IEEE Transactions on Nuclear Science, Vol. NS-34, No. 4, August 1987

UPGRADE OF MHD DATA ACQUISITION SYSTEM FROM ISX-B TO ATF*

J. D. Bellt and V. K. Pare

P.O. Box Y, Oak Ridge National Laboratory, Oak Ridge, TN 37831

The data acquisition system assembled to study magneto-hydrodynamic (MHD) activity on the Impurity Study Exper-iment (ISX-B) tokamak at Oak Ridge National Laboratory(ORNL) is being revised for use on the Advanced Toroidal Fa-cility (ATF). The new hardware and software architectures arebased on ISX-B experience and will feature (1) different modesof operation for storing various subsets of available data, (2) auser interface that requires less routine activity than the earliersystem, and (3) continued support of calibration and testingmeasurements used on ISX-B. The new hardware organizationand software components are described in detail.

Description of the Plasma DiagnosticsTwo different kinds of sensors are used to detect activ-

ity arising from interaction of external and internal magneticfields and the experimental plasma ("MHD activity"). Mirnovloops are wound-wire coils that respond to changes in themagnetic field along their axis. Mounted just beyond theplasma boundary, they measure magnetic fluctuations at ornear the edge of the plasma. (The skin effect of the conduct-ing ptlasma prevents their sensing events in the plasma inte-rior.) We also use surface-barrier diodes that are manufacturedand shielded to respond to "soft" X rays (SXR) with energiesof 0.5-10 keV. Total X-ray emission in this range increaseswith increasing plasma temperature and density, so signals are

not only observed from throughout the plasma, they are alsosteeply peaked at the center of the minor cross section. Thesesignals respond quickly to internal fluctuations and are helpfulin tracking the location and gross behavior of the plasma.

Figure 1 shows a 3-loop Mirnov assembly, and Fig. 2 showsa 20-channel SXR array. The diagnostics are shown at theirrespective locations inside the ATF vacuum vessel, along withthe shape predicted for the plasma. Much complexity of con-struction and analysis arises from the fact that the ATF tor-satron's plasma cross section rotates 360° poloidally (the shortway around the machine) for every 300 toroidally (the long wayaround the machine).

The CAMAC System

Figure 3 is a single-channel schematic of the CAMAC hard-ware system that will be used on ATF. It is very similar to theone that served on the ISX-B tokamak experiment at ORNLfrom 1981 to 1984. The 'start of shot" signal from the ATFdevice will come into a trigger recognition module. For somedata signals, this trigger will go directly to waveform digitizers.Other data signals will be digitized during a short window inthe shot, and the trigger for them will pass through a Jorway220 delay generator to the digitizers. Signals from the sen-

sors will come into programmable gain/filter modules (PGFs),whose gain level can be set by CAMAC command. The digitiz-ers will be LeCroy 8210s, 2264s, and 8212s, running at speedsranging from 1 kHz to 1 MHz, with various numbers of auxil-iary memory modules attached to them. We will use KineticSystems (KS) crate controllers and LAM encoders and will beconnected to a KS 2053 branch driver inside a DEC VAX com-

puter. A time-sharing terminal will serve as the operator'scontrol and display station.

*Research sponsored by the Office of Fusion Energy, U.S. Departmentof Energy, under contract DE-AC05-840R21400 with Martin Marietta En-ergy Systems, Inc.

tComputing and Telecommunications Division, Martin Marietta En-ergy Systems, Inc.

Fig. 1. Mirnov loop assembly in ATF, with vacuum vesseland flux surfaces.

Fig. 2. Soft X-ray array on ATF, with vacuum vessel andflux surfaces.

0018-9499/87/0800-0768S01.00 1987 IEEE

768@sJK0 4tS@-rw

769

ATF SHOT TRIGGER

SIGNAL PROGRAMMABLE-FROM GAIN FILTERSENSOR MODULE

CRATECONTROLLER

LAM ENCOOER

L _J

Fig. 3. Single-channel schematisystem.

ic of ATF MHD CAMAC

ISX-B Results and Observations

Since many features of the ATF system were selected inlight of our experience on ISX-B, it is worthwhile to discussthat system's achievements and problems. Both Mirnov loopand SXR signals were staples of tokamak operation and con-

trol, helping the tokamak operators characterize and tune theplasma conditions between shots. In the realm of experimentalphysics results, the Mirnov loops observed various phenom-ena, including "m = 2" oscillations that were phase-lockedwith the internal, "m = 1" SXR signals and short bursts ofnoise at >100 kHz just after internal disruptions. After inten-sive research led to the conclusion that these bursts were notdirectly related to ISX-B's energy confinement problem, theloops' high-frequency background spectra appeared to corre-

late with confinement results. The SXR signals were crucialin matching experiment with a theory that explained the cou-

pling of internal m = 1 and exterior m = 2 oscillations fortoroidally rotating plasmas. Another experiment used them toconfirm the importance of electron thermal diffusion in ISX-Benergy confinement. These signals also produced an experi-

mental two-dimensional (2-D) movie of an internal disruption.Specific physics lessons were learned about maximizing the

utility of the MHD diagnostics on ISX-B. For the Mirnov loops,we now know that background fluctuation levels are important,so we should start measuring these levels early in a device'slife, both as waveforms and statistically. For the SXR signals,radial profile and diffusion studies can be done with one array.

Tomography based on invertible polynomials is valid only inthe interior of the plasma, because the polynomial fitting willemphasize the high signals of the central region without regardto the much lower edge signal. For both diagnostics, we needto store and study fewer waveforms from more shots.

Certain operational limitations were particularly notice-able. The stand-alone computer system had on-line storagefor only about 20 shots at a time, so only a few shots could bestored from each day's run. Since the ISX-B system was not ona central computer, MHD data were isolated from those takenby diagnostics not plugged directly into the system. The suiteof data acquisition programs required at least three operator

commands for each shot archived and two for each shot ex-amined. Operators also took oscilloscope photographs of theSXR and Mirnov traces used by the tokamak operators. Sincethe MHD operator had to take scope photographs and decidewhether to save a given shot before the next shot occurred, asingle operator was kept very busy.

Consideration of these items led us to make a variety ofchanges that will increase the value of the MHD system. First,we have moved to a central computer system with more storagecapability, and we will use ATF-wide data management and su-pervisory systems. This will allow us to store data for manyor all shots and will ease comparison with other data. In viewof the limited value of the mass of data stored for SXR tomog-raphy on ISX-B, we will use only one SXR array on ATF, for1-D studies. We will collect a few channels of high-frequencywaveform and statistical data for as much of the experimentallife of ATF as we can manage. As an aid to effective use ofthe MHD system, we will need to relieve the operators of sometedious, routine tasks.

Configuration Changes for ATF

Since fast and slow waveforms and statistical data needto be acquired, but we cannot process huge masses of datafrom every shot, we have chosen to implement multiple modesfor data acquisition, storing more or fewer data per shot asappropriate. The hardware will be rearranged to allow multi-ple digitizing of some waveform signals at widely varying timerates and passing of some magnetic loop signals through ana-log bandpass filter/rectifier circuits to obtain time histories ofthe frequency content of a shot.

There will be three modes of data acquisition, each servinga different purpose for the experimenters:1. Monitor mode provides a general picture of the entire shot

(useful for tuning operation of ATF) and accumulates fre-quency content information over many shots for turbulencestudies.

2. Standard mode adds time histories of a radial scan of SXRemission to the monitor data. This information is useful intracking plasma position and gross behavior in the plasmainterior.

3. Detail mode, picking a short selected window out of theshot for fast digitization, will be used to study transientphenomena (such as internal and edge fluctuations lastingfrom 0.01 to 10 ms) and time-domain turbulence.

Figure 4 summarizes the sensor/signal components of each dataacquisition mode. Table I gives an indication of the amountof data that we expect to store in the various data acquisi-tion modes and compares that amount with the correspondingnumbers from ISX-B.

ATF Data Acquisition Code

A new data acquisition computer code is being written togo with the new hardware scheme and address some of theproblems of the old system. A control-flow diagram for thiscode is given in Fig. 5. The basic idea is that the ATF Syn-chronizing and Monitoring System (SAMS) utility [11 and theVAX/VMS SYS$QIO, used in their asynchronous-system-trapand event-flag-setting modes, allow data acquisition, operatortuning, and physics display to occur independently in a singlecode. The detailed sequence of execution proceeds as follows:1. A configuration file with the location of CAMAC modules,

module settings, and signal paths is read from disk andloaded into a FORTRAN COMMON block. Then, one-time CAMAC module setups are performed.

2. An asynchronous-system-trap call to SAMS is issued, ask-in; that (a) process execution be interrupted and a VMS"event" flag be set when the ATF "-30 seconds" signal isdetected and (b) a subroutine that starts the LeCroy digi-tizers be executed before control is returned to the portionof code that was interrupted.

770

DATA STORAGE

Fig. 4. ATF MHD data flows, from sensors to digitizers tostorage.

Table I. Amount of data acquired

Type of signal

Data acquisition mode

Monitor Standard Detail

ATF data load (1-s pulse)Signature (32K)SXR diodesLow-frequency Mirnov loops

Monitor (1K)SXR diodesBand-separated Mirnov

loop signalsDetail (8K)SXR diodesLow-frequency Mirnov loopsHigh-frequency Mirnov loops

Total signalsTotal points

11

22

22

2 20 2010 10 10

201030

14 34 9476K 158K 638K

ISX-B comparison load(300-ms pulse, typical 16-ms window)

SXR diodes (8K)Low-frequency Mirnov loops (8K)High-frequency Mirnov loops (8K)

Total signalsTotal points

724

12

88704K

771over time, but initial options will include: (a) Change modeof data acquisition among Monitor, Standard, or Detail.(b) Change the delay time for the Detail window in theshot. (c) Scan the stored data for a shot to adjust the PGFgains (see GSET below). (d) Select and plot stored datafrom previous shots. (e) Exit the data acquisition code.When the non-exit commands are complete (or aborted dueto the "-30 seconds" flag), the code goes back to step 2above.

Fig. 5. MHD data acquisition code control flow. (Dashedlines are "wait to execute." Dotted lines show flow of eventflags.)

3. A SYS$QIO ("proceed I/O") call is made and waits for ter-minal input, setting another event flag when the operatorenters a command at the terminal.

4. A VMS system call is then made that causes the programto wait until one of the event flags is set. When that occurs,the code decides which flag was set and branches accord-ingly.

5. If the "-30 seconds" flag was set, an ATF shot is occur-ring and those data need to be collected. User input isdisabled, and then a call is made to a suite of subroutinesthat wait for the digitizers to finish, read data from them,write those data into the ATF Data Management System[2], and display selected signals on the operator's terminal.A no-shot digitizer cycle will also be run to determine thedc offset for each signal. After these subroutines are done,the code goes back to step 1 to get ready for the next shotcycle or user command.

6. If the "user input" flag was set, another set of subroutineswill be called to decode and execute the operator's displayor adjustment command. The list of commands will grow

The result is that the operator can adjust data acquisitionparameters or study recent (or old) data during ATF operation,while the data acquisition itself cycles automatically. It is ourexpectation that none of the user commands will take longenough to interfere with the readout of CAMAC data; thedigitizing of these data cannot be interfered with.

Utilities to DevelopA number of utilities developed for ISX-B were useful

enough to merit being revived for ATF. They are summarizedhere.

1. CALDC does a least-squares fit to digitized data from alow-frequency triangle wave input to each channel, producinga final calibration factor for each channel.

2. DVM cycles the digitizer and PGF module for a givenchannel to display pre- and post-PGF offsets in real time, whilethe PGF internal offset is being adjusted.

3. GSET reads stored data for a shot and counts the num-ber of values greater than binary fractions of the maximumdigitizer value. The PGF gain is then turned up or down bypowers of 2 to get each signal to be at least half-scale on thenext shot. (This was so useful that it will be an option withinthe data acquisition code.)

4. For FRESP, the experimenter feeds a sequence of se-lected sine waves into the PGF modules. The code digitizesat each sine-wave frequency and computes the DFT amplitudeand phase relative to the test signal. Seventh-order polynomialfits are stored as the frequency response function for the PGFchannel.

To help make full use of the data acquired, various special-purpose display codes will be needed, beyond the capabilitiesof the data acquisition code. One will be a stand-alone versionof the plotting section of the data acquisition code. Otherswill plot time histories of groups of signals ordered by sensorposition, rather than by digitizer groups. Also, slice-in-timeradial SXR profiles will be needed.

SummaryA data acquisition system is being developed for MHD di-

agnostics on ATF. The system is based on a previous systemused on ISX-B, modified both by lessons learned from ISX-Band by new ATF requirements. The system will store a flexiblemix of data and will allow the operator to acquire and examinedata simultaneously. Different modes of operation for the dataacquisition system will address the various plasma control andturbulence studies to be undertaken on ATF.

References[1] D. E. Greenwood, "SAMS-The Synchronizing and Moni-

toring System for ATF," this conference.[2] K. L. Kannan and L. R. Baylor, "The ATF Data Manage-

ment System," this conference.