Title: Designing a CODAC for Compass
1Designing a CODAC for Compass
- Presented by
- André Sancho Duarte
2Outline
- Introduction to the CODAC concept
- Compass Tokamak
- CODAC in modern fusion experiments
- Issues
- Needs
- Solutions
- CODAC implementations
- Firesignal
- Other examples
- Application to Compass
3CODAC System
- Control, Data Access and Communications
- System for
- Control
- Experiment configuration
- Support systems configuration
- Data Acquisition and Retrieval
- Communications
- Remote Participation
4CODAC Diagram for ITER
5Compass Tokamak
- Major radius 0.56 m
- Minor radius 0.18 0.23 m
- Plasma current lt 350 kA
- Magnetic field 1.2 or 2.1 T
- Triangularity 0.5 - 0.7
- Elongation 1.8
- Pulse length lt 1 s
- PLH, 1.3 GHz lt 0.4 MW
- PNBI 2 0.3 MW
6 CODAC for Compass
- The development of a control and data acquisition
system for Compass represents an opportunity to
test ITER relevant solutions - The following areas are planned to test in
Compass - Remote maintenance/upgrade of the control
software and re-programmable hardware. - Automatic/interactive installation and deployment
of instrumentation hardware. - Formal self-description of plant systems,
including diagnostic systems, using the XML set
of technologies. - Fast, real-time multivariable (MIMO) plasma
controllers. - Online data reduction as an option or in parallel
to raw data storage on large memories.
7Modern Fusion Experiments
- Pulse duration over 1 second
- Expectation of human intervention
- Around 50 diagnostics, some very complex
- Over 100 MB/s of data per diagnostic
- Example Rogowsky coils in Compass can produce
256 MB/s (32 channels of 4 bytes _at_ 2 Msamples/s) - Small number of pulses during a campaign
- Constant monitoring of the machine and its
envolving
8Typical Experiment Flow Chart
9Desired Experimental Chart
10Issues- Collected Data (1/3)
- The size of the data collected can cause data
transport and storage issues and increment of the
operation cycle-time beyond the machine
constrains - Implement faster data transport to comply with
machine cycle-time (use of new generation faster
data transport networks) - Higher-speed real-time pulse processing both
during and after shot? - Implement event-driven data acquisition operation
- Data is acquired or actions performed (e.g.
change acquisition rate) only when relevant
events occur - Provide data compression capability into the
diagnostics (less data to store and faster data
transfer)
11Issues- Collected Data (2/3)
- Some diagnostics require high sampling
frequencies current technical capabilities may
be exceeded when operating for large periods - Use of standards-based fast data transfer on the
data paths (e.g. PCIe) - Use of local fast memory with sizes of several GB
and bandwidth of GB/s - Use of data compression when bandwidth
bottlenecks still remain
12Issues- Collected Data (3/3)
- Data reduction techniques
- Data Compression
- Lossless algorithms
- Keep all the data
- Fast compression and decompression available
- Typical data can be highly compressed
- Loss algorithms can provide extra compression
- Can provide extra compression for specific data
- Variable acquisition rates
- Good for events localized in time
- Data loss for unexpected events
13Issues RT Data Processing (1/2)
- Higher RTC processing power required for local
data compression or reduction, monitoring of
diagnostic output and generation of plasma
control variables - Use of processors with parallel processing
capabilities, high-throughput and low latency
(multi-core CPUs, FPGAs, DSP ) - hardware processors included on the digitizers
can process and manage RTC high throughput data
flow and perform preliminary basic algorithms or
data compression/reduction - Use of data processing units where various boards
are interconnected through a full-mesh topology
network having low-latency and high bandwidth
14Issues RT Data Processing (2/2)
- New diagnostics and plasma controllers may
require an updated real-time control and
monitoring infrastructure. - Higher algorithm complexity and higher number of
input signals - Lower loop delays for time-critical real-time
control and distribution of plasma variables and
events (sometimes under 10 µs) - Better timing, synchronization and RT messages
networks.
15Issues Digital Instrumentation
16Innovation on Instrumentation
- The referred requirements reveal the importance
of a platform capable of providing - High-throughput real-time hardware signal
processors at the acquisition level - Low-latency serial gigabit full-mesh
interconnection between cards - Integrated RTC event-based acquisition, operation
and storage - Integrated synchronism of all digitizer
- Presently the ATCA based instrumentation is a
good candidate - ATCA systems are expected to become the backbone
of the CODAC in Compass
17Existing CODACs for Long Pulses (1/2)
- LHD (Japan)
- Based on PC cluster
- Communication through TCP/IP
- VXI based systems
- Data Streaming (10 s slices)
- Lossless data compression (ZLIB and JPEG-LS)
- Two stage backup
- Web interface for data analysis
18Existing CODACs for Long Pulses (2/2)
- EAST (China)
- Distributed data system
- Communications via TCP/IP network
- CAMAC and PCI based systems
- Data streaming (5 s slices)
- Data compression with LZO
- Windows software for data analysis
19 The Firesignal System
- Modular client/server approach with XML plant
description/ systems integration. - Standalone operation or interfaced with other
CODACs. - Event-driven/Steady State Operation on absolute
time. - User friendly interface with remote management
and participation control room spread over
campus/web. - Easy and universal integration (Matlab, IDL,
SciLab, C, Java, Python...)?. - Modules connected through CORBA run in various
OS. - PlugPlay and HotSwap of hardware
20Conclusions
- Modern fusion experiments share common needs and
issues regarding control and data acquisition - Technological developments in hardware and
software allow us to address them efficiently - Existing CODACs have implemented with success
many of these technologies - Compass provides an excellent platform for
deploying and testing the ideas here presented. - It is desirable for the new CODAC to be flexible,
in order to accommodate new developments
21 Improvements on Firesignal
Issues Improvements
Data transmission bottleneck Data transmitted through TCP/IP Support to data streaming Distributed server Variable acquisition rates Data compression
Assumes all data with same size and equally spaced in time Improved support for other types of data
Event support added later ? somewhat poor event management Event support from start
Designed mainly for data acquisition More flexible support of mixed acquisitions and real-time control boards
Hardware clients need to be restarted after hot-swap Intrinsic support to hot-swap
22Support Slides
23Data Compression
JETs Fast Camera. Results provided by Jesús Vega
(CIEMAT/ES)?
L.Ying, L. Jiarong, L. Guiming, Z. Yingfei, L.
Shia, The EAST Distributed Data System, Fusion
Eng. Des. 82 (2007) 339 - 343