White Rabbit for long-haul fiber-optic distribution of high-precision clocks for VLBI

Size: px

Start display at page:

Download "White Rabbit for long-haul fiber-optic distribution of high-precision clocks for VLBI"

Alisha Day
5 years ago
Views:

1 White Rabbit for long-haul fiber-optic distribution of high-precision clocks for VLBI Tjeerd J. Pinkert (VU) Henk Peek (Nikhef) Peter Janswijer (Nikhef) Paul Boven (JIVE) Arpad Szomoru (JIVE) Erik Dierikx (VSL) Rob Smets (SURFnet) Eduardo Ros (Univ. Gr.) Javier Diaz (Univ. Gr.) Jeroen C. J. Koelemeij (VU) This project has received funding from the European Union s Horizon 2020 research and innovation programme under grant agreement No Institute for Lasers, Life and Biophotonics Vrije Universiteit Amsterdam

2 Optical T-F transfer in the Netherlands Amsterdam (VU) Groningen (RUG) Amsterdam (Nikhef) Delft (VSL) Optical frequency transfer WR T-F transfer Westerbork (WRST) Dwingeloo (CAMRAS) WR T-F transfer

3 Delft Amsterdam: link topology Demonstrator link of 200 km during tour on Wednesday Connect VSL and Nikhef Double WR link with Grandmaster and Slave Two quasi bidirectional optical amplifiers (commercially available from OPNT B.V.)

4 Delft Amsterdam: time domain view

5 Delft Amsterdam: Frequency transfer Internal WR data: dms - setp

6 Delft Amsterdam: Frequency transfer E.F.Dierikx et. al., UFFC, doi: /tuffc

7 Delft Amsterdam: Time transfer E.F.Dierikx et. al., UFFC, doi: /tuffc

8 Delft Amsterdam: Calibration efforts Try to calibrate WR gear with highest possible accuracy Measurements of systematic effects Long haul SFPs with sensitive detectors (can be blown up) Attenuator(s) Power detectors external WDMs Delay calibration of all system components Connector reproduceability Receiver power dependence systematics Link restart reproduceability Temperature dependence (see [1,2]) Test setups always involve: Aim at link accuracy better than 1 ns Setups involved In situ chromatic dispersion determination Measurement scheme [1] H. Li et al., IEEE Trans. on Nuclear Science, 2014 (arxiv: ) [2] M. Lipinsky, May 2013, Torture report (ohwr.org)

9 Connector reproduceability High accuracy delay measurements [3] using cross correlation on Gbit ethernet signals [3] N. Sotiropoulos et al., Opt. Express 21, (2013)

10 SFP receiver optical power dependence Determine the minimum power at which the White Rabbit link stays in lock (to avoid restart spread on measurements) Keep system in lock while measuring! Do this for every SFP in your setup Reproduceability after optical power cycle typically ~5 ps

11 SFP receiver optical power dependence

12 Attenuator comparison Attenuator length change has to be measured independently! Influenced by wavelength dependent systematics of attenuators Mainly important for SFP's with large wavelength difference and internal WDM's

13 WR SPEC restart accuracy SPEC firmware version 2 Locked 8 ns off No lock realised but system says locked no locks observed Link restart gives a different time transferred System regularly fails to lock Expected to find systematics for bitslide pairs

14 WR SPEC restart accuracy SPEC firmware v.3 pre-release Henk Peek will talk more about short term frequency stability Link restart gives a different time transferred Pre-release has failed to lock in 4 out of ~7000 restarts Large spread is possibly a short term frequency stability issue

15 Remarks about relocking Detect locking failure, reset physical connection Investigate cause of large spread in accuracy of relocks mid term jitter of crystal oscillators and PLL's are suspects Relocking currently the largest and most time consuming part in calibrating components. TX power cycle, or be able to always lock correctly About 60 relocks give STE in the order of 5 ps Each relock takes about 60 seconds to gather enough data Each relock takes about seconds for the link restart Then produces about seconds of calibration data SPEC relock currently easiest and most reliable Use crontab: #TJP: reload WR modules for reconnect test... #reload every minute: 0 59/1 0 59/1 * * * * root modprobe r fmc trivial spec fmc; modprobe spec; modprobe fmc trivial gateware=fmc/spec demo.bin Switch relocking: wrs_port_tx_control (together with wrs_pps_control for external PPS)

16 Setups: attenuator length determination Accuracy ~12 ps Correct data for power dependence of photodetectors Average out link restart spread leave attenuators at fixed attenuation due to wavelength dependent delays

17 Including rx power and restarts: equations become a bit nastier. Preliminary, no temperature depedence

18 Setups: WDM length determination Accuracy ~12 ps Correct data for power dependence of photodetectors Average out restart spread

19 Golden calibrator calculation Preliminary

20 Golden calibrator calculation Accuracy ~30 ps Preliminary, no temperature depedence

21 In-situ chromatic dispersion determination Scheme by Henk Peek and Peter Jansweijer Assumes system fibres have equal (enough) dispersion Assumes equal wavelengths of CWDM lasers in SFPs Well suited for long-haul links where fibre pairs are common

Frequency and time transfer for VLBI H-maser over fiber <10-12 frequency stability at 1 s <10-14 frequency stability at > 1000 s translates roughly into < 10 ps

22 Frequency and time transfer for VLBI H-maser over fiber <10-12 frequency stability at 1 s <10-14 frequency stability at > 1000 s translates roughly into < 10 ps timing jitter/drift at 1000 s Piggyback on existing fibre networks outside of C-band bidirectional fibre channels bidirectional optical amplifiers Patent no: WO A1

23 White Rabbit instead of H-masers See also talk of Paul Boven

24 Clock requirements for VLBI VLBI datapoints for 22 GHz Even a H-maser is not yet an ideal clock for VLBI our goal: VLBI at < 5GHz Rb oscillator is good enough Our aim: make White Rabbit as good as H-maser Ideal clock: σy at 0.1 to 1000 seconds σy < 10-11/τ at > 1000 seconds figure: Rogers et al., IEEE Trans. Instr. Meas. IM-30, (1981)

25 Clock requirements for VLBI VLBI datapoints for 22 GHz Even a H-maser is not yet an ideal clock for VLBI our goal: VLBI at < 5GHz Rb oscillator is good enough Our aim: make White Rabbit as good as H-maser Ideal clock: σy at 0.1 to 1000 seconds σy < 10-11/τ at > 1000 seconds figure: Rogers et al., IEEE Trans. Instr. Meas. IM-30, (1981)

26 White Rabbit for VLBI, WR internal data passive link stability slave output stability

27 White Rabbit for VLBI, WR internal data short term stable oscillator locked to WR link output passive link stability slave output stability

28 White Rabbit for VLBI, WR internal data short term stable oscillator locked to WR link output external measurement on short fiber passive link stability slave output stability

29 White Rabbit for VLBI, WR internal data WR is not up to spec on short term frequency stability. Better short term stability gives better long term stability short term stable oscillator locked to WR link output external measurement on short fiber Option 1: Use cleanup oscillators > a piece ADEV: < s passive link stability slave output stability Option 2: Improve short term locking stability of WR equipment Large bandwidth feedback on the quartz oscillators Improved performance of clock recovery circuits Better board layouts (Henk Peek will talk more about this) Which strategy is best for long-haul performance? or is a combination the solution?

30 Conclusions Systematics investigated for high performance time transfer: Measurement setups have been developed Connectors: SMA good, BNC 'bad' SMC, LEMO-00 not tested yet Link restarts: Need averaging out Photodetectors: Power dependence is significant Amsterdam Delft link: Calibration of components almost ready In-situ chromatic dispersion measurement in preparation Experience and learning for Westerbork Dwingeloo link Westerbork Dwingeloo link: Components for physical link have been ordered Work on strategy for high performance frequency transfer progresses (see also talks of Henk Peek and Paul Boven)

31 Outlook Amsterdam Delft with < 100 ps accuracy? Absolute calibration (O E and E O of SFPs) Setups for high-volume and high-accuracy delay characterisation Phase locked oscillators on physical layer H-maser replaced by WR (slave), do VLBI Needed: Calibration parameters that need to be included: system level delay asymmetries (static, dynamic) system temperatures (dynamic) receiver optical powers (dynamic) System level management architecture? updating upon change of hardware (dynamic) e.g. when changing SFPs, FMCs, SPECs or WR Switches System level measurement architecture e.g. for near real time dissemination of TAI/UTC

32 Questions?

33 Challenges in large network calibration Agregator Calibration tables

34 Challenges in large network calibration Characterisation of dependencies: Online monitoring of parameters: temperature, power, wavelength, others schemes have been worked out how to do this. agregate all parameters and determine dependent delays needs a component capable of doing this might need a standardized scheme? Dynamic updating of endnode calibration parameters: is it already possible? (Yes?, see [1]?) might need a standardized scheme?

Methods for data, time and ultrastable frequency transfer through long-haul fiber-optic links

Methods for data, time and ultrastable frequency transfer through long-haul fiber-optic links Jeroen Koelemeij, Tjeerd Pinkert, Chantal van Tour (VU Amsterdam, NL) Erik Dierikx (VSL Delft, NL) Henk Peek,