Vladimír Smotlacha, CESNETvs@cesnet.cz

Accurate Time Transfer over Optical Network

6th CEF Networks Workshop

Prague

13 September 2010

Time and frequency standards

• pendulum – frequency standard until 1930s

• quartz crystal

• Cesium (Rubidium) clock – elements of IA periodical table group (single electron in level s)

– hyperfine levels transition emits microwave frequency photon

• Hydrogen maser

• near future: quantum logical clock– transition (in optical frequency) of isolated ion kept in

electromagnetic field trap (Al, Hg, Sr, ...)

Time transfer

• 1 ns represents 30 cm light path in vacuum, ~20 cm in cable or fiber

• cable, optical link– short distance, negligible environmental influence (e.g. thermal

dilatation)

• radio broadcast systems• satellite navigation systems (GPS, ...)

– time broadcast – tens of nanoseconds noise– timestamping of local clock – “common-view”, “all-in-view”,

similar signal propagation in geographically close localities, accuracy below 1 ns (e.g. GTR50)

• two-way satellite transfer (TWSTFT)– assumes equal propagation delay in both directions

• two-way optical link time transfer

Time transfer over fibre

Goal

• Design device and method for accurate time transfer – alternative to satellite based methods

• Comparable or better accuracy and stability on range ~1000 km

• Use existing DWDM all-optical networks

Adapters

Features

• Two-way transfer

• Optical signal modulation for 1-pps encoding

• Uses SFP transceivers

• Based of FPGA Virtex-5

• Requires two time interval counters

Time transfer principle

xA = TrA – TA xB = TrB – TB

εA = TsA – TA εB = TsB – TB

δAB = (xA – εB – ΘAB ) δBA = (xB – εA + ΘAB ) assume δAB = δBA

ΘAB = ((xA – xB ) + (εA – εB )) / 2

Adapter prototype

Cesnet DWDM network

Experiments

We made 3 experiments:• Optical loop measurement

• Time transfer Cesnet – BEV (Prague – Vienna)

• Comparison with GPS time transfer

Participants:• IPE (Institute of Photonics and Electronics), Czech national time and

frequency laboratory, Prague

• BEV (Bundesamt für Eich - und Vermessungswesen), Austrian national time and frequency laboratory, Vienna

• ACOnet, Austrian NREN, Vienna

• West Bohemian University, Plzen

Optical loop experiment

• Both endpoints in one laboratory, common clock

• Bidirectional optical loop length 744 km

• DWDM production network

• 4 segments, 3 fiber providers

• 12 optical amplifiers

• Various optical elements (Cisco, CzechLight)

• Segment Praha – Hradec Kralove on top of electricity distribution poles

Optical loop - geography

Praha - Brno                          284 km Brno - Olomouc                             113 km

Hradec Králové - Olomouc            197 kmPraha - Hradec Králové                  150 km

------------

                                   744 km

Optical loop - results

One-way delay in both directions• fluctuation ~130 ns (temperature changes about 12 °C)• aerial fiber on top electricity distribution poles • residual asymmetry < 2 ns (resp. TDEV 8.7 ps / 500 s)

Optical loop – results II

Prague – Vienna experiment

Time transfer between Cesnet and BEV

• Site A: Rb clock in Cesnet, Prague, GPS disciplined

• Site B: Rb clock, BEV (resp. Vienna university), free

running

• 506 km, DWDM in production network, (Prague – Brno –

Vienna)

Prague – Vienna results

Step in one-way delay (2.3.2010 10:13 UTC)– direction to Prague +72 ns (cca +14 m)– direction to Vienna +16 ns (cca +3 m)

Free running Rb clock: relative frequency offset 8.08 *10-12

Optical x GPS time transfer

Comparison with GPS time transfer

• Site A: free running Rb clock in Plzen (West Bohemian

University)

• Site B: Cs clock in IPE – UTC(TP)

• 150 km of fibre, WDM, production network

• GTR50 installed in both sites

• 95 km geographical distance

Optical x GPS transfer - results

• 10 days measurement• comparison with GTR50 – exact calibration system based on GPS • difference in range ± 2 ns• short time TDEV (time stability) 120 ps

Optical x GPS transfer – results II

• Optical transfer is more stable for averaging time up to 100 s then “common-view“ method

• Equal stability for longer averaging time interval (influenced by Rb clock rather then the measurement method)

Conclusions

• Adapters prototypes successfully tested and method verified

• No influence to other DWDM traffic

• Fiber length dilatation cancels in two-way transfer (residual asymmetry less than 1 ns at 744 km fiber in 6 day test)

• TDEV 8.7 ps / 500 s at 744 km fiber

• Better TDEV then “common-view” method (averaging interval up to 100 s, distance about 100 km)

Future work

• Build all-optical path between BEV and IPE

• Convert experimental method to service – timescale

comparison

• TIC integration into FPGA structure

• Design adapters suitable for time distribution – no data

processing at server side

Thank you