e-VLBI at the NRAO R. Craig Walker NRAO, Socorro NM Connecting the Global VLBI Array in the New Era of High-Speed Networks Haystack, April 2002 INTRODUCTION VLBA Upgrades (Desirable, but no definite plans or budgets): – – – Recording system upgrade Connect current systems and correlator at ~1 Gbps Long term integration with EVLA at up to 96 Gbps Real time data link projects at NRAO: • Pie Town Link: In active use. Analog, 200 MHz. • ALMA and EVLA: Funded projects – ~20 km. 96 Gbps. Similar technology for both. – – • Observatory owned and installed fibers. No regeneration. Technology not directly applicable to VLBI. NMA: ~8 new antennas for EVLA. To be proposed. – – – eVLBI April 2002 ~300 km. 96 Gbps. Phone company fibers. Technology partially applicable to VLBI. Craig Walker The VLBA Ten 25m Antennas, 20 Station Correlator 327 MHz - 86 GHz National Radio Astronomy Observatory A Facility of the National Science Foundation WHY CONSIDER REAL TIME FOR THE VLBA? • Recording system forces large operations cost – – – – – Separate correlator operations group Correlator job preparation and “sniffing” Shipping and tape library Maintenance of many systems (about 50 now) Tape changing at sites • Real time gives more robust operation of instrument – – – – – Real time feedback on performance No lost or damaged tapes No marginal playback Recordings have always been the weak link in VLBI More interferometer testing (pointing etc) • A switch to real time is operations rather than science driven – But may help geodesy (rapid EOP) and observations of transients • But give up multi-pass processing unless data are buffered eVLBI April 2002 Craig Walker POSSIBLE INITIAL VLBA REAL TIME SYSTEM • Direct replacement for recording system – Or use recording system for short term buffer • Use commercial fiber networks and data formats • Use current correlator – 512 Mbps max per station, 20 stations (1 or 2 OC12 lines) – 1 Gbps for 10 stations • Partial step - move some recorders to correlator • Expect mixed recording and real time operation for some time eVLBI April 2002 Craig Walker REAL TIME VLBA CONSIDERATIONS • Network backbones can handle even EVLA bandwidth – Some doing 1.6 Tbps per fiber (many wavelengths) – But costs would be prohibitive now. Going down • Last mile problems for VLBA not too bad – – – – PT, NL, BR, VLA have fibers at site KP, MK have fibers on the mountain HN, FD, SC, OV, LA have fiber within 3-10 km Others like GBT and Arecibo might be a problem • May need to deal with many companies • Need plan to deal with global observations – Need international standards – Need access to intercontinental lines eVLBI April 2002 Craig Walker REAL TIME VLBA QUESTIONS • Cost and access are likely to be the problems, not technical feasibility. • What is best model for acquiring access to the fiber? – – – – Leased bandwidth Leased fiber “Customer owned fiber” Other • When will the cost be affordable? – Standard leased bandwidth costs would be way to high now – Special deals may be possible • VLBI has relaxed requirements for error rates and down time • Only need wide band in one direction eVLBI April 2002 Craig Walker EVLA & ALMA DATA TRANSMISSION • EVLA and ALMA will use the same system • Fibers to be installed by the observatory. • Data rate is 96 Gbps – 16 GHz bandwidth, 32 Gsamp/s, 3 bit/sample • OC192 (10 Gbps) transmitters and receivers • Will use 12 channels • Special formatters and data format – No error correction • Single channel prototype working in lab • Not appropriate for commercial networks eVLBI April 2002 Craig Walker NEW MEXICO ARRAY • Extend the VLA resolution by a factor of 10. – Part of the EVLA completion phase – This phase of the EVLA project is not yet funded • About 8 new antennas with baselines 35-350 km. – Will also use the VLBA LA and PT antennas. • Will use telephone company fibers – Small rural companies – several – May lease just fiber. NRAO to provide regenerators etc. Cost appears to be reasonable. • Will need 16 Sonet OC192 channels for 96 Gbps plus overhead. – Will not need the high availability and low error rates required by most internet users • Technology may be appropriate for VLBI eVLBI April 2002 Craig Walker VLBA/EVLA INTEGRATION • NMA fills the gap in uv spacings between VLA and VLBA • Some VLBA/EVLA integration is required because 2 VLBA stations will be part of NMA (LA and PT) – But integration is not yet in the completion phase budget • EVLA will have 500 times the VLBA current sustained bandwidth (16 GHz vs 32 MHz)! – Reasonable integration requires VLBA bandwidth upgrade. – More bandwidth would enhance VLBA/NMA science • Observe microJy sources • Improved phase referencing with weaker and closer calibrators • Study thermal sources with high resolution eVLBI April 2002 Craig Walker Resolution vs. Frequency VLBA/EVLA INTEGRATION: CORRELATOR • Being built by HIA group in Penticton, Canada • Basic capabilities: – – – – 32 Stations in EVLA first phase, 40 or 48 later with NMA 16 GHz bandwidth, 3 bit samples 96 Gbps from each antenna 16384 spectral channels at maximum bandwidth • Can handle VLBI baselines – Could interface recorders, but prefer real time – Can trade stations and bandwidth – Uses VLBI standard bandwidths (N×1 MHz) • EVLA correlator could serve VLA/NMA/VLBA – Becomes one large instrument eVLBI April 2002 Craig Walker EVLA/VLBA INTEGRATION: STATION ENHANCEMENTS • A very wide band data transmission system will require upgrades in other areas • EVLA/NMA will have complete frequency coverage over 1-50 GHz and will use 8 GHz in each polarization • VLBA receivers and IFs more limited, but better than current VLA • At 1 GHz bandwidth, could sample current VLBA IFs and use current receivers • For more bandwidth, must upgrade LO/IF systems and some receivers eVLBI April 2002 Craig Walker SUMMARY • Real time links would be a boon for VLBA operations. – Could use up to 1 Gbps with current correlator etc. • NRAO is developing very wide band links for distances up to a few hundred km for NMA. • The NMA will require much of the same technology needed for wideband eVLBI. • Eventually the EVLA and VLBA must be integrated. – Links of 16-96 Gbps needed. – Must be compatible with rest of global VLBI • The big question: How do we do it at an affordable cost in a reasonable time scale? eVLBI April 2002 Craig Walker