For sources near the optical axis of the telescope, the size of the point spread function is smaller than the size of the ACIS pixels. Applying the energy-dependent sub-pixel event-repositioning (EDSER) algorithm can improve the image quality of ACIS data for such sources.

This document provides guidance on how to reprocess the data with the EDSER algorithm and analyse the results. Please read the technical details available in the ACIS Sub-Pixel Event Repositioning why topic as well.

About the PSF Calibration

The CXC believes that the sub-pixel event-resolution (SER) algorithm does improve the ability to separate small scale structure, but we have not yet calibrated ACIS on sub-pixel scales, so obviously the potential for artifacts exists. As reported last year to the Chandra Users' Committee (Report, PDF), we have a plan to enhance our simulation tools to represent the SER-enhanced PSF. This is a significant calibration, modelling, and development effort since we need to model the grade distribution and have a much more detailed internal model of the instrument. Refer to the Chandra Ray Trace caveats for some other limitations on the PSF modelling tools.

In principle, sub-pixel analysis should not affect the HRMA ray tracing. However, at the sub-arcsecond scale, HRMA calibration is still on-going (e.g. Probing higher resolution: an asymmetry in the Chandra PSF). Therefore, there may be uncalibrated mirror-related effects at the subarcsecond scale in addition to the unmodelled ACIS detector pixel effects.

1. Reprocessing the Data

The first step in sub-pixel data analysis is to ensure that the data has been reprocessed with pix_adj=EDSER. EDSER became the default parameter value in CIAO 4.3, and has been the default in standard data processing (SDP) since version DS 8.4 (28 June 2011).

Before the use of EDSER, the coordinates of ACIS events were randomized by +/- 0.5 pixel to avoid possible aliasing affects associated with the CCD pixel grids. Information on that correction is available in the pixel randomization why topic.

The chandra_repro reprocessing script runs all the data processing threads for ACIS imaging data. The steps are also described in detail in the ACIS Data Preparation analysis guide.

The tool acis_process_events was modified in SDP version DS 10.4.2.1 (30 September 2015) and in CIAO 4.8 to make it possible to use the subpixel algorithm for continuous-clocking mode datasets. Since the subpixel adjustments affect not only the coordinates, but also the times of such datasets, and since the algorithm can introduce non-astrophysical features in the times of arrival, the adjustment is not used in the standard data pipeline.

2. Run ChaRT to create a PSF

Dither and PSFs at Sub-Pixel-Scales

At present, ChaRT does not account for the spacecraft dither, with information contained in the aspect solution file of an observation; thus limiting the usefulness of ChaRT-generated PSFs at sub-pixel resolution that are trying to match observations where the EDSER algorithm is applied.

At sub-pixel resolutions, without the dither, the discrete set of chip locations that result from applying the EDSER algorithm translates into discrete sky positions, causing an aliasing effect which is smoothed out by the aspect solution. SAOTrace v2 and MARX v5 can account for the dithering.

Using MARX to generate a PSF accounting for dither is demonstrated in the Using MARX to Simulate an Existing Observation thread.

The Chandra Ray Tracer (ChaRT) simulates the best available point spread function (PSF) for a point source at any off-axis angle and for any energy or spectrum. Technical details are available from the About ChaRT page.

It is important to read the caveats before running ChaRT and using the resulting PSF simulations in analysis.

3. Create a Sub-Pixel PSF Image

Thread: Creating an Image of the PSF

The projected event file created with ChaRT and Marx contains X and Y values without any pixel quantization, allowing a 2-D histogram image can be made at any desired sub-pixel scale.

To create a sub-pixelated image with dmcopy, specify a binning factor that is less than 1. This example uses 0.1, and the limits have been changed to +/- 25.6 pixels to create a 512 square pixel image:

unix% dmcopy \
      "marx_psf.fits[bin x=4678.41:4729.61:0.1,y=3674.33:3725.53:0.1]" \
      psf_subpix.fits

4. Convolve the Data

The sub-pixel PSF image can be used as a convolution kernel for fitting and modelling in Sherpa. It may also be used to convolve the data by running one the CIAO's convolution tools aconvolve.