skip to content
ASCA Guest Observer Facility

Initial Report on the SIS Internal Background

--Keith C. Gendreau, MIT


We present an initial report on the instrumental background of the Solid State Imaging Spectrometers aboard ASCA. Factors which affect the intrumental background include the local magnetic rigidity and CCD clocking mode. The spatial, spectral, and temporal dependence of the background will also be addressed.

A Brief Description of the Intrumentation

The two Solid State Imaging Spectrometers (SIS) aboard ASCA are X-ray CCD cameras made at MIT. In order to understand some of the instrumental background features of the SIS, we review here a basic description of the detectors.

Each detector consists of a hybrid array of four-edge abuttable frame-store CCDs made at MIT's Lincoln Laboratory. Each CCD has 27x27 micron pixels in the imaging array and 18x24 micron pixels in the framestore array. Both the imaging array and framestore array have 420x420 pixels.

Each CCD is approximately 380 microns thick, of which about 25-30 microns are depleted for charge detection. The charge cloud produced by an event occuring inside this front 25-30 micron depletion region will be completely collected and registered (usually) as a single pixel or two-pixel event. The remaining 350-355 microns of silicon will still participate in photoelectric and other ionizing events with X-rays, g-rays, and high energy particles, but the charge clouds produced will expand freely with time until either recombination occurs or the charge enters a depletion region. Events which interact deep from the front surface in the CCD tend to be higher energy photons or particles. The expanding charge clouds from these deep background events can encompass several pixels and thus tend to get graded as multiple pixel events (grades 6 and 7; see the AO description of event grading or Ricker, et al. (1994)).

A factor which affects some of this grade rejection of the background is a feature on the CCD called the "Back Diode." The Back Diode is plate attached to the back side of the CCD which can be set at a high potential. The plate then can deplete the silicon reducing the 350-355 micron field- free region by about 200 microns. The effect is to reduce the maximum size that an individual deep-interacting event can affect, thus reducing dead area. On ASCA, the back diode has two states: a high voltage and a low voltage state. These result is different back depletion region depth, and thus a possible effect on the grading of deep-interacting background events. Figure 1 shows a cross section of a part of the CCD with the various depletion and field-free regions shown.

A more complete description of the CCDs can be found in Burke, et al. (1991).

The CCD is mounted on a ceramic plate inside of a gold coated kovar (iron, nickel, cobalt alloy) package. The package is mounted within a gold coated aluminum housing, which is surrounded by a polyethylene shield. Figure 2 shows a schematic of the arrangement. The polyethylene shield and aluminum housing provide protection against radiation damage in the CCD (Gendreau, et al. (1993)). The optical block filter shown in the figure above the CCD is extremely thin (1000 angstom Lexan + 800 angstrom aluminum). A beryllium door is also near the CCD (but not in a clear line of sight). The door was closed for the first few days after the SIS turn on and is now permanently open.

Figure 1: Cross section of CCD (not to scale) showing X-rays (X) illuminating the CCD from above. The X-rays pass through a thin deadlayer gate structure consisting of polysilicon, silicon oxide, and silicon nitride. X-rays which interact within the front side depletion region usually produce charge clouds which are entirely collected within 1 or 2 pixels (event "S"). Higher energy photons and particles interact deeper in the silicon. Events which originate in the field free region (event "D") will produce expanding charge clouds which affect several pixels. Particles which interact even deeper in the "back diode depletion region" will be entirely collected by the back diode ("R") and not register as an event. All dimensions in the figure are approximate.

Figure 2: Cross section of CCD camera (not to scale) showing X-rays (X) illuminating the detector from above. The figure to the left is slice through the whole camera. A polyethylene shield (A) surrounds the gold plated aluminum housing (B) of the camera. Other material around the CCD include: a beryllium door (C) now permanently open, an aluminized lexan optical blocking filter (F), and a gold plated kovar block (D). The figure to the right is a blow-up of the cross section of the region around the CCDs. The CCDs sit on a ceramic plate inside of a gold plated kovar package. A gold-plated kovar framestore shield is mounted on the package allowing X-rays to illuminate the imaging arrays of the CCDs.

A Two Component Model for the Origin of the Instrumental Background and the Effects of CCD Operating Mode

The internal background of the SIS can be divided into two components:
  1. internal background falling on the imaging array and
  2. internal background falling on the framestore array.
The first component is proportional to the live time (the sky exposure time), while the second component is proportional to the number of readouts. The number of readouts (Nr) is proportional to the live time (Tl) by the following formula:

Nr = Tl/Tm (1)

where the frame exposure time (Tm) is a function of clocking mode:

Tm = 4s for 1CCD, 8s for 2CCD, 16s for 4CCD (2)

Thus if you calculate internal background rates based on sky exposure time, the contribution of the internal background on the framestore array will be higher as you go from 4- to 2- to 1-CCD mode.

In general, there will be a position dependence (in detector coordinates) to the internal background. For the imaging array, it is not easy to see where the background is coming from since the geometry is rather complicated. For the component due to the internal background on the framestore array, the rate will increase with row number, since the “exposure time” of a row in the framestore area increases with row number. Considering this, below is an expression for the background counting rate per row (R(r)) read out:

R(r)=D(r)+I(r)+(InaughtXRrow X (r/Tm)XPa) (3)


is the sky background (0 when looking at Dark Earth) in counts s-1 row-1,
is the internal background per row on imaging array in counts s-1 row-1,
is the intrinsic internal background rate per row on the framestore region in counts s-1 row-1,
is the readout time per row (9.4 msec)
is the ratio of the pixel areas in the imaging and framestore region (= 0.59 )

Thus we should see a slope to the internal background with row.

When equation (3) is multiplied by the live time, Tl, then it is easy to see that the sky and image area components scale with time, and that (via equation (1)) the framestore component is proportional to the number of readouts.

If we integrate equation (3) over 420 rows and then divide by 420 rows, then we will get the counting rate per CCD per second (Cm). We can use the last equation to determine the ratio of internal counting rates for different CCD clocking modes. This could also be used to predict total background rates (Sky+internal) for different modes. For now, consider looking at the dark Earth, then we may approximate that

D(r)=0 (5)

If we make the additional assumption that

I(r)=Inought (6)

then equation (4) yields the following results for 1, 2, and 4-CCD mode operation:

C1=1.29I, C2=1.14I, C4=1.07I (7)

In particular, the predicted ratio of internal background rates for both 4-CCD to 1-CCD mode and also 4-CCD to 2-CCD mode observations per CCD should be:

C1/C4 = 1.203, C2/C4 = 1.068 (8)


Event Grade Rejection: We normally do science analysis with data of grade less than 5. This practice results in a rejection of about 98 percent of the all cosmic ray induced events. The higher grade events occur with a frequency of about 1 count s^(-1) cm^(-2) of detector. The two component model applies here, and thus the rate is clocking-mode dependent and row dependent. There is some evidence that the high grade event rate in the framestore array is larger than it is in the imaging array. This may be a result of both showering effects from the framestore shield as well as by grading differences due to the smaller pixels in the frame store array. Since we can reject the high grade events so easily, we will not discuss them further here.

Dark Earth Data Selection and Cleaning: To handle the remaining 2 percent of the internal background, 110 ksec of 4-CCD mode data were collected from April to September 1993 during satellite night while ASCA was looking at the dark side of the Earth. We took this "dark Earth" data only during satellite night to avoid optical light leakage problems.

The SIS is sensitive to optical light. Optical light contamination affects the SIS in two ways: it will lead to high grade events which often results in telemetry saturation (in particular with chips 2 and 3 of SIS-S0), and it will lead to gain offsets for normal X-rays. An additional effect that comes with optical light contamination in proximity to the bright Earth is the associated atomic oxygen K emission line (0.54 keV) from the bright Earth atmosphere. The complications due to optical contamination will be discussed elsewhere.

Hot and Flickering pixels were removed from the data by disregarding any pixels which registered an event twice. This is a reasonable thing to do considering that each CCD has over 160,000 pixels and during the 110 ksec each accumulates about 1000 internal background events. If the high frequency flickering pixels (hot pixels) are left in, then several "lines" will appear in the resulting pulse height spectra. The inclusion of the very lowest frequency flickering pixels (e.g. including pixels which registered events 2 or 3 times during this 110 ksec) results in a steep power law feature in the spectra. The flickering pixels will be discussed elsewhere in more detail.

Row Dependence: Equation (3) describes how the background counting rate should vary with row due to the extra exposure time in the framestore region. Figure 3 shows the variation in grade 0-4 counts (in 0.4-12 keV) with row for the dark Earth data. The slope is consistent with what would be expected for 4-CCD mode operation with identical intrinsic rates over the imaging and framestore regions.

Spectral Features: The spectra (Figure 4) of the cleaned dark Earth data sets can be characterized by a nearly flat continuum, fitted without using the telescope and detector responses, with fluorescence lines due to iron, nickel, aluminum, silicon, and gold. Table 1 lists the line features as well as the possible origin for most of the lines. Due to geometric considerations (see Figure 2), the origin of the iron and nickel emission lines is most likely the particle induced fluorescence in the kovar framestore shield. The X-rays from framestore shield will be landing mostly on the framestore array of the CCD leading to strong spatial gradients and clocking mode effects by the two component model described above.

Figure 3: The gradient in grade 0-4 events with row for the dark Earth data. The events represent those from pulse heights for 400 eV to 12 keV. All eight chips of the two SISs are combined here to show the effect. 4-CCD mode was employed.

Figure 4: The spectrum grade 0-4 events for the dark Earth data. The events represent those from pulse heights for 400 eV to 10 keV. All 8 chips of the 2 SISs are shown. 4-CCD mode was employed.

Table 1. Spectral Features in the Instrumental Background of the SIS

There is also an oxygen line present, which we consider to be most likely of atmospheric origin since there is no Si-K line of comparable strength seen in the spectrum while the fluorescence yield of oxygen is more than an order of magnitude smaller than that of silicon and there is much more silicon around the sensitive part of the detector. The average spectrum of the darkEarth data set from the two SIS sensors has a total intensity of 7.3x10^(-4) ct s-1 keV-1 per CCD (6x10-4 ct s^(- 1) keV^(-1) cm^(-2)) when in 4-CCD mode. More detailed rates will be given below.

Magnetic Rigidity Effects and Counting Rates: The background rates have a dependence on the local magnetic rigidity in which the satellite is located. Figure 5 shows a histogram of fractional exposure as a function of magnetic rigidity. In this figure, you can see that for a typical observation, the local magnetic rigidity ranges from 4 to 15 GeV/C with a large hump peaked at about 12 GeV/C. To examine the rigidity effects on the background, we have divided the dark Earth data set into low (less than 12 GeV/C) and high (greater than 12 GeV/C) subsets. Table 2 shows some of the count rates and model parameters found for these two rigidity ranges. The values quoted in table 2 are mean values for the listed sensor.

Figure 5: A histogram showing the fraction of an observation that is done while the satellite is within a region of given local magnetic rigidity. For the plot above, two histograms are shown: one for the collection of dark Earth data taken and one for the cosmic XRB data taken. The similarity of the curves helps to justify the use of the dark Earth data as background for the XRB data.

The model parameters refer to "/b" models found in the XSPEC fitting package. We fit power laws and multiple Gaussian models to the data using the "/b" (background) option which allows us to use the detector response matrix without using the effective area information (i.e. we only use the spectral redistribution information and gain relation). For more information type help background in XSPEC.

Table 2. Rigidity Dependencies and Sensor to Sensor Variations
There are a few things to notice in Table 2:
  1. The total counting rate of SIS-S1 is 12 percent larger than that of SIS-S0.
  2. The iron line intensity on S1 is about 5 times larger than it is on S0 while having very similar nickel intensities. This may indicate that the gold plating on the framestore shield of S0 is thicker than it is for S1. The iron and nickel lines most likely originate in the kovar (an Fe/Ni alloy) framestore shield which is plated with a nominal thickness of 2 microns of gold. The 1/e length of a 6.4 keV photon in gold is about 1.5 microns, so the transmitted intensity would vary greatly with variations of the gold plating thickness.
  3. The aluminum line intensity is about 2 times bigger on S0 than it is on S1. This may be due to variations in the thickness of gold plating on the aluminum camera bodies of the sensors.
  4. The fluorescent line intensity is about 2 times bigger for the lower magnetic rigidity data than for the higher rigidity data. This may just reflect that the lines are a result of particle- induced fluorescence and the particle intensity increases as the local magnetic rigidity decreases.
For most of the analysis we have done, we choose data taken with magnetic rigidity greater than 8 GeV/C. Table 3 shows the fit parameters for each of the individual CCDs with this rigidity selection. Again, the parameters apply to "/b" models found in XSPEC. Also, all lines are assumed to have a width (sigma) of 1eV. Line intensities are given in counts s-1CCD-1. We used the XSPEC model mo = powerlaw/b + gaussian/b + gaussian/b + gaussian/b + gaussian/b + gaussian/b + gaussian/b.

Table 3. Background Model Parameters for Dark Earth Data with Rigidity Greater than 8 GeV/C
Typical errors (90 percent confidence) on the powerlaw index and normalization are: +/- 0.1 and +/- 3x10^(-4). The errors on the bright lines (the Ka and La lines) are typically about +/- 1x10^(-4).

Temporal Stability of the Instrumental Background: For a typical observation, the local magnetic rigidity defines the short term intensity of the instrumental background. Although, the number of flickering pixels may change suddenly during passages across the night/day terminator of the orbit as well as near the bright Earth. This will be discussed elsewhere.

To address the long-term stability, we have looked at the intensity of the 8-12 keV spectrum from all of the deep survey fields observed with ASCA. The instrumental background dominates over the cosmic X-ray background in this energy range (Gendreau, et al. (1994)). The observations occurred at various times from March to September 1994. No significant variations in the 8-12 keV flux was measured.


We have presented a two component model to the instrumental background on the SIS. This model should hold for any CCD used in frame transfer mode. Two important results of this model are that the measured instrumental background rates should change with clocking mode and that there should be a gradient in the instrumental background with row. This model can be exploited to try and understand the spectral differences of the instrumental background on the framestore array and imaging array. This type of analysis still needs to be done. A knowledge of the differences could be used to evaluate schemes for reducing the background by modifying the framestore array. Some modifications could be made with the SIS now on ASCA (at the cost of extra calibration), while others would be reserved for future missions with CCDs. Possible modifications include:
  1. Changing the clock voltages on the framestore array to reduce the quantum efficiency of the array for particle induced events. This is currently possible with the SIS, but some recalibration may be necessary. This may also modify hot pixel (flickering pixel) growth rates.

  2. Modification of split threshold level and grades in event detection. The pixels are smaller and shaped differently in the framestore region leading to differences in the grading of events. This is partly doable with the SIS, but would require a huge recalibration effort.

  3. Modification of pixel shapes and sizes in the framestore array to enhance rejection by grade of particle induced events. This would be something to consider for future missions. A possible modification would be to make the framestore pixels really small so that even X-rays produced by particle interactions in materials near the framestore array appear as large events.

  4. Modification of the framestore shield to reduce unwanted background features. Again, this would be something to consider for future missions. Possible modifications would be to increase the gold plating thickness so that the iron and nickel lines do not come through (if kovar is still used).

Finally, we must consider where the instrumental background stands compared to the cosmic X-ray background. Figure 6 (Gendreau et al, 1994) shows the data from several deep field observations made during the ASCA PV phase. The models fitted to the dark Earth data of Figure 4 are drawn into Figure 6. For energies less than about 5 keV, the sky background is more important than the instrumental background. Although, the aluminum K line at 1.5 keV might pose a problem in some cases.

For a future report, more dark Earth data will be collected in 1-, 2-, and 4-CCD mode. This data may come from the dark Earth crossings during the various AO cycles. Particular emphasis will be made on 1- and 2-CCD mode data since the growth in the number of flickering pixels will be causing telemetry saturation problems for 4-CCD mode observations.

Figure 6: The spectrum of the Cosmic X-Ray background shown against the instrumental background. All data taken with magnetic rigidity greater than 8 GeV/C.


Burke, B. E., Mountain, R. W., Harrison, D. C., Bautz, M. W., Doty, J. P., Ricker, G. R., and Daniels, P. J. 1991, IEEE Transactions on Electron Devices, 38, 5, 1069-1076.

Cash, W., 1979, ApJ, 228, 939-947.

Gendreau, K. et al, 1994, submitted to PASJ.

Gendreau, K., Bautz, M. and Ricker, G., 1993, Nuc. Instr. and Meth. in Phys Res A, 335, 318-327.

Ricker, et al. 1994 (in preparation).

Next Proceed to the next article Previous Return to the previous article

Contents Select another article