Open Access
Add to BibSonomy
Abstract
We demonstrate a new optical pulse amplitude modulation (PAM) scheme where joint ultrastable time-frequency and gigabit ethernet data transfer with the same laser wavelength is realized. Time transmission is compatible with the White Rabbit (WR) based on gigabit ethernet networks, and frequency transmission is achieved by using 100MHz radio frequency (RF) modulation and the round-trip compensation methods. The laser is on-off keying (OOK) modulated by the WR signal, the RF and WR signal are modulated by optical PAM in a Mach-Zehnder interferometer modulator (MZM), and the local and remote site are connected by 96km urban fiber in Shanghai. The experimental results demonstrate that the frequency instabilities are 5.7E−14/1 s and 5.9E−17/104s, and the time interval transfer of 1 pulse per second (PPS) signal with less than 300fs stability after 104 s are obtained. This novel scheme can transmit frequency signals at hydrogen-maser-level stability in the gigabit ethernet network.
© 2021 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
The synchronization of time and frequency is of great important importance in atom clocks comparisons [1], Global Navigation Satellite System (GNSS) [2], very long baseline interferometry (VLBI) [3], and some commercial applications (e.g., metrology and timescale development). With the development of atom clocks, such as hydrogen masers, cesium fountains, and optical clocks, the frequency instability is better than 1E−13 at 1 s for RF signal [4]. However, the traditional satellite-based time-frequency transmission method is greatly affected by the terrain and weather conditions [5–6], it can no longer meet demands of the increasing accuracy of application requirements. Due to the advantages of low loss and high safety of optical fiber, time-frequency transmission with higher stability over fiber-optic has been developed rapidly [7–12].
Time-frequency transfer in optical fiber has been widely discussed, various schemes have been proposed. For the applications which require higher performance, advanced techniques have been developed. One classical solution is transferring time-frequency with different wavelengths which are combined or separated by Dense Wavelength-Division Multiplexing (DWDM). A typical scheme is to use two wavelengths for time-frequency transfer, the 9.1GHz RF frequency stability is 7.0E−15/1s and 4.5E−19/1day at 80 km fiber link, and the time [one pulse per second (PPS)] synchronization is about 50 ps [13]. The other are transfer ultrastable optical frequency and time and Internet data over 540 km using a public telecommunication optical fiber network with three wavelengths [8]. In order to reduce the transmission wavelength, the author modulated time and 1GHz RF signals on the same laser wavelength with phase and intensity modulation by independent electro-optic modulators, the frequency instabilities of 1.7E−14/1 s and 5.9E−17/104 s and time interval transfer of 1PPS signal with sub-ps stability after 1000s are obtained [11]. The 1 PPS signal is coded by specific phase modulation on the 10/100 MHz signal with the same laser wavelength in another electronically stabilized fiber-optic time and frequency (T&F) distribution system, the frequency stability in the range of 1.7E−17 for 105 s averaging and time calibration with accuracy well below 50 ps [9]. To meet the demand for fiber access to precise time and frequency standards in areas such telecommunication, internet, metrology with larger scalability, data layer approaches is proposed. Network time protocol (NTP) [14] and precision time protocol (PTP) [15], are able to synchronize time and frequency in the public telecommunications network, but the accuracy is limited to a few milliseconds. The WR project [16–19] is a fully deterministic Ethernet-based network which can synchronize over 1000 nodes with sub-ns time accuracy, and the frequency instabilities of the WR is 2E-11 at τ=1s [20] due to FPGA implementation. The enhanced WR with an additional daughter-board can achieves a tenfold improvement in terms of phase noise, jitter and short-term stability with respect to the current WR performance [21], frequency MDEV stabilities down to 6E-13 at 1s are achieved(BW=50Hz). And with the help of low-noise clean-up oscillators and daughter-board, frequency stabilities (ADEV) averaging down from 7E-13 at 1 s to 3E-16 at 7,000 s are achieved(BW=0.5Hz) [22]. However the improved WR scheme still cannot transmit frequency signals at hydrogen-maser-level stability.
In this letter, we combine ultrastable time-frequency transfer methods and data layer approaches, and propose a novel scheme to transmit time-frequency signal in gigabit ethernet networks over fiber link with the same wavelength in one direction. This novel scheme has two merits that one is compatible with Ethernet protocol, the other are transfer ultrastable frequency signals at hydrogen-maser-level stability. The rate of WR is 1.25Gbps, while the RF modulation signal is 100MHz. Time-frequency transfer over 106 km fiber(10km spools and 96km urban fiber) based on the proposed scheme is demonstrated. The relative stabilities of 5.7E−14/1 s and 5.9E − 17/104 s for the frequency and time interval for 1 PPS with below 300fs stability after 104 s are obtained. It represents a significant step towards ultrastable time-frequency network distribution in the public telecommunications network.
2. Principle
To mitigate the propagation delay caused by the fiber link from the ambient environment, the common round-trip time-frequency transfer method is adopted [23]. Time transmission is based on the WR which is compatible with IEEE 802.3z Ethernet protocol and IEEE1588v2 [15]. Frequency transmission is achieved by using 100MHz RF modulation [10] as shown in Fig. 1. The DFB laser is OOK modulated by the WR signal, the RF and WR signal are PAM modulated in MZM [24].
Fig. 1. Time-frequency transfer schematic diagram. PPS, Pulse Per Second; RF, radio frequency; WR, white rabbit; DFB distributed feedback; MZM, Mach-Zehnder Interferometer Modulator; PAM Pulse Amplifier Modulation; DWDM, Dense Wavelength-Division Multiplexing; ODL optical delay line; OC, optical coupler; PD, photodetector; COMP, comparator; BPF, Band-Pass Filter.
WR system consists of the WR grand master and the WR slave, the grand master can be synchronized with external 1PPS and frequency signals, and the frequency of the slave is locked to the grand master via the phase lock loop (PLL) based on the synchronous Ethernet technique. With the help of Digital Dual Mixer Time Difference (DDMTD) phase detector for fine phase measurement, and in turn phase adjustment, WR can achieve phase synchronization between the grand master and the slave. To obtain the precise time delay from the grand master to slave, the PTP is performed to calculate the coarse link delay through exchanging the time-stamps message, then the time-stamps resolution is enhanced by phase synchronization and sub-ns accuracy is finally achieved [25].
The 1.25Gbps WR data sent to DFB laser is encoded by 8B/10B, in 8B/10B encoding each byte of data is assigned a 10 bit code, the 10 bit code must contain either five ones and five zeros, or four ones and six zeros, or six ones and four zeros, so the number of “0” and “1” sent by WR signal is basically the same. Assuming that the data sent by WR are all “01” sequences, the WR signal that was OOK modulated on an optical carrier of λ1, can be expressed as a sequence of rectangular pulse signal $\textrm{S}_\tau (t)$ with a repetition frequency of 1.25G. $\textrm{S}_\tau (t)$ is described as:
$\textrm{d}_\tau (t)$ is a rectangular pulse which the high level is $\tau$, and ${T_s}$ is the period of the WR signal. At the LS, the reference 100MHz RF signal can be expressed as: ${\varphi _{{0^{\prime}}}}$ is the initial phase of RF signal. In the LS, RF and WR are optically modulated by PAM in MZM to generate analog/digital hybrid signals ${I_1}(t )$, it can be expressed as:3. Field test results and analysis
3.1 Field test and system setup
We construct an evaluation system as shown in Fig. 2, WR signal is modulated on the optical carriers with the wavelength λ1 = 1548.5 nm (C36) and the wavelength λ2 = 1547.8 nm (C37) at the LS and RS via DWDM SFP transceivers. SFP is a compact transceiver that consists of Distributed Feedback Laser (DFB) laser, photodiode, amplifier (AMP), here it is used to convert Gigabit electrical signals into optical signals.
Fig. 2. Experimental setup of time-frequency and Gigabit Ethernet data transfer. Rb, Rubidium clock; ASG, Analog signal generator; PD1, phase detector; PS, power splitter; SFP, Small Form-factor optical module; P.I.D. Proportional-integral-derivative; PD2 photodetector;
The analog-digital hybrid signal at C36 carrier is generated by MZM at LS and transmitted to the RS through optical fiber. MZM is set to quadrature point of positive slope for linear analog modulation. The output optical power is 7dBm, the 100MHz RF modulation power is 8dBm, and the PAM is 0.41.
The link includes 10km ODL fiber spools and 96km urban fiber. One terminal of urban fiber is in Shanghai Institute of Optics and Fine Mechanics (SIOM), Chinese Academy of Sciences, and the other terminal is in Chinese Academy of Sciences Shanghai Branch. The length of fiber cable between the two sites is 48km. For the ease of performance evaluating, both the LS and RS are settled in SIOM, while a bidirectional erbium-doped fiber amplifier (Bi-EDFA) is placed at Shanghai Branch for signal amplification. Bi-EDFA is a dedicated quasi-bidirectional optical amplifier which is based on two anti-parallel EDFA in the spirit of the design [27]. The output power of this amplifier is 6 dBm, and the input optical power is greater than -25dbm. The delay asymmetry of this amplifier was calibrated by WR.
At the RS, after a forward transmission, the C36 analog/digital hybrid signal is divided into two parts, one part recovers the 1.25Gbps WR signal through the SFP receiver, and the other part recovers the 100MHz RF signal V2 through the photodetector(PD) and BPF. To block the effects from the ambient environment, half of the received frequency signal is directed back to the LS modulated on C37. At the LS, the round-trip frequency signal V3 is compared with the other branch from the 100MHz frequency source V1 by a phase detector (PD1) to follow up the fiber link propagation delay variation. The ODL, which is controlled by PID, contains a 10 km temperature-controlled fiber-optic spools and a PZT in order to cancel out the slow and fast propagation delay caused by the fiber link.
Rubidium clock generates 1PPS and 10MHz time-frequency signal, 1PPS is locked by WR Grand Master, 10MHz is locked by analog signal generator (ASG) to generate 100MHz RF reference signal, then the RF reference signal is divided into three channels by power splitter (PS). AMP is used to adjust the RF signal to the appropriate power.
The frequency transmission instability is tested by phase-noise measurements (Microsemi5125), and the result is characterized by Overlapping ADEV. The WR time transmission stability is tested by a high-speed digital oscilloscope (Agilent DSO91304A, one-shot resolution is at ∼500fs), and the stable32 software is used to calculate TDEV.
3.2 Time and phase drift, ADEV and TDEV
The black line in Fig. 3 shows the time delay drifts when the 106km fiber link is open-loop. In the open-loop state, the ODL does not work, when in the closed-loop state, the ODL is driven by the PID controller to compensate the slow and fast propagation delay variations. The time delay drift is mainly due to the influence of temperature on the fiber link. During the test, the weather temperature is 12°C-1°C with a change of 11°C, and the fiber link delay fluctuation range is about 5ns. The G.652D fiber delay variation is 37km/ps/°C, and the urban fiber link length is 96km. Because the urban fiber cable is a buried cable, so it changes about 1.4°C. according to the result of the delay change. The red line and the blue line respectively show the time delay drift and frequency phase drift in the closed loop condition, where the time delay drift is reduced to a peak-to-peak value less than 50ps, and the frequency phase drift is less than 2ps.
Figure 4(a) describes the frequency transfer performance (BW=5Hz) of the new scheme under the 106km fiber link. The black line is the open-loop performance of 100MHz RF signal. Due to the temperature and vibration noise of the fiber link, the frequency instabilities is 3.1E−13 at 1s, and the instabilities deteriorates to 1.2E-13 at 104 s. The pink line is the result of turning on the noise compensation system. Users can obtain 5.7E−14 at 1 s and 5.9E−17 at 104 s RF signal at the RS, which proves that the noise compensation system can well suppress the temperature and vibration noise effects of the fiber link. There is a great improvement when comparing the performance of the closed loop with the open loop. The green dashed line is active hydrogen maser (Microsemi MHM 2010) frequency instabilities (BW=0.5Hz), it will be worse in case of BW=5Hz [22]. The red line is the performance of standard WR frequency transmission with 106km fiber link. The user can obtain 4E-12 at 1 s and 7E-16 at 104 s RF signal at WR slave node. The stability of the standard WR frequency transmission instabilities is one order of magnitude worse than the hydrogen maser before the thousand-second average, which is due to FPGA implementation [21]. The short-term frequency stability of the new scheme is improved nearly two magnitudes compared with the standard WR. More importantly, the frequency transfer stability outperforms hydrogen maser stability within 1-10000s, demonstrating that the novel scheme is very suitable for transmission at hydrogen-maser-level stability accompanying with the Ethernet networks.
Fig. 4. (a) Measured 100MHz Overlapping Allan deviation (ADEV) (b)Measured 1PPS time deviation (TDEV)
Figure 4(b) shows the time transfer performance. The red line is the short fiber time transfer performance only with WR, and the performance is almost consistent with the time transfer performance in the new system (blue line), which proves that optical PAM does not interfere with the WR signal. For the case of the closed loop over 106km fiber link, the time deviation (TDEV) of the new system is 5.7E-12/1s and 2.1E-13/104 s, which is almost the same as the short fiber.
4. Discussion
When the WR transfer time in the proposed scheme, MZM and OC devices are added to the WR link. The same MZM and OC devices are selected to ensure the same delay between the master and the slave. The delay caused by MZM and OC in the local site is 35881ps, and the delay in the remote site is 35795ps, the time difference between two sites is 86ps. The Bi-EDFA has also been calibrated, two unidirectional EDFAs delay are 850425ps and 850378ps, and the difference is 47ps. In order to verify whether PAM modulation has an effect on the accuracy of WR time transfer, the time delays between the master and slave nodes are measured under different conditions. The experimental results are shown in the Fig. 5, the time delay is -42.3 ps with short fiber, and the time delay is -8.6ps with 96km link, which indicates that PAM modulation has no effect on WR time accuracy.
Figure 6(a) red line describes the close-loop frequency instabilities (BW=5Hz) of the new scheme under the short fiber, the black line is the close-loop performance of 100MHz RF signal which use DFB laser to replace SFP module. The degradation of the frequency stability of the new scheme is mainly due to the interference of the WR signal, and MZM has less impact.
Fig. 6. (a) The influence of PAM modulation on frequency transmission (b)The relationship between PAM depth and ADEV, PAM depth and BER
Because of the optical carrier is intensity modulated by WR and RF signal at the same time, The PAM depth will affect the transmission stability of the RF signal and the bit error rates (BER) of the WR data.
In Fig. 6(b), red line shows the relationship between RF transfer stability and PAM depth in the case of closed loop with a short fiber. In the experiment, the PAM depth is achieved by adjusting the output power of ASG. When the power of 100MHz RF output is 11dBm, the PAM depth is 0.51. When the modulation depth is greater than 0.51, WR stops working due to too high BER. When the modulation depth is 0.23-0.51, the RF transmission performance does not change because the limitation is from the system noise floor. When the PAM depth is less than 0.23, the RF transmission performance begins to deteriorate quickly. The relationship between PAM depth and RF transfer stability is consistent with the theoretical calculation [27].
Because WR data BER is affected by the modulation system, and WR cannot generate the random code signal required for BER testing. In order to measure the relationship between PAM depth and WR data BER, an Anritsu BER tester was used to generate 1.25Gbps PRBS7 random code. The input of modulation system mentioned in Fig. 2 red box is connected to the PRBS7 random code generated by Anritsu BER tester, then BER tester measures the BER of the output port of the modulation system. The relationship between PAM depth and BER is shown as the green line in Fig. 6(b) showing that when PAM depth is less than 0.43, the BER is less than 1E-12. The BER is 0 when the PAM is less than 0.43 which is not shown in Fig. 6(b). In summary, when PAM modulation depth is between 0.23 and 0.43, the performances of both the ADEV of frequency transmission and the BER of WR data can be guaranteed.
The WR frequency distribution is based on the SyncE (Synchronous Ethernet). The 125MHz frequency signal is recovered from the 1.25Gbps data as the slave clock reference, and then reference generates output frequency signals such as 10MHz, so there is no harmonic interference between the 10MHz and the 100MHz RF PAM signals.
WR is compatible with Ethernet, and its clock can be transmitted and recovered in all nodes, so time can be distributed in long links through WR, as shown in [12]. The cascaded method is a good choice for ultrastable frequency dissemination, and some experiments on cascaded systems have been carried out with good performance [10]. Since the SFP input power needs to be greater than -18dBm, the MZM output power is 5dBm, and the insertion loss of DWDM, OC, and ODL is 8dBm, so when the fiber link loss is greater than -15dBm, Bi-EDFA is required to increase the transmission distance.
The new time-frequency joint transmission scheme based on PAM optical modulation improves the performance of frequency transmission in Gigabit Ethernet networks. With the 106km fiber link, the frequency transmission performance is 5.7E-14/1s, which is better than hydrogen maser frequency stability within 1-10000s. In inheriting the advantages of WR's large-scale networking and data transmission, and the frequency compensation accuracy is improved by adding ODL high-precision delay adjustment devices. Compared with the standard WR system, the complexity of the system hardware is increased. This problem can be solved by optimizing the structure in the later stage, such as deep integration with WR, and using electrical compensation solutions.
5. Conclusion
In this letter, we propose and demonstrate a new scheme which can transfer ultrastable time-frequency with one wavelength in Gigabit Ethernet networks. The experimental results demonstrate that the frequency instabilities is 5.7E−14 at 1 s and 5.9E−17 at 104 s and time interval transfer of 1PPS signal with less than 300fs stability after 10000 s are obtained over 106km urban fiber links. This solution not only inherits the advantages of WR's large-scale networking and data transmission, but also improves the performance of frequency transmission. It can transmit frequency signals at hydrogen-maser-level stability in Ethernet networks, and it represents a significant step towards ultrastable time-frequency network distribution in the public telecommunications network.
Funding
National Key Research and Development Program of China (2020YFB0408300, 2020YFB0408301); Strategic Priority Research Program, Chinese Academy of Sciences (XDB21030200); National Natural Science Foundation of China (61535014, 61805260); Youth Innovation Promotion Association of the Chinese Academy of Sciences (YIPA2021244).
Disclosures
The authors declare that there are no conflicts of interest related to this article.
References
1. H. Marion, F. P. Dos Santos, M. Abgrall, S. Zhang, Y. Sortais, S. Bize, I. Maksimovic, D. Calonico, J. Grunert, C. Mandache, P. Lemonde, G. Santarelli, P. Laurent, A. Clairon, and C. Salomon, “Search for variations of fundamental constants using atomic fountain clocks,” Phys. Rev. Lett. 90(15), 150801 (2003). [CrossRef]
2. D. Borio, L. Camoriano, S. Savasta, and L. L. Presti, “Time-Frequency Excision for GNSS Applications,” IEEE Systems J. 2(1), 27–37 (2008). [CrossRef]
3. D. Calonico, C. Clivati, M. Frittelli, A. Mura, M. Zucco, and F. Levi., F. Perini, C. Bortolotti, M. Roma, R. Ambrosini, and G. A. Costanzo, “Time and frequency optical fiber links for space metrology,” in 2015 IEEE Metrology for Aerospace (MetroAeroSpace), (2015), 204–208.
4. S. Weyers, V. Gerginov, M. Kazda, J. Rahm, B. Lipphardt, G. Dobrev, and K. Gibble, “Advances in the accuracy, stability, and reliability of the PTB primary fountain clocks,” Metrologia 55(6), 789–805 (2018). [CrossRef]
5. W.-H. Tseng, S.-Y. Lin, K.-M. Feng, M. Fujieda, and H. Maeno, “Improving TWSTFT Short-Term Stability by Network Time Transfer,” IEEE Trans. Ultrason., Ferroelect., Freq. Contr. 57(1), 161–167 (2010). [CrossRef]
6. W. Lewandowski, J. Azoubib, and W. J. Klepczynski, “GPS: primary tool for time transfer,” Proc. IEEE 87(1), 163–172 (1999). [CrossRef]
7. P. Krehlik, L. Sliwczynski, L. Buczek, and M. Lipinski, “Fiber-Optic Joint Time and Frequency Transfer with Active Stabilization of the Propagation Delay,” IEEE Trans. Instrum. Meas. 61(10), 2844–2851 (2012). [CrossRef]
8. O. Lopez, C. Kamj, P. E. Pottie, D. Rovera, J. Achkar, C. Chardonnet, A. Amy-Klein, and G. Santarelli, “Simultaneous remote transfer of accurate timing and optical frequency over a public fiber network,” Appl. Phys. B 110(1), 3–6 (2013). [CrossRef]
9. P. Krehlik, L. Sliwczynski, L. Buczek, J. Kolodziej, and M. Lipinski, “ELSTAB—Fiber-Optic Time and Frequency Distribution Technology: A General Characterization and Fundamental Limits,” IEEE Trans. Ultrason., Ferroelect., Freq. Contr. 63(7), 993–1004 (2016). [CrossRef]
10. Q. Liu, S. Han, J. Wang, Z. Feng, W. Chen, N. Cheng, Y. Gui, H. Cai, and S. Han, “Simultaneous frequency transfer and time synchronization over a 430 km fiber backbone network using a cascaded system,” Chin. Opt. Lett. 14, 072602 (2016).
11. J. L. Wang, C. L. Yue, Y. L. Xi, Y. G. Sun, N. Cheng, F. Yang, M. Y. Jiang, J. F. Sun, Y. Z. Gui, and H. W. Cai, “Fiber-optic joint time and frequency transfer with the same wavelength,” Opt. Lett. 45(1), 208–211 (2020). [CrossRef]
12. N. Kaur, F. Frank, P. Pottie, and P. Tuckey, “Time and frequency transfer over a 500 km cascaded White Rabbit network,” in 2017 Joint Conference of the European Frequency and Time Forum and IEEE International Frequency Control Symposium (EFTF/IFCS), (2017), 86–90.
13. B. Wang, C. Gao, W. L. Chen, J. Miao, X. Zhu, Y. Bai, J. W. Zhang, Y. Y. Feng, T. C. Li, and L. J. Wang, “Precise and Continuous Time and Frequency Synchronisation at the 5 ( 10(19 Accuracy Level,” Sci. Rep. 2(1), 556 (2012). [CrossRef]
14. J. Levine, “A review of time and frequency transfer methods,” Metrologia 45(6), S162–S174 (2008). [CrossRef]
15. T. Neagoe, V. Cristea, and L. Banica, “NTP versus PTP in Com puter Networks Clock Synchronization,” in 2006 IEEE International Symposium on Industrial Electronics(IEEE-ISIE), (2006), 317–362.
16. J. Serrano, P. Alvarez, M. Cattin, E. Garcia Cota, J. Lewis, P. Moreira, and T. Wlostowski, “The white rabbit project,” in The International Conference on Accelerator and Large Experimental Physics Control Systems (ICALEPCS), Kobe, Japan, (2009), 93–95.
17. P. Moreira, P. Alvarez, J. Serrano, I. Darwezeh, and T. Wlostowski, “Digital dual mixer time difference for sub-nanosecond time synchronization in Ethernet,” in 2010 IEEE International Frequency Control Symposium(IEEE-FCS), (2010), 449–453.
18. D. W. Allan and H. Daams, “Picosecond time difference measurement system,” in Proc. 29th Annu. Symp. Freq. Control, (1975), 404–411.
19. White Rabbit: Users of White Rabbit Technology. [Online]. Available: https://www.ohwr.org/project/white-rabbit/wikis/WRUsers, (2009)
20. M. Rizzi, M. Lipiński, T. Wlostowski, J. Serrano, G. Daniluk, P. Ferrari, and S. Rinaldi, “White rabbit clock characteristics,” in 2016 IEEE International Symposium on Precision Clock Synchronization for Measurement, Control, and Communication (ISPCS), (2016), 1–6.
21. M. Rizzi, M. Lipinski, P. Ferrari, S. Rinaldi, and A. Flammini, “White Rabbit Clock Synchronization: Ultimate Limits on Close-In Phase Noise and Short-Term Stability Due to FPGA Implementation,” IEEE Trans. Ultrason., Ferroelect., Freq. Contr. 65(9), 1726–1737 (2018). [CrossRef]
22. Sub-Nanosecond Time Accuracy and Frequency Distribution through White Rabbit Ethernet, [Online]. Available: https://library.nrao.edu/public/memos/ngvla/NGVLA_22.pdf, (2019)
23. L.-S. Ma, P. Jungner, J. Ye, and J. L. Hall, “Delivering the same optical frequency at two places: accurate cancellation of phase noise introduced by an optical fiber or other time-varying path,” Opt. Lett. 19(21), 1777–1779 (1994). [CrossRef]
24. M. Wang, L. Zhou, Y. Zhou, J. Chen, and Ieee, “BPSK and PAM Modulation in a Single-drive Push-pull Silicon Michelson Interferometric Modulator,” in 2016 Conference on Lasers and Electro-Optics (2016).
25. X. Yuan and B. Wang, “Using single wavelength light to improve the synchronization accuracy of the White Rabbit system,” Chin. Opt. Lett. 15(10), 101202 (2017). [CrossRef]
26. O. Lopez, A. Amy-Klein, C. Daussy, C. Chardonnet, F. Narbonneau, M. Lours, and G. Santarelli, “86-km optical link with a resolution of 2 × 10-18 for RF frequency transfer,” Eur. Phys. J. D 48(1), 35–41 (2008). [CrossRef]
27. M. Amemiya, M. Imae, Y. Fujii, T. Suzuyama, F. Hong, and M. Takamoto, “Precise Frequency Comparison System Using Bidirectional Optical Amplifiers,” IEEE Trans. Instrum. Meas. 59(3), 631–640 (2010). [CrossRef]
