Introduction

Fast 2-D cameras on NSTX look at a variety of important aspects of the plasma edge and in-vessel components. One of these applications is the Gas Puff Imaging (GPI) Diagnostic. Data is taken from a Phantom camera and saved in "cine" files. A variety of tools have been developed at PPPL for viewing and analyzing GPI data, such as for "blob tracking" (also see Visualizing and Quantifying Blob Characteristics on NSTX ) and other purposes.

A popular viewing option for the cine files from Phantom Fast Cameras is the IDL routine FCplayer.pro. "Blobs," related to turbulence in plasmas, may be tracked in FCplayer.pro by using selections near the bottom of the Special Menu:

FCplayer can be invoked with a command like:

    IDL> fcplayer, file='nstx_5_137582.cin', $
	           aveT1=0.329, aveT2=0.333, minXsize=400, FrameAveDeltaMS=0

Once image processing options and interesting times are determined, blob data can be entered into an SQL database (currently hosted on sql2008.pppl.gov) using the IDL routine loadblobs.pro.

Blob Speed Validation Procedure

This page details the validation procedure for the calculations of blob speed. The steps are:

1. Create test files with known speeds
   We don't have code to write to Cine files, so we create NetCDF files that have simulated blob images. FCplayer and other IDL software reads Cine files with fastcam_frame.pro (and fastcam_times to get the times).

      IDL> image = fastcam_frame( filename, frame )
   

   fastcam_frame.pro will read Cine files or NetCDF files that follow a convention, and HDF files produced by Jim Myra from his simulations. gennetcdfblobs.pro will produce NetCDF that follow the expected convention and include 10 frames each that have blobs moving at different speeds. As of 6-Nov-2013, it produced 5 types of files by:

      IDL> for i=1,5 do gennetcdfblobs, i
   

   These files have the following names and characteristics:

      'circoneup_1.nc'      one circular blob moving upwards
   	 xVel & yVel =   0       3 pixels/frame
   			 -5.86e-06    0.00941 Km/s
   
      'circoneout_2.nc'     one circular blob moving outwards
   	 xVel & yVel =   3       0 pixels/frame
   			 0.0109  -3.56e-05  Km/s
   
      'circmany_3.nc'       many circular blobs moving in different directions
   	 xVel & yVel =   0.60    0.378 pixels/frame
   			 0.00214	 0.001318 Km/s
   
      'elipseone_4.nc'      one circular blob moving upwards and slowly outwards
   	 xVel & yVel =   1       3 pixels/frame
   			 0.00374	 0.00947 Km/s
   
      'elipsemany_5.nc'     three elliptical blobs moving outwards, but up and down
   	 xVel & yVel =   1.50    0.333 pixels/frame
   			0.00545	0.000916 Km/s
   

   The variables and types within these files can be seen with the IDL routine ncdf_cat:

      IDL> ncdf_cat,'circoneup_1.nc'
      Dimensions		3
          x	  64
          y	  80
          z	  10(Unlimited dim)
   
      Variables
      Time (FLOAT ) Dimension Ids = [ 2 ]
      	   Attribute units=Sec
      	   Attribute long_name=Seconds since T0
      Frames (LONG ) Dimension Ids = [ 0 1 2 ]
      	   Attribute units=Intensity
      	   Attribute long_name=Pixel Values
      xVel (FLOAT ) Dimension Ids = [ 1 ]
      	   Attribute units=Pixels/frame
      	   Attribute long_name=velocity in X direction
      yVel (FLOAT ) Dimension Ids = [ 1 ]
      	   Attribute units=Pixels/frame
      	   Attribute long_name=velocity in Y direction
   

   If you view one of these NetCDF files in fcplayer, e.g.,

      IDL> fcplayer, '/u/bdavis/cvs/idl_cvs/elipsemany_5.nc', horiz=0, minx=400
   

   and turn on "blob tracking" and "history" (in the Special Menu), and step through 10 frames, you will see:
   
   the plus signs show the previous locations of the centers of the elliptical fits to the blobs.

   You may download the following:

   The NetCDF files:       circoneup_1.nc       circoneout_2.nc       circmany_3.nc       elipseone_4.nc       elipsemany_5.nc

   QuickTime Movies:       circoneup_1_0_9.mov       circoneout_2_0_9.mov       circmany_3_0_9.mov       elipseone_4_0_9.mov       elipsemany_5_0_9.mov

   The NetCDF files can be read with the IDL routine readnetcdfblobs.pro.

2. Blobs from these NetCDF files can then be loaded into the 'blobs' SQL database.

      IDL> loadblobs, 'circoneup_1.nc', t1=0., t2=0.010, horiz=0, norm=0
      IDL> loadblobs, 'circoneout_2.nc', t1=0., t2=0.010, horiz=0, norm=0
      IDL> loadblobs, 'circmany_3.nc', t1=0., t2=0.010, horiz=0, norm=0
      IDL> loadblobs, 'elipseone_4.nc', t1=0., t2=0.010, horiz=0, norm=0
      IDL> loadblobs, 'elipsemany_5.nc', t1=0., t2=0.010, horiz=0, norm=0
   

   The above commands will produce intermediate text files like circoneup_1_blobs_0-9ms.txt, circoneout_2_0-9ms.txt, etc. that show the location and size of the blobs, in pixels.
   If loadblobs.pro is not given a shot number, it tries to extract one from the filename; the shot #'s for the above files go from 1-5. The corresponding tables in the blobs database are SH1-SH5.

   populateblobs.pro, called from loadblobs, converts from pixels per frame to Km/s, and writes to the database table. Stewart Zweben says that horizontal and vertical pixel distance is approximately the same. The conversion between pixels and centimers is 23.5cm/64 pixels.

3. DbAccess can be used to see what was loaded into the table

      IDL> dbaccess, 'blobs'	
   

   1. click on database table SH1 in the left column
   2. Move the following column names to the "Work Area" and click "Select" to see:

   

   Average speeds from the database can also be seen by running /u/bdavis/cvs/idl_cvs/checkvaves.pro

   IDL> for i=1,5 do checkvaves,i
   
     Reading from database table SH1
   Average x & y velocities for shot 1=  -5.9610376e-06    0.0098332996 Km/s
   Average x & y velocities for shot 1=   -0.0016012501       3.1088427 pixels/frame
   
     Reading from database table SH2 
   Average x & y velocities for shot 2=     0.011117898  -3.7208858e-05 Km/s
   Average x & y velocities for shot 2=       2.9864827    -0.011763751 pixels/frame
    
     Reading from database table SH3
   Average x & y velocities for shot 3=    0.0021750933    0.0013772208 Km/s
   Average x & y velocities for shot 3=      0.58427216      0.43541465 pixels/frame
    
     Reading from database table SH4
   Average x & y velocities for shot 4=    0.0038055245    0.0098963027 Km/s
   Average x & y velocities for shot 4=       1.0222375       3.1287614 pixels/frame
    
     Reading from database table SH5
   Average x & y velocities for shot 5=    0.0055449076   0.00095672607 Km/s
   Average x & y velocities for shot 5=       1.4894695      0.30247333 pixels/frame
   

   The above compare well with the values printed from the file-generation step, in 1., above.

Checking speeds for an NSTX shot in a database table

	IDL> dbaccess, 'blobs'
	
	select table, SH137582, move Column Names to Work Area and click on Select

    IDL> fcplayer, file='nstx_5_137582.cin', $
	           aveT1=0.325, aveT2=0.335, /horiz, /RotCC,  $
	           colorTable=3, minXsize=400, FrameAveDeltaMS=0  

• Under the Special menu, select "Turn ON Blob Tracking"
• Enter 330.0637 in the "Frame Time" box, above slider
• Mouse over the plus sign, and note the X loc and Z loc, in fields at the bottom. E.g., 35.0 & 66.9
• Click the "Advance-->" button 10 times, to go to 330.0888 ms
• Mouse over the plus sign, and note the X loc and Z loc, in fields at the bottom. E.g., 37.4 & 54.8.
• So in 0.02511 ms, X pixels=2.4 & Y pixels = -12.06
• So the average X Speed calculation between 330.0637 and 330.0888 would be:

       
         2.4 p   * 23.5cm * 1000ms *  1 Km
       ---------   ------   ------   --------  = 0.35 Km/s
       0.02511ms    64 p     sec     100*1000
  
  IDL> print, 2.4/.02511*23.5/64./100
        0.350956
  
  

  and the average Y speed would be

     
     -12.06 p  * 23.5cm * 1000ms *  1 Km
     ---------   ------   ------   --------  = -1.76 Km/s
     0.02511ms    64 p     sec     100*1000

IDL> print, -12.06/.02511*23.5/64./100
      -1.76355

• you could save to an IDL .sav file, restore and:

  IDL> inds = where( SH137582_TIME ge 330.062 and SH137582_TIME le 330.09)
  IDL> help, inds
  INDS            LONG      = Array[11]
  IDL> print, average( SH137582_xvel[inds])
        0.35266491
  IDL> print, average( SH137582_yvel[inds])
        -1.9471798
       
  

• To check velocities between adjacent frames, at times 330.06366 & 330.06619, say, velocity is computed from current position to the position at the last frame (so blobs with no parent will have a velocity of zero). So to compute the velocity recorded for time 330.06619:

                  X         Y      Xvel     Yvel
  330.06619    35.134    65.895   0.176   -1.423
  330.06366    35.014    66.870   
               ------    ------
  	      0.120    -0.974
  
  For Xvel
        0.120 p  * 23.5cm * 1000ms *  1 Km
       ---------   ------   ------   --------  = 0.175
       0.002511ms   64 p     sec     100*1000
  
  IDL> print, 0.120/.002511*23.5/64./100
       0.175478
  
  For Yvel
        -0.974 p  * 23.5cm * 1000ms *  1 Km
       ---------   ------   ------   --------  = -1.42430
       0.002511ms   64 p     sec     100*1000
  
  IDL> print, -0.974/.002511*23.5/64./100
     -1.42430
     
  these values show up in the file produced by populateblobs.pro, and that
   get written to the database:
  
  IDL> it=nearesti(time,330.06619)
  IDL> print, xvel[it], yvel[it]
       0.175974     -1.42316
  
  

Possible Future Validation Work

• The articially created files used here did not have a large average "hump" of signal near the plasma edge, as actual shots do. As such, "normalization" was not needed, and, in fact, would have degraded these results. Validation cases that have a realistic hump of average signal should be generated.
• Irregular-shaped blobs were not included in the above work. Small amounts of noise (~3%) had no effect on the above results, but larger variations should be tested.
• We should create cases of blobs splitting and of blobs merging.

If you have questions about this page, please send email to szweben.

updated: 1-Jul-2015
by: Bill Davis