GammaLib

Here the time intervals of the different files:

The SKYDR3 validity intervals are set from the start of the first OAD and the stop of the last OAD superpacket that is used:

C             set the start and end validity interval
              IF(INIT.EQ.1) THEN
                INIT=0
                VALSTA(1)=OATSRT(1)
                VALSTA(2)=OATSRT(2)
                VALEND(1)=OATEND(1)
                VALEND(2)=OATEND(2)
              ELSE
                VALEND(1)=OATEND(1)
                VALEND(2)=OATEND(2)
C               avoid ticks > 1 day
                IF(VALEND(2).GT.691 199 999)VALEND(2)=691 199 999
              ENDIF

              OATEND(1)=CURTIM(1)
              OATEND(2)=CURTIM(2) + 131071

The GammaLib DRE and DRG times are set by the GCOMDri::use_superpacket() method and reflect the start time of the first and the stop time of the last superpacket that is used, hence the times should be equivalent to those of SKYDR3 is the same OAD files are used.

The SKYDR3 start time corresponds to row 2132 of m22366_oad.fits, while the stop time corresponds to row 2152 of m22387_oad.fits, hence in principle the same OAD records are accessible to GammaLib. The method GCOMOads::read() sets the OAD stop time in the same manner as SKYDR3, hence in principle identical times should be achievable. It is possible a difference in the TIM that can explain the difference.

Using m14033_tim.fits instead of the GammaLib database TIM produces the exact same start and stop times. However it turns out that the number of superpackets read is smaller than for the SKYDR3 run, and that the OAD file for TJD 9125 is missing.

After Werner sent me the missing OAD dataset I reprocessed the data with the full list of OAD files.

An initial comparison of the SKYDR3 and GammaLib DRIs showed some structure that indicated some difference in the WCS projection between the DRIs. Reproducing the coordinate transformations in SKYDR3 and GammaLib using the script check_dri_projection.py confirmed that the script dri2fits.py that I used to add WCS projection information to the FITS header was not correct. After correcting the FITS headers (and after replicating the same logic in check_dri_projection.py) the script produced the result below. Each row corresponds to one event. Columns with CO correspond to COMPASS (i.e. SKYDR3), columns with GA to GammaLib. The first step of columns show the X/Y pixel index for both code followed by the difference. Pixel indices are now identical. The second set of columns shows the Galactic longitude and latitude of the events, also here the results are very close. Note that while GammaLib exploits directly the Galactic longitude and latitude information in the EVP file, SKYDR3 performs coordinate transformations, using the event’s theta and phi angles.

************************
* Check DRI projection *
************************

   #      CO_X     CO_Y     GA_X     GA_Y  CO-GA_X  CO-GA_Y |     CO_L     CO_B     GA_L     GA_B  CO-GA_L  CO-GA_B
====  ======== ======== ======== ======== ======== ======== | ======== ======== ======== ======== ======== ========
   0:  43.6003  34.0330  43.5958  34.0331  0.00443 -0.00004 | 194.1003  -8.4670 194.0958  -8.4669  0.00443 -0.00004
   1:  53.5884  35.6523  53.5826  35.6512  0.00579  0.00109 | 204.0884  -6.8477 204.0826  -6.8488  0.00579  0.00109
   2:  38.5077  31.6222  38.5036  31.6225  0.00408 -0.00032 | 189.0077 -10.8778 189.0036 -10.8775  0.00408 -0.00032
   3:  29.6426  16.8326  29.6371  16.8324  0.00545  0.00017 | 180.1426 -25.6674 180.1371 -25.6676  0.00545  0.00017
   4:  45.0548  31.9575  45.0500  31.9575  0.00480 -0.00003 | 195.5548 -10.5425 195.5500 -10.5425  0.00480 -0.00003
   5:  28.6003  63.7477  28.6017  63.7496 -0.00131 -0.00192 | 179.1003  21.2477 179.1017  21.2496 -0.00131 -0.00192
   6:  27.4380  20.9859  27.4332  20.9856  0.00482  0.00031 | 177.9380 -21.5141 177.9332 -21.5144  0.00482  0.00031
   7:  36.0730  58.9957  36.0722  58.9962  0.00078 -0.00049 | 186.5730  16.4957 186.5722  16.4962  0.00078 -0.00049
   8:  50.3808  44.0255  50.3761  44.0242  0.00476  0.00127 | 200.8808   1.5255 200.8761   1.5242  0.00476  0.00127
   9:  38.2415  19.7424  38.2361  19.7428  0.00537 -0.00037 | 188.7415 -22.7576 188.7361 -22.7572  0.00537 -0.00037
  10:  57.5948  11.2286  57.5868  11.2291  0.00796 -0.00048 | 208.0948 -31.2714 208.0868 -31.2709  0.00796 -0.00048

Finally the plots below show a comparison of the DREs for the four energy bands. Since the projects of the DREs are not identical I did some random resampling of the events into the same projection, which of course can not reproduce an identity. Nevertheless the agreement is very satisfactory. Only the 10-30 MeV band shows a significant difference, which comes from the different ToF selections used by Werner and myself.

Here are the comlixfit results for the four energy band in comparison to the SRCFIX results. While for SRCFIX the results are not corrected for ToF and lifetime, GammaLib includes these corrections. I therefore applied the corresponding corrections to Werner’s results for comparison (1.40, 1.27, 1.17 and 1.05 for the ToF correction of the four energy bands, 1.03 for the lifetime). For the analysis, the UNH simulated response files were used.

Below a comparison of the number of counts that were attributed to both sources.