A “Universal Time” system for ASDEX upgrade

Share Embed


Descrição do Produto

Fusion Engineering and Design 66 /68 (2003) 947 /951 www.elsevier.com/locate/fusengdes

A ‘‘Universal Time’’ system for ASDEX upgrade Gerhard Raupp a,*, R. Cole c, K. Behler a, M. Fitzek c, P. Heimann a, A. Lohs a, K. Lu¨ddecke c, G. Neu a, J. Schacht b, W. Treutterer a, D. Zasche a, Th. Zehetbauer a, M. Zilker a, ASDEX Upgrade Team a

b

Max-Planck-Institut fu ¨ r Plasmaphysik, EURATOM Association, Boltzmannstrasse 2, D-85748 Garching, Germany Max-Planck-Institut fu ¨ r Plasmaphysik, EURATOM Association, Walther-Rathenau-Strasse 49a, D-17489 Greifswald, Germany c Unlimited Computer Systems, Seeshaupterstrasse 15, D-82393 Iffeldorf, Germany

Abstract For the new generation of intelligent controllers for plasma diagnostics, discharge control and long-pulse experiment control a new time system supporting steady state real-time operation has been devised. A central unit counts time at nanosecond resolution, covering more than the experiment lifetime. The broadcast time information serves local units to perform application functions such as current time readout, trigger generation and sample time measurement. Time is treated as a precisely measured quantity like other physical quantities. Tagging all detected events and sampled values with measured times as [value; time]-entities facilitates real-time data analysis, steady state protocolling and time-sorted archiving. # 2003 Elsevier B.V. All rights reserved. Keywords: ASDEX Upgrade; Steady state protocolling; Time-sorted archiving; Time measurement

1. Introduction Time is the key quantity to properly coordinate large distributed experimental devices and to understand complex physical processes. The primary task of a time system is to provide a stable time base for the reconstruction of the proper sequence of all control and measurement actions so that the evolution of plasma and machine parameters and phenomena being investigated can be analyzed and understood. As experimental

* Corresponding author. Tel.: /49-89-3299-1715; fax: /4989-3299-1313. E-mail address: [email protected] (G. Raupp).

devices and equipment evolve, models and methods to synchronize actions and handle time information will change. Despite progress in measurement technology from oscilloscopes to ADCs the simple operation model for lab equipment remained. External triggers (usually time-driven) start and synchronize the measurement hardware. When acquisition terminates (at the specified number of samples) data can be read out. Sampling times must be calculated after a discharge from trigger start time and sample frequency: with analog trigger lines, precise knowledge of topology and local delays is needed to avoid errors. If a digital timing system is installed, a relative time count value is available for the start of the trigger generation.

0920-3796/03/$ - see front matter # 2003 Elsevier B.V. All rights reserved. doi:10.1016/S0920-3796(03)00381-8

948

G. Raupp et al. / Fusion Engineering and Design 66 /68 (2003) 947 /951

With such a method data acquisition is not flexible, data are not accessible during discharge, and sampling times are not immediately available. For next generation fusion machines this operation model is not adequate. Future machines will investigate discharge scenarios required for a burning plasma, and increase discharge times towards steady state. Control and data acquisition must be more sophisticated [1]. They will have to be more intelligent (take own decisions) and need to be integrated (exchange data): plasma controllers must access real-time sensor data sampled and processed by real-time diagnostics to improve control quality. Diagnostics must access real-time plasma and machine state information provided by discharge controllers to improve the significance of data sampled. Diagnostics operating in such a system will perform real-time data acquisition (including tuning of front ends to plasma state) and immediately access, analyze and process sampled data. Results produced are then sent via a low-latency real-time network at moderate bandwidth to plasma controllers, whereas raw and processed data are visualized or transferred to archives with moderate latency at higher bandwidth. Plasma controllers will execute supervision and feedback control processes. While data sampling is shifted to diagnostics, command value output will remain a controller task. A novel time system for such a distributed system must support real-time operation. Time representation must be unique and common for all equipment on-site. Time information must be easy to access and process. When data are sampled or calculated, associated times must be measured concurrently and added to form self-describing [value; time]-entities. Time resolution must be adequate for the fastest processes to be diagnosed (physics, computing) and the coverable time span sufficient for steady-state operation, or even experiment lifetime. Finally, the unit price must be low as large numbers are needed. Synchronization is NOT a requirement with ASDEX Upgrade’s process model: it is not necessary to enforce synchronized data sampling among different diagnostics and controllers. Instead, it is sufficient to make sure that all front-end

data sampled are immediately tagged with a precise universal timestamp. In this way we can guarantee that all data can be properly accessed and processed, even if intelligent control and diagnostics processes are scheduled by the realtime operating system at times unknown a priori. This is why we call the system a time system, rather than a timing system.

2. System structure The system topology is conventional: a central time counter (CTC) unit outputs to a unidirectional network, and local units located at host computers access and process the distributed time information (Fig. 1). The CTC generates the time information. Started once, it operates autonomously. Its core is a precise quartz oscillator. While future versions are intended to run at clock rates up to 1 GHz, cost (fiber optics distribution) and technological (FPGA counter speed achievable) reasons currently restrict the initial implementation to 50 MHz. This clock drives a 64 bit time counter. The least significant bit represents 1 ns, and with each 50 MHz clock cycle a 20 ns value is added. The counter covers 500 years. Periodically (e.g. each millisecond), the current time count, an ID number identifying the CTC and redundant CRC information are Manchester encoded onto the clock stream. Time generation is compatible with that of the Wendelstein 7-X time and trigger system [2]. Time is distributed via a dedicated unidirectional optical Time Network. The central unit’s optical output is connected to a single height 19 in. broadcast module where the optical input is sent to 16 fiber optics outputs. Modules can form cascaded networks. Distributed throughout the experiment site are local Universal Time-to-Digital-Converter (UTDC) units. They offer basic time-related functions. CTC and UTDC have the same board layout: Each UTDC has an onboard CTC for test and troubleshooting. Boards intended for use as CTCs, however, are equipped with an addi-

G. Raupp et al. / Fusion Engineering and Design 66 /68 (2003) 947 /951

949

Fig. 2. UTDC (and CTC) internal structure.

Fig. 1. Structure of new time system and interface to old timing system.

tional optical output (shaded ‘‘opto out’’ in Fig. 2). UTDCs receive and decode the optical input Manchester signal and separate clock frequency, current time count and system ID. CRC information, system ID and current time count value are checked. If the actually received time count value is correct (i.e. equal to that sent before and clocked with the base clock transmitted continuously) it is latched as new valid time count. (In case of a difference indicating a failure, a host interrupt is generated; the latch is not touched to show the last reliable time count value before the failure. While

the UTDC performs degraded operation with the local quartz, the host can read that last valid time count and reset the board to automatically resynchronize when the next actual time count value is broadcast.) If no failure occurred, a correction value (settable) is added to the valid time count to compensate for transmission path differences, and the time count is clocked continuously. The current time count value and the base clock drive all UTDC time functions. To support real-time processes the host can read the current time count value at any time. This serves to add time stamps to data calculated, or to monitor process execution times. The host can define target times on UTDC to generate interrupts to start time-driven processes. To support data acquisition, UTDC translates current time into trigger and vice versa.

950

G. Raupp et al. / Fusion Engineering and Design 66 /68 (2003) 947 /951

The Time-to-Digital-Converter (TDC) operates similarly to ADC channels. In a diagnostic system (Fig. 3) an external trigger causes an ADC to freeze and convert an external analog sensor signal into a digital value, and it causes a TDC to freeze an external digital time signal any copy the digital value into the Timestamp Buffer. When a trigger train with n pulses is applied to m ADC channels and one TDC channel we receive n/m data samples and 1 /m time samples. Information overhead is moderate, the benefits, however, are considerable: self-describing [value; time]-entities can be immediately processed or archived. Physics information is represented in a natural way. Tagging of ALL measured data with measured time stamps gives highest precision. New strategies for archiving are feasible, which is ideal for long

running or steady state discharges, and better supports continuous machine operation (interdischarge parameter modifications, preparation procedures, failures, error logs). All TDC functions match those of standard ADCs, i.e. support pre/post-trigger measurements, FIFO or ring buffers, and generation of interrupts for empty/half/ full buffers. TDCs produce a 64 bit time stamp and a 16 bit identifier for trigger sources (eight external trigger channels and eight internal event codes). The inverse function to TDC is trigger generation at given times. Two Trigger-Out-Generator channels can be configured to issue a number of equidistant trigger pulses (plain or gated clock, set by host). Four event sources may start such a pulse train (in order of priority): external triggers (i.e. externally event-driven), internal commands (i.e. process-driven), occurrence of a target time (i.e. time-driven), or detection of an event code issued by the CTC (internally event-driven, e. g. start of plasma phase, start of experiment operation, . . .). Pulse train programs (described by priority, characteristics, event source, follow up program; final stop) are handled by the Time Trigger Processor unit, and may be concatenated to form complex sequences. The UTDC is a standard PCI board, supporting 64 bit/66 MHz and 32 bit/33 MHz operation. DMA from UTDC Timestamp Buffer to computer memory works with ASDEX Upgrade’s PC and SUN. To facilitate maintenance the UTDC I/ O can be extended to a single unit 19 in. front panel.

3. System operation

Fig. 3. Diagnostic setup with UTDC operating ADC.

ASDEX Upgrade’s control system is being renewed with real-time diagnostics and controllers. Many CAMAC based diagnostics will not be modified: the old timing system [3] must still run together with the new time system which provides a 10 MHz output to clock the old timing system (Fig. 1). A new real-time application process will be programmed to generate and issue all those event and command codes (including pre-discharge time counter reset) required by the old timing system’s CT. This connects the old dis-

G. Raupp et al. / Fusion Engineering and Design 66 /68 (2003) 947 /951

charge related 32-bit time count to the new 64-bit absolute time count. ASDEX Upgrade’s new time system was devised as a standalone system to increase availability and reliability. External time resources providing generally accepted time standards such as DFC and GPS may be temporarily down or transmission disturbed. However, we intend to measure and monitor ASDEX Upgrade’s local time against external time references. All interface boards to access those time references and one UTDC board are installed in a diagnostic and trigged synchronously. The internal versus external time representation we get helps to detect failures or shutdown periods or to compare local times at different locations with the help of an external time standard. Although the simple structure of ASDEX Upgrade’s time system guarantees high reliability, a failure might occur (e.g. CTC power failure). To avoid confusion by doubly defined times and to clearly tag unreliable time periods the CTC should be re-started with a new secure future time value. This time value set by host should be higher than the last valid time count value before failure plus time period until being restarted plus maximum tolerances of local UTDCs running freely since occurrence of the failure.

4. Conclusion The novel time system of ASDEX Upgrade was integrated with the real-time data acquisition and

951

control system. It locally provides a stable unique time representation, makes time a measurement quantity itself, linked closely to all measured data and events as [value; time]-entities. It supports all standard data acquisition methods of real-time controllers and fulfils steady-state requirements. The simple system structure guarantees reliable operation. The method to measure the trigger times of data sampled gives unmatched precision. The self-describing natural representation of process quantities as [value; time]-entities facilitates real-time processing in a distributed system. The clear concept opens new strategies for steady state protocolling and time-sorted archiving for future experiment devices.

References [1] W. Treutterer, K. Behler, R. Cole, J. Hobirk, M. Jakobi, A. Lohs, K. Lu¨ddecke, G. Neu, The new ASDEX Upgrade real-time control and data acquisition system, Proceedings of the 22nd SOFT (this conference), Helsinki, 2002, Fus. Eng. Des. 66 /68 (2003) 755 /760. [2] J. Schacht, H. Niedermeyer, C. Wiencke, J. Hildebrandt, A. Wassatsch, A trigger-time-event system for the W7-X experiment, Fusion Eng. Des. 60 (3) (2002) 373. [3] G. Raupp, H. Richter, The timing system for ASDEX Upgrade experiment control, in: Proceedings of Seventh Conference Real Time on Computer Applications in Nuclear, Particle and Plasma Physics, Aachen (D), 1991, p. 462.

Lihat lebih banyak...

Comentários

Copyright © 2017 DADOSPDF Inc.