next up previous contents
Next: Feasibility: GIS+XRT Up: ASCA AO7 Appendix E Previous: GIS

Subsections

SIS

Introduction

At the most abstract level, CCD detectors work the same way as proportional counters: an X-ray is absorbed in the body of the detector and produces electrons whose total charge is proportional to the original photon energy. CCDs differ from proportional counters in the way the charge is created, collected and transferred.

CCD structure

The SIS CCDs are frontside-illuminated, frame-transfer CCDs. Frontside illumination means that the electrodes which carry the electronic charge are on the surface of the CCD above the silicon substrate. Frame transfer, as opposed to line transfer, means that after X-ray photons are detected in the imaging region, the resulting electronic charge is temporarily stored in the accumulation region and passed to the horizontal transfer register where electrons are read line by line.

The way such a CCD works is best appreciated by first considering its vertical structure. From the surface, the SIS CCDs comprise:

When an X-ray photon hits the surface of the CCD and is absorbed in the silicon substrate, it creates one electron per 3.65 eV of energy. If the X-ray photon is absorbed in the depletion layer, the resulting cloud of electrons is moved into the potential well located between the embedded channel and the depletion layer. By applying pulses of voltage across the electrodes, the electrons are transferred or ``clocked'' from one potential well to the next and eventually read out by the pre-amplifier. This transfer of electrons is not without loss: as the charge makes its way across the CCD, some of it may be trapped by defects associated with impurities or by small potential wells created by cosmic ray hits. This effect is parameterized by the charge transfer efficiency (CTE) or charge transfer inefficiency (CTI =1- CTE).

Properties of the SIS CCDs

The two identical X-ray CCD cameras on board ASCA are known as the Solid-state Imaging Spectrometers (SIS) and have been provided by a hardware team from MIT, Osaka University and ISAS. Each CCD camera head is based around four $420\times422$ pixel2 MIT Lincoln Laboratory CCD chips abutted side by side, with four preamplifiers, front-side illuminated. The pixels measure $27\times27$ $\mu{\rm m}^{2}$, and the resistivity is 6500 $\Omega$ cm; measured system readout noise levels of 4-5 electrons RMS are characteristic of the flight instruments. When cooled to a modest temperature of -60 degrees Celsius, a FWHM energy resolution of 2 per cent at 5.9 keV has been achieved and has been verified in flight during the PV phase. Using a $\approx 40 \mu{\rm m}$ thick depletion layer, the entire energy range from 0.4-12 keV can be covered, allowing the spectroscopic study of high-ionization lines from oxygen through nickel. The field of view of each camera is $22\times 22$ arcmin2. The configuration of the CCD chips is shown in Figure 4.5.

SIS Camera

Figure 7.3 shows a side view of the SIS camera. The body of the camera is made from aluminum and is designed support a vacuum. It is continuously vented through a vent pipe attached to the body to prevent degradation of the CCDs. The body is attached to the satellite with legs made of lexan which has a low thermal conductivity. A copper mesh is attached between the bottom of the camera and a heat pipe. Inside the body, a Peltier cooler (Thermal Electric Cooler) is installed to cool the CCDs. A flexible printed circuit board is used to wire the CCDs to lower the thermal conduction from the body. The top side of the CCDs is covered with MLI (Multi-Layered Insulator) to prevent the intrusion of thermal radiation. This MLI also works as an optical light blocker. A film of 100 nm lexan coated with 80 nm of aluminum is also used as an optical light blocker at the top. The measured optical transmission of this filter (at 630 nm) is $4\times
10^{-6}$. A thin film-type resistance heater is attached on the bottom side of the CCD. This heater, controlled by a thermostat, is used to remove materials contaminating the surface of the CCDs.

Several techniques are used to shield the camera from radiation damage. The baffle, which extends from the top of the camera, blocks high-energy X-ray background radiation, while X-rays which try to enter though the sides and back of the camera are blocked by the satellite body itself and by tin films attached on it. The camera also has a proton shield to protect it from the high proton fluxes during passages through the South Atlantic Anomaly.

Detection and Classification of Events

Charges generated by X-ray photons in the CCDs are concentrated in a small number of neighboring pixels (most often 1 or 2, possibly up to 3 or 4); charged particle events are often spread over many pixels. It is important therefore to judge which pixels contain the charges from a given event, as well as the total amount of charges collected from it. Note also that adding signal from multiple pixels also increases the read out noise, which is a limiting factor of the energy resolution.

On-board processors search for pixels that have a pulse height exceeding the event threshold; if, moreover, this pulse height is higher than any of its 8 neighbors, then it is judged to be an event center. The processors then concentrate on the 3$\times$3 pixel area of the chip. In Faint Mode, the pulse heights of the $3\times 3$ pixel block will be sent to the ground. In Bright Mode, the eight pixels around the central pixel will be compared with the split threshold; the many possible combinations of which of the 8 neighboring pixels exceed the split threshold is then reduced to 8 basic patterns, called Grade (see Figure 7.4b). Charges collected in corner pixels (except in the case of square, or Q, events) are unlikely to be from the X-ray event and are not included in the total pulse height. Pulse heights of non-corner pixels exceeding the split threshold are then added, and the total pulse height is telemetered to the ground along with the Grade information. Faint Mode data are processed similarly on the ground, but with some additional sophistication.

From the early PV phase to AO-3, the event threshold was fixed at 100; starting from AO-4, the event threshold for each chip has been adjusted to be $\sim$0.4 keV, somewhat higher than the original value. Except in some early PV phase observations, split threshold has been fixed at 40; since the choice of split threshold changes the SIS response, it is inadvisable to choose any other values. Currently, Bright Mode events having Grades between 0 and 6 are telemetered to the ground; from launch up till 1993 Nov 30, only Grades 0 through 4 were telemetered to the ground. The SIS response is best-calibrated for a combination of Grades 0, 2, 3, and 4. These and Grade 6 events are mostly X-ray events, whereas Grades 1, 5 and 7 are mostly charged particle events.

If multiple X-ray photons were detected in the same, or neighboring, pixels during the same exposure, the classification software has no way of recovering the pulse heights of each photon. This is called photon pile-up. Such an event may appear as a photon of higher energy, or may classified into a Grade that would suggest a charged particle origin.

The above description applies to the imaging (Faint and Bright) modes; in Fast Mode, a 1-dimensional version of Grade is calculated (single, 0, or split, 1).

Observation Modes

Various modes are available for observing a source with the SIS, the choice depending on source brightness, the extent of the source and the time resolution desired. For an observation, two concurrent modes have to be set: the clocking mode, which determines how the CCD is read out, and the data mode, which determines how the charge cloud created by a photon is described.

CCD clocking modes

The SIS CCDs are operated in what is known as ``frame-transfer'' configuration. Each CCD is divided into two areas, the imaging area, which is exposed to radiation, and the readout area, which is shielded. Each imaging-area pixel has a corresponding pixel in the readout area to which the charge is transferred (in 16 ms). Once transferred, the charge resides in the readout pixel until the pixel takes its turn to be read out. Readout is effected by applying pulses of voltage (``clock'' pulses) across the CCD which move the charge packets--row by row, pixel by pixel--to the amplifier. Reading out the whole CCD in this way takes 4 s (the slower the readout, the lower the noise). Each SIS camera has four CCDs, and the time resolution is determined by the ordering of the imaging/read-out cycles of the four CCDs. The various CCD clocking modes are:

CCD data modes

Address and level discrimination

There is a provision in the ASCA on-board software to ignore events that were detected in certain areas on the chip. This then frees up more slots in the telemetry for events detected elsewhere on the chip. This capability is currently being used for two purposes:

1.
To combat light leaks. When SIS0-chip2 and/or chip3 are in use, address discrimination, i.e., masking, is used on these chips to reduce the number of light-leak induced events in telemetry. Light leaks occur whenever the spacecraft points close to the limb of the bright Earth (currently within 25 degrees). When this technique is applied, about a third of each chip is not taking data.
2.
To combat telemetry saturations and photon pile-ups in bright sources. The core of the PSF in this case is masked off, with some leeway for pointing uncertainties, so that the source count rate is reduced by about a factor of 2.

It is also possible to apply level discrimination, to ignore (on-board) any events of central pulse height less than specified values. This is also useful for reducing the risk of telemetry saturation.

Mode choice

SIS mode specified in the proposal form is not regarded as final. The GOF will contact successful PIs before their observations take place to discuss the choice of mode. However, for some bright and/or extended sources, the desired combination of mode and bit rate might not be feasible, especially given the problem of hot and flickering pixels. In this case, the proposal can be rejected on technical grounds. As this is a feasibility issue, the choice of mode is discussed in Chapter 9.

Inflight SIS Performance

Energy resolution

Although the energy resolution of the SIS after launch achieved its expected value of $\sim 2\%$ at 5.9 keV (FWHM), it has now been degraded to $\sim 4\%$ (1-CCD mode) at 5.9 keV. This is due to radiation damage and in 1-CCD mode is mostly connected with CTI non-uniformity. In 2-CCD and 4-CCD modes the effect RDD (see §7.7.2) degrade the energy resolution even further. Further details can be found in ASCA Newsletter #5 (Dotani et al. 1997) in which line-width measurements of Cas A are presented as a function of time since launch. See also §9.6.2. As far as the SIS team has been able to determine, the initial energy resolution of the SIS as derived from in-flight calibration matches almost exactly what was expected from ground measurements. That is, about 2 per cent at 5.9 keV (FWHM). For the effects of radiation damage, see below.

Background

The total background rates (diffuse X-ray background & particle-induced) in the 1-8 keV band are 0.09 count s-1 sensor-1 for SIS0 and 0.10 count s-1 sensor-1 for SIS1. Both values have been derived from night-time observations. For the recommended pairing of chips for 1-CCD mode, i.e., SIS0-chip1 & SIS1-chip3, these rates are 0.026 count s-1 chip-1 and 0.025 count s-1 chip-1, respectively, day-time or night-time, provided ASCA is pointed no closer than 25 degrees from the Earth's bright limb. [Within the approximate 25-degree limit, both light leakage--see below--and atmospheric oxygen K emission raise these values.] Of the night-time background rate, particle-induced background comprises about 20 per cent in the 1-8 keV band.

In addition, flickering pixels contribute to the background. Due to the sporadic nature of flickering pixels, those with low duty cycles cannot be eliminated from the data. The overall increase in flickering pixels (particularly in 2 and 4 CCD mode data) is reflected in the increased `background'. It can be recognized by its characteristic steep increase towards the lowest channels.

Limiting sensitivity

The sensitivity (5$\sigma$ level of confidence) achieved by the SIS in a 105 second observation is $\sim 1.0\times 10^{-3}$ count s-1 in a fixed detection cell of 3 arcmin diameter in the range 0.5-10 keV.

Optical and UV light leakage

The SIS cameras experience some degree of light leakage for observations at low (<80 $\deg$) Sun angles and when pointed at or near (i.e., within 25 degrees) the bright Earth. SIS0 is affected the most, with SIS0-chip2 and SIS0-chip3 showing the most pronounced effect. These effects are due to both an increased overall bias level and apparent reflection effect near the co-linear outside edges of SIS0-chip2 and SIS0-chip3. Within a few minutes in time of the day/night transition, the on-board bias estimator can be in error. This effect is a strong function of clocking mode: most severe in 4-CCD mode, only mildly noticeable in 1-CCD mode.

Echo

The analog electronics that read out the CCDs have been found to misassign a small fraction of the charge in one pixel to the next pixel. The fraction is different between SIS-0 and SIS-1, and is slightly time-dependent. This effect leads to different branching ratios among various Grades from that expected from the considerations of physical interactions of X-ray photons and the CCD. For Bright Mode data, response matrices that treat echo effects on a statistical basis should be used; Faint Mode data can be processed on the ground exactly as Bright Mode data, or echo correction can be applied to individual photons.

Hot and flickering pixels

Each CCD chip in the two SIS cameras has a small number of catalogueable pixels (approximately 10 out of 175,000) which have higher than normal dark current. One of the eight chips, SIS0-chip0, has about 50 such ``hot'' pixels. Even though the number of these pixels is small as a percentage of the array total, each such pixel produces a false ``event'' which is not rejected by the onboard digital processor and is telemetered (i.e., passed on to the bubble data recorder).

In addition to the hot pixels, there is also a population of ``flickering'' pixels, with a variety of duty cycles. Some stay active for many weeks, some for a much shorter period; even when they are active, they often do not consistently produce charges above the event threshold. The pulse height distribution of flickering pixels is strongly peaked towards the lowest detectable (i.e., event threshold) level. All hot pixels and the more active flickering pixels can be removed during the data analysis stage.

By occupying telemetry, hot and flickering pixels reduce the amount of telemetry available for true events. Operationally, this imposes restrictions on which combinations of mode and bit rate can be used. Regardless of source count rate, low bit rate cannot be used at all, and 4-CCD mode is not recommended even in high bit rate because other problems are too severe (see below). Since this is a feasibility issue, it is also covered in Chapter 9.

Radiation Damage Monitoring and Predictions for the AO-7 Phase

The SIS team has been monitoring the effects of charged particles on the CCD chips. The two main manifestations of the radiation damage are the creation of charge traps and the distortion of the dark current distribution. These in turn result in an increased number of flickering pixels, lower detection efficiency, shifts in energy scale and degraded spectral resolution. These effects depend on CCD clocking mode: 4-CCD mode is virtually unusable whereas 1-CCD mode remains the least affected. Here, we present a summary of expected SIS performance during the AO-7 period. Most up-to-date information can be obtained on the WWW and via e-mail bulletins.

CTI

Traps caused by radiation damage impede the transfer and read-out of charges from the CCDs (Charge Transfer Inefficiency, or CTI: ${\rm
CTI} = 1 - {\rm CTE}$). This effect has now been measured. It causes position-dependent changes in energy-scale. The Poissonian distribution of traps alone would cause some variation in the CTI applicable to individual pixels around the mean; however, the measured non-uniformity of CTI on a column to column scale, as well as across the chip, appears to be greater than expected from a Possionian distribution.

The SIS team is continuously monitoring the position dependent gain changes; the energy scale calibration therefore appears not to pose a major problem for AO-7 observations. However, the non-uniform CTI is manifesting itself in the degradation of spectral resolution. Even though we now have the capability to create response matrices with time-dependent resolution, this may nevertheless limit what can be achieved with the SIS (see section 9.6.2).

It should also be noted that the calibration observations are optimized for the calibration for the standard point source pointing locations on the standard chips. The CTI calibrations are less accurate at non-standard locations and non-standard chips, which may affect observations of extended or serendipitous sources.

Residual Dark Distribution

The ASCA on-board processor maintains a coarse map of the dark levels and subtracts these from individual pixel values. The dark frame map is calculated assuming a symmetric noise around the mean dark level. Unfortunately, the cumulative effects of radiation damage have created a population of active pixels, which skews this distribution, and thus the on-board dark level subtraction is imperfect. This effect is called the Residual Dark Distribution, or RDD.

RDD interacts with the way events are detected and classified. In addition to adding noise, events are misclassified as having a higher grade, including ones not recognized as a photon event. Thus the effect of RDD is a complex interplay of the analog electronics, on-board digital processing and off-line processing, and involves intrinsic randomness of these active pixels. It effectively lowers the quantum efficiency of the chips, distorts the energy scale and degrades the energy resolution. RDD is clocking-mode dependent: the longer you integrate, the bigger RDD becomes. Therefore, it is most pronounced in 4-CCD mode (16 s integration), and less important in 1-CCD mode (4 s integration).

It has been found that the old version of faintdfe, the software used to fine-tune the dark level for Faint mode data, was working poorly in the presence of RDD. While the RDD effects alone reduce the detection efficiency and degrades the spectral resolution, the combination of RDD and the old faintdfe made the situation worse. A new version of the program, based on the algorithm developed by Dr. Otani, is available and appears successful in reducing, but not eliminating, the effects of RDD for Faint mode data.

The SIS response generator, sisrmg, contains a model of how RDD is degrading the spectral resolution. However, the current version of sisrmg does not have a model of how RDD is degrading the detection efficiency. The SIS team hopes to develop a quantitative model for this and impliment it into sisrmg during 1998.

In addition, we now support a new tool, correctrdd. This uses frame mode data, taken by the SIS team at quasi-regular intervals, to correct for the deficincies of the on-board dark level map. For this tool to have the full effect, the data need to be taken in Faint mode; in this case, the spectral resolution and the detection efficiency can be restored to a considerable extent. Recent tests suggest, however, that a full restoration of SIS performance may not be feasible: correctrdd can only correct for the mean change in the dark currents, not for the increased readout-to-readout variance in the dark currents. Moreover, the calibration of RDD-corrected data remains uncertain at the moment. For the Bright mode data, this tool attempts the correction of a much reduced scope; correctrdd cannot be expected to restore the full performance of, e.g., 4-CCD Bright mode.

Predicted performance of SIS during AO-7

4-CCD mode

4-CCD mode suffers severely from both hot and flickering pixels and from RDD effects. correctrdd can reduce the RDD effects for Faint mode data; however, this requires a level discrimination of >1 keV at Medium bit rate. Such a high level discrimination can be avoided if Bright mode is used at Medium bit rate, but this prevents the use of correctrdd. Moreover, even after correctrdd, recent 4-CCD mode data appear to show appreciable degradation in performance. We therefore do not recommend 4-CCD mode for general use.

2-CCD mode

With no level discrimination, 2-CCD Faint mode will not be available at Medium bit rate. 2-CCD Bright mode at Medium bit rate will probably require level discrimination at 0.48-0.55 keV level. Without correctrdd, 2-CCD mode data suffer appreciably from RDD effects (count rates reduced by $\sim$15%, excess ``NH'' of 1-2 $\times 10^{21}$ cm-2, as well as degraded spectral resolution). We therefore do not recommend 2-CCD mode for point source observations.

Status of 1-CCD mode

This is the least affected mode of all. Telemetry saturation due to hot/flickering pixels is not a major issue. RDD effects are small; the spectral resolution will continue to slowly decline due to non-uniform CTI.

Predicted degradation in detection efficiency

The RDD effect, by leading to misclassification of events, leads to a loss of detection efficiency. For 1-CCD mode data, however, this effect remains small and the efficiency is expected to remain close to 100% (relative to the efficiency at launch) through the AO-7 period. For 2-CCD mode data, the efficiency will have dropped to the 80-85 recovered somewhat by using the new tool correctrdd.

Warm CCDs

The above statements apply generally to the times when the CCD temperatures are regulated to be within the nominal range. For more than half the observations, however, the CCDs are warmer. This increases the flickering pixel rate: to avoid telemetry saturation, level discrimination by $\delta$E of $\sim$0.05 keV is necessary. Moreover, this induces additional uncertainties in the energy scale.

Status of Fast mode

Fast mode is the least affected by RDD effects; however, it is severely affected by CTI. CTI is dependent on the detector coordinates at which a photon was detected; however, the position information is not telemetered in Fast mode. Thus Fast mode data suffers more than the imaging modes due to CTI. It is rather less well calibrated than the imaging modes, and the lack of positional information will limit the level of calibration that we can eventually achieve. Therefore it is recommended that Fast mode be used only when there is a compelling scientific rationale (for extremely bright sources or for sub-4 second timing), and even then some Bright mode data be taken to provide a check of the Fast mode calibration.

Acknowledgement

Material from this section has been drawn from the ASCA Interim Report (1992, ISAS, in Japanese, translated by Ms.Takako Loewenstein and Dr.Takashi Isobe). Further technical information about the SIS may be found in Gendreau 1995 (PhD thesis, MIT) and references therein. It can be found on the WWW at URL

http://asd.gsfc.gsfc.nasa.gov/users/kcg/www/thesis.html

Figure Captions

Figure 7.3: A side view of the SIS camera. The vent pipe and radiation shield are omitted.

figure7_4b
Figure 7.4b: The definition and examples of SIS Bright mode grades.


next up previous contents
Next: Feasibility: GIS+XRT Up: ASCA AO7 Appendix E Previous: GIS
Michael Arida
1998-06-02