Bugs: wavdetect

Three significant issues have been identified in the CIAO 4.0 version of wavdetect:

• The falsesrc parameter should be left at the default value
• The lower limit on the sigthresh value is 1e-8
• Limitations in the PSF information

Caveats

1. Calibrated range: 0.1 to 1e-8 false sources per pixel (12 Dec 2007)

Bugs

1. There cannot be two sets of brackets ("[ ]") in the input file specification.

2. Autonaming of ASCII region files uses the .fits extension instead of .reg as would be expected.

3. Zero-length radius in source region (23 Mar 2007)

4. Comma-separated values for scales parameter (16 May 2007)

5. Wavdetect deletes the input file (10 Oct 2007)

6. The falsesrc parameter should be left at the default value (12 Mar 2008)

7. The lower limit on the sigthresh value is 1e-8 (12 Mar 2008)

8. Limitations in the PSF information (12 Mar 2008)

Caveats

1. Calibrated range: 0.1 to 1e-8 false sources per pixel (12 Dec 2007)

   The calibrated range of the new wavdetect is 0.1 to 1e-8 false sources per pixel. Depending on the significance settings and the input image size, users can go outside of this range in which case the results become unpredictable.

Bugs

1. There cannot be two sets of brackets ("[ ]") in the input file specification.

   An example of an unacceptable syntax is

   unix% wavdetect "evt2.fits[energy=500:8000][bin X=4275:4925:1,Y=3775:4425:1]"
   

   The tool will exit with an error:

   grep: RE error 49: [ ] imbalance.
   parammatch : can't find parameter X in reference list
   parameter error?
   

   Workaround:

   Include the parameter name before the file specfication:

   unix% wavdetect infile="evt2.fits[energy=500:8000][bin X=4275:4925:1,Y=3775:4425:1]"
   

2. Autonaming of ASCII region files uses the .fits extension instead of .reg as would be expected.

   In general, autonaming only works for simple cases.

3. Zero-length radius in source region (23 Mar 2007)

   If the source region determined by wavdetect is only one pixel wide, then the axis in that direction is equal to 0. If the entire region is one pixel square, then both the semi-major and minor axes are (0,0). A region with a zero-length yradius looks like this in the output ASCII region file:

   ellipse(4164.500000,3885.033333,5.986652,0.000000,90.000000)
   

   If the source list is used as stack input to dmextract - e.g. to create radial profiles - these zero-length regions will result in NaN values in the output file, usually because the source area is zero.

4. Comma-separated values for scales parameter (16 May 2007)

   The values given for the scales parameter must be space-separated, e.g "scales='1.0 1.414 2.0 2.828 4.0 5.656 8 11.3'". If commas are used instead, the tool only recognizes the first scale and disregards the rest.

5. Wavdetect deletes the input file (10 Oct 2007)

   If the input filename does not contain a "." (e.g. ".fits"), the tool deletes the input file and exits with an error.

6. The falsesrc parameter should be left at the default value (12 Mar 2008)

   There are problems in the extrapolation which make the results potentially incorrect with any falsesrc value. The default value of "-1" indicates that wavdetect should use the sigthresh value instead, replicating the CIAO 3.4 syntax of the tool. The CXC is investigating a more robust algorithm for use with falsesrc.

7. The lower limit on the sigthresh value is 1e-8 (12 Mar 2008)

   Independent of the image size, users should not set the sigthresh value lower than 1e-8. There are extrapolation issues with the significance lookup tables at lower values. Note that there is still statistical noise at 1e-8. 1e-7 is better, 1e-6 is where the statistics get "clean".

   This bug was introduced in CIAO 4.0. It is still recommended that this version of wavdetect be used with sigthresh set above the limit. If you choose to run CIAO 3.4 wavdetect instead, understand that the significance lookup tables are not as accurate as those in CIAO 4.0.

8. Limitations in the PSF information (12 Mar 2008)

   Beyond 20 arcminutes, wavdetect significantly underestimates source properties. This is because of lack of PSF information. In this case, the PSFRATIO and PSF_SIZE values in the source list will be negative. Even if the counts look reasonable, users should check these values for sources far off-axis.