1. 1/12 A. Luchetta 17 th Real-Time Conference, 25 May 2010, Lisboa, Portugal. Data Acquisition in the ITER Ion Source Experiment Adriano Luchetta, Gabriele Manduchi, Antonio Barbalace, Anton Soppelsa, Cesare Taliercio Consorzio RFX – Euratom-ENEA Association, Padova, Italy adriano.luchetta@igi.cnr.it Summary  Introduction  Requirements  Software frameworks  EPICS/MDSplus integration  Real-time performance  Conclusions CAD view of SPIDER Vessel

2. 2/12 A. Luchetta 17 th Real-Time Conference, 25 May 2010, Lisboa, Portugal. Introduction (1) - Context  Tokamaks require additional heating to reach fusion-relevant parameters, as ohmic heating is limited by instability at a given toroidal magnetic field value.  Additional heating adds controllability:  Plasma rotation, current profile control  MHD instability control  Additional heating is provided by:  Radio Frequency coupled to plasma (ion and electron cyclotron, lower hybrid)  Heating H 0 /D 0 Neutral Beams  Heating H 0 /D 0 Neutral Beams (HNB) injected into plasma neutral to reach plasma core without magnetic deflection  ITER will have #2 HNBs Injectors with option for 3. Fig.1. JET toroidal chamber.

3. 3/12 A. Luchetta 17 th Real-Time Conference, 25 May 2010, Lisboa, Portugal. Fig.2. CAD view of ITER HNB. Introduction (2) - Neutral Beam Test Facility  Required parameters for ITER HNBs are well beyond achievements in devices developed so far. Neutral Beam Test Facility  To develop full-size HNB and test it up to maximum performance, ITER approved construction of ad-hoc Neutral Beam Test Facility H°D° Beam Energy1MeV Beam Power16.5MW Beam-on time3600s Table I.Main parameters of ITER HNB.  It is under construction in Padova, Italy, and will comprise 2 test-beds  Full-size Ion Source – op. 2013 Called SPIDER Called SPIDER  Full-size HNB ITER CODAC compatible

4. 4/12 A. Luchetta 17 th Real-Time Conference, 25 May 2010, Lisboa, Portugal. Introduction (3) – Ion Source Experiment Fig.3. Operating principle of SPIDER.Fig.4. CAD view of SPIDER. 6m 4m Water-cooled Beam-dump UnitHD Beam energykeV100 Maximum Beam Source pressurePa<0.3 Uniformity%±10 Extracted current densityA/m2>355>285 Beam-on times3600 Co-extracted electron fraction (e-/H- or e-/D-) <0.5<1 Table II. Main parameters of SPIDER.

5. 5/12 A. Luchetta 17 th Real-Time Conference, 25 May 2010, Lisboa, Portugal. Data Acquisition Requirements (1) SPIDER Diagnostics DescriptionMeasurement(s) ThermocouplesTemperature Electric MeasurementsVoltage, current, power CalorimetryCoolant flow, temperature, pressure, thermal power Electrostatic ProbesElectron temperature and density Source SpectroscopySource plasma parameters Beam TomographyBeam position, shape, uniformity Beam SpectroscopyBeam uniformity, beam divergence, stripping losses Cavity Ring Down Spectr.Ion source H - density Neutron/X Ray DiagnosticsNeutron flux and X-Ray Instrumented Calorimeter Beam uniformity and divergence (limited up to 10s) Diagnostics ImagingVisible beam images and beam dump temperature Table III. SPIDER Diagnostics Systems.

6. 6/12 A. Luchetta 17 th Real-Time Conference, 25 May 2010, Lisboa, Portugal. Data Acquisition Requirements (2) - Quantity Table V. Estimated data amount. Table IV. Channel number and estimated data throughput. Data Class Sampling rate Chan. No. Unit Throughput Beam dump Throughput Instr. Calor. PLC-based≤ 20 S/s136kB/s2.02 Continuous≤ 20 kS/s649MB/s3.113.23 Event-driven≤ 10 MS/s74MB/s1.14 Burst≤ 100 MS/s1MB/s3.00 Diagn. Images≤ 150 fps76MB/s88.93121.07 TOTAL 936MB/s96.18129.07

7. 7/12 A. Luchetta 17 th Real-Time Conference, 25 May 2010, Lisboa, Portugal. Software Frameworks (1)  Control & data acquisition will be implemented by open source, collaborative software frameworks:  EPICS(ITER-driven)  EPICS for control (ITER-driven)  MDSplus  MDSplus for data acquisition  MDSplus is a set of data management tools:  Data acquisition system (hardware configuration, data read-out)  Remote data access system  Data visualization and analysis system data available via FORTRAN, C, C++, Java, idl, matlab, visual basic, labview, php, python data available via a unified object model in python, java, c++, matlab  Data archival system based on a shared record store (pulse file) Hierarchical; Simple API Does not distinguish between classes of data (python-like)

8. 8/12 A. Luchetta 17 th Real-Time Conference, 25 May 2010, Lisboa, Portugal. Software Frameworks (2) - MDSplus  There have been over 8000 downloads of MDSplus installation kits  Sites using complete MDSplus system: RFX (Euratom/ENEA - Italy),  TCV (EPFL - Switzerland), RFX (Euratom/ENEA - Italy), Heliac (ANU - Australia),MST (U. Wisconsin), HIT, TIP, TCS and ZAP (U. Washington), PISCES (UCSD), CHS (NIFS - Japan), LDX (MIT), HBT-IP and CTX (Columbia U.), Alcator C-Mod (MIT)  NSTX (PPPL) and KSTAR (NFRI - S. Korea) use MDSplus and EPICS.  Sites using MDSplus remote data access:  JET, ASDEX-Upgrade, Tore Supra, DIIID  Physics codes  EFIT, TRANSP, GS2  Integrated Tokamak Modeling Task (EFDA)  ITPA collaborative data archives Fig.5. MDSplus sites.

9. 9/12 A. Luchetta 17 th Real-Time Conference, 25 May 2010, Lisboa, Portugal. EPICS/MDSplus integration (1)  We want a high level of integration between EPICS and MDSplus. EPICS Channel Archiver  New MDSplus-based EPICS Channel Archiver (JCA-based). See G. Manduchi Poster PCM-16 Fig. 6. Remote data access time to EPICS and MDSplus archivers. Table 6. Percentage of lost samples in EPICS Channel Archiver. 2.4 GHz quad-core Linux workstation 4 GB RAM and SATA disk controller

10. 10/12 A. Luchetta 17 th Real-Time Conference, 25 May 2010, Lisboa, Portugal. EPICS/MDSplus integration (2) New EPICS Records  mdsput provides direct storage of EPICS IOC data into MDSplus-based Channel Archiver. Mdsplus actions  mdsaction allows to command ‘ Mdsplus actions ’ by an EPICS IOC.  MDSplus actions  MDSplus actions execute operations. INIT method INIT method : reads set-up information from pulsefile and configures hardware. STORE method STORE method : reads samples from ADC and stores them into pulse file.  mdsevent implements reception of ‘ MDSplus events ’ (asynchronous communication).  MDSplus events can also carry data.  Channel Access Server for MDSPlus PV mdsaction INITSTORE mdsevent(UDP) Channel Archiver (CAC) Ch. Access mdsip (TCP) Fig. 7. EPICS and MDSplus data flow. ADC MDSplus CAS MDSplus Pulse File mdsput EPICS IOC wave

11. 11/12 A. Luchetta 17 th Real-Time Conference, 25 May 2010, Lisboa, Portugal. Real-time Performance (1)  Measurements on real-time (RT) performance of EPICS and the latency ‘fingerprint’ of Linux k.2.6 and RT patches (tollerable latency: a few hundreds  s). Fig. 8. Measured latency values. Reference application MARTe EPICS Reference application+ EPICSX MARTe* test 1 kHz Latency (  s) Occurrence graph bin width 0.1  s Fig. 9. Sample application. Performance See A. Barbalace Poster PFE-13 MARTe See A.C. Neto Poster PFE-4 NI6255 x86 Intel Core 2 Duo 2.66GHz 3MB cache and 3GB RAM Linux 2.6.29.6 rt-24

12. 12/12 A. Luchetta 17 th Real-Time Conference, 25 May 2010, Lisboa, Portugal. Conclusions  Requirements of SPIDER data acquisition are not trivial due to the long duration of the SPIDER beam-on time, the large data amount and, finally, the real-time constraints.  System software architecture will rely on the integration of EPICS and MDSplus, for which specific tools have been implemented and profiled, such as the MDSplus Channel Archiver, interface EPICS records and, in progress, the MDSplus Channel Access Server.  The real-time characteristics of EPICS and Linux kernel 2.6 with real-time patches satisfy the real-time requirements of the SPIDER data acquisition and fast real-time control. Thank you for your attention.