Controlling the Sun’s Center
In Livermore, Calif., early in the next decade, a "science factory" with an astounding array of control technologies will aim up to 192 lasers at one point, another step toward using lasers to "ignite" fusion. The blast of energy aims to fuse hydrogen isotopes of deuterium and tritium into helium nuclei—with energy output higher than energy input.
In Livermore, Calif., early in the next decade, a “science factory” with an astounding array of control technologies will aim up to 192 lasers at one point, another step toward using lasers to “ignite” fusion. The blast of energy aims to fuse hydrogen isotopes of deuterium and tritium into helium nuclei—with energy output higher than energy input.
The National Ignition Facility (NIF) is being constructed by Lawrence Livermore National Laboratory (LLNL), in conjunction with Sandia National Laboratories, Los Alamos National Laboratory, and the University of Rochester. France and the U.K. have been collaborating with LLNL under bilateral agreements with the U.S. LLNL is operated by University of California for the U.S. Department of Energy (DOE). Groundbreaking was in May 1997.
Human efforts have achieved fusion previously. Unconfined fusion occurs in weaponry. Controlled and contained fusion to date by nonlaser methods has consumed more energy than it produced—not optimal for the eventual goal of a clean, energy producing plant or for high-energy space propulsion.
Applications, 500 TW
NIF, at a cost of just over $1 billion, will support the DOE’s Stockpile Stewardship Program. The program’s goal is to assure safety and reliability of the U.S. nuclear weapon stockpile in absence of nuclear tests. Other applications have resulted from research associated with predecessors to NIF’s 192 ultraviolet (Nd-glass) lasers. Past and anticipated future applications from related laser research include the first soft X-ray laser, efficient visible lasers, radar use, medical lasers, EUV lithography, and ultra-short laser pulses for cutting. Commercial nuclear fusion electric generating plants may take until 2050 or longer, predict NIF scientists.
Energies expected at the focal point of the NIF laser array are enormous, if brief. In a pulse length of 3 x 10-9sec, NIF will produce 1,800 kilojoules of energy, at peak power of 500 teraWatts (500 trillion W). For a sliver of a moment, the energy produced will exceed total U.S. electric generating capacity by more than 1,000 times!
NIF lasers will contain more precision optics (33,000 ft2) than all the telescopes in the world combined, requiring precise control of lasers, optics, and related equipment so each beam travels 600 feet, through amplifiers, and into the target chamber center, hitting within a 50 micron target in a 30 picosecond span.
Further requirements, according to Paul J. Van Arsdall, system engineer, integrated computer control system for NIF, include 24-hour by seven-day availability (with 7.5 days of unscheduled maintenance per year), 30-year project life, and ability to use direct and indirect laser targeting methods.
LLNL served as system integrator in the project. “We’re using a lot of new technology in challenging ways. We were concerned about performance of a fully deployed system, versus lab tests, so we worked on a program to simulate critical control scenarios using discrete modeling techniques, scaling up lab designs to ensure we weren’t being naive about performance of the full-scaled system,” says Mr. Van Arsdall, project leader for the integrated control system.
Controls and related labor, at some $96 million, is running about 8% of total project cost. Unlike other projects, the “goal is to future-proof the work and create portability of code with long-term maintainability. This is a longer-term product than most industries construct.”
Mr. Van Arsdall agreed to name some of the hundreds of vendors involved, but says use of a particular vendor doesn’t mean that the project participants or the U.S. government endorses any particular manufacturer. [Here’s what the attorneys require: Neither LLNL or the U.S. government makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, product, or process disclosed in this article.]
NIF lasers and control system, inside a 704 x 403 x 85 ft high building (football stadium-sized), require approximately 9,500 stepper motors for laser positioning, Wind River Systems’ (Alameda, Calif.) VxWorks operating system, and PCI bus embedded system to serve 10-15 years within a multi-operating system environment.
Software for the integrated control system, which Mr. Van Arsdall calls the system’s soul, is expected to take 150 person-years of development time.
Software helps a dozen engineers control 50,000 points, 600 cameras, 3,000 actuators, analyze 3,000 images, measure offsets, and move steppers as needed, without operator intervention, unless a control loop doesn’t stabilize. The automated process must occur in under an hour and may include 48,000 messages between front-end processors and the supervisory system. While planning for a firing may begin weeks in advance, configurations may vary widely depending on the experiment. Three shots a day will occur with an intricate choreography for each, including elaborate backups in case of an aborted shot.
Software is expected to gather about 400 MB of data per firing with an array of sensors. The highly distributed architecture had to be flexible enough to allow upgrades in controls, diagnostics, computers, and design, during construction and within its 30-year anticipated life, Mr. Van Arsdall says.
The integrated computer control system architecture is described as an abstract object-oriented framework for constructing distributed control systems. NIF selected international software standard Ada 95 as the central NIF software language, also used for air-traffic control and military command and control systems.
Code is written in frameworks, modular pieces built and tested in increments, allowing problem-solving without writing a lot of detailed code. In September, the team completed the shot coordination and lifecycle framework, the third on the six steps code-writing journey to a full system operation.
The control system has front-end processors and a supervisory system. The distributed control system used for system operation has 354 computers—35 Sun Microsystems (Palo Alto, Calif.) UltraSPARC computers running eight supervisory applications, 264 PowerPC-based computers with VxWorks, and 55 other UltraSPARC computers for 14 front-end processor applications. The supervisory system layer, hosted on workstations, runs Sun Solaris UNIX.
Highlights include helping 400 computers communicate using object technologies of CORBA (Common Object Request Brokerage Architecture by the 500-company consortium Object Management Group, Framingham, Mass.) For Ada 95 codes, NIF uses ORBexpress from Objective Interface Systems (Reston, Va.)
A simulation included study of CORBA, looking at impact on network and processor performance, by varying transaction rate and message size (200-300 bytes). The project team wrote the services code themselves, to improve ability to maintain or modify, and to make it more lean than CORBA services. For example, notification by exception from a chosen setpoint reduces network traffic.
With configuration and development environment in multiple operating systems, Sun Java technology was chosen.
“Staff had some experience with it and developed a Java-based thin-client graphical user interface for multiple stations, mixing languages and operating systems so control is in the software, using CORBA to communicate to the rest of the control system,” Mr. Van Arsdall says. Java programming tools and programmers are readily available. Programmers develop on Microsoft (Redmond, Wa.) Windows NT in an office environment then transfer to the Solaris environment for the control room, with “some front-end Windows NT,” Mr. Van Arsdall says.
NIF was among early adopters of Rockwell Automation (Milwaukee, Wis.) Allen-Bradley Control Logix PLCs; a related task was to work out the interface details to communicate with Solaris workstations.
Echelon (Palo Alto, Calif.) LonWorks is being used for distributed laser diagnostic sensors. Self-designed front-end circuitry required a customized chip-level solution.
Ethernet will be used with no shared segments. ATM (155 Mb/sec, Asynchronous Transfer Mode) will be used to transport digitized motion video from sensors to process equipment and operator stations. Worst-case network deployment scenarios were studied with Opnet (Annandale, Va.) Modeler simulations, mixing ATM and Fast Ethernet (100 Mb/sec) networks.