1. E-VLBI Activity in NICT National Institute of Information and Communications Technology (NICT) Kashima Space Research Center M. Sekido, 、 Y. Koyama, M. Kimura, H. Takiguchi, T. Kondo, T. Hobiger, T. Ikeda, H. Harai, M. Hirabaru

2. E-VLBI Activity in NICT E-VLBI Project JGN2 Sympo07 Demo: real-time e-VLBI with Haystack Realtime e-VLBI(Kashima,Tsukuba,Gifu,Yamaguchi,Usuda) Ultra-rapid UT1 measurement (Onsala,Tsukuba,Metsahovi) 6Gbps S/X observation on Kashima-Koganei(100km) baseline. E-VLBI demonstration (ATNF-Kshima-Seashan) for today Hardware development Two types of K5 system(K5/VSSP, K5/VSI) ADS-3000plus(DBBC) development Benefit of Standard

3. Mark5B JGN2Sympo2007 with Haystack

4. Ultra-rapid UT1 Experiment Metsahovi14m (Finland ） Kashima34m （ NICT) JGN2 Onsala20m (Sweden ） Geant2 NorduNET Abiline/ Internet2 TransPAC Tsukuba32m （ GSI) Within 5 min after the session, UT1 has derived! Collaborators: R.Haas(Onsala), J.Ritakari, J.Wagner(Metsahovi), S.Kurihara, K.Kokado(GSI) APAN E-VLBI Projct(1)

5. DAS, Transport, Processing Work shared collaboration Observation and data transport at European side. Observation and Correlation processing at Japense side. Metsahovi14m (Finland ） Onsala20m (Sweden ） Mark5 VSIB (Metsahovi) E-VLBI Projct(1) K5/VSSP 128/ 256/ 512Mbps Mk5 K5 Software Converter (K5 package) Correlation processing by K5 software correlator on cluster of PCs Ultra-rapid UT1 measurement Tsunami/UDP

6. Prediction (BulletinA),EOPc04, and e-VLBI E-VLBI Projct(1)

7. Rapid Solution (Bulletin-A) and e-VLBI – EOPc04 E-VLBI Projct(1)

8. Success of 8Gbps real-time VLBI E-VLBI Projct(2) 1 １０ km VOA200 JGN2 10G-Ethernet VSI-H x4 Koganei 11m Headquater NICT, Tokyo Kashima 34m NICT,Kashima VSI-H x4

9. VSI-H, VSI-S, VSI-E Sampler DIMDOM VSI-H VSI-S VSI-E Correlator E-VLBI VSI-H RTP/RTCP Mark5B

10. 6Gbps routine VLBI for V773tau VOA200 JGN2 10G-Ethernet Koganei 11m Kashima 34m Name:V773 Tau A d=148pc Binary stars M:1.5M 1.3M a=0.4AU P=51 days Monitoring of Flux S/X dual Freq. It is expected flare up with binary orbiting period. Variation of Spectrum index associated with orbiting period. E-VLBI Projct(2)

11. The first real-time e-APT JGN2Plus CSTNET Parks 64m Telescope Sydney Los Angeles Seattle Kashima 34m Telescope Shanghai 25m Telescope ATCA 22m Telescope Mopra 22m Telescope AARNet E-VLBI Projct(3)

12. A brief History of VLBI development in NICT From K3-system to K5-system K3 System K4 (KSP) System 1983~ Longitudinal Recorder Open Reel Tapes Hardware Correlator 1990~ Rotary Head Recorder Cassette Tapes Hardware Correlator e-VLBI with ATM(256M) K3 Correlator (Center) K3 Recorder (Right) K4 Terminal K4 Correlator K5 System K5 Data Acquisition Terminal 2002~ PC based system Hard Disks Software Correlator e-VLBI with IP

13. K5-System （１） : K5/VSSP K5/VSSP K5/VSSP32 K5/VSSPK5/VSSP32 Sampling Speed40, 100, 200, 500kHz, 1, 2, 4, 8, 16 MHz 40, 100, 200, 500kHz, 1, 2, 4, 8, 16, 32, 64MHz, Sampling Bits1, 2, 4, 8 No. Channels1 or 4 (16 with 4 PCs) Max. Data Rate128 Mbps (512 Mbps with 4 PCs)256 Mbps (1024Mbps with 4PCs) InterfacePCI (Full Height)USB2.0 VSSP = Versatile Scientific Sampling Processor

14. ADS3000 ADS1000 ADS2000 ADS1000ADS2000ADS3000ADS3000Plus Sampling Speed1024Msps64Msps2048Msps ～４ Gbps Sampling Bits1 bit or 2 bits 8 bits2/4/8 bit No. of Input11612 No. Channels116Programmable Max. Data Rate2048Mbps 4096Mbps8192Mbps InterfaceVSI-H (2 ports) VSI-H (4 ports) K5 System （２） : K5/VSI ADS3000Plus

15. ADS1000 (1024Msample/sec 1ch 1 or 2bits) ADS2000 (64Msample/ch·sec, 16ch, 1 or 2bits) PC : Data Acquisition Correlation Correlator other DAS Internet PC-VSI Board (~2048Mbps) ADS3000 (2048Msample/sec 1ch 8bits + FPGA) K5/VSSP32 Unit (~32Msample/ch·sec, ~4ch, ~8bits) K5 System Mark5B sampler (64Msample/ch·sec, 16ch, 1 or 2bits) VSI-H

16. K5/VSI DBBC AD サンプラー： ADS3000 4Gbps (2GHz, 2sps) 初フリンジ （ 2006 年 1 月）

17. ADS3000 plus under development Sampling Mode ModeTotal Rate/ch Sample (Msps) Quanti zation Clock (MHz) A1Gbps1288bit32 B2Gbps10242bit32 C2Gbps5124bit32 D4Gbps20482bit64 E4Gbps10244bit64 F4Gbps5128bit64 Dual Channel sampler For S/X observation DBBC Function VSI-H Compliant

18. ADS3000 + PC-VSI Recorder 4096Mbps recording for １７ hours

19. Summary VSI-H was quite successful, and useful. Mark5B-K5 ＠ JGN2Symp2007 Mark5-VSIB@Ultra-rapid UT1 NICT-NAOJ Devices(ADS-X000,VOA100/200,Software Corr(VERA) Sharing standard interface reduces the cost and brings freedom and benefit for VLBIers. E-VLBI era is good chance to make all VLBI system fully compatible/connectable. Handling data via software/network is great advantage. We may discuss about standard or framework for taking compatibility. Defining Standard Data Transport protocol

20. Thank you for attention!

21. Accuracy and latency of UT1 As an example of requirement: 10usec of accuracy(100m at Mars) can enables use of ar dragging and greatly reduce fuel weight for the space prove. E-VLBI Projct(1)

22. What will happen when global e- VLBI is operated routinely? At present, No problem for one stream Europe-Japan with shared network. Because of no congestion. What about multiple streams? In case of congestion, data rate drastically decreases, especially in TCP/IP. Even Tsunami(UDP) tries to send complete data set by re-transmission mechanism, and … finally fails.

23. What is the Optimal transport protocol for real-time VLBI? Features of Realtime VLBI are Fixed Data Rate or slightly adjustable. Relatively large error rate(<0.1) is acceptable, if padding and flagging is done. Only Header must be preserved. “Optimal protocol for realtime e-VLBI” may looks like Sending data in fixed rate regardless congestion Just sending out by UDP (without re-transmission) VLBI data may win the resource race, because it accept largest error rate than any other application. This may be good for VLBI, but….

24. Conflict between Network and VLBI Network is assuming Faire users, who shares network bandwidth equally. ex).TCP reduces transmitting rate in congestion condition. Video stream may adjust the rate by changing the compression or resolution of image. Rate control is expected. But VLBI cannot change the rate in case of real- time VLBI. “Optimal protocol for VLBI” may defeat other data streams, but it may be criticized and kicked out from network. ⇒ Not good for VLBI

25. Possible Solutions are Bandwidth on demand connection Realized by GMPLS and started in operation in SINET3. Using Disk/Memory buffer to enable rate control… but it will not guarantee the success of realtime VLBI. Switching to offline or best effort basis VLBI. ⇒ This may work with software correlator. Bandwidth on demand connection will become available in global network in future.