Development of ORBIT Data Generation and Exploration Routines

1
Development of ORBIT Data Generation and
Exploration Routines

• G. Shelburne
• K. Indireshkumar
• E. Feibush

2
Outline

• Overview of ORBIT
• Our Goals over the Summer
• Methods Used, and Accomplishments
• Conclusion
• Future Work

3
What is ORBIT?

• ORBIT is a program written in Fortran that
  numerically integrates the Hamiltonian equations
  of guiding center motion in axisymmetric
  magnetically confined plasmas.
• Developed by R.B. White

4
Issues

• ORBIT Data
• ORBIT in its previous form produced a single data
  file, readable only by NCAR
• In some instances, depending on user discretion,
  it produced minimal ASCII data
• NCAR
• outdated graphics software
• little or no interaction with data
• requires the use of ctrans to convert the
  metafile
• limited capabilities

5
Goals

• Save ORBIT computed data
• Easily understandable conventions
• Flexible for use with any option
• Portable netCDF format
• Replace the NCAR graphics software
• SM (formerly SuperMongo)
• ElVis

6
The NCAR graphics package

• Until recent developments, NCAR was the main
  visualization device used in conjunction with
  ORBIT.
• SM was also used, however the use of SM required
  data files.

7
Graphical Limitations

• NCAR does not directly view data
• Poor visualization devices
• NCAR X windows are essentially "pictures"
• Data cannot easily be compared to one another
• Data cannot be explored
• This would require changes to NCAR protocols
• re-compiling, and re-running the program for each
  change

8
ORBIT Data

• Initially, the little data being produced were
  only being produced in select sets and for select
  modes of operation.
• Data were going directly to NCAR, then to the
  metafile.

9
Solution

• Bypass NCAR graphics calls, instead write the
  data in portable binary format. We use netCDF
  format for writing our data.
• Data can then be accessed by multiple users, and
  allows easy comparison across ORBIT
  configurations. Can be publicly viewed.
• Object oriented, scientific data standard.

Next Step Find out how data is being used.
10
How data in ORBIT is being used
Orbsub sets initial equilibrium values
stores information in C files
retrieves particle scatter information
(gyro-kinetic values) creates post-scatter
particle plots and histograms
accesses spline interface (MHD values) creates
field equilibrium plots
11
Program Structure

• Data are computed and passed through various
  subroutines within the ORBIT program structure.
• Data are sent from these routines into NCAR
  graphics calls.
• There are no central locations storing the
  initial data values and the final data values.
• We must retrieve these values according to where
  they are calculated
• We should therefore intercept data going to NCAR

12
Method

• Bypass NCAR graphics calls
• Originally placed CDF write routines within the
  code.
• This soon became too cumbersome
• We decided to make one subroutine within ORBIT

13
Why insert another subroutine in ORBIT?

• ORBIT is literally a work in progress.
• Spline calculations are not in the final stage
• New and better methods are being developed
• So, we want adaptability.
• Much easier to simply call a subroutine
• Can easily be moved to an entirely new program

14
What we need to write a CDF file

• Rank 1 arrays
• variable name (weight, pitch, helicity,...)
• variable value (xx, aa, rr,...)
• long description (equilibrium helicity,
  density,...)
• units (m, keV, MA/m2,...)
• array size
• value type (real, complex, logical,...)
• Rank 2 arrays
• same as rank 1 with another dimension

15
Data Generating Subroutine

• Takes in data that would otherwise go to an
  NCAR call
• Uses the Ezcdf library
• user friendly
• no arrays larger than 3 ranks
• Separates 1-D and 2-D variables into 1-D and 2-D
  arrays.
• Writes the arrays into a large binary file that
  can be easily accessed by any user.
• Labels the arrays so as to keep track of them
  later.

16
Other considerations labels

• We need to label each and every variable
  carefully and accurately, so we know what we are
  looking at when it is graphed later.
• Gradient in what coordinate? a, theta, psi
• Equilibrium, Initial, Final?
• How many particles? 50, or 50000
• Flux surface? a 1, 0

17
Visualization ElVis

• Plots a variety of data formats
• f(x) type two column ASCII data
• f(x,t) type three column ASCII data
• compatible with SM multi-column ASCII data
• plots directly from CDF format
• Allows interactive exploration of data
• zoom scroll functions
• multi-window capability allows easy data
  comparison
• produces publication quality graphics
• programmed in Java
• portable across platforms
• available
• http//w3.pppl.gov/efeibush/elvis.html

18
(No Transcript)
19
(No Transcript)
20
Example
21
End Game?

• "I prefer SM," "Will it work with Gnuplot?"
• Conversion program reads binary CDF files and
  writes multi-column text data
• Allows the user to
• Specify which variables to write
• What format to write the data to
• What to title it
• Simple, independent of ORBIT

22
(No Transcript)
23
Conclusions

• ORBIT now writing portable, data files for all
  options.
• Subroutine is flexible enough to be adapted to
  future changes.
• CDF data can be plotted directly with ElVis,
  entirely portable.
• Binary data can be converted to user-specific
  text format

24
Future Work

• Send data to ElVis directly
• Upgrade output file format to HDF5

25
Thanks.

• I would like to thank
• K. Indireshkumar (Kumar), and E. Feibush for
  their support and help.
• S. Ethier for the occasional advice
• R. White for allowing me to play with his code