Detect Pipeline
The detect pipeline is run for each calibrated observation interval (OBI) from the calibrate pipeline. In the detect pipeline, full-field data are used to determine the total (cosmic plus instrumental) background during the observation, and a wavelet-based source detection algorithm identifies candidate X-ray sources.
Create high and low frequency background maps
The ACIS background is split into two parts: a high-frequency "streak map" and a low-frequency mean background. HRC data only has a low-frequency component.
The high-frequency background component (Figure 1) is determined by creating a per-chip sky image in each energy band. The image is rotated to align with the y-axis. Source-free rows are found and the counts information in these rows is used to exclude rows which have Nσ above the minimum counts. The source-free rows are projected onto the x-axis and the mean value is replicated along the streaked columns. Finally, the image is rotated back to sky coordinates.
The low-frequency background component (Figure 2) is created by sliding a 129x129 pixel box across the image and computing a histogram of pixel values. The mean value of pixels within +/- 1.5 of the peak is then used as the map value.
The background maps are combined (in the ACIS case) and exposure-corrected; the final product is included in the source detection.
![[ACIS high-frequency background map]](imgs/det_hf.png)
Figure 1. ACIS high-frequency background map. The readout streaks are caused by bright sources.
![[ACIS low-frequency background map]](imgs/det_lf.png)
Figure 2. ACIS low-frequency background map.
Detect source candidates with wavdetect
Source detection candidates are identified by a wavelet transformation method, using the wavdetect tool. (The version of wavdetect used for CSC is a slightly modified version of the CIAO 4.0 tool.)
An independent run of wavdetect is done for every combination of source detection energy band and blocking factor; e.g. for ACIS, 4 source detection bands and 3 blocking factors generate 12 wavdetect runs. Five wavelet scales are used (1, 2, 4, 8, and 16 pixel radii), and the significance threshold for identifying a pixel as belonging to a source is set to 2.5e-7.
The exposure maps created in the calibrate pipeline and the combined, exposure-corrected background map for the OBI are both included in the source detection.
The output source list is filtered to eliminate detections with fewer than 6 net counts or a detect significance lower than 2.5.
Combine detections
The detections are combined by comparing the smallest blocking factor (i.e. highest resolution) results first across all energy bands. If the detection appears in multiple energy bands at the same blocking factor, the entry with the highest detect significance is chosen.
The detections are then compared across blocking factors. There may be cases, for instance, where there are two candidates identified at a location at bin=1, but only one resolved at the same location at bin=4. The results at the smaller binning take precedence over those at the coarser resolution.
Calculate the source and background regions
The spatial regions defining a source and its corresponding background are determined by scaling and merging the individual source detection regions that result from all of the spatial scales and source detection energy bands. The result is a single elliptical source region and a single, co-located, scaled, elliptical annular background region.
Then the modified source region and modified background region for each source is calculated. These regions are defined as the areas of intersection of the source region and background region for that source with the field of view, excluding any overlapping source regions. The modified regions are recorded as separate blocks (HDUs) in the region file (reg3.fits) file.
This source is then considered a candidate for inclusion in the CSC.
Calculate the limiting sensitivity
The limiting sensitivity is a measure of how faint the source would have to be to not be detected.
The background map is used to simulate counts with a Poisson mean at each location. The PSF size of each pixel is determined. Given a region the size of the PSF and the background counts in that region, the upper limit for a 3 σ detection is computed.
Figure 3. shows that a source with a relatively low flux which might be detected on-axis would have to be brighter to be detected farther off-axis. The PSF gets larger as you move off-axis, so the counts are more spread out and the source may not be detected against the background. The same source on-axis with a smaller PSF could be detected.
![[Image of Limiting Sensitivity]](imgs/det_sensitivity.png)
Figure 3. Image of Limiting Sensitivity [photon/cm²/s].
Output data products
The detect pipeline produces these data products:
region file (reg3.fits): the modified source region and modified background region for the source.
full-field background-subtracted source image (img3.fits) for each science energy band and blocking factor.
full-field background image file (bkgimg3.fits) for each science energy band and blocking factor.
full-field sensitivity map (sens3.fits) for each science energy band with a blocking factor of 4 for ACIS-I and ACIS-S, 12 for HRC-I, and 25 for HRC-S.
Post-pipeline Tasks
After this pipeline runs, some source candidates are automatically identified for removal. When a source is removed, it is not included in the database and no further processing is done.
- regions where the radius is too large or too small
- regions which are too elliptical
- detections which have too few counts or the region area is too small
Manual review of the pipeline output may identify additional detections for removal:
- the low-frequency background rate is too high
- the local source density is too high
- too many total sources were detected in the field