Analysis of Time Synchronization in PTN

    With the development of mobile telecommunications technologies, precise network time synchronization has become increasingly important. CDMA2000, Time Division Synchronous Code Division Multiple Access (TD-SCDMA), and Time Division Long Term Evolution (TD-LTE) base stations all require high precision time synchronization. In TD-SCDMA systems, for example, the time synchronization index is ±1.5 μs. Such precision cannot be achieved through a free running oscillator, or even through the frequency synchronization network. On the other hand, installing a Global Positioning System (GPS) for each Time Division (TD) base station is difficult to engineer, costly, and insecure. Therefore, time synchronization protocol has become an important technology for transmitting high precision time synchronization signals in fiber systems.

    Using synchronization protocol to transport high precision time synchronization signals via optical fiber systems is expected to be a leading technology in the future.

    Terrestrial transmission of high precision time synchronization signals involves two key technologies: Precision Time Protocol (PTP), and compensation technology for transmission delay and jitter. Networks using Network Time Protocol (NTP) synchronization are only accurate to dozens of milliseconds, which does not meet the requirements of TD-SCDMA systems. Based on a delay request-response mechanism, IEEE 1588v2, also known as Precision Time Protocol (PTP), uses synchronization messages to calculate the time error between the slave and master clocks. Moreover, by using a hardware-embedded time stamp as well as a Boundary Clock (BC) or Transparent Clock (TC) to compensate delay and jitter incurred by network components or protocol stacks, IEEE 1588v2 achieves precision of sub-microseconds.

    IEEE 1588 is a PTP originally designed for synchronization between industrial computers. Before it can be applied to large-scale communication networks, further research must be undertaken into precision, networking mode, maintenance, and Best Master Clock Algorithm (BMCA). This paper analyzes the factors that may affect time synchronization precision in a Packet Transport Network (PTN) adopting IEEE 1588v2[1-5].

1 Key Factors Affecting Time Synchronization Precision
IEEE 1588v2 adopts a master/slave clock configuration. The master clock distributes clocks periodically, and the receiver—taking advantage of network link symmetry—measures time offset and delay. Synchronization of frequency, phase, and absolute time of master and slave clocks is thereby realized. Transmission of IEEE 1588v2 packets may require each node to process a time stamp, which incurs delay and jitter. As a result, the number of hops in a transport network impacts the precision of time synchronization signals. Because a store-and-forwarding mechanism is used in PTN, Packet Delay Variation (PDV) may have a significant impact on precision. Furthermore, network protection switching, signal degradation, temperature, and frequency synchronization may also impact timing precision.

1.1 The Number of Hops
In BC mode, each node terminates PTP packets, so PDV in the last node is not accumulated. The main factor affecting timing precision is the queue at the output port. In addition, because slave clocks at all levels are synchronized to the master clock, drift is incurred during clock recovery. This low-frequency drift is accumulated. Figure 1 illustrates a test platform that is designed for evaluating the impact of the number of hops on timing precision. On this platform, all PTN nodes are configured in BC mode and are interconnected by Gigabit Ethernet (GE) interfaces. The device under Test 1 (DUT1) is connected to the GPS receiver with a
1 Pulse Per Second (1PPS) + Time of Day (ToD) interface. Timing test equipment is used to measure the timing difference between the DUTn and GPS receiver.

    Figure 2 shows timing differences in tests with different hops. In a nine hour test with 10 hops, the timing difference ranges from -120.3 ns to 131.5 ns with a peak-to-peak value of 252 ns. In a nine hour test with 20 hops, the timing difference ranges from -61 ns to 192 ns with a peak-to-peak value of 253 ns. In a four hour test with 30 hops, the timing difference ranges from -239.3 ns to 26.8 ns with a peak-to-peak value of 266 ns. These test results show that timing precision changes only slightly in proportion to the number of hops, and the noise model approximates random distribution.

1.2 PDV
PDV mainly impacts timing precision in the purely transparent transport mode. As shown in Figure 3, all PTN nodes in the test platform in Figure 1 are set to transparent transport mode. Timing difference is measured where there is no load, and a 90% load with packet lengths of 64 bytes, 576 bytes, and 1 518 bytes. The peak-to-peak timing differences in each case are 250 ns, 450 ns, 3??200 ns, and 10 μs respectively. It can be seen that the heavier the load or longer the packet, the greater the impact of PDV on timing precision. Before transparent transport mode can be used to provide high precision time synchronization signals, much optimization must be done.

1.3 Network Changeover
Changeover of time source, link, or clock board may lead to frequency
and/or phase shift, affecting timing precision. Test results show that time source changeover causes a timing difference of 6 ns, optical link changeover causes a difference of
26 ns, and clock board changeover causes a difference of 13 ns, as shown in Figure 4.

    Generally, the timing difference arising from network changeover is within 30 ns, enough for a 50 ns allowance in the parameter setting.

1.4 Signal Degradation and Temperature Change
Signal degradation may cause the rate of packet loss to increase, while temperature change may affect clock performance. In either case, timing precision may be affected. When bits are inserted with a Bit Error Rate (BER) of 1× 10-3, the peak-to-peak timing difference is within 110 ns, similar to the case where there is no bit error. When the temperature of PTN nodes is raised from -10℃ to 50℃, the peak-to-peak timing difference is within 40 ns, and timing does not change proportionally with temperature.

1.5 Frequency Synchronization
PTN can obtain the clock from the physical layer via synchronous Ethernet, or recover the clock from 1588v2 packets. Tight coupling occurs when PTN realizes frequency synchronization using synchronous Ethernet, and timing and phase synchronization using 1588v2 packets. Loose coupling occurs when PTN realizes frequency, time, and phase synchronization using only 1588v2 packets. Time synchronization in the cases of tight and loose coupling is illustrated in Figure 5.  Under normal conditions, the peak-to-peak timing difference in tight coupling is 22 ns, while that in loose coupling is 67 ns. Synchronization in tight and loose coupling is therefore similar. 

 
    Figure 6 shows the impact of frequency synchronization on timing precision in tight coupling mode. When the frequency is in Keep state, with an offset of 5×10-9, the peak-to-peak timing difference is 100 ns. When the frequency is in Free state, with an offset of 3.8×10-8, time synchronization signals are lost.

    The two coupling modes have their own advantages and disadvantages. In the case of tight coupling, frequency synchronization signals are degraded or lost, which may affect time synchronization signals. In the case of loose coupling, each local PTN clock may take a long time to trace the 1588v2 clock. Also, in loose coupling mode, the frequencies of 1588v2 packets must be increased, which increases the network load. Tight coupling is therefore preferred.

2 Long-Term Stability of Time Synchronization in BC Mode
To verify the stability of a large PTN adopting 1588v2, part of the existing network is selected for testing. The test network includes three core nodes, three aggregation nodes, and 58 access nodes. Each access node has one TD-SCDMA base station, as shown in Figure 7. All PTN nodes adopt BC mode and all fiber segments compensate for delay asymmetry. Frequency synchronization is implemented using synchronous Ethernet. Time interfaces include
1PPS+ToD interface and FE/GE interfaces.

    When FE is used, the peak-to-peak timing difference over 72 hours is 65 ns, as shown in Figure 8. The Maximum Time Interval Error (MTIE) and Time Deviation (TDEV) results are shown in Figure 9. These results show that the timing is stable, there is little noise, and the noise model approximates random distribution.

3 Conclusions
For TD-SCDMA and TD-LTE systems, using GPS to implement time synchronization is costly, insecure, and difficult to engineer. Time synchronization protocol has therefore become the preferred technology for transmitting high precision time synchronization signals in fiber systems. High precision time synchronization in a PTN adopting IEEE 1588v2 has been proven feasible. In existing networks, many fiber segments are asymmetric, so compensation for delay and jitter must be done segment by segment. Both BC and TC eliminate the impact of PDV, but BC is simpler. Standards for time synchronization in telecom networks are still immature, and experience is limited. High precision time synchronization will first be deployed in PTN, and then in Optical Transport Networks (OTN) and Passive Optical Networks (PON) in the future.

References
[1] Report of the Lannion SG15/13 interim meeting [R]. WD03, ITU-T SG15 Q13. 2009.
[2] LI Han, HAN Liuyan. Analysis of time synchronization performance using PTP in hybrid transport network [R]. CMCC, WD64, ITU-T SG15 Q13. 2010.
[3] LI Han, HAN Liuyan. Proposal of accelerating MTIE and TDEV model study for time synchronization [R]. CMCC, WD63, ITU-T SG15 Q13. 2010.
[4] LI Han, WANG Lei. Test and analysis of time synchronization using 1588v2 for transport network [R]. CMCC, C599, ITU-T SG15 Q13. 2009.
[5] LI Han, WANG Lei. Time distribution model using 1588v2 for performance indication [R]. CMCC, C601, ITU-T SG15 Q13. 2009.

[Abstract] For Time Division Synchronous Code Division Multiple Access (TD-SCDMA) and Time Division Long Term Evolution (TD-LTE) wireless systems, using Global Positioning System (GPS) for time synchronization is problematic. In these wireless systems, GPS is costly, insecure, and difficult to deploy. Nowadays, transportation of high precision time/phase synchronization signals via fiber, and based on Precision Time Protocol (PTP) has become mainstream technology. This paper analyzes the main factors affecting time synchronization in a Packet Transport Network (PTN) adopting IEEE 1588v2. Laboratory experiments and field tests prove the feasibility of transporting high precision time synchronization signals through a PTN adopting IEEE 1588v2. A comparison is also made between different networking modes for a PTN adopting IEEE 1588v2.