[Date Prev][Date Next][Thread Prev][Thread Next][Search] [Main Index] [Thread Index] [HEASARC Archives]

Updated ASCA "Calibration Uncertainties"

Dear ASCA PIs and Archival Researchers:

An updated version of the "Calibration Uncertainties" is now available.
You can find it on the ASCA GOF WWW home page, the URL of which is:


The document contains an account of the current problems and limitations
of ASCA data and calibrations. Everyone who analyzes ASCA data should
read it.

For those without access to WWW, a text-only version is appended to this


Please send comments and questions to ascahelp@athena.gsfc.nasa.gov



Analysis of extended sources.

Guest observers interested in doing spatially-resolved spectroscopy are 
referred to the report by Takahashi et al. available at the ftp site 
ftp.astro.isas.ac.jp in the file asca/report/xrt_US.ps.


Although GIS data can have time stamps precise to 0.061 msec, the actual
accuracy is considerably less. The time assignment at KSC is probably
accurate only to 0.2-0.3 msec. Other limiting factors are inaccuracies 
in: (i) the on-board clock drift correction, (ii) the orbit-to-ground 
data transmission delay, and (iii) the delay within the ground station. 
As a result, data between two consecutive KSC contacts may keep relative 
accuracy of time down to about 0.1 msec, while two data sets, reproduced 
at different KSC contacts, may have relative shift of time up to 0.5 
msec. However, absolute time seems to have a systematic offset of at 
least a few tens msec. These limits have been derived from observations 
of the Crab and PSR 0540-69 pulsars.

GIS gain: intensity dependence.

It is found that the GIS gain is dependent on the source flux when
the flux is very large (> 0.1 Crab, or about 100 counts per second). For 
bright sources, the gain becomes larger up to 1.5 per cent according to 
the source flux.

GIS gain: spatial dependence.

The GIS team now believes that the determination of how the gain varies 
spatially across the GIS field of view is uncertain by about 1 per cent. 
According to the GIS team, this uncertainty may actually represent a 
fundamental limit based on the properties of the instrument. That is, 
further calibration may not improve the accuracy of the spatial gain 
calibration. Users should be aware of this uncertainty when analyzing 
their data. Methods for dealing with it are being investigated. In 
addition to the spatial uncertainty, the GIS team has found that the 
current calibrations are not sufficient to track the gain variations 
which arise as the satellite experiences variations in solar 
illumination. The GIS team estimates that the amplitude of these gain 
fluctuations is about 0.5 per cent. The spatial gain calibration of GIS3 
is less accurate than that of GIS2.

SIS gain calibration.

The gain calibration of the two default chips (SIS0-chip1 and SIS1-
chip3) agree to within 0.5 per cent. The other chips are not as well 
calibrated, and may differ by up to 2 per cent. Due to radiation damage, 
the gain is changing with time. This drift can be compensated for by 
using PI rather than PHA channels. The PI column in event files can be 
populated by using the FTOOL SISPI for 1-CCD mode data. However, in 4-
CCD mode this does not work well.

SIS low-energy response.

>From various reports it appears that SIS spectral fits generally 
overestimate the column density by up to 2E20, indicating a probable 
overestimate of the effective area at low energies (<1 keV). In 
addition, gain and offset in energy scales and the variation of the 
Galactic soft X-ray background over the sky may contribute to 
uncertainties. (For extended sources, where blank field data must be 
used for background subtraction, the latter is a serious and unavoidable 
problem.) Therefore, column measurements with the SIS should not be 
believed at about the 2E20 level.

SIS energy scale offset.

Due to light leaks, the energy scale of both SIS has a variable offset 
(Dark Frame Error, or DFE) of up to 30 eV. It is noticeable when: (i) 
the Sun angle is less than 90 degrees during satellite day; and (ii) the 
elevation angle above the Earth's bright (Sun-lit) limb is less than 
30 degrees. For Faint mode data, this can mostly be corrected by using 
the FTOOLS FAINTDFE and FAINT, except when the DFE is changing rapidly 
(such data should be excluded from analysis). However, FAINTDFE is not 
perfect, particularly for recent 4-CCD mode data: a 1-2 ADU (~5 eV) 
offset may remain. This can cause a spurious features at various edges 
where instrument throughput changes suddenly, such as 0.54 keV oxygen 
Users can also investigate the effects of the energy scale offset on 
their data by using the "gain" command in XSPEC. Since this command 
works by shifting the energy boundaries of the response matrix, rather 
than by changing the matrix elements themselves, it should be regarded 
as an approximation. However, it does provide a simple first step toward 
addressing the problem. The "gain" command takes as its arguments:
 o The datafile number.
 o The factor by which the slope of the gain is multiplied.
 o The offset, in keV, by which the gain is shifted.
Thus the command:

   XSPEC> gain 1 1 0.020
will multiply the slope of gain of the response of the first datafile by 
a factor of 1 - i.e., leave it unchanged - but shift the offset by +20 
eV. In practice, users should first fit their data with the original 
gain, and then iterate toward the correct gain by adjusting the gain and 
refitting the model until parameter values, especially of sharp 
features, agree with their expected values and with their GIS values. 

GIS and SIS response: gold edge residuals.

Both GIS and SIS have an artificial hump feature at around 2 keV, near 
the gold edge energy (the reflecting surfaces of the XRT are coated with 
gold). This feature has been incorporated into the ARF generator for the 
SIS and GIS but may not have been completely modelled away. Please be 
cautious when interpreting possible features around 2 keV.

GIS response: residuals in Crab spectrum.

The Crab nebula was observed for 50 ksec in September 1994, and 375 
million photons(!) collected. The energy spectrum shows a clear 
deviation from a power-law, which originates not in the data, but the 
model (response). A wavy structure appears at 5 to 9 keV with the 
amplitude +/- 4 per cent. This is the current limit of the GIS response.

GIS RMF: RTI cleaning always assumed.

The current GIS RMF is made for data which have passed through the 
PHA-dependent RTI (Rise Time Invariance) window (data filtered with the 
'gisclean' command in XSELECT). When RISE_TIME is not available (e.g., 
fast timing observations), gisclean is not usable, and events outside 
the PHA-dependent RTI window cannot be rejected. Applying the current 
GIS RMF to the data without the RTI filter is not precisely correct, and 
will lead to a flux apparently larger by 1 to 5 per cent in the 1-10 keV 
band (larger in higher energies). A modification to ASCAARF will be made 
available soon to deal with this limitation.

GIS background uncertainty.

As a first approximation, GIS internal (non-X-ray) background is 
proportional to the magnetic cut-off rigidity (COR). Blank sky data and 
night-earth data sorted by COR have been released for GOs to 
estimate/subtract GIS background. However, using only COR does not 
perfectly describe time variation of GIS background. That is, there are 
other minor parameters which control the GIS background. For a given 
COR, the GIS non-X-ray background can be different by 20-30 per cent. 
The GIS team is investigating the nature of the GIS background, and 
trying to model the background more accurately.
XRT PSF: azimuthal dependence

The XRT point response function has an azimuthal energy dependence. Not 
much quantitative information is available at the moment: there appears 
a factor of two difference at 10 keV between the lobes of the "Maltese 

GIS efficiency at low energy.

Both GIS have greater efficiency in the range 0.6-0.9 keV than predicted 
by the response matrix. The difference is ~20 per cent at maximum and 
will be corrected in the next version of the response.

GIS-SIS cross-calibration.

The GIS tends to give larger (~20 per cent) absolute flux than the SIS. 
The SIS flux seems to be reliable, so the problem is probably in the GIS 
response (the plasma shield is the likely origin). GIS team is now 
revising the GIS response.

SIS absolute efficiency.

The ground calibration of the SIS established the absolute uncertainty 
in its efficiency to be 13 per cent.


Underestimated XRT effective area above 6 keV.

In the early phase of the mission the XRT model had less effective area 
above 6 keV than is observed. The difference is ~40 per cent by 10 keV. 
This discrepancy led to an apparent hard tail but has now been 
incorporated into ASCAARF.

GIS gain map now covers full FOV.

The latest GIS gain maps, which are incorporated into ASCALIN, are now 
accurate for the full GIS field of view (50 arcmin diameter). Data from 
the edge of the FOV can be reliably analyzed, although particle 
background remains high at the edge.

Gold edge fudged in XRT response.

The gold edge which appears around 2 keV has been incorporated into 
ASCAARF for both the GIS and SIS. However, users should still be 
cautious when interpreting spectral features around 2 keV.

SIS gain drift can be corrected for 1-CCD mode.

The FTOOL SISPI corrects the secular gain drift of the SIS by converting 
PHA to PI. The correction works for 1-CCD data but is of limited 
accuracy for 2-CCD and 4-CCD data.