Skip to the navigation links
Last modified: 23 February 2010

URL: http://cxc.harvard.edu/ciao4.2/why/cti.html

ACIS CTI Correction

Data Analysis

The tool acis_process_events includes a charge transfer inefficiency (CTI) adjustment procedure that can be used to compensate for most of the effects of CTI. This adjustment can significantly improve the spectral resolution of the data. As of CALDB 4.1/CIAO 4.1 (15 December 2008), a number of CIAO tools require that the event files have a CTI_APP keyword; the ACIS CTI_APP Keyword Required section explains the details and how to add the keyword.

There is CTI calibration data for the entire ACIS detector (chips I0-3 and S0-5). Only the -120 C focal plane temperature is calibrated.

If the data have gone through Reprocessing III, the CTI correction has been applied. Repro III started with DS 7.6.7. For instructions on how to check the processing version and apply the CTI correction to your data, see the Reprocessing Data to Create a New Level=2 Event File thread.

How the Data are Affected

There are a number of values in the event file that are affected by applying the CTI correction. The columns that can be changed include FLTGRADE, GRADE, PHA, ENERGY, PI, and STATUS. Here is an outline of what takes place:

Additionally, the STATUS column can be modified if the adjustment does not converge in the specified maximum number of iterations. An explanation is provided in the apply_cti parameter description for acis_process_events; also see the max_cti_iter and cti_converge parameters.

ACIS CTI_APP Keyword Required

CALDB 4.2 requires that all ACIS event files have a CTI_APP header keyword to indicate whether the CTI correction has been applied. The older CTI_CORR keyword is no longer used. The CTI_APP header keyword was added to standard data processing at version DS 7.6.10 and to data reprocessed in CIAO since version 3.4.

CIAO 4.2 requires CALDB 4.2.0 to work correctly. The following CIAO 4.2 tools and scripts are affected by this change:

These tools will either fail or return incorrect results if the CTI_APP header keyword is missing.

Adding the CTI_APP keyword

If your dataset is old enough that it doesn't have a CTI_APP keyword, consider running the Reprocessing Data to Create a New Level=2 Event File thread to be sure that the newest calibration has been applied to the file.

If you don't wish to reprocess, run the CIAO contributed script, check_ctiapp.sh, to update the file header:

unix% check_ctiapp.sh acis_evt2.fits
Setting CTI_APP value to PPPPPNPNPP

It is also simple to edit the file header to add a CTI_APP keyword; the following is the same as running check_ctiapp.sh.

The CTI_CORR value is a boolean: "TRUE" (or "1") if the CTI has been applied and "FALSE" (or "0") if it hasn't.

Comparison of CTI_CORR and CTI_APP header keywords

Prior to CIAO 3.4 and DS 7.6.10, the CTI_CORR header keyword was used to record whether the CTI correction had been applied to a dataset. The value of the keyword was a boolean: "TRUE" (or "1") if the CTI had been applied and "FALSE" (or "0") if it hadn't:

unix% dmkeypar acis_1838_evt2_yes.fits CTI_CORR echo+
1

unix% dmkeypar acis_1838_evt2_no.fits CTI_CORR echo+
0

The updates to CTI support in the CIAO 3.4 and DS 7.6.10 releases include a new, more descriptive header keyword, CTI_APP. The value of this keyword is a ten-character string which records the CTI applied to each of the ACIS chips (0-9, left to right). Allowed character values are:

Here are some examples:

Note that all ten spaces in the value are always populated, even though there are only a maximum of six chips on at a time.

The CTI_CORR keyword is retained in the output for backward-compatibility, but all CIAO tools have been updated to use the value of CTI_APP when checking for the CTI correction.

Additionally, the CTIFILE header keyword stores the name of the CTI calibration file applied to the data:

unix% dmkeypar acis_1838_evt2.fits CTIFILE echo+
acisD2000-01-29ctiN0006.fits

Reapplying or removing the CTI correction

If the input file has already been CTI-corrected and "apply_cti=yes", the CTI correction is recalculated. It is not, however, applied on top of the first correction; since the PHAS column remains unadjusted, the new correction will start from there to redetermine the PHA, ENERGY and PI values.

If the input file has already been CTI-corrected and "apply_cti=no", the CTI correction is removed (PHA, ENERGY and PI values are recomputed using the original, pre-CTI-corrected data).

Caveat: Frame times, subarrays, and CC mode

The CTI adjustment has been fully tested and calibrated on -120 C observations in TIMED mode with frame time of 3.2 s. The calibration group knows that the CTI process is slightly different in modes which have longer or shorter frame times, such as subarrays. This is because different amounts of cosmic ray and other background charge can partially fill traps. They expect that CTI-corrected data will still be an improvement, and will be as accurate as our calibration base allows at the present time. Please let us know (via Helpdesk) if you have evidence of specific discrepancies due to different frame times.

In CC mode there is also a change in parallel transfer rate; the rate for TE mode is 40 microsec/row, and for CC mode it's 2.8 millisec/row. Different traps become more or less important for the different transfer rates. There is very little calibration data in this mode, so it is suggested that CTI correction be compared with uncorrected data and the results carefully considered.


Technical Details

Introduction

When X-rays (and cosmic rays) deposit charge in an ACIS CCD, the charge is read out using one of four sets of read-out electronics. Each read out is used for a specific 256 pixel x 1024 pixel subset (node) of the CCD. Since charge is read out at only one location on a node, the charge at all other locations must be moved to the read out. Charge is moved both vertically (i.e. in the negative CHIPY or "parallel" direction) and horizontally (i.e. in the positive or negative CHIPX or "serial" direction). The total number of pixels through which charge must be moved depends on the location at which charge is deposited on the CCD. As charge is moved, some may be lost to charge traps that are distributed across the detector. The mean fractional amount of charge lost per pixel transferred is called the charge transfer inefficiency (CTI). At launch, the values of CTI were < 1 x 10-6 and < 3 x 10-6 for parallel and serial motion, respectively, on a front-illuminated CCD and = 1-3 x 10-5 and = 8-16 x 10-5 for parallel and serial motion, respectively, on a back-illuminated CCD (at 5.9 keV and -120 C). Due to the accumulated effects of cosmic radiation damage, the number of charge traps (and, hence, the CTI) on the CCDs is increasing with time. As of September 1, 2002, the values of CTI are about 1-2 x 10-4 and < 4 x 10-6 for parallel and serial motion, respectively, on a front-illuminated CCD and about 2-3 x 10-5 and 6-14 x 10-5 for parallel and serial motion, respectively, on a back-illuminated CCD (at 5.9 keV and -120 C).

As of CALDB v3.1.0 (23 June 2005), parallel CTI calibration products are available for the ACIS-I0, I1, I2, I3, S0, S2, S4, and S5 CCDs. Parallel and serial calibration for the back-illuminated chips (ACIS-S1, S3) were released in CALDB v3.3.0 (18 December 2006).

Gain Shift and Spectral Resolution

CTI affects the measured spectral distribution of astrophysical sources in two ways:

  1. Since some of the charge is trapped, the amount of charge read out is less than the amount of charge deposited. This effect causes the measured pulse-height distribution for a source to be shifted to lower pulse heights (i.e. results in an apparent gain shift).

  2. For a variety of reasons, CTI causes a degradation in the energy resolution of a CCD. The measured pulse-height distribution of a monoenergetic source (or a line feature) is broadened.

These effects are functions of the location where an X-ray interacts in a CCD because they depend on the number of traps through which charge is moved. Therefore, it is necessary to calibrate the gain and spectral response of several separate regions on each CCD.

An algorithm has been developed to estimate the amount of charge deposited on a CCD for an event from the amount of charge read out and the location of the event on the detector. This algorithm is implemented in the tool acis_process_events. Use of the CTI adjustment eliminates nearly all of the apparent gain shift and can significantly improve the energy resolution of a detector. (The energy resolution is not fully restored because charge trapping is a stochastic process and we do not know the different charge trapping histories of each event.)

Grade Migration and Detection Efficiency

Charge captured by a trap is typically released on a short time scale. A significant amount of the trapped charge is released into the pixel immediately following the pixel from which it was trapped. As a result, the distribution of charge in a 3 pixel x 3 pixel event island is "smeared out" in the read-out direction. The GRADE (and FLTGRADE) of an event is a numerical representation of the distribution of charge in the event island. If enough charge is added to a pixel to yield a pulse height that is greater than or equal to the split threshold (e.g. 13 adu), then the GRADE associated with the measured distribution of charge at the read out may be different than the GRADE associated with the distribution of charge produced at the location where the event interacted with the detector. This effect, called "grade migration," depends on the amount of charge deposited (i.e. the energy of an X-ray) and the location of the event on the CCD. Since events whose GRADEs are changed from a "good" value (0, 2, 3, 4, or 6) to a "bad" value (1, 5, or 7) are excluded from Level 2 event files, grade migration results in an apparent reduction in the detection efficiency of a CCD. The apparent reduction in the detection efficiency was calibrated for data obtained at -120 C. This information is contained in the QEU file.

When the CTI correction is applied to the BI ACIS chips, a number of low-energy events are recovered, because their grades had migrated to BAD grades as a result of loss of charge at the readout. This presents a change in the QE uniformity (QEU) with the CTI correction as compared with the QEU without CTI correction. (For the FI chips, this effect is negligible due to their much-reduced low-energy QE.) For the BI chips, however, it was necessary to create a new QEU file (version N0006) for -120 C observations.

References


Last modified: 23 February 2010