Data selection post-sortac

Occasionally, it is necessary to reject data after examining the data cubes and time series produced by sortac (e.g. contamination by Earth-scattered X-rays). The program select allows the user to reject data from cubes and time series when any HK parameter falls outside a given range. It can apply cuts based on any HK parameter, which means that most of the cuts in sortac are available in select and there is no need to run sortac again. Note, however, that select only has the HK data available at the time resolution used to sort the data, so that results obtained from applying the same cuts in select and sortac may not be identical.

After a cube has been created it should nearly always be cleaned using cleanall. Flares in the soft electron background can cause the SUD and the PI_MONI rates to depart from their normal linear relationship. Cleanall removes such flares by deleting data points which lie CLN_NSIGMA001 standard deviations from the best-fit linear relationship.

For background data (and non-varying source observations) a similar relationship exists between the SUD and LAC count rates. CLN_NSIGMA002 specifies the number of standard deviations from the best-fit linear relationship. Its use can be dangerous, as the SUD/LAC relationship is only approximately linear, and if CLN_NSIGMA002 is changed from its default the file can easily be overcleaned. Over-cleaning an off-source file means that a background parameterisation subsequently obtained using that file will be incorrect. CLN_NSIGMA002 = 0.0 should be used for variable sources.

Finally, the LAC count rates in adjacent spectral channels are also compared to identify spectral data ``spikes'' (the relatively poor spectral resolution of the LAC ensures that this does not remove valid data). CLN_NSIGMA003 specifies the number of standard deviations from the best-fit linear relationship. An example of cleanall is given below:

qcl>cleanall

** CLEANALL Version 1.5 Mon Jun 16 18:34:59 1997 **
 
 CLEANALL removes some types of bad data from Ginga time series and
cubes by
 searching for data which does not obey a linear SUD v PI_MONI, SUD v
LAC or
 LAC v LAC relationship.
 
CLN_DATIN Cube to be cleaned =mkn335_91a.cube
CLN_DATOUT Cleaned data cube = mkn335_91a.cube_cl =
CLN_NSIGMA001 SUD/PI data acceptance factor (? for help) = 3.00000 =
CLN_NSIGMA002 SUD/LAC data acceptance factor (? for help) = 4.00000 =0
CLN_NSIGMA003 LAC/LAC data acceptance factor (? for help) = 5.00000 =4
CLN_DUMP Write output to screen? = No =
 
 Points deleted using the SUD/PI correlation:            8
 Points deleted using the SUD/LAC correlation:           0
 Points deleted using the LAC/LAC correlation:           1
qcl>

Clearly, the values of the CLN_NSIGMA00n parameters determine the level of ``cleaning'' applied. For each correlation, cleaning can be turned off by setting the appropriate value of CLN_NSIGMA00n to zero. Recommended values of CLN_NSIGMA00n for source and background observations are given in table 2.

Finally, after inspection of the cleaned data cube or time series (see section 2.7) it is sometimes necessary to manually select and remove suspect data. The QCL procedure set_times.qcl displays a time series, and enables periods to be flagged interactively. The program zapts deletes the flagged points from either data cubes or time series. Examples of set_times.qcl and zapts are given below:

qcl>.set_times
QCL_CL HDS file name = HDS_NODE =mkn335_91a.cube_cl_sq
QCL_CL Graphics device = xw =
Starting /soft1/xanadu/OSF1_4.0_alpha/bin/pgxwin_server.
    Graphics device /xw open
QCL_CL Mode: ZAP points or SET times = ZAP =
QCL_CL Ident channel to plot = 1 =1
QCL_CL PHA channel to plot = 1 =1
    HDS_NODE mkn335_91a.cube_cl_sq
    GRP_DATYPE TIME            
    GRP_ISEL            1
    GRP_IDLO            1
    GRP_IDHI            1
    LABX  Time (s)                                                    
    LABY  Count rate                                                  
    Number of data points          393
    Range of data points
    X  0.000000E+00 to   50304.0    
    Y   1.00000     to   32.6304    
QCL_CL Start time = 0.0000000E+00 =
QCL_CL End time = 50304.000000000 =
QCL_CL Plot YMIN = 0.0 =
QCL_CL Plot YMAX = 39.156480000000 =
    Mapping   0.0000000E+00   50304.00      0.0000000E+00   39.15648
    Position  0.000000E+00   37.1987    
    Position   50304.0       37.1987    
    Position  0.000000E+00   1.95782    
    Position   50304.0       1.95782    
 
  Select time
 
    Position   29738.9       30.6951    
  Relative time = 29738.878906250
  Absolute time(MJD) = 48439.575449987
  Time bin = 233
  Data value: 30.123980000000
 
  Select time
 
    Position   31255.8       30.6951    
  Relative time = 31255.800781250
  Absolute time(MJD) = 48439.593006953
  Time bin = 245
  Data value: 0.0
 
  Select time
 
    Position   24249.1      0.997767    
QCL_CL finished with plotting [Y/N] = Y =y
qcl>

The screen is divided into three sections: points are flagged by hitting the cursor in the middle panel; to advance to the next section of the light curve, hit the top panel; to stop the procedure, hit the bottom panel. The start and end times are in seconds, specified with respect from the start time used in the program sortac. The response ZAP to the QCL prompt will allow the user to flag single data points; SET will allow the user to flag sets of points.

qcl>zapts
 
** ZAPTS Version 1.1 Thu Oct 16 16:52:29 1997 **
 
ZAP_DATIN Data file name =mkn335_91a.cube_cl
 
 Number of time samples: 393
 
ZAP_MODE ZAP_BINS or ZAP_RANGES = ZAP_RANGES =
        ZAP_RANGES option chosen 1 ranges going ...
ZAPTS Complete
qcl>

The program zapts has two modes of operation. The response ZAP_RANGES to the parameter ZAP_MODE will remove inclusive bin ranges specified by the procedure set_times.qcl. Alternatively, the response ZAP_BINS will delete individual data points.