Suzaku Flash
Research News
Tech Description
Accepted Targets
Help Desck
Public Data
Top > Suzaku > Observation Planning > XIS modes

Availability and restriction of the XIS modes

*** Notice ***
The spaced-row CI is always used from AO2 (April, 2007).
Thus, SCI is not a user-selectable option any more.

May 8, 2006 T. Dotani and the XIS team
revised on February 21, 2007 by H. Mori

This document summarizes the availability/restriction of various XIS modes as of 2007 February. Here, XIS modes mean both the clocking modes of CCD and the editing modes of the data, including the Spaced-row Charge Injection (CI). All the XIS modes described in the Technical Description are usable except for the timing/2x2 modes of the BI sensor (XIS1). However, some of the XIS modes are not well calibrated yet, and the data may include extra systematic errors other than those of the normal mode. Thus, those who plan to use these XIS modes should be aware of the current status described in this document.

The table below summarizes the recommendation on the usage of the XIS modes. The XIS modes which are not mentioned in this table can be used without problems.

Editing/clocking modeStatus/Recommendation
Spaced-row CIRecommended to use except for a combination with a very short burst option
1/4 window modePI should be aware of the systematic errors
1/8 window modeUsable at PI's own risk
2x2 mode (BI sensor)Unavailable
2x2 mode (FI sensor)PI should be aware of the systematic errors
Timing mode (BI sensor)Unavailable
Timing mode (FI sensor)Usage strongly discouraged
Burst optionSmall difference in the background

1. Spaced-row CI

Principle of the spaced-row CI is found in the Suzaku technical description ( We suggest to use the spaced-row CI whenever possible, as it recovers the degraded energy resolution. Spaced-row CI is applicable to the normal mode, normal mode with an window option. Burst option (even combined with the window option) is also compatible with the CI. However, we do not recommend to use CI when exposure of the burst option is very short (<0.5 sec), because a fraction of the out-of-time events becomes significant. Merits/Demerits of the Spaced-row CI are summarized below. We also note that the analysis tools and calibration data for CI are not available now, and need to wait for a few months before they become ready.

(a) Energy resolution

Energy resolution is improved from ~200-220 eV to ~140-150 eV at the iron band with the application of the spaced-row CI. From the preliminary analysis of E0102.2-7219, the improvement of the energy resolution is also verified at low energies.

(b) Dead area

Dead lines are produced every 54 row. The width of the dead line is 3 (or 4) rows. The dead lines are fixed in RAWX, RAWY coordinate and are common to all the sensors. This means that the dead area will be 5.6% (7.4%) of the image.

(c) Out-of-time events

Out-of-time events will increase. It takes about 25 ms to transfer the image area data to the frame-store area when no CI is implemented. This will increase to 156 ms in the normal mode when spaced-row CI is used. This means that the fraction of the out-of-time events increases from 0.3% to 2.0%. The fraction becomes even larger when window/burst options are employed. We should notice that a thickness of the depletion layer is changed during the charge transfer to the frame-store area. In the normal mode, the out-of-time events can be subtracted as a background. However, the appropriate background region cannot be extracted when window option is applied. Furthermore, for the window + <0.5 sec burst option, the fraction also exceeds 17%. Therefore, we recommend not to use the spaced-row CI in case you select the burst option with such a short exposure.

(d) Analysis of the CI data

It may take at least a few months for the XIS team to prepare the software and the calibration database necessary to analyze the data with the CI. Meanwhile, proper data analysis is not possible. (Of course, some quick-look analysis is possible.) Note that the proprietary time will start when both the analysis software and the calibration database become available. For the processing of the CI data, please see the announcement in the following URL:

2. Window mode

At present, 1/4 and 1/8 window options are available. These window options may be combined with the burst option. Following is the summary of the possible problems of the window options (regardless of the burst option) in data analysis.

(a) Attitude fluctuation

It is known that the Suzaku attitude can fluctuate with its orbital period. The fluctuation can be as large as 40-50 arcsec (in full amplitude). Because the window widths are only 4.5 arcmin (1/4 window) and 2.2 arcmin (1/8 window), the fluctuation is not negligible especially for the 1/8 window option. The fluctuation might produce an artificial variation in the light curve of the X-ray flux, although it is easy to distinguish due to its exact period as the orbital period. We believe that the artificial variation is within an acceptable range for the 1/4 window option. If you are not interested in the absolute flux or the long-term flux variations (time scale of 100 min), 1/8 window option may be still usable.

(b) CTE correction

Current version of the CTE correction tool (xispi, in the HEAsoft package v.6.1.2) correct the CTE based on the ACTY and RAWX coordinate. This works fine for the normal mode data. However, numbers of the frame-store and the parallel transfer are different when the window option is applied, even if the ACTY is same. Because the CTEs associated to the frame-store and the parallel transfer are different, this introduces a slight inaccuracy in the CTE correction. This means a systematic error in the energy scale becomes larger with the window option compared to the normal mode. The magnitude of the systematic error is not yet quantified, mainly because the data of the calibration source, which illuminates corners of the CCD, are not available in the window mode.

(c) Absolute flux

Because the standard event extraction region (a radius of 6mm or 4.3 arcmin) is larger than the width of the window modes, the pre-calculated arf files are not usable for the data analysis. They might reproduce the spectral shape relatively well, but will give a wrong absolute flux. To estimate the flux accurately, it is necessary to calculate the arf file corresponding to the event extraction region of the window mode. This task can be done by "xissimarfgen", the Suzaku FTOOLS in the HEAsoft package. A brief document of the "xissimarfgen" are in the following URL:
In order to understand the proper usage of this tool, users should also refer to the paper by Ishisaki et al. (PASJ, 59, SP1, S113 (2007), astro-ph/0610118).

(d) Detector background

It is expected that the background rate due to the charged particles is generally lower than that of the normal mode. Thus the background database XIS team is now developing may not be usable for the window mode. However, background rate in the window mode is not studied yet. We believe that the window modes are mostly used for bright sources, and the background subtraction is not important.

3. 2x2 mode

Because the event information is limited in the 2x2 mode, it is difficult to apply full CTI correction in this mode. This introduces some systematic errors in the data. Because these systematic errors are rather large for the BI sensor (XIS1), XIS team decided not to use the 2x2 mode for the BI sensor. The 2x2 mode is not supported for the BI sensor regardless of the clocking mode. Thus 2x2 mode is used only in the FI sensors (XIS0,2,3) for bright sources, when the telemetry saturation is concerned. In the actual XIS operations, we usually switch between 3x3 and 2x2 modes depending on the telemetry rate (high or medium). This means that both the 3x3 and 2x2 mode are obtained from the source. It is sometimes possible to correct the systematic error in 2x2 mode mentioned below referring to the 3x3 mode data.

(a) Energy scale

CTI correction tool (xispi) corrects two effects of CTI: charge trail and the total loss of charge. It is possible to correct the total loss in the same way for the 5x5/3x3/2x2 modes. However, charge trail correction is not possible for the 2x2 mode. This means that the 2x2 mode has systematically lower gain than the 5x5/3x3 modes. Although the gain reduction may be time dependent, it is approximately ~20 eV at 5.9 keV for recent data.

(b) Quantum detection efficiency

Because the charge trail affects the grade branching ratio, the 2x2 mode can have slightly different quantum detection efficiency compared to the 5x5/3x3 modes. Charge trail generally increases the grade. Thus the 2x2 mode may have systematically lower efficiency than the 5x5/3x3 modes.

4. Timing/P-sum mode

Timing mode is not usable for the BI sensor (XIS1). It works fine for the FI sensors (XIS0,1,2), but the mode has significantly different performance in terms of the energy scale, energy resolution, quantum detection efficiency and the background level. In spite of these differences, timing mode is poorly calibrated. Furthermore, time assignment of the data (in both absolute and relative accuracy) is not calibrated yet. Because little observations were carried out so far in the timing mode and the existing (flight) calibration data are limited, we are afraid that the calibration of the timing mode will not reach the accuracy of the normal mode in future.

(a) Unavailability in the BI sensor

There is a number of bad columns in XIS1, which produces variable hot pixels. These hot pixels appear and disappear frequently. Because of this variable nature, these hot pixels are difficult to remove completely on board. In the case of the normal mode, such variable hot pixels are mostly harmless. On the other hand, a hot pixel can produce at most 1024 false events in the timing mode. Thus only a few remained hot pixels can easily fill the telemetry. This makes it difficult to obtain the X-ray events in timing mode from the BI sensor (XIS1).

(b) Performance differences in timing mode

Timing mode works OK for the FI sensors. However, performance of the timing mode may be significantly different from that of the normal mode in many ways. Various CTI-related corrections are not possible for the timing mode. This makes the energy scale of the timing mode different from the normal mode. Different clocking scheme may also contribute to the difference of the energy scale. The latter also affects the energy resolution. Because we have only 1-dimensional information, the event classification scheme is simplified in the timing mode. This makes the background rejection efficiency worse. We have one or two orders of magnitude larger particle background in the timing mode.

(c) Calibration of the time assignment

Timing mode is designed to have a time resolution of 7.8 msec. However, time assignment to the events are not calibrated yet even in a sense of relative accuracy. Furthermore, a software to assign the photon arrival time to the events with the 7.8 msec resolution is still under development. Thus, the full timing capability of the timing mode is not available yet.

(d) Recommendation

Timing mode is in principle useful for the observations of a bright point source, when large telemetry capacity and/or a high time resolution are needed. However, considering the current status of the timing mode, XIS team strongly discourage the use of this mode. Even if an observation were made with the timing mode, the XIS team may not be able to supply calibration information necessary for the data analysis. Moderately large telemetry capacity and better time resolution are obtained with the window option. Thus the window option should be considered as the primary mode for bright point sources.

5. Burst option

A burst option, even if combined with the window option, does not introduce extra uncertainties in the energy scale or quantum detection efficiency. However, background level may be different. If the background is measured in unit of the net exposure time, background level may become higher when the burst option is combined. It might be lower if the background level is measured in unit of the number of frames. It is also noted that relative contribution of the out-of-time event becomes larger. However, because the burst option is used only for a bright source, these differences in the background may not be important.

ISAS/JAXA Department of High Energy Astrophysics

Last Modified: Tuesday, 10-Apr-2007 17:01:14 JST