Time Assignment Error of XIS (rev 1.2/1.3)
2007 Apr 26 by the XIS Team, edited by the GOF
There is a systematic error in the XIS time assignment in rev 1.2 and 1.3 pipeline products. Because of this systematic error, XIS time is shifted compared to the HXD time by 6 sec (THXD=TXIS + 6 sec) when 1/4 window option is used and 7 sec (THXD=TXIS + 7 sec) when 1/8 window is used. Although not directly measured, the nature of the error indicates that XIS data taken in normal mode without a window option should not be affected. We ask the users to note this XIS time assignment error when conducting timing analysis with the XIS data. This time assignment error will be corrected in Version 2.0 processing.
In what follows, we explain how the time assignment error was measured and how we can correct it.
2. Calibration method of XIS time assignment
Time assignment of Suzaku HXD is calibrated using the Crab pulsar. Absolute time assignment of HXD is known to be better than a few hundred micro seconds. Because the time resolution of XIS is of an order of seconds in normal mode, we calibrated the XIS time assignment by referring the HXD PIN data.
We used data from bright X-ray binaries and took a cross-correlation between XIS and PIN light curves to determine the offset of XIS time assignment compared to PIN. If possible, we directly compared the XIS and PIN light curves.
We used the XIS data with 1/4 window or 1/8 window options. Although we tried to calibrate the normal mode data without options, we could not find appropriate data, which had large variability and good statistics in both XIS and PIN. Thus the current calibration is restricted to the data with 1/4 window or 1/8 window option.
3. Revision of data
We used the revision 1.2 pipeline products. Because there was no change in the XIS time assignment between the revisions 1.2 and 1.3 products, the XIS time offset should be same for both revisions of products. However, it was different in rev 0.7 or earlier. Thus, the results in this page are not applicable to the rev 0.x products.
4. Results: 1/8 window option
XIS data have a time resolution of 1 sec when 1/8 window option is applied. We made the XIS light curve in 1-sec bin and calculated a cross-correlation with the PIN light curve also made with 1-sec bin. We calculated the cross-correlation in both the total energy band and in the overlapping energy range of 8-12 keV. This was to check the effect of energy dependent time lag intrinsic to the source. However, we found that the results in the total band and 8-12 keV were consistent in all cases.
(1) Cyg X-1
Cyg X-1 is a proto-type black hole candidate, and is well known for its rapid and large time variations. The source was observed in 2005 Oct 5 with 1/8 window option. Cross-correlation functions between XIS and PIN are shown below for XIS2 and XIS3 and for the total band and the 8-12 keV band. Here, a positive time las means that PIN lags behind XIS. An offset of +7 sec is very clear in both 8-12 keV band and in the total band. This means that the hard lag intrinsic to the source is negligible in the current analysis.
4U1626-67 is an accretion-powered X-ray pulsar with a pulse period of 7.7 sec. It was observed in March, 2006, with 1/8 window option. Cross-correlation functions are shown below in both the 8-12 keV and the total bands separately for XIS1 and XIS2. Because the source is a pulsar of 7.7 sec, the cross-correlations have peaks separated by 7.7 sec. However, another peak is clearly seen at +7 sec, which may be originated from the random variations of the source.
GRS1915+105 is one of the black hole candidates accompanying relativistic jets, often referred to as micro-quasars. The source is known to show rapid and violent time variations. It was observed in October 2005 with 1/8 window option. In addition to the window option, a burst option (0.1 sec) was also applied to XIS1 (the BI sensor) to prevent the possible telemetry saturation. Thus we used only the data from the FI sensors. Calculated cross-correlation functions is shown below for XIS0. Although the data statistics are poor in the 8-12 keV band, a clear peak is seen at +7 sec offset in the total band.
5. Results: 1/4 window option
XIS data have a time resolution of 2 sec when 1/4 window option is applied. We made the XIS light curve in 2-sec bin and calculated a cross-correlation function with the PIN light curve also made with 2-sec bin. We calculated the cross-correlation in both the total energy band and in the overlapping energy range of 8-12 keV.
(1) IGR J17497-2821
This is a hard X-ray transient discovered by INTEGRAL. Suzaku made a TOO observation of the source in Sep 25-26, 2006. The source had a hard spectrum, extending well above 100 keV, typical for the low/hard state of a black hole candidate. We observed the source applying both 1/4 window and 1-sec burst options to avoid the photon pile-up. Because the effective exposure of 1-sec covered the latter half of the 2-sec window in this case, the burst option introduced an additional systematic shift of -0.5 sec between XIS and PIN data. We show the calculated cross-correlation functions below. Although the statistics are poor for the data in 8-12 keV band, both the 8-12 keV and the total band data are consistent with +6 sec offset. The peak in the total band data shows slight asymmetry around +6 sec. This may be interpreted due to the additional -0.5 sec offset introduced by the window option.
This is a low-mass X-ray binary of a moderate luminosity showing both dipping and bursting activities. An X-ray burst usually shows a sharp rise. Thus we directly compared the XIS and PIN light curves to quantify the time assignment error of XIS. The source was observed in January 2007 using the 1/4 window option. XIS and PIN light curves of one of the X-ray bursts are sown below. Note that XIS2 did not operate during the observation. Because of the spectral evolution of the burst, XIS and PIN light curves are somewhat different. However, if we compare the onset of the burst, an advance of XIS time by 6 sec compared to PIN is clear.
6. How to correct the XIS time assignment error
One can modify the header keyword in the event files to correct the XIS time assignment error. The origin of time is defined by the keyword "MJDREF?" in the event files as shown below:
<< An example of the header keywords >>
MJDREFI = 51544 / MJD reference day MJDREFF = 0.00074287037037037 / fractional part of the MJD reference (64.184 s)MJD is split into the integer and decimal parts. You can add 7 sec or 6 sec (8.102E-5 or 6.944E-5) to "MJDREFF" to correct the time assignment error. Note that you need to correct the keyword for all the extensions.
If you have any questions concerning Suzaku, visit the Feedback form.