliability: Sync-E does one thing (frequency locking) and does it properly. The frequency and phase synchronization in Sync-E is most reliable and highly scalable. 1588 can be used for frequency, phase and time synchronization but its reliability is yet to be proven over different networks and scenarios. GPS is riddled with interference problems, atmospheric and multi-path effects.

Comparison – SyncE, NTP, GPS, IEEE-1588

Sync-E

(Layer 1)

1588/NTP

(Layer2/3)

GPS
N/w Load No Yes No
Scalable Yes Limited Yes
Support Ethernet Everywhere Everywhere
Special Equipment DPLL H/w Time Stamping GPS receiver
ToD No Yes Yes
Reliability Guaranteed Not proven Interference

Here are some of the comparison numbers between 1588 and NTP for Clock synchronization:

PXI Backplane IEEE 1588 NTP on IP
Resolution ~0.01 ns ~50 ns <1×107 ns
Jitter ~0.002 ns ~100 ns ~3×106 ns
Distance ~0.5 m <400 m Worldwide
Sample Rates 100s of MHz <100 KHz <10 Hz
Cabling n/a CAT 5 Ethernet Ethernet, etc.
Topology User-defined Auto-resolve, master/slave Peer-to-peer

 Table: Compare synchronization alternatives
(Source: National Instruments)

Combination

IEEE 1588 and Synchronous Ethernet together make a synergetic combination. This is because:

  1. Sync-E is extremely reliable at L1 and not affected by network congestion, so two nodes can be synchronized in frequency and phase accurately. This mean, my watch is now, neither faster nor slower than yours. Also, it means a one second in my watch is precisely equal to the one second in your watch till the last nanosecond.
  2. IEEE 1588 can deliver a highly accurate Time-Of-Day using 64-bit time-stamps and measurement of mean-path delay between nodes. If clocks are already syntonized, 1588 can phase-synchronize them much more accurately, and with a lesser demand on the network bandwidth for trying to achieve the same accuracy if it were used alone.

Combination Synchronous Ethernet and IEEE-1588

Figure: Sync-E and 1588 combined
(Source: National Semiconductor)


Author 

Jagmeet Singh Hanspal 

Jagmeet is a Senior Technical Leader and has worked with various organizations like Ericsson, Juniper Networks, TranSwitch Semiconductors in the field of Telecommunications and Embedded Systems. You can contact him via his Linkedin profile. His interests include Linux, Micro-controllers, Programming, Parallel Processing, Data Visualization, Networks, Synchronization protocols, Statistical analysis etc. You can find more about his profile from Author’s personal page.

2 thoughts on “Comparison – SyncE, NTP, GPS, IEEE-1588

  1. Very well written. I have few comments:

    1.Network load will not have an impact with the Phy timestamping feature where resident time inside the DUT can be updated in the correction filed, which can be used by SERVO for correcting the frequency and phase offsets.

    2. Scalability , where master needs to support the number of clients are dictated by customer requirements. Existing solutions supports 256 and 512 clients at different rates based on the constraint w.r.t the s/w and h/w resources. I don’t thinks so these exchange of the PTP packets going to congest the network because these packet rates are very less compare to the the 1G,10G,100G line rates.

    3. Reliability, I feel PTP is highly reliable for time synchronisation as the ITU-T standards are moving to support 5G requirements.

    1. Thanks… Some clarifications though

      (1) The PDV is bad for achieving the accuracy even when PHY is time-stamping.
      (2) It was in comparison to Synchronous Ethernet ESMC multicast. In 1588, as clients increase, so does the packet exchange for each peer.
      (3) For Phase-acquisition, yes 1588. For frequency, the physical layer PLL lock of sync-ethernet is much superior.

      And thanks for the comment and knowledge share. Do subscribe for more articles and share us forward.

Leave a Reply