Frequently asked questions

  • 1) Question : When should 'regmode' be set to RADEC and when should it be set to DET in aharfgen ?
    Answer: For the SXS the 'regmode' parameter should always and only be set to DET and the selection region should always be in DET coordinates. Any other 'regmode' is not allowed and would result in an incorrect ARF. The selection DET region should include all the pixels that are to be used and partial pixel are not allowed. The calibration pixel is always ignored. For SXI and HXI the 'regmode' parameter should be set to RADEC and the region should be in RA, DEC (following standard SAO region formats). Here is a summary of the allowed 'regmode' for different instruments:
    Instrument:	        SXS	SXI	HXI		
                     	DET	RADEC	RADEC
                            	DET	DET
  • Q: 2) Can I make an SXS spectrum by selecting on PIXEL in XSELECT?
    Answer: The SXS spectrum may be extracted by selecting on PIXEL in XSELECT. However it is not reccomended since 'aharfgen' requires the region in DET to generate the ARF. If the spectrum is extracted in PIXEL, the DET region required by 'aharfgen' must contain exactly the pixel selected.
  • Q: 3) When calculating an exposure map file how should I choose the number of attitude bins and the how should be set the parameters 'delta' and 'numphi'. How many attitude bins are needed for 'aharfgen' ?
    Answer: If the attitude is stable to better than ~15 arcsec or so, valid in most cases, 1 attitude bin is sufficient when creating the exposure map file to be used in aharfgen. To make an exposure map file with 1 attitude bin, 'numphi' should be set 1, and 'delta' set to 20 (in arcmin) is sufficient for all Hitomi instruments since much larger than any expected attitude variation. If the attitude is varies by more than ~15 arcsec during the observation, or if the core of the source PSF is near the edge of the active detector region (e.g. SXI, MZDYE mode) then more than 1 attitude bin gives a more accurate ARF. For example setting delta=0.25 and numphi=4, 'ahexpmap' creates several attitude bins. The actual number of attitude bins corresponds to the number of rows in the 1st extension of the exposure map.
  • Q: 4) How should be determined the value for paramater 'numphotom' in 'aharfgen' and what effect does it have on the statistical error of the ARF?
    Answer: If N is the number of attitude bins in exposure file made by 'ahexpmap' (= number of rows in extension 1 of the exposure map file), reasonable values are numphoton = A/N, where A=300,000 for SXS, SXI, and A=10,000 for HXI. These values result in run times less than an hour or so. These rough estimates are for standard selection regions (SXS: full array; SXI: 2.5' circle; HXI: 3' circle) and smaller region sizes require larger values of A. The statistical errors on the effective area depend on energy and the dependence can be seen in Fig. 4.1a in the Hitom Data Analysis Guide. The plots show the maximum statistical error that is achievable with the current CALDB files and numphoton=10,000,000.
  • Q: 5) When I run aharfgen, I see this message on the screen "There are an insufficient number of photons to make the ARF". What should I do?
    Answer: There are several possibilities (one or more may apply):
    (a) The coordinates of the center of the region in the region file are wrong;
    (b) The region size is too small;
    (c) The value of 'numphoton' is too small.
  • Q: 6) The tool aharfgen is taking many hours to run. What is wrong?
    Answer: Check the following:
    - Value of numphoton
    - Number of attitude bins: should not be more than ~10. If there are too many, check the coordinates of the center of the selection region are correct and that the 'source_ra' and 'source_dec' input parameters are correct.
    - If sourcetype=image the image should not be too large (generally the maximum linear dimension should not be more than ~20 arcmin).
  • Q: 7) In aharfgen image mode (sourcetype=image), the effective area in the ARF appears to be more than an order of magnitude too small.
    Answer: It is likely that the values of the input parameters 'source_ra' and 'source_dec' are not correct. They should correspond to the RA and DEC of the centroid of the source respectively. If the source is diffuse with no obvious center then 'source_ra' and 'source_dec' should be set to the values at the center of the image.
  • Q: 8) When I make an HXI flat field using hxirspeffimg there appear to be an anomalous patch along one edge of the image. What is wrong?
    Answer: The anomalous spot is due to shadowing by the Americium calibration source.
  • Q: 9) My background spectrum shows negative counts in some energy channels in XSPEC. What is wrong?
    Answer: There are two possibilities:
    (a) The value of the BACKSCAL keyword in the spectrum file is wrong, or
    (b) The source is very weak, in which case the background should be modeled, and included in the source model, and the spectrum without background subtraction should be fitted with the combined model.
  • Q. 10) I don't see a source in the image I made from a clean events file. What is wrong?
    Answer: The two most likely causes are:
    (a) The pointing of the satellite deviated from the intended pointing during the time interval covered by the cleaned event files. This happened on several occassions during the Hitomi mission. Please consult the observation log in Appendix H of the data analysis guide, which gives such details on all the Hitomi observations. The data analysis guide can be found at
    (b) The energy range selected for making the image can significantly affect the counts in the image because the spectrum may be diminished at low energies (e.g. due to absorption, or in the case of the SXS to the gate valve). or high energies (e.g. if the source is "supersoft"). The remedy is to make an image covering a wider energy band.
  • Q: 11) How do I subtract background if there is no appropriate region in the field from which to extract a background spectrum?
    Answer: In some energy ranges, depending on the instrument and the brightness and spatial extent of the source, the background may be negligible enough to ignore. If that is not the case then one can use the tool ahnxbgen to generate a non-X-ray (instrumental) background. The cosmic X-ray background must be included in the model of the source. The Hitomi data analysis guide at has a section on background treatment for each instrument. Further examples can found in the "step-by-step" guide (also linked from the above WWW page), for example in the section on the source G21.5-0.9.
  • Q: 12) What is the difference between the different types of SXS RMF?
    Answer: Four types of RMF files may be created. The small (s/S) accounts only for the Gaussian core of the line-spread function (LSF). The medium (m/M) includes also the exponential tail. The large (l/L) includes the escape peaks and the x-large (x/X) adds the continuum down to the minimum energy given by the 'emincont' parameter. Note that the x-large may exceed the 2.1 GB default limit for the FITS file structure used in response files. Thus the successful creation and application (e.g., in XSPEC) of x-large RMF files cannot be assured, and this option is not recommended for normal usage.
  • Q: 13) I made an 'extra large' (XL) RMF for the SXS but XSPEC crashes when I try to load the RMF into XSPEC.
    Answer: Unfortunately, on some machines, the extra large RMF is too large for XSPEC. Currently the only option, aside from trying another machine, is to use a smaller RMF.
  • Q: 14) In the antico lightcurve files, I see that the time values start at 0. I suppose that 0 corresponds to TSTART in the header, that the number in each bin covers the next 20 seconds, and that the rates are per-second, one-side. Is all of that right?
    Answer: You are partially correct. Rates are indeed given in count/s and correspond to one side, A. The integration time covers 20 seconds. The time is not offset from TSTART but from TIMEZERO and is calculated as TIMEZERO + Time(n) where TIMEZERO is given as a header keyword and Time(n) is the value in the TIME column for row n.
  • Q: 15) If we input in 'aharfgen' a chandra image after the background subtraction, the image contains minus numbers in some pixels. How does the software treat the minus numbers?
    Answer: 'aharfgen' called the simulator 'heasim' that uses the image. 'heasim' treats the image as a probability distribution, and assigns pixels with negative values as having zero probability -- the same as pixels with no counts.
  • Q: 16) What the meaning source_ra/dec arguments for 'aharfgen' in the case of sourcetype=image for diffuse source ?
    Answer: In aharfgen in image mode source_ra and source_dec are used in a factor that is the cosine of the off-axis angle in order to calculate a flux normalization factor for the ARF. If there is no obvious source in the image then source_ra, source_dec should be set to the center of the image. The values of source_ra, source_dec do not affect the raytracing file. If the values of source_ra and source_dec correspond to a point that is very far from any point in the image, the off-axis cosine factor may be significantly inaccurate. Note that in image mode, aharfgen calculates linear approximation formulae for coordinate conversions, using the nominal pointing, but the source_ra and source_dec parmeters are not used for this.
  • Q: 17) I understand how the source_ra/dec are used in the aharfgen in the image mode. Is it possible to estimate degradation of conversion accuracy at an offset angle from source_ra/dec? For example in Perseus cluster, if (source_ra, source_dec) = (49.9507, 41.5128) as specified in the step-by-step guide, what is the accuracy at 5 arcmin offset angle from that source_ra/dec? It might have a significant impact when I estimate the contamination from the PSF tail. (say contamination from the bright central part to the obs1 region)
    Answer part 2: This factor that is calculated from the values of source_ra and source_dec is an overall normalization so does not impact the PSF analysis. Regarding the accuracy of the coordinate transformation approximation, we estimate that at the RA & DEC in your example, in a square grid +/-9 arcmin centered around the optical axis, the worst error in RA or DEC is not greater than ~0.15 arcmin.
  • Q: 18) I tried to multiply SXS-ARF and extra large SXS-RMF of the Perseus data using the marfrmf ftools task, but it failed by segmentation fault. When I tried the same procedure with large or medium RMF, it finished successfully. Therefore, I suspect that the cause of the segmentation fault is the very large file size of the extra large RMF. Is there any way to work around this problem?
    Answer: The marfrmf has not a consistent behaviour across platform and not always seg fault. Note the rsp created by marfrmf is in general twice in size of the original so for the xlarge extended that will 4GB and may create other problems. A way around is to use sxsmkrmf and use the paramater to create an rsp as for the exmaple below sxsmkrmf infile=in.evt.gz outfile=out.rmf resolist=0 regmode=det regionfile=SXS_det.reg whichrmf=x nchanin=32768 dein=1.0 eminin=0.0 outrsp=yes outrspfile=out.rsp arfinfile=in.arf Last I am sure you know but there is not need to create an rsp and using arf and rmf in xspec work as well.