Methods for data, time and ultrastable frequency transfer through long-haul optical fiber links Jeroen Koelemeij LaserLaB & Depart. Physics and Astronomy VU University Amsterdam, The Netherlands Outline • Why time & frequency through optical fiber? • (Ultra)stable fiber-optical frequency transfer • Accurate fiber-optical time transfer • Integration into high-capacity fiber-optical telecom infrastructure and application to VLBI • (Ultra)stable fiber-optical frequency transfer Partners/collaborators in the Netherlands: Tjeerd Pinkert VU Amsterdam Chantal van Tour VU Amsterdam Wim Ubachs VU Amsterdam Kjeld Eikema VU Amsterdam Roeland Nuijts SURFnet Rob Smets SURFnet Oliver Böll KVI Groningen Lorentz Willmann KVI Groningen Klaus Jungmann KVI Groningen JK Optical path length stabilization Compensation of frequency fluctuations due to length fluctuations*: *L.-S. Ma, P. Jungner, J. Ye, J.L. Hall, Opt. Lett. 19, 1777(1994) 1.5 mm clock laser roundtrip contains 2× noise! Partial reflector Clock laser + noise power power power Noise detection PLL +compensation Optical fiber (~ 100 km) Laser frequency in Laser Laser frequency frequency out out Example: 920 km link PTB group (Braunschweig, Germany): K. Predehl et al., Science 336, 441 (2012) 1840 km link: S. Droste et al., Phys. Rev. Lett. 111, 110801 (2013) Free-running link Germany H-maser Transport through telecom fiber • Fiber attenuation: 20 dB/100 km, need amplifiers! • Issue: bi-directional optical amplifiers needed, but telecom amplifiers are uni-directional (to avoid lasing) Scattered EDFA • Two approaches: 1. Dark fiber (no other signals, us bi-di amp) 2. Dark channel (bi-di ‘bypass’ amplifier) optical isolators (Paris groups, O. Lopez et al., Appl. Phys. B 110, 3 (2012)) Location A Both approaches work Both approaches sacrifice telecom capacity Approach 2: additional insertion loss Telecom operators often reluctant Location B Bidir amp Part of the solution: out-of-band channels • Use out-of-band wavelength channels – C-band: 1530 nm – 1565 nm erbium-doped fiber amplifier (EDFA) gain spectrum – Use semiconductor optical amplifiers (SOAs) for signal amplification <1530 nm – Ease of wavelength multiplexing with standard components … but does it work for optical frequency transfer? Lab test on 5 km spooled fiber (Amsterdam) EDFA SOA Max. gain [dB] 25-30 20-25 Max. bi-di gain [dB] <25 <25 Noise Figure [dB] 6-8 8-10 Nonlinearity (keep Pin low) Results H-maser 5 km link + SOA 5 km link YES SOA adds a small amount of noise, but link stability still far below the stability of optical clocks (and masers)! Work in progress: compare performance SOAs with EDFAs From lab to field: SURFnet optical fiber link • • • • Link part of SURFnet DWDM network Length 317 km, round trip 635 km Single l-channel (1559.79 nm) Fiber carrying live data traffic • Optical clocks under development at both ends of fiber link • Fiber connects to JIVE Dwingeloo • Future: bi-directional fiber link • Accurate fiber-optical time transfer Partners/collaborators in the Netherlands: Nikos Sotiropoulos TU Eindhoven Chigo Okonkwo TU Eindhoven Huug de Waardt TU Eindhoven Tjeerd Pinkert VU Amsterdam Roeland Nuijts SURFnet Utrecht Rob Smets SURFnet Utrecht Martin Fransen VSL Delft Erik Dierikx VSL Delft Henk Peek NIKHEF Amsterdam JK Time transfer – the state of the art Method Distance Accuracy Ref. GNSS >1000 km 3 – 50 ns TWSTFT >1000 km 1 ns T2L2 >1000 km 200 ps expected White Rabbit (fiber) (1 Gpbs Ethernet, PTP) 10 km 0.1 - 1 ns Optical fiber (20 Mbps PRBS) 540 km 100 - 250 ps Lopez et al., Appl. Opt. (2012) Optical fiber (20 Mbps PRBS) 73 km 74 ps Rost et al., Metrologia (2012) Dedicated optical fiber (10 MHz + 1pps) 69 km (480 km) 8 ps (35 ps) Sliwczynski et al., Metrologia (2013) Fridelance et al., Exp. Astr. (1997) www.ohwr.org Approach LaserLaB VU – TU Eindhoven • Collaboration funded by SURFnet, setup at TU Eindhoven • Find delays via XCOR of 10 Gb/s bit streams through 75 km fiber link Quasi-bidirectional amplifier (Amemiya et al., IEEE IFCSE 2005) 25 km 50 km Two round-trip delays measured: t12 (l1, l2) and t13 (l1, l3) Advantages: • Transmit 10 Gb/s data, no telecom capacity sacrificed • Time + data transfer • Compatible with existing telecom methods & equipment PRBS signals and correlation 50 GS/s 75 km 12.5 GS/s 150 km Results OWD - tAB(t) [ps] Time difference= <OWDestimate> - <OWDdirect> Estimated accuracy: 4 ps (agrees with observations) 75 km link Bit-error rate (BER) below 10-9 : Error free communication at 10 Gb/s Measurement number 25 255(1) 50 405(1) 75 552(1) [ps] [ps] [ps] I 3.4 3.4 3.4 DPO time base stability 0.8 1.0 1.7 Fit uncertainty 1.5 1.5 1.0 VOAs 1.0 1.0 1.0 PMD correction 0.6 1.0 0.6 Wavelength measurement 0.2 0.5 0.7 XCOR interpolation 0.3 0.3 0.3 Estimate n′′′ 0.05 0.1 0.1 SPM and XPM <0.1 <0.1 <0.1 Fast fiber length fluctuations <0.1 <0.1 <0.1 Sourcea b Link length uncertainty Total <10 -4 4.0 <10 -4 4.1 75 km -log BER Link length [m] 25 km 0 km 50 km <10-4 4.2 Received power [dBm] Delivery of 10 Gb/s optical data with 4 ps accuracy over 75 km distance 75 km link Measurement number 75 km -log BER OWD - tAB(t) [ps] Results 25 km 0 km 50 km N. Sotiropoulos et al. (submitted) Received power [dBm] Time transfer – the state of the art Method Distance Accuracy GNSS >1000 km 3 – 50 ns TWSTFT >1000 km 1 ns T2L2 >1000 km 200 ps expected White Rabbit (fiber) (1 Gpbs Ethernet, PTP) Optical fiber (20 Mbps PRBS) State-of-the-art 10 km delay determination 0.1 - 1 ns + Error-free optical data 540 km 100 - 250 ps transfer at 10 Gbps Ref. Fridelance et al., Exp. Astr. (1997) www.ohwr.org Lopez et al., Appl. Opt. (2012) Optical fiber (20 Mbps PRBS) 73 km 74 ps Rost et al., Metrologia (2012) Dedicated optical fiber (10 MHz + 1pps) 69 km (480 km) 8 ps (20 ps) Sliwczynski et al., Metrologia (2013) Cross correlation of 10 Gbps optical data 75 km 4 ps Sotiropoulos et al. (submitted) Speed bonus • Delay determination/synchronization requires a single shot of 10 Gb/s data lasting less than 1 ms – For comparison: state-of-the-art methods require 10-100 s of averaging to achieve 4 ps stability • Integration into high-capacity fiber-optical telecom infrastructure and application to VLBI Use out-of-band wavelengths integrate time and frequency transfer in hardware for high-capacity optical telecom Data out T&F out Fiber in Will require involvment of manufacturers of optical telecom network equipment and NRENs… … AND a convincing test case! eVLBI using fiber-optical synchronization? Application to eVLBI? Disclaimer: not necessarily limited to Europe! • 10 Gb/s channel for antenna signal transport • Synchronize LO’s at telescope sites through fiber to 4 ps = (1/5) of a 50 GHz cycle – Useful for initial calibration? • Phase-lock 10 Gb/s to stable ‘Master clock’ and distribute through stabilized fiber links – Phase lock LO to recovered clock at remote sites Master • Use low-noise TCXO/OCXO for short-term clock stability • Use recovered clock for long-term stability – Do away with expensive H-masers? Special thanks to Paul Boven and Arpad Szomoru of JIVE for insightful discussions about eVLBI Work in progress… • Demonstrate time transfer VSL-VU-SARA-NIKHEF • Ultrastable frequency transfer VU – JIVE Dwingeloo – KVI • Test new techniques that do not affect/sacrifice telecom capacity and performance • Demonstrate an optical GPS-timing backup system • Develop terrestrial optical4 ps 2.4 mm accuracy (4D positioning) wireless positioning with cm accuracy (with TU Delft Aperture synthesis through mobile handsets? SuperGPS Thanks!