The Advanced CCD Imaging Spectrometer (ACIS) offers the capability to simultaneously acquire high-resolution images and moderate resolution spectra. The instrument can also be used in conjunction with the High Energy Transmission Grating (HETG) or Low Energy Transmission Grating (LETG) to obtain higher resolution spectra (see Chapters 8 and 9). ACIS contains 10 planar, 1024 x 1024 pixel CCDs (Figure 6.1); four arranged in a 2x2 array (ACIS-I) used for imaging, and six arranged in a 1x6 array (ACIS-S) used either for imaging or as a grating readout. Currently any combination of up to 6 CCDs may be operated simultaneously. However, see the discussion in Section 6.20.1 where we encourage users, where possible, to specify as many as 5 chips as optional. Prior popular combinations include an extended ACIS-I imaging mode, using chips I0-I3 plus S2 and S3; and an ACIS-S imaging mode, using chips S1-S4 plus I2 and I3. Operating six chips enhances the chance of serendipitous science, but at the price of increased total background counting rate and therefore a somewhat enhanced probability of saturating telemetry. Another penalty for 6 operating CCDs is an increased power load (Section 6.20.1). Two CCDs are back-illuminated (BI) and eight are front-illuminated (FI). The response of the BI devices extends to energies below that accessible to the FI chips. The chip-average energy resolution of the BI devices is better than that of the FI devices.
The original Instrument Principal Investigator for ACIS is Prof. Gordon Garmire (Pennsylvania State University). ACIS was developed by a collaboration between Penn State, the MIT Kavli Institute for Astrophysics and Space Research and the Jet Propulsion Laboratory, and was built by Lockheed Martin and MIT. The MIT effort was led by Dr. George Ricker. The CCDs were developed by MIT's Lincoln Laboratory.
ACIS is a complex instrument having many different characteristics and operating modes. Radiation damage suffered by the FI chips has had a negative impact on their energy resolution - the BI devices were not impacted - thus impacting the basic considerations as to how to make best use of the instrument. We discuss the trade-offs in this chapter. Software methods for improving the energy resolution of the FI CCDs are discussed in Section 6.7.1. The low energy response of ACIS has also been affected by the build-up of a contaminant on the optical blocking filters and this is discussed in Section 6.5.1.
Many of the characteristics of the ACIS instrument are summarized in Table 6.1.
|Focal plane arrays:|
|I-array||4 CCDs placed to lie tangent to the focal surface|
|S-array||6 CCDs in a linear array tangent to the grating Rowland circle|
|CCD format||1024 by 1024 pixels|
|Pixel size||23.985 microns (0.4920±0.0001 arcsec)|
|Array size||16.9 by 16.9 arcmin ACIS-I|
|8.3 by 50.6 arcmin ACIS-S|
|On-axis effective Area||110 cm2 @ 0.5 keV (FI)|
|(integrated over the PSF||600 cm2 @ 1.5 keV (FI)|
|to > 99% encircled energy)||40 cm2 @ 8.0 keV (FI)|
|Quantum efficiency||> 80% between 3.0 and 6.5 keV|
|(frontside illumination)||> 30% between 0.7 and 11.0 keV|
|Quantum efficiency||> 80% between 0.8 and 5.5 keV|
|(backside illumination)||> 30% between 0.4 and 10.0 keV|
|(QEs do not include contaminant|
|Charge transfer inefficiency(parallel)||FI: ∼ 2×10−4; BI: ∼ 2×10−5|
|Charge transfer inefficiency(serial)||S3(BI): ∼ 7×10−5; S1(BI): ∼ 1.5×10−4;|
|FI: < 2 ×10−5|
|System noise||< ∼ 2 electrons (rms) per pixel|
|Max readout-rate per channel||∼ 100 kpix/sec|
|Number of parallel signal channels||4 nodes per CCD|
|Pulse-height encoding||12 bits/pixel|
|Event threshold||FI: 38 ADU ( ∼ 150-350 eV)|
|BI: 20 ADU ( ∼ 150-220 eV)|
|Split threshold||13 ADU|
|Max internal data-rate||6.4 Mbs (100 kbs ×4 ×16)|
|Output data-rate||24 kb per sec|
|Minimum row readout time||2.8 ms|
|Nominal frame time||3.2 sec (full frame)|
|Allowable frame times||0.2 to 10.0 s|
|Frame transfer time||40 μsec (per row)|
|Point-source sensitivity||4 ×10−15ergs cm−2 s−1 in 104 s|
|Detector operating temperature||−90 to −120°C|
A CCD is a solid-state electronic device composed primarily of silicon. A "gate" structure on one surface defines the pixel boundaries by alternating voltages on three electrodes spanning a pixel. The silicon in the active (depletion) region (the region below the gates wherein most of the absorption takes place) has an applied electric field so that charge moves quickly to the gate surface. The gates allow confined charge to be passed down a "bucket brigade" (the buried channel) of pixels in parallel to a serial readout at one edge by appropriately varying ("clocking") the voltages in the gates.
The ACIS front-illuminated CCDs have the gate structures facing the incident X-ray beam. Two of the chips on the ACIS-S array (S1 and S3) have had treatments applied to the back sides of the chips, removing insensitive, undepleted bulk silicon material and leaving the photo-sensitive depletion region exposed. These are the BI chips and are deployed with the back side facing the HRMA.
Photoelectric absorption of an X-ray in silicon results in the liberation of a proportional number of electrons: an average of one electron-hole pair for each 3.7 eV of energy absorbed. Immediately after the photoelectric interaction, the charge is confined by electric fields to a small volume near the interaction site. Charge in a FI device can also be liberated below the depletion region, in an inactive substrate, from where it diffuses into the depletion region. This charge may easily appear in two or more pixels.
Good spectral resolution depends upon an accurate determination of the total charge deposited by a single photon. This in turn depends upon the fraction of charge collected, the fraction of charge lost in transfer from pixel to pixel during read-out, and the ability of the readout amplifiers to measure the charge. Spectral resolution also depends on read noise and the off-chip analog processing electronics. The ACIS CCDs have readout noise less than 2 electrons RMS. Total system noise for the 40 ACIS signal chains (4 nodes/CCD) ranges from 2 to 3 electrons (rms) and is dominated by the off-chip analog processing electronics.
The CCDs have an "active" or imaging section (see Figure 6.1) which is exposed to the incident radiation and a shielded "frame store" region. A typical mode of the ACIS CCD operation is: (1) the active region is exposed for a fixed amount of time (full frame ∼ 3.2 s); (2) at the end of the exposure, the charge in the active region is quickly ( ∼ 41 ms) transferred in parallel into the frame store; (3) the next exposure begins; (4) simultaneously, the data in the frame store region is transferred serially to a local processor which, after removing bias (see Section 6.13), identifies the position and amplitude of any "events" according to a number of criteria depending on the precise operating mode. These criteria always require a local maximum in the charge distribution above the event threshold (see Table 6.1). The position and the amount of charge collected, together with similar data for a limited region containing and surrounding the pixel are classified ("graded") and then passed into the telemetry stream.
Since the CCDs are sensitive to optical as well as X-ray photons, optical blocking filters (OBFs) are placed just over the CCDs between the chips and the HRMA. The filters are composed of polyimide (a polycarbonate plastic) sandwiched between two thin layers of aluminum. The nominal thicknesses of the these components for the two arrays are given in Table 6.2. Details of the calibration of these filters may be found in the ACIS calibration report at http://www.astro.psu.edu/xray/docs/cal_report/node188.html. These calibrations do not include the more recent effects of molecular contamination. This is discussed in Section 6.5.1.
|ACIS-I||Al/Polyimide/Al||1200Å 2000Å 400Å|
|ACIS-S||Al/Polyimide/Al||1000Å 2000Å 300Å|
The threshold for optical contamination (a 1 ADU (3.4 eV) shift in the bias level) is based on on-orbit calibrations of a number of stars with different optical spectra. The threshold for detectable visible light contamination varies according to source color and is lowest for red stars observed on ACIS-S The detection threshold for an M star on the ACIS-S array is V ∼ 8.1 for the nominal 3.2 second exposure or V ∼ 6.3 using a 0.4s frame time and a 1/8 chip subarray. The thresholds are about 5 visual magnitudes brighter for the ACIS-I array. While the impact of the OBF molecular contamination has not been studied in detail, a preliminary estimate (using typical optical constants for organic materials) suggests little change in the broadband optical transmission. See Section 6.19 for a discussion of the observation of optically bright targets such as solar system objects.
Measurements of ACIS include laboratory calibrations, a system-level ground calibration of the HRMA and ACIS at the X-Ray Calibration Facility (XRCF) at MSFC, and on-orbit calibration using celestial and on-board radioactive X-ray sources. The ACIS external calibration source (ECS) consists of an Fe-55 source and a target made of aluminum and titanium. The source emits five strong lines (Al K-α at 1.49 keV, Ti K-α and β, at 4.51 and 4.93 keV, and Mn K-α and β at 5.90 and 6.49 keV. A number of weaker lines are also present.
The on-orbit calibration of ACIS is a continuing activity. All calibration data are, or will be, described in detail, at http://cxc.harvard.edu/cal/. The user is urged to consult the WWW site and its links for the latest information.
The quantum efficiencies near the readout for the ACIS CCDs for the standard grade set, including optical blocking filters and molecular contamination, are shown in Figure 6.3. Note that the quantum efficiency for the FI chips varies somewhat with row number (not shown), and decreases by 5-15% farthest from the readout at energies above about 4 keV. This is due to the migration of good grades to bad grades produced by charge transfer inefficiency (CTI), which varies with row number. The quantum efficiency (QE) variation with position for the BI chips is much smaller.
Cosmic rays tend to cause large blooms on FI chips, and much smaller ones on BI chips. This results in a 2-4% decrement of the QE for FI chips and ∼ 0.5% for BI chips.
The combined HRMA/ACIS on-axis effective areas are shown in Figure 6.4 (log energy scale) and 6.5 (linear energy scale). The effective areas are for an on-axis point source and a 20-arcsecond-diameter detection cell. The ACIS effective areas include a correction that accounts for the build-up of molecular contamination on the ACIS filters (see the discussion in Section 6.5.1). Figures 6.4 and 6.5 show the predicted ACIS effective areas for the middle of Cycle 14 based on the current time-dependent ACIS contamination models.
There is a count rate dependent effect, currently under investigation, that may impact a subset of observers. The problem is that the effective area calibration at the Si K edge (1.842 keV; 6.74 Å) appears to be somewhat incorrect, but only at very high count rates. At high rates, the edge will appear somewhat higher than it should be. Here high count rate is: 1.9×1010 erg cm−2 s−1 (ACIS) and 1.1×10−9 erg cm−2 s−1 (HETG/ACIS). Of course at these counting rates pileup will likely dominate ACIS observations, but not necessarily HETG/ACIS observations. In very extreme cases, the effective discrepancy between the fluxes determined by the nominal effective area and observation may be as large as 10%. The effect manifests itself in strong residuals around the edge region for sources with moderate absorption columns ( < 3×1021 cm−2), an underestimate of the actual source column at higher absorption columns, and a change in effective area at higher energies.
Figure 6.6 shows the vignetting (defined as the ratio of off-axis to on-axis effective area) as a function of energy at several off-axis angles. These data are from a calibration observation of G21.5, a bright supernova remnant. The detector was appropriately offset for each off-axis angle so that the data were obtained at the same focal position, minimizing the effects of any spatially-dependent variations in the CCD response.
Astronomical observations and data acquired from the on-board ACIS external calibration source (ECS) show that there has been a slow and continuous degradation in the ACIS effective area at low energies. Our best explanation is that this is due to the build-up of out-gassed material condensing on the cold ACIS filters. The HRC operates at a warmer temperature and shows no sign of contamination build-up. The degradation in effective area is most severe at low energies. The build-up of contamination has been monitored by ECS observations before and after each perigee passage and by periodic observations with the gratings. The grating observations have shown that the contaminant is dominated by carbon, with smaller amounts of oxygen and fluorine. The ECS is used to monitor the degradation in effective area at energies above the C-K edge. The depth of the contaminant increases from the center of the detectors toward the edges where the filters are cooler.
Over the past few years we have performed annual raster scans of the rich cluster Abell 1795 to monitor the build-up of contamination on the ACIS filters. Analysis of the ECS and Abell 1795 produce consistent results for the optical depth of the contaminant at 700 eV. Since the ECS continues to fade due to the 2.7 year half-life of 55Fe, we will be relying more heavily on these Abell 1795 observations to normalize the optical depth of the contaminant in the future.
An initial model of the contamination on the ACIS filter was developed for use in CIAO and other data reduction packages and released in CALDB v3.0.0 (Dec. 2004). Beginning in 2006, there were some systematic differences in the actual build-up rate of the contaminant on the ACIS filters compared with the predictions of the 2004 model. These differences were initially noticed from discrepancies in the optical depth of the contaminant as measured from ECS data and gratings observations of AGN. As a result, updated versions of the ACIS-I and ACIS-S contamination models were released in CALDB 4.2 (Dec. 2009). A further update to the ACIS-I contamination model was released in CALDB 4.4.1 (Dec. 2010). Analysis of calibration (both imaging and gratings) and ECS observations taken over the last few years show that the condensation rate of the molecular contamination onto the ACIS filters has increased. The inflection point in the growth curve of the ACIS contamination is consistent with the date when one of the ACIS detector housing heaters was turned off in April 2008 to help stabilize the ACIS focal plane temperature and detector gain. Also, since 2008, the difference between the optical depth at the edges and centers of the detectors has been increasing at a greater rate. As a result of these changes in the behavior of the ACIS contamination, an updated ACIS contamination model was released in CALDB 4.4.10 on May 30, 2012. This new model accounts for the greater condensation rate and spatial dependence of the molecular contaminant onto the ACIS filters.
The spatial resolution for on-axis imaging with HRMA/ACIS is limited by the physical size of the CCD pixels (24.0 μm square ∼ 0.492 arcsec) and not the HRMA. This limitation applies regardless of whether the aimpoint is selected to be the nominal point on I3 or S3 (Figure 6.1). Approximately 90% of the encircled energy lies within 4 pixels (2 arcsec) of the center pixel at 1.49 keV and within 5 pixels (2.5 arcsec) at 6.4 keV. Figure 6.7 shows an in-flight calibration. There is no evidence for any differences in data taken with either S3 or I3 at the nominal focus. The ACIS encircled energy as a function of off-axis angle is discussed in Chapter 4 (see Section 4.2.2 and Figure 4.13).
The ACIS FI CCDs originally approached the theoretical limit for the energy resolution at almost all energies, while the BI CCDs exhibited poorer resolution. The pre-launch energy resolution as a function of energy is shown in Figure 6.8. Subsequent to launch and orbital activation, the energy resolution of the FI CCDs has become a function of the row number, being near pre-launch values close to the frame store region and substantially degraded in the farthest row. An illustration of the dependence on row is shown in Figure 6.9.
|ccd||chipx||chipy||1.49 keV||5.9 keV|
|I3, no offset||941||988||155.757||284.542|
|I3, w offset||972||964||131.488||256.309|
|S3, no offset||224||490||95.3451||147.033|
|S3, w offset||206||521||96.7391||151.302|
In late 2006, a 10′′ shift in the alignment of the aspect camera occurred when its primary focal plane temperature was cooled from −15C to −19C. In 2011 July, Chandra entered a safemode, and since that time the aimpoint has drifted back to near the 2007 position. However, its short term instability has increased in the past two years or so, especially after the safemode. In combination with the long-term alignment drift, the ACIS aimpoints using the previous default Y and Z offsets have shifted. In order to place the aimpoint so that targets are at the optimal positions with respect to CHIP geometry and telescope focus, we have updated the default offsets for ACIS-S to ∆Y = +9′′ and ∆Z = −15′′, to avoid dithering across the node boundary on the S3 chip. For ACIS-I the default offsets of ∆Y = −12′′ and ∆Z = −15′′ are unchanged. For grating observations an offset toward the readout on the ACIS-S array is often recommended (i.e. in the negative Z direction - see Table 6.5). Note that standard ACIS subarrays (Section 6.12.1) will not center zeroth order if the recommended aimpoints are selected for a grating observation. In this case, a custom subarray is necessary (e.g. see Section 8.5)
Further information can be found in the memos linked from this page: http://cxc.harvard.edu/cal/Hrma/OpticalAxisAndAimpoint.html
Lastly, it should be kept in mind that the observatory is typically dithered about the aimpoint with an 8′′ half-amplitude (see Section 6.11, and 6.5 for recommended offsets).
|ACIS-I :||(941, 988) in I3||no offsets|
|ACIS-I :||(972, 964) in I3||with offsets of ∆Y=−12′′, ∆Z=−15′′|
|ACIS-S :||(224, 490) in S3||no offsets|
|ACIS-S :||(206, 521) in S3||with default offsets ∆Y=+9′′, ∆Z=−15′′|
|Observation Mode||SIM-Z Offset||Source Position (w/Offsets|
|ACIS-S w/ HETG TE mode:||-3mm = -1.02417′||(206, 396)|
|ACIS-S w/ HETG CC mode:||-4mm = -1.36556′||(206, 354)|
|ACIS-S w/ LETG TE mode:||-8mm = -2.73112′||(206, 187)|
|ACIS-S w/ LETG CC mode:||-8mm = -2.73112′||(206, 187)|
A timed exposure refers to the mode of operation wherein a CCD collects data (integrates) for a preselected amount of time - the Frame Time. Once this time interval has passed, the charge from the 1024 x 1024 active region is quickly ( ∼ 41 ms) transferred to the framestore region and subsequently read out through (nominally) 1024 serial registers.
Frame Times - Full Frames
Frame times are selectable within a range of values spanning the time interval from 0.2 to 10.0 seconds. If the data from the entire CCD are utilized (full frame) then the nominal (and optimal!) frame time is 3.2 s, Topt. Selecting a frame time shorter than the nominal value (e.g. to decrease the probability of pileup - Section 6.15) has the consequence that there will be a time during which no data are taken, "deadtime", as 3.2 s are required for the frame store readout process regardless of the frame time. The fraction of time during which data are taken is simply the ratio of the selected frame time to the sum of the nominal frame time and the 41 msec transfer time - e.g. for a new frame time of t ( < 3.2) secs, the fraction of time during which data are taken is t/(Topt+0.041). We note, strictly speaking, the full-frame time depends on how many CCD s are on - see the equation below - but the differences are very small. Finally, we note that selecting a frame time longer than the most efficient value increases the probability of pileup occurring and is not recommended.
Frame Times & Subarrays
It is also possible for one to select a subarray - a restricted
region of the CCD in which data will be taken. A subarray is fully
determined by specifying the number of rows separating the subarray
from the framestore region (q) and the number of rows in the subarray
(n). Examples of subarrays are shown in Figure 6.13. The
nominal frame time for a subarray depends on (q), (n), and the total
number of CCDs that are activated (m) - see Table 6.6.
The nominal frame time is given by:
As with full frames, selecting a frame time less than the most efficient value results in loss of observing efficiency. Frame times are rounded up to the nearest 0.1 sec, and can range from 0.2 to 10.0 sec
When operating with only one chip, subarrays as small as 100 rows are allowed (this permits 0.3 sec frame times which pay no penalty in dead time). For multichip observations, the smallest allowed number of rows is 128.
|Subarray||ACIS-I (no. of chips)||ACIS-S (no. of chips)|
It takes 40 μsec to transfer the charge from one row to another during the process of moving the charge from the active region to the framestore region. This has the interesting consequence that each CCD pixel is exposed, not only to the region of the sky at which the observatory was pointing during the long (frame time) integration, but also, for 40 μsec each, to every other region in the sky along the column in which the pixel in question resides. Figure 6.14 is an example where there are bright features present, so intense, that the tiny contribution of the flux due to trailing is stronger than the direct exposure - hence the trailed image is clearly visible. Trailed images are also referred to as "read out artifact" and "out-of-time images". The user needs to be aware of this phenomenon as it has implications for the data analysis - e.g. estimates of the background. In some cases, the trailed image can be used to obtain an unpiled spectrum and can also be used to perform 40 microsecond timing analysis (of extremely bright sources).
In some instances, it is desirable to have both long and short frame times. If the exposure time is made very short, pile-up may be reduced, but the efficiency of the observation is greatly reduced by the need to wait for the full 3.2 seconds (if six chips are clocked) for the framestore array processing.
With alternating exposure times, all CCDs are clocked in unison, but have two exposure times. One (typically short) primary exposure of length 0.2 < tp < 10.0 sec is followed by k standard exposures ts (for example, 3.2 seconds if six chips are clocked). Permissible values of k range from zero (the standard single exposure mode) to 15. The short exposures are used to reduce photon pileup, and the long exposures are useful for studying the fainter objects in the field of view. For example, a typical choice of long and short exposure times might be 3.2 and 0.3 seconds. If k = 3, ACIS would perform one 0.3 second exposure, followed by three exposures of 3.2 seconds, repeating until the total observing time expires.
If the duty cycle of long exposures is 1:k (short:long), the observing efficiency η is then
The continuous clocking mode is provided to allow 3 msec timing at the expense of one dimension of spatial resolution. In this mode, one obtains 1 pixel x 1024 pixel images, each with an integration time of 2.85 msec. Details as to the spatial distribution in the columns are lost - other than that the event originated in the sky along the line determined by the length of the column.
In the continuous clocking mode, data is continuously clocked through the CCD and framestore. The instrument software accumulates data into a buffer until a virtual detector of size 1024 columns by 512 rows is filled. The event finding algorithm is applied to the data in this virtual detector and 3 x 3 event islands are located and recorded to telemetry in the usual manner (Section 6.14.1). This procedure has the advantage that the event islands are functionally equivalent to data accumulated in TE mode, hence differences in the calibration are minimal. The row coordinate (called CHIPY in the FITS file) maps into time in that a new row is read from the framestore to the buffer every 2.85 msec. This does have some minor impacts on the data. For example, since the event-finding algorithm is looking for a local maximum, it cannot find events on the edges of the virtual detector. Hence CHIPX cannot be 1 or 1024 (as in TE mode). Moreover, CHIPY cannot be 1 or 512. In other words, events cannot occur at certain times separated by 512*2.85 msec or 1.4592 sec. Likewise, it is impossible for two events to occur in the same column in adjacent time bins.
The TIMEs in continuous-clocking mode Level 1 event data files are the read-out times, not the times of arrival. The differences between the read-out times and the times of arrival are in the range 2.9 to 5.8 s. The differences depend on the nominal location of the source on the CCD and the dither of the telescope. Code to compute the times of arrival at the spacecraft from the read-out times has been developed and was released as part of the tool acis_process_events in CIAO 3.0. Data processed with this version of acis_process_events will have Level 2 files with corrected times.
In general, the CCD bias, the amplitude of the charge in each pixel in the absence of external radiation, is determined at various times - every change of instrumental parameters or setup when ACIS is in place at the focus of the telescope. These bias maps have proven to be remarkably stable and are automatically applied in routine data processing.
The bias maps for continuous-clocking mode observations can be corrupted by cosmic rays. If a cosmic ray deposits a lot of charge in most of the pixels in one or more adjacent columns, the bias values assigned to these columns will be too large. As a result, some low-energy events that would have been telemetered will not be telemetered because they do not satisfy the minimum pulse height criterion and the spectrum of a source in the affected columns will be skewed to lower energies. The BI CCDs are relatively insensitive to the problem. A bias algorithm was implemented in Cycle 6 to mitigate the problem.
Occasionally a cosmic ray produces an artifact in a bias map. The pipelines search for these artifacts, and, when found, replace the bias map with another from the same epoch. Work is in progress to use long-term average bias maps, either when there are artifacts in the observation-specific bias map, or for all observations.
During the first step in the algorithm for detecting X-ray events, the on-board processing examines every pixel in the full CCD image (even in the continuous clocking mode (Section 6.12.3)) and selects as events regions with bias-subtracted pixel values that both exceed the event threshold and are greater than all of the touching or neighboring pixels (i.e., a local maximum). The surrounding 3×3 neighboring pixels are then compared to the bias-subtracted split-event threshold; those that are above the threshold establish the pixel pattern. On the basis of this pattern, the event is assigned a grade. Depending on the grade, the data are then included in the telemetry. On-board suppression of certain grades is used to limit the telemetry bandwidth devoted to background events (see Section 6.16.1).
The grade of an event is thus a code that identifies which pixels, within the 3×3 pixel island centered on the local charge maximum, are above certain amplitude thresholds. The thresholds are listed in Table 6.1. Note that the local maximum threshold differs for the FI and the BI CCDs. A Rosetta Stone to help one understand the ACIS grade assignments is shown in Figure 6.15, and the relationship to the ASCA grading scheme is given in Table 6.7.
|ACIS Grades||ASCA Grade||Description|
|0||0||Single pixel events|
|64 65 68 69||2||Vertical Split Up|
|2 34 130 162||2||Vertical Split Down|
|16 17 48 49||4||Horizontal Split Right|
|8 12 136 140||3||Horizontal Split Left|
|72 76 104 108||6||"L" & Quad, upper left|
|10 11 138 139||6||"L" & Quad, down left|
|18 22 50 54||6||"L" & Quad, down right|
|80 81 208 209||6||"L" & Quad, up right|
|1 4 5 32 128||1||Diagonal Split|
|33 36 37 129||1|
|132 133 160 161||1|
|3 6 9 20 40||5||"L"-shaped split with corners|
|96 144 192 13 21||5|
|35 38 44 52 53||5|
|97 100 101 131||5|
|134 137 141 145||5|
|163 166 168 172||5|
|176 177 193 196||5|
|24||7||3-pixel horizontal split|
|66||7||3-pixel vertical split|
|All other grades||7|
|Mode||Format||Bits/event||Events/sec*||Number of Events|
|in full buffer|
Faint format provides the event position in detector coordinates, an arrival time, an event amplitude, and the amplitude of the signal in each pixel in the 3 x 3 island that determines the event grade. The bias map is telemetered separately. Note that certain grades may be not be included in the data stream (Section 6.16.1).
Graded format provides event position in detector coordinates, an event amplitude, the arrival time, and the event grade. Note that certain grades may be not be included in the data stream (Section 6.16.1).
Very Faint format provides the event position in detector coordinates, the event amplitude, an arrival time, and the pixel values in a 5 x 5 island. As noted in Table 6.8, this format is only available with the Timed Exposure mode. Events are still graded by the contents of the central 3 x 3 island. Note that certain grades may be not be included in the data stream (Section 6.16.1). This format offers the advantage of reduced background after ground processing (see Section 6.16.2) but only for sources with low counting rates that avoid both telemetry saturation and pulse pileup.
Studies (see http://cxc.harvard.edu/cal/Acis/Cal_prods/vfbkgrnd) of the ACIS background have shown that for weak or extended sources, a significant reduction of background at low and high energies may be made by using the information from 5×5 pixel islands, i.e. very faint mode, instead of the faint mode 3×3 island. This screening results in a 1-2% loss of good events. CIAO 2.2 and later provides a tool to utilize the VF mode for screening background events. Please note that the rmf generation is the same for very faint mode as it is for faint mode. See the "Why Topic" http://cxc.harvard.edu/ciao/why/aciscleanvf.html.
It is important to realize that VF mode uses more telemetry; the limit is ∼ 68 cts/s, which includes the target flux and the full background from all chips. Check the calibration web page (http://cxc.harvard.edu/cal/Acis/Cal_prods/bkgrnd/current/background.html) for a discussion of background flares and the telemetry limit. In particular, review Section 1.3 of the memo "General discussion of the quiescent and flare components of the ACIS background" by M. Markevitch
In order to reduce the total background rate and the likelihood of telemetery saturation, VF observations should consider using no more than 5 CCDs and an energy filter with a 12 keV upper cutoff. Starting from Cycle 11, the default upper energy cutoff has been decreased from 15 keV to 13 keV. But VF mode observers should consider reducing this threshold further to 12 keV. If the target is brighter than 5-10 cts/s, one has to take more drastic steps, such as turning off more chips or using Faint mode.
This mode should not be used for observing bright sources (see the discussion at the end of Section 6.16.1 for more detail).
Obviously the effects of pileup are most severe when the flux is highly concentrated on the detector. Thus, the core of the PSF suffers more from pileup induced effects than the wings. Figure 6.17 illustrates this point.
It is clearly important in preparing a Chandra observing proposal to determine if the observation will be impacted by pileup, and if so, decide what to do about it (or convince the peer review why the specific objective can be accomplished without doing anything). There are two approaches to estimating the impact of pileup on the investigation. The most sophisticated uses the pileup models in XSPEC, Sherpa, and ISIS to create a simulated data set which can be analyzed in the same way as real data. A less sophisticated, but very useful, approach is to use the web version of PIMMS to estimate pileup or to use Figures 6.17 and 6.18.
Simple Pileup Estimates
The pileup fraction is the ratio of the number of detected events that consist of more than one photon to the total number of detected events. An estimate of the pileup fraction can be determined from Figure 6.18. The algorithm parameterizes the HRMA+ACIS PSF in terms of the fraction of encircled energy that falls within the central 3×3 pixel event detection cell, and assumes that the remaining energy is uniformly distributed among the 8 surrounding 3×3 pixel detection cells. The probabilities of single- and multiple-photon events are calculated separately for the central and surrounding detection cells and subsequently averaged (with appropriate weighting) to obtain the pileup fraction as a function of the true count rate - the solid line in Figure 6.18. The model was tested against data taken on the ground under controlled conditions - also shown in Figure 6.18.
As a general guideline, if the estimated pileup fraction is > 10%, the proposed observation is very likely to be impacted. The first panel (upper left) in Figure 6.19 qualitatively illustrated the effect on a simulated astrophysical X-ray spectrum. However, the degree of pileup that is acceptable for a particular objective will depend on the particular scientific goals of the measurement, and there is no clear-cut tolerance level. If one's scientific objective demands precise flux calibration, then the pileup fraction should probably be kept well below 10%.
The PIMMS tool provides the pileup fraction using the algorithm described here, both for direct observation with ACIS and also for the zeroth-order image when a grating is inserted.Simulating Pileup
John Davis at MIT has developed an algorithm for modeling the effects of pileup on ACIS spectral data. The algorithm has been implemented as of XSPEC V11.1 and Sherpa V2.2. The algorithm can be used to attempt to recover the underlying spectrum from a source, or to simulate the effects of pileup for proposal purposes.
The algorithm has been tested by comparing CCD spectra with grating spectra of the same sources. Care should be taken in applying the algorithm - for example, using the appropriate regions for extracting source photons and avoiding line-dominated sources. A description of the algorithm can be found in Davis 2001 (Davis, J.E. 2001, ApJ, 562, 575). Details on using the algorithm in Sherpa are given in a Sherpa "thread" as of CIAO V2.2 on the CXC CIAO web page: http://cxc.harvard.edu/ciao/.
There are three components to the on-orbit background. The first is that due to the cosmic X-ray background (a significant fraction of which resolves into discrete sources during an observation with Chandra). The second component is commonly referred to as the charged particle background. This arises both from charged particles, photons, and other neutral particle interactions that ultimately deposit energy in the instrument. The third component is the "readout artifact" which is a consequence of the "trailing" of the target image during the CCD readout; it is discussed in Section 6.12.1.
The background rates differ between the BI and the FI chips, in part because of differences in the efficiency for identifying charged particle interactions. Figure 6.20 illustrates why.
Beginning in September 2002, observations have been carried out with the ACIS in the stowed position, shielded from the sky by the SIM structure, and collecting data in normal imaging TE VF mode at -120C. Chips I0, I2, I3, S1, S2, S3 were exposed. The SIM position was chosen so that the on-board calibration source did not illuminate the ACIS chips. This allowed us to characterize the non-celestial contribution to X-ray background (i.e., from charged particles). The resulting spectra from different chips are shown in Figure 6.21. Chip S2 is similar to the ACIS-I chips (denoted I023 in the figure) and not shown for clarity.
In addition, in July-September 2001, Chandra performed several short observations of the dark Moon, which blocks the cosmic X-ray background. The dark Moon and stowed background spectra were indistinguishable (except for short periods of flares and variable Oxygen line emission in the Moon observations). We have not observed any background flares in the stowed position. Thus, the ACIS-stowed background is a good representation of the quiescent non-X-ray background in the normal focal position and can be used for science observations.
|Bkgrd rates (cts/s/chip)|
|Upper E cutoff||15 keV||15 keV||15 keV||15f keV||13 keV||12 keV||10 keV|
|Chip S2 (FI)||10||6.3||10.7||7.1||5.8||5.7||5.2|
|Chip S3 (BI)||11||7.7||14.7||10.9||7.4||6.7||4.8|
|f Scaled from 13 keV rate.|
FI chips: 25 krads 2,500,000 cts/pix BI chips: 625 krads 62,500,000 cts/pixIf your observation calls for observing a bright point-like source close to on-axis, we suggest you use the MARX simulator (with the parameter DetIdeal=yes & dither, typically, on) to calculate whether your observation may reach 1% of the above mission limits in any one pixel. If so, please contact the CXC HelpDesk in order to custom design an observational strategy which may accommodate your science aims, while maintaining the health & safety of the instrument.
Number and Choice of CCDs
Up to six CCDs can be operated at once, if the science objectives require 6 CCDs. Previously in the mission, the CXC advised GOs to use 6 CCDs if possible. Given the changes in the thermal environment on the spacecraft around ACIS, the CXC now recommends that GOs use 5 or fewer CCDs if there is no impact on their science. Please see the discussion under "Choosing Optional CCDs" that follows for the reasons why we wish the observer to request fewer than 6 if feasible. The identity of the CCDs and the desired aimpoint must be specified.
Choosing Optional CCDs
The observer may specify that a given CCD must be on for an observation by entering "Y" for that CCD at the appropriate place in the RPS form. If a CCD must be off, one enters an "N". Finally, one may specify rank-ordered optional CCDs which will be turned on or off at the discretion of the mission schedulers depending on the specific thermal status of the Observatory at the time of the observation. This is done by entering "OPT1-OPT5" ("O1-O5" in RPS). The first to be turned off, if necessary, would be designated by "OPT1", the second to be turned off would be designated by "OPT2", etc. Observers are encouraged to use 5 or fewer CCDs if their science objectives are not significantly affected. Starting in Cycle 14, the RPS forms for proposal submission will no longer allow the proposer to specify "Y" for 6 CCDs. If the observer's science objectives nevertheless require 6 CCDs to be on, the observer should set 5 CCDs to "Y" and 1 CCD to "OPT1". If the proposal is selected, the observer should discuss the configuration with the user uplink support scientist and the optional CCD may be turned on. The CXC will do its best to schedule 6 CCD observations, but proposers should be aware that satisfying the thermal requirements of the Observatory may make the scheduling of such observations more complicated if not impossible.
Using fewer than 6 CCDs is beneficial in keeping the ACIS focal plane and electronics temperatures within the required operating ranges. Because of changes in the Chandra thermal environment, the ACIS Power Supply and Mechanism Controller (PSMC) has been steadily warming over the course of the mission. Under current conditions, and assuming an initial PSMC temperature of less than +30 C, observations at pitch angles less than 60 degrees, longer than 50 ks, and with 6 CCDs operating, are likely to cause the PSMC to approach or exceed the thermal limit. For pitch angles larger than 130 degrees, the ACIS focal plane, the Detector Electronics Assembly (DEA), and the Digital Processing Assembly (DPA) temperatures can also warm outside of the desired range. As above, each of these temperatures can be reduced by reducing the number of operating CCDs.
If an observer requires the most accurate gain calibration for their observation (provided by a cold and stable focal plane temperature), they should use 5 or fewer CCDs. This is done by setting 5 CCDs to "N", and 5 CCDs to "Y" or 1 CCD to "Y" and 4 other CCDs to "OPT1-OPT4". For example, if the observer is using the ACIS-I array for imaging, they could select the four I array CCDs and one of S2 or S3. If the observer is using S3 for imaging, they could select S2, S3, S4, I2 and I3 and turn S1 off or they could select just S1, S2, S3, and S4.
If no optional CCDs are selected, and six CCDs are on and the observation is not constrained in such a way as to prohibit it, the observation is likely to be scheduled at a time for which the pitch angle is between 60 degrees and 130 degrees. The user should be aware that such an observation may be listed in the long term schedule (LTS) at a date for which the pitch angle for this target is less than 60 degrees or greater than 130 degrees. However, when the observation finally appears in the short term schedule, it will be at a date for which the pitch angle will be in the allowed range.
Some Recommended Chip Sets
Observers should specify the chip set that is best for their primary science. The following suggestions have proven to be popular, and would facilitate a more useful and homogeneous archive.
ACIS-I (5 CCDs) imaging, nominal aimpoint ------- I0 I1 Y Y I2 I3 Y Y S0 S1 S2 S3 S4 S5 N N Y N N N ACIS-I (5 CCDs) imaging, nominal aimpoint ------- I0 I1 Y Y I2 I3 Y Y S0 S1 S2 S3 S4 S5 N N N Y N N ACIS-I (6 CCDs) imaging, nominal aimpoint ------- I0 I1 Y Y I2 I3 Y Y S0 S1 S2 S3 S4 S5 N N O2 O1 N N
ACIS-S (5 CCDs) imaging, nominal aimpoint ------- I0 I1 N N I2 I3 Y Y S0 S1 S2 S3 S4 S5 N N Y Y Y N
ACIS-S (5 CCDs) imaging, nominal aimpoint ------- I0 I1 N N I2 I3 N Y S0 S1 S2 S3 S4 S5 N Y Y Y Y N
ACIS-S (6 CCDs) imaging, nominal aimpoint ------- I0 I1 N N I2 I3 O1 O2 S0 S1 S2 S3 S4 S5 N Y Y Y Y N
ACIS-S (6 CCDs) spectroscopy, nominal aimpoint ------- I0 I1 N N I2 I3 N N S0 S1 S2 S3 S4 S5 O1 Y Y Y Y O2
Further information can be found at the CXC website http://cxc.harvard.edu/acis/optional_CCDs/optional_CCDs.html.
ACIS-I Count Rate and Number of Spectral Lines
We want to know the maximum number of counts from the brightest source in the field (not necessarily the target), and if the proposer believes that there will be lines in the spectrum from this source. Both questions are being asked to trigger assessment by the CXC of the sensitivity of the experiment to the gain calibration and so that appropriate scheduling may be employed to avoid thermally-induced gain drifts that might impact the observation. If the GO is interested in the optimal spectral performance of the ACIS CCDs, they should seriously consider using only 4 CCDs as discussed above.
There are only two choices: Timed Exposure (Section 6.12.1) or Continuous Clocking (Section 6.12.3).
Timed Exposure Mode
The timed exposure mode with the default nominal (and optimal) frame time of 3.2s is the typical mode for ACIS observations. Note that the option of selecting frame times shorter than nominal reduces observing efficiency, and hence the number of photons collected for a given observation time.
Continuous Clocking Mode
The Continuous Clocking mode is useful when timing data are so critical and/or pileup is such a problem that the sacrifice of one dimension of spatial data is warranted. The use of continuous clocking may also lead one to consider specifying a particular satellite roll orientation (see Chapter 3) in order to avoid having two different sources produce events in the same CCD column. (See Section 6.20.4 below.)
This option applies only to Timed Exposures. The parameters specifying an alternating exposure are:
Frame times and efficiencies in TE mode are discussed in Sections 6.12.1 and 6.12.1. The Alternating Exposure option is discussed in Section 6.12.2.
It is possible to remove events from the telemetry stream, and thus avoid telemetry saturation, by specifying an energy acceptance filter within which detected events will be telemetered. The default discards events above 3250 ADU (nominally 13 keV). The total per-chip background rates for different upper energy cut-offs are in Table 6.11.
Starting September 2006, a new energy to PHA conversion is used for observations with energy filters. Two sets of conversions are used, depending on the aimpoint of the observation. Observations with ACIS-S at the aimpoint use a conversion tailored for the BI CCDs and those with ACIS-I at the aimpoint use FI CCD-specific conversions. The BI and FI specific conversions are more accurate for each type of CCD than the conversion used in previous cycles. The assumption is that it is desirable to have the most accurate gain conversion for the CCD on which the HRMA aimpoint falls. Note that the conversion only impacts the on-orbit energy filtering. Ground data processing will always apply the appropriate PHA to energy conversion.
The observer should be aware that for observations which mix CCD type (i.e. BI and FI CCDs on), the selected conversion (based on aimpoint as above), will nevertheless apply to all selected CCDs. This will not affect the observation if the low energy threshold for the energy filter (the "Event filter: Lower" parameter) is 0.5 keV or less as the use of either conversion at these energies results in essentially no difference in the number of accepted events. However, above 0.5 keV, the conversions are significantly different. Finally, observations which apply an energy filter with a low energy threshold greater than 0.5 keV will automatically be assigned spatial windows that allow the FI CCDs to use the FI conversion and the BI CCDs to use the BI conversion regardless of aimpoint. Proposers who need an energy filter above 0.5 keV are encouraged to contact the CXC HelpDesk to discuss their plans with an instrument scientist.
A more sophisticated approach to removing data from the telemetry stream, and thus avoiding telemetry saturation, is by the use of a Spatial Window. This option offers a good deal of flexibility. One may define up to 6 Spatial Windows per CCD. Each window can be placed anywhere on the chip. Note there is a significant difference between a Spatial Window and a Subarray (Section 6.12.1): Subarrays affect the transmission of CCD data to the on-board ACIS processors; Spatial Windows select events detected by the processors and only impact the telemetry rate. The user may also specify the window energy threshold and energy range.
Spatial windows can specify the sample rate for events inside them. A sample rate of 0 excludes all events; the default rate of 1 includes all events; a rate of n > 1 telemeters one out of every n events in the window. A spatial window could be used to eliminate a bright, off-axis source that would otherwise overwhelm the telemetry stream. The order in which the spatial windows is specified is important if they overlap. The earliest specified window including a given pixel will be applied to events at that pixel.
There are a small number of additional parameters that need to be considered in specifying an observation with ACIS: (1) the off-axis pointing (if required), which reduces the flux, and spreads out the image; (2) the roll angle (Chapter 4); (3) time constraints (if any); and (4) time monitoring intervals (if any).
In the past, the continuous clocking (CC) mode (see Section 6.12.3) has been used to mitigate pileup in very bright sources (see Section 6.15.2). ACIS has offered two standard telemetry formats for observations performed in CC mode, faint (F) and graded (G) (see Section 6.14.2). While the CC-F mode choice has been used on occasion, the CC-G mode has been in the program for over 2 Msec of Chandra observing time, primarily to accommodate HETG spectra of bright X-ray binaries. The objective was to use CC mode in order to effectively mitigate pileup and conserve discrete structures such as emission and absorption lines, and edges in the dispersed spectrum. G mode also reduces the possibility of telemetry saturation.
Spectra in CC-mode, however, carry a number of currently uncalibrated artifacts ranging from a wrong spectral shape at wavelength larger than 10 Å, a smeared Si K edge, and charge losses below 3 Å. For bright sources up to about 300 mCrab we recommend the use of TE faint mode with four CCDs S1 - S4 and a 350 row subarray, at the expense of the spectral range above 12 Å ( < 1 keV), some pileup ( < 10%), and some moderate amount of frame drops. The latter can be mitigated by using TE graded mode in extreme cases.
We now have developed methods to calibrate modes similar to, but not identical to, the G mode. Starting with Cycle 12, the G mode has been altered to include some of the flight grades that were previously rejected on board. ACIS now telemeters all flight grades except 24, 107, 127, 214, 223, 248, 251, and 255.
Here are a few guidelines for the use of CC-mode and the HETG. Please note, that in any case spectra above above 12 Å ( < 1 keV) are always affected by some enhanced background:
For even brighter sources and in the case of HETG spectra, other mitigating actions can be taken, such as shutting off CCDs or moving half the source and dispersion off the array (i.e. using Z-SIM = -11mm). In these cases sources of over 300 mCrab can be accommodated.
For sources significantly exceeding such flux levels, CC-graded mode may be used. Like in the case of CC-faint mode, ASCA grade 7 events are no longer entirely rejected and similarly additional flight grades enter the telemetry stream causing frame drops to increase by about 10% compared to the previous definition. Mitigating strategies similar to those described above may be used.