GIS

General Principles

The ASCA GIS (Gas Imaging Spectrometer) is an imaging gas scintillation proportional counter (IGSPC). X-ray photons, focused by the mirrors, enter the detector through the thin entrance window (made from beryllium) into the ``drift region'' containing the detector gas: xenon with ten per cent of helium. The xenon atoms are ionized by the X-rays, and the photoelectrons produced then drift into the ``scintillation region'' which contains the same mixture of gases as the drift region. Here, the electrons are accelerated by a strong electric field the voltage of which is such that the electrons gain enough energy to excite xenon atoms but not to ionize them. When the atoms de-excite, UV scintillations are produced with intensity proportional to the energy of the original X-rays. These scintillations are then detected by the position sensitive phototube (PSPT). The electric field is parallel to the entrance window so the two-dimensional position of the scintillation corresponds to the position of the incident X-ray photon.

The Detector

Detector assembly

The structure of the focal plane assembly of the flight-model GIS is shown in Figure 6.2a. The total weight of the focal plane assembly is 3.8 kg, and the X-ray sensitive area has a diameter of 50 mm. It consists of the detector housing and the high-voltage supply units. The IGSPC--comprising the gas cell, the position-sensitive phototube (PSPT), and the front-end electronics--is placed in a housing cast from magnesium alloy with a diameter of 140 mm and a length of 580 mm, including a small radiation-belt monitor (RBM). The detector housing is placed on the optical axis of the X-ray reflective mirror which has a focal length of 3.5 m. The top section is a hood which prevents background X-rays and particles from entering along paths parallel with the mirror axis but outside the field of view of the GIS. The middle section accommodates the gas cell and the phototube, and the bottom section holds the front-end electronics. To improve the mechanical strength against vibration, a number of ribs run along the housing for support.

Two high voltage units are used for GIS. One produces a maximum of +1400 V for the phototube, and the other applies a negative potential of -8000 V to the gas cell. Since the photocathode of the IPMT is held at the ground potential, the potential of the X-ray entrance window, which is in principle exposed to open space, needs to be set at -7000 V. At an altitude of $\sim 550$ km, the orbit of ASCA is in the Earth's ionosphere where space plasma entering the GIS system could cause a high-voltage breakdown. To assess this effect, a simulation experiment was carried out in the space chamber at ISAS. Based on this study, the GIS has a thin ( $\sim 1\ \mu$m) plastic window with a conductive coating placed within the hood. Also, the electric current running in the high voltage network will be monitored continuously, giving a quantitative measure of the safety level.

The Radiation Burst Monitor (RBM) is a small solid-state detector, similar to the one used for Ginga, which continuously monitors the strength of the radiation environment and automatically controls the high voltage of the GIS.

Gas cell

The structure of the IGSPC in the GIS system is shown in Figure 6.2b. The gas cell is made from a ceramic tube with a beryllium entrance window and a quartz exit window. The volume is divided by a mesh electrode into two parts: the drift region at the top and the scintillation region at the bottom. X-rays entering through the window, the electric potential of which is about -8,000 V, are absorbed in the drift region. The electron cloud created by the photo-ionization slowly drifts to the first mesh (-7000 V) and is then accelerated by the strong field in the scintillation region toward the ground mesh placed just above the quartz window. In this process, the electrons excite the xenon atoms which then produce UV scintillations. Seen through the quartz window, the UV scintillations are detected with the PSPT and their position and intensity ( $\propto E_X$, the energy of the original X-ray photon) determined.

For the X-ray entrance window, a newly developed $10 \mu$m-thick beryllium foil, produced by the Yamaha Corporation, has been employed for the first time in space. With this window, 10 per cent transmission can be achieved at an energy as low as $E_X \sim 0.7$ keV--the lowest energy achieved by a sealed-off gas detector. The helium leak rate of the window is $\ ^{\displaystyle <}_{\displaystyle \sim}\ 1 \times 10^{-9}$ torr l s^-1 through an area 66 mm in diameter, an acceptable level for sealed-off gas detectors. The X-ray transmission of the beryllium foil has been tested at various positions on the surface with a narrow X-ray beam at different energies, and proved to be uniform (within experimental error) over the surface.

The structure of the window support, developed specially for the GIS, is shown in Figure 6.2c. It consists of a thin metal grid and a mesh, supporting the entrance area of 50 mm diameter. The molybdenum grid has 7 mm pitch and 3.5 mm height, with a wall thickness of 0.1 mm. The grid itself can withstand a pressure of several atmospheres and causes negligible shadowing on the image since the position resolution of IGSPC is $\sim 0.5$ mm. The grid also supports the mesh underneath against rupture. The stainless steel mesh has 1.2 mm pitch, 87 per cent transmission, and 0.3 mm thickness. Since the rigidity against deflection is a strong function of thickness, the supporting mesh used for the GIS is actually made from two identical meshes bonded together. The loss of the effective area due to the welding is negligible because the wire patterns of the two meshes overlap each other smoothly. The beryllium foil is then bonded onto a kovar flange using a high-temperature epoxy resin. The window support restricts the linear expansion of the beryllium to less than 0.1 per cent under a pressure of 1.5 atm, inside the limit of damaging the foil.

The quartz window is 2.5 mm thick with a diameter of 78 mm, and is bonded to the cupro-nickel flange using indium and a small amount of high-temperature epoxy resin as an additional coating. The sealing is successful over a temperature range from $-25^\circ$C to $100^\circ$C under a pressure of 1.5 atm and causes negligible stress to the quartz plate.

The prime content of the gas is xenon with the addition of 10 per cent of helium to improve the rise-time performance for X-ray signals (the helium increases electron drift velocities and can be used to check leakage). A small getter is attached to the gas cell and keeps the gas free from contamination. The gas mixture and the getter material are essentially same as those used for the Tenma SPC, therefore their performance is already known.

Phototube

The position-sensitive phototube (PSPT) is a Hamamatsu Photonics R4268 type, equipped with a 3-inch quartz window and 10-stage dynodes. The dynodes in the PSPT have a plane geometry, and in each dynode plane runs a number of parallel grids of triangular cross section. The direction of the grids is rotated by $90^\circ$ for each successive dynode stage. This dynode configuration keeps the electron path from the photocathode to the anode plane straight and the cloud size small. The anode is a cross-wire grid with 16 wires running in each X and Y direction at a separation of 3.75 mm. The position information is obtained from the anode signals, and the pulse-height is measured with the last dynode. A pencil-beam input to the PSPT shows the distribution of the output charge to be $7\pm1$ mm FWHM. The size of the sensitive area and the uniformity of the output signal have been measured by mapping the whole cathode area with a blue LED ( $\lambda = 450 \pm 30$ nm). The peak-to-peak gain variation is less than 10 per cent over the 60 mm length of the photocathode diameter, except for the dynode support 10 mm from the center. Such a small gain variation does not influence position determination since the light distribution from the gas cell has a large positional spread ( $\ ^{\displaystyle >}_{\displaystyle \sim}\ 15$ mm FWHM).

The 2-dimensional X-ray position is calculated from the signals of the 32 anode wires. The front-end electronics attached to the phototube accommodate 33 charge-sensitive preamplifiers, including the one for the last dynode. For the anodes, using operational amplifiers saves weight and space. Since, at $\sim 2\ \mu$s, the rise time of X-ray signals is rather slow, the op-amps work well as pre-amplifiers. These 32 anode signals are handled in four groups, i.e., even and odd numbered anodes in the X and Y directions. Each group carrying eight signals is handled by an analog multiplexer which picks up the signal and transfers it to an 8-bit flash ADC in the main electronics. Thus, only four flash ADCs can handle all the anode signals. The decay time of the preamplifier--an e-folding time of 100 $\mu$s--is set long enough to hold the signal level during the switching process which lasts $\sim10\ \mu$s. The switching at the multiplexer is controlled by the main electronics, and the duration, 1.2 $\mu$s for each anode, is long enough for the flash ADCs to convert the signal. By contrast, the pulse-height, rise-time, and lower-discriminator information are obtained from the last dynode, which produces the total-charge signal with inverse polarity. The use of the last dynode is simpler than adding all the anode outputs. The charge-sensitive preamplifier for the last dynode consists of discrete components to assure high signal-to-noise.

The Data System

On-board data processing

Signals from the IGSPC are first processed in the front-end electronics and then analyzed in the main electronics of the GIS. The pulse-height signal from the last dynode is first examined by the upper and the lower discriminators and then processed by the pulse-height and rise-time ADCs. The rise-time discriminator, the same as the one employed for the Tenma SPC, efficiently removes most of the background events. After filtering in this way, events are then analyzed by the on-board CPU, a 16-bit processor 80C286 operated at a clock rate of 10 MHz with a memory of 32 kB. The position signals sent from the multiplexers are converted with 8-bit flash ADCs, before transfer to the CPU for position determination. The two main functions of the CPU are the calculation of event position and background rejection.

Position determination

Since the photons from the gas cell spread radially according to an inverse square law while the dynodes have a plane geometry, the distribution shows a long extended tail. In addition, when an event occurs near the edge of the window, the profile becomes asymmetric. A simple calculation of the ``center of gravity'' of the light distribution thus does not yield the true incident position. We find, however, that a Lorentzian distribution, with only three parameters, can describe the light distribution adequately. For the actual CPU algorithm, we employ a linearized approximation of the Lorentzian function to avoid iteration and to quicken the calculation. A simulation of the position determination method has been carried out by generating an approximate light distribution based on the measured distribution of 6-keV X-rays at various incident positions. The calculated position is linear out to a radius of 30 mm, and the position resolution is better than 0.5 mm within a radius of 20 mm and scales approximately with $1/\sqrt{E}$. The processing time is $\sim 5$ ms per event. Considering the image size produced by the ASCA mirror (HPD $\sim 3$ mm), the position resolution of IGSPC thus obtained is satisfactory for an instrument in the focal plane of the XRT.

For the observation of strong X-ray sources (up to $\sim 0.5$ Crab), faster position calculation may be required. A simple alternative method is to take the square of the light intensity in each anode and then to calculate the mean. This procedure emphasizes the peak position and makes the result relatively insensitive to the extended tail. Although the position linearity and the usable area are not as good as for the Lorentzian fit, the calculation is three times faster.

Background rejection

The method of background rejection is briefly reviewed here. First, a background particle which deposits more energy in the gas cell than can be associated with an X-ray photon can be immediately rejected by the upper discriminator. In addition, non X-ray background can be distinguished from the proper X-ray signal in various ways, even when inexcessive energy is released in the gas cell. In the IGSPC, the characteristics of a background event can be recognized by any of the following four features:

1.

Frequent occurrence near the ceramic wall of the gas cell

2.

A long electron cloud track along the drift direction (perpendicular to window and mesh)

3.

A long electron track perpendicular to the drift direction

4.

The penetration of the event into the emission region.

The effect of the background rejection can be seen in terms of as the change of the ⁶⁰Co background count rate. The first method used is the rise-time discriminator (RTD) which proved successful in Tenma. The discrimination is effected by a hard-wire analog circuit that can operate even under very high counting rates (> 10³ count s^-1). RTD is particularly good at rejecting background types 1, 2 and 4 (above). The background rate within the pulse-height range of 1-10 keV is reduced to about 10 per cent after the RTD. The second method, carried out by the CPU, is to remove events occurring outside the X-ray sensitive region (50 mm diameter). Even though RTD removes most of the events near the detector wall, the residual image still indicates that the events tend to concentrate in the outer region. The removal of the outer region results in a further reduction by a factor of 2-3 in the background rate. Finally, type-3 background can be rejected using the width of the light distribution. This results in a 20 per cent reduction of the final background level. Therefore, the use of CPU results in the background reduction by a factor of 3 compared with the case of hard-wired electronics only. We also note that the background rejection does not seriously affect the X-ray efficiency. The counting rate for 6-keV X-rays decreases only by 5 per cent within 20 mm from the center after background rejection.

Observation Modes

The GIS has four different modes: pulse height (PH) mode, multi-channel pulse count (MPC) mode, position calibration (PCAL) mode, and memory-check mode. In practice, only PH and MPC mode are used for Guest Observations:

• PH mode: In this mode, the on-board CPU calculates the event position and discards background events using rise-time and position information. Only accepted X-ray events are sent to the ground. Pulse-height spectra of up to 1,024 channels for each detector are produced.
• MPC mode: This mode is meant for bright sources, optimizing temporal resolution at the expense of dispensing with positional information: only pulse-height information is recorded. This is also a back-up mode against CPU failure, since the data are processed without the intervention of the CPU. Pulse-height spectra of 256 channels for each detector are produced.

In almost all cases, PH mode will be used. Although MPC mode may be used for studying short time variations of bright sources, proposers must be aware that a precise spectral response is not available for the MPC mode.

Inflight GIS Performance

Energy resolution

The energy resolution of the GIS as derived from in-flight calibration matched almost exactly what was expected from ground measurements. That is, about $7.8\sqrt{(5.9/E_{keV}}$ per cent (FWHM).

Background

The internal background rate measured from night Earth observations is $\sim 6\times 10^{-4}$ count s^-1 cm^-2 keV $^{-1}\sim 0.1$ count sensor^-1 with on-board background discriminators effective. This is not very different from the measured laboratory background rate of $3\times 10^{-4}$ count s^-1 cm^-2 keV^-1 and the Ginga LAC in-flight background rate of $4\times 10^{-4}$ count s^-1 cm^-2 keV^-1. The diffuse X-ray background is $\sim 0.4$ count sensor^-1, hence the 5$\sigma$ detection limit is $\sim 0.01$ count sensor $^{-1}\sim 10 \mu$Crab for an exposure of 10⁵ seconds. On the other hand, the confusion limit for the XRT half power diameter (3 arcmin) is $\sim 10^{-4}$ count s^-1 (XRT HPD) $^{-1}\sim 0.4 \mu$Crab.

Useable field of view

Due to the nature of the detector, the performance of the GIS depends on radial position. When an X-ray is captured near the edge of the gas cell, some of the resultant UV photons cannot be detected by the phototube. Consequently, the energy gain becomes lower and the light distribution on the anode wires becomes asymmetric near the rim of the detector. The latter effect results in the radially ``shrunk'' position response. These effects are taken into account in the GIS response function and are correctable. According to current calibrations the GIS performance does not vary within 64 per cent ($\sim$15 arcmin) of the full 25 mm radius of the phototube. Data taken in the range 15-20 arcmin can still be used for most purposes, but beyond 20 arcmin the calibrations are unreliable and beyond 22 arcmin the particle background completely dominates the signal. Note that it is estimated that the GIS gain is uncertain by $\sim 1\%$ within a radius of 15 arcmin; by $\sim 2\%$ between 15 and 20 arcmin; and up to $\sim 4\%$ outside 20 arcmin. Note that only events inside the 22 mm arcmin radius (as a default) are tagged wth the corresponding sky coordinates derived from the attitude solution. Energy resolution is flat within a radius of 20 arcmin.

The other effect on the useable field of view is the presence of the supporting grid. The supporting grid structure yields non-uniformity of the efficiency over the FOV. The efficiency can be degraded by as much as by 30% on the grid. Figure 6.4 shows the grid superimposed on the field of view.

Acknowledgements

The material in this chapter has been drawn almost entirely from Imaging Gas Scintillation Proportional Counters for ASCA by Ohashi et al. (1991, SPIE Proceedings Series, 1549, 9). Further information can be found in the following references.

Tashiro, M. et al. 1995, "In-orbit performance of the GIS instrument on board ASCA (Astro-D)", SPIE, 2518, 2

Kohmura, Y. et al. 1993, "Calibration of imaging gas scintilation proportional counters on Astro-D", SPIE, 2006, 78

Figure Captions

Figure 6.2a: The focal-plane assembly of the GIS. The total length is 580 mm.

Figure 6.2b: The structure of the IGSPC for the GIS. The depth of the drift and scintillation regions are 100 mm and 150 mm, respectively, with an inner diameter of 92 mm.

Figure 6.2c: The structure of the beryllium window flange for the IGSPC. The flange body is made of kovar.

Figure 6.4: A bright earth image taken with GIS2. The data were accumulated over a period of one year. The calibration source is included (upper-right). The support grid structure is clearly seen. The plot has a linear scale from 50 to 170 cts/pixel. This image can be accessed at

ftp://heasarc.gsfc.nasa.gov/asca/gis_information/earth2.img.