Skip to the navigation links
Last modified: 14 April 2021


Caveats and Limitations

While the Chandra Source Catalog (CSC) is capable of supporting many diverse scientific investigations, users should carefully consider the limitations of the CSC when assessing the efficacy of the catalog for their particular line of inquiry.

Chandra Source Catalog 2.0

This page describes the caveats for release 2.0 of the Chandra Source Catalog. Although we recommend people to use this release, the caveats and limitations for release 1.1 should be referred to if using the older data.

  1. Catalog Data:
    1. General
    2. Specific Caveats and Limitations applicable through CSC release 2
    3. Specific Caveats and Limitations for certain ObsIds
    4. Flux Correcting Very Bright Sources
    5. Issues for users of the preliminary release of CSC 2
  2. Catalog Interfaces
    1. CSCview
    2. The contributed scripts

Catalog Data


In particular, users of the CSC should be aware that there may be fundamental and significant selection effects that restrict the source content of the catalog and which therefore may limit scientific studies that require an unbiased source sample. The CSC is constructed from pointed observations obtained using the Chandra X-ray Observatory; it is not an all-sky catalog, and does not include sources detected to a uniform depth. Users should also be aware that the detection of sources with spatial extent more than ~30 arcsec is done with a separate algorithm from that for compact sources (the sources have names which end in an 'X' to distinguish them from compact sources).

Note that for some sources in observations with short exposures, there are zero counts in the local background region for some bands. In these cases, the source flux is not computed.

Specific Caveats and Limitations applicable through CSC release 2

  • The true astrometric position error should be consistent with the reported position error for ~90% of the sources. For the ~10% of sources for which this is not the case are not distributed randomly. There are a small number of stacked observations for which the absolute error may be significantly larger than reported (up to a maximum of ~2.0 arcseconds).
  • The model independent flux_aper values are computed from the events' observed energy. In narrow energy bands, especially the ultra-soft (0.3-0.5keV) band, the spectral response can change rapidly with energy. This means a small change in observed energy from an event's true energy (as described by the RMF) can result in large changes in the flux. If there are few events this can result in large, undetermined systematic errors. In the other energy bands where spectral response generally changes slowly with energy this is less of a problem. It also becomes less of a problem when there are a large number of events as the Gaussian-like RMF is well sampled and the uncertainty balances out.
  • Caveat on inter-observation variability: This has since been resolved in CSC 2.0.1. Users should be aware that in CSC 2.0 the master-level inter-observation variability properties (var_inter_flag, var_inter_prob, var_inter_sigma, var_inter_index, var_inter_hard_flag, var_inter_hard_prob, var_inter_flag_sigma) were estimated using only a fraction of all observations contribution to the master source, i.e., only those in the best time-ordered Bayesian block for a particular master source. This did not include all the observations of a particular source. As a result, about 10% of the truly variable sources at the master level had unreliable variability properties. Users interested in selecting variable sources were advised to use the stack-level variability properties, that were estimated using all observations in a given stack. This did not affect intra-observation variability properties at the master level.
  • Some data products such as PHA files (pha3) contain 'pixel masks', a new feature. Users of the Chandra Interactive Analysis of Observations (CIAO) data analysis package prior to version 4.10 may encounter problems using these files with some tools.
  • Users of the XSPEC X-ray spectral fitting package may need to manually split the source region PHA spectrum file (pha3) into separate source region and background region spectra prior to analysis in XSPEC; see the page "Using the L3 Data Products" for instruction. Users of Sherpa can just read in the file with load_pha or load_data, which will read in the background automatically, and so do not need to split up the file.

Specific Caveats and Limitations for certain ObsIds

  • Some ACIS observations were performed with energy filters applied. Depending on the particular observation, this could remove all or most of the sensitivity for certain CSC energy bands. These data will have NULL values for some of the flux properties and possibly the hardness ratios. More details can be found in the ACIS event energy threshold caveat.

    This caveat also applies to the properties calculated for the convex-hull sources, athough in this case only fluxes are calculated (no spectral fits, hardness ratios, or timing analysis), so the number of affected columns is less than for compact sources.

Flux Correcting Very Bright Sources

For some very bright sources, such as 2CXO J150842.1+332802, the method used to correct the energy flux in the source aperture for background contributions, using energy flux in the background aperture, may fail, resulting in NaN's for flux_aper or flux_aper90. Other flux quantities, such as photflux_aper, or src_rate_aper, are unaffected.

Issues for users of the preliminary release of CSC 2


Version 2.0 of the Chandra Source Catalog has now been finalised, as of October 24 2019. Prior to this date the CSC 2.0 data was made available to the community as the "current database" in CSCview. The following section is only relevant for people who used the CSC 2 data before the October release.

The "current database" view of the catalog saw several changes over time as the CXC performed validation and verification steps on the data. The following section lists the major changes made to the data. Please note that due to the complex nature of some of the changes, and the fact that these changes were not all made at the same time, there is no easy way to identify if any particular data request was affected. It is strongly suggested that queries are repeated against the finalised 2.0 release is if there is any concern about potential changes.

  • The limiting-sensitivity maps were incorrect.

  • A small number of per-observation PSF files were wrong.

  • The background components of the per-observation light curve files were invalid.

  • Some Master Source energy flux columns were empty even though the source significance was high. This happened for sources with more than 10 contributing ObsIDs.

  • For a small number of master sources (6,648), the aperture photometry values for the master source best estimate, master average, and detect stack average were either missing or in error. Bayesian blocks data products for these sources could be in error. The cause was two-fold:

    1. a bug in the algorithm for combining data from ten or more ObsIDs
    2. a bug in the algorithm for combining data from any number of ObsIDs, if the intensity-space of the probability distributions for individual ObsIDs did not overlap sufficiently
  • For a number of master sources (30,874) aperture photometry values for individual observations (ObsIDs) were wrong, due to errors in computing the encircled energy fractions in the source apertures.

  • The upper and lower limits on the spectral-fit parameters - such as powlaw_gamma_hilim and powlaw_gamma_lolim - were incorrect for certain values.

  • The extent flag and the Source Extent parameters were not reliable for compact sources identified as being larger than the Chandra Point-Spread Function.

Catalog Interfaces

Access to a limited set of data products related to convex-hull sources will be made available soon. The affected products are the master-source poly3 files and the source images.


The software-requirements page contains current instructions for downloading and running the CSCview java application.

Interoperability with other tools

  • Users are warned that there is a limit to the amount of data which may be imported into CSCview from a remote, SAMP-connected client (such as Topcat), using the crossmatch User table --> Received Table option in order to perform a crossmatch of this data against the CSC. This limit is 250000 rows in a table of data being imported, e.g., a table containing a full catalog of astronomical data to be crossmatched against the CSC. This means that the number of source matches reported by CSCview will be incorrect, as only the first 250000 rows of the input table will have been used in the query. It is important to note that when an incoming table of data is truncated by CSCview, no warning is issued.


The quick-search web page is new in CSC 2.0, and allows you to search for a fixed set of master-source properties, and the limiting sensitivity, for one or more positions on the sky. The limiting-sensitivity search is limited in the radius over which it will search.

The contributed scripts

The search_csc and obsid_search_csc scripts must be updated to at least version 4.11.5 of the contributed scripts package otherwise they will fail with an error message similar to the following:

# search_csc (06 March 2018): ERROR: Problem accessing CSC site.  Server responded with the following message:
Query error: master_obi_assoc not found. Specify owner.objectname or use sp_help to check whether the object
exists (sp_help may produce lots of output).