Characterization of Flight Detector Arrays for the Wide-field Infrared

Characterization of Flight Detector Arrays for the Wide-field Infrared Survey Explorer Amy Mainzera, Mark Larsenb, Maryn G. Stapelbroekc, Henry Hoguec, James Garnettd, Majid Zandiand, Reed Mattsone, Stacy Masterjohnc, John Livingstona, Nicole Lingnera, Natali Alsterf, Michael Resslera, Frank Mascig a Jet Propulsion Laboratory/California Institute of Technology 4800 Oak Grove Dr., Pasadena, CA USA 91109 b Space Dynamics Laboratory/Utah State University, Logan, UT 84341 USA c DRS Sensors & Targeting Systems, 10600 Valley View St. Cypress, CA 90630 USA d Teledyne Imaging Sensors, 1049 Camino Dos Rios, Thousand Oaks, CA 91360 USA e MOSET Corp. f University of California, Berkeley, 366 Le Conte Hall, Berkeley, CA 94720 USA g Infrared Processing and Analysis Center/California Institute of Technology 770 S. Wilson Ave. Pasadena, CA 91125 USA ABSTRACT The Wide-field Infrared Survey Explorer is a NASA Midex mission launching in late 2009 that will survey the entire sky at 3.3, 4.7, 12, and 23 microns (PI: Ned Wright, UCLA). Its primary scientific goals are to find the nearest stars (actually most likely to be brown dwarfs) and the most luminous galaxies in the universe. WISE uses three dichroic beamsplitters to take simultaneous images in all four bands using four 1024x1024 detector arrays. The 3.3 and 4.7 micron channels use HgCdTe arrays, and the 12 and 23 micron bands employ Si:As arrays. In order to make a 1024x1024 Si:As array, a new multiplexer had to be designed and produced. The HgCdTe arrays were developed by Teledyne Imaging Systems, and the Si:As array were made by DRS. All four flight arrays have been delivered to the WISE payload contractor, Space Dynamics Laboratory. We present initial ground-based characterization results for the WISE arrays, including measurements of read noise, dark current, flat field and latent image performance, etc. These characterization data will be useful in producing the final WISE data product, an all-sky image atlas and source catalog. Keywords: infrared, survey, detectors 1. INTRODUCTION 1.1. WISE mission We describe here the preliminary results of characterization testing of the infrared detector focal plane arrays (FPAs) selected to be flown on board the Wide-field Infrared Survey Explorer (WISE) mission.1,2,3 Scheduled for launch in November, 2009, WISE will conduct an all-sky survey in the wavelength range from 2.8 to 26 microns with sensitivity up to three orders of magnitude beyond that achieved in 1983 by the IRAS survey. Science goals for WISE include a wide variety of studies ranging from the discovery of the nearest star to our Sun to the detection of the most luminous galaxies in the universe. Asteroids, protoplanetary debris disks, the structure of our Milky Way galaxy, and the star formation history of normal galaxies are also among the many WISE science targets. WISE will employ a scan mirror moving in a sawtooth pattern to “freeze-frame” sections of sky on the detectors while the spacecraft slews with an opposite velocity. The arrays will integrate during the 8.8 sec frame time and will be read out and reset during the 1.1 sec scan mirror flyback period. The entire six month survey will be conducted in this single mode, greatly simplifying test requirements both for operations and detector characterization. Further information on WISE can be found at http://wise.ssl.berkeley.edu/.

High Energy, Optical, and Infrared Detectors for Astronomy III, edited by David A. Dorn, Andrew D. Holland, Proc. of SPIE Vol. 7021, 70210X, (2008) · 0277-786X/08/$18 · doi: 10.1117/12.789585

Proc. of SPIE Vol. 7021 70210X-1 2008 SPIE Digital Library -- Subscriber Archive Copy

1.2. WISE arrays Using a solid-hydrogen-cooled telescope with a 40-cm primary mirror, WISE will survey the entire sky in four photometric bands with nominal wavelengths centered at 3.3, 4.7, 12, and 23 µm (hereafter referred to as bands W1 – W4). The four bands are imaged simultaneously using three beamsplitters and four 1024x1024 FPAs. WISE will be the first space mission to fly infrared detector arrays of this size. The two shortest wavelength channels, W1 and W2, use mercury-cadmium-telluride (HgCdTe) detector arrays manufactured by Teledyne Imaging Sensors of Thousand Oaks, CA that are indium bump-bonded to Teledyne’s HAWAII-1RG cryogenic readout. Teledyne delivered the flight focal planes and flight spare to DRS Sensors & Targeting Systems of Cypress, CA. The two long wavelength channels (W3 and W4) employ arsenic-doped silicon (Si:As) Impurity Band Conduction (IBC) or Blocked Impurity Band (BIB) detector substrates that are indium bump-bonded to readout/multiplexer circuits that were developed especially for the WISE program. The Si:As FPAs were manufactured by DRS, along with the cables, mechanical packaging and electronics for all four focal plane module assemblies (FPMAs; see Figure 1). The HgCdTe arrays are sensitive from ultraviolet wavelengths out to ~5.4 µm; the Si:As arrays are sensitive to wavelengths as long as 28 µm. All four WISE arrays have been anti-reflection coated to maximize their in-band responses. The nominal operating temperature for the HgCdTe arrays on WISE is 32 K; their temperature will be maintained by using active heaters. The Si:As will operate at 7.8 K. This temperature is passively set by making a good thermal connection between the Si:As arrays and the primary tank of the WISE hydrogen cryostat. Table 1 summarizes the measured WISE detector performance for all four bands. Parameter Wavelength Range (µm)

HgCdTe Performance (W1 and W2) 2.8 – 3.8 (W1) 4.1 – 5.2 (W2) 32 ± 0.1 1024x1024 >70

Si:As Performance (W3 and W4) 7.5 – 16.5 (W3) 20 – 28 (W4) 7.8 ± 0.5 1024x1024 >60

Operating Temperature (K) Array Format Quantum Efficiency (mean over band, with AR coating) (%) 18 18 Pixel Pitch (µm) Pixel Operability >90% >90% Dark Current (mean, @ operating 100,000 Power Dissipation (mW) 6.7 3.7 Outputs 16 4 Table 1: HgCdTe and Si:As FPA measured parameters. The need to operate the Si:As arrays at 7.8 K necessitated the development of a new cryogenic multiplexer circuit in the WISE 1024x1024 format. This new readout was developed and manufactured by DRS Sensors & Targeting Systems. At present, all four focal planes have been delivered to the WISE payload contractor, Space Dynamics Laboratory/Utah State University, along with two flight spare arrays (one of each type). The arrays have undergone extensive characterization testing prior to delivery, and more testing at the payload level of integration is planned4. The arrays have been integrated into the WISE beamsplitter assembly (BSA; Figure 2), which in turn has been integrated into the WISE telescope assembly. In addition, many characterization tests have been performed at the NASA-Ames Infrared Laboratory using two non-flight Si:As arrays, one of which was chosen from the same detector wafer as the flight units5. In this work, we describe the results of tests that have been done on the four flight arrays. The results of these tests are being used to initialize the data processing pipeline being written by the WISE Science Data Center (WSDC) at the Infrared Processing and Analysis Center at Caltech. Many of the data products described below will be updated using on-orbit data, but in some cases (e.g. dark current measurements for W3 and W4), the ground-based tests will yield the best results, emphasizing the need for a thorough pre-launch array characterization program.

Proc. of SPIE Vol. 7021 70210X-2

2. DESCRIPTION OF DEVICES The HgCdTe hybrid detector arrays to be flown in the WISE system have been adapted from those being produced for the James Webb Space Telescope (JWST). The WISE arrays are using Teledyne Imaging Sensor’s HAWAII-1RG readout/multiplexer, which is bonded to the detector material using Teledyne’s balanced composite structure to account for the different thermal expansion coefficients between detector and readout. The HgCdTe arrays are each read out through 16 outputs arranged in vertical stripes with an ~60 kHz pixel clock rate. The ratio of mercury to cadmium (the “x” in HgxCd1-xTe) was varied to produce cutoff wavelengths appropriate to the respective W1 and W2 bandpasses. In general, lower cutoff wavelengths result in lower bulk current densities within the detector material, yielding better noise and dark current performance.6 The cutoff wavelength of the W1 array was set to 4.2 µm, appropriate for its 2.8 – 3.8 µm bandpass, while the W2 array cuts off at 5.4 µm, suitable for its 4.1 – 5.2 µm bandpass. The flight spare array also cuts off at 5.4 µm so that it can be used for either W1 or W2 if necessary. Because the HgCdTe arrays have had their CdZnTe substrates removed, the arrays are sensitive from their cutoff IR wavelengths into the ultraviolet. To avoid carrier freeze-out, the HgCdTe FPMAs are actively heated to 32 K. The arrays are thermally isolated from the hydrogen-cooled beamsplitter assembly by means of a G-10 composite stage assembly. The WISE payload is not designed to heat the HgCdTe arrays to the FPA ~300 K that would be necessary to thermally anneal them.

—sIege Assembly

Base Assembly

I

I

I

Figure 1: FPMA construction (left), and an actual flight HgCdTe array (W1; right).

p—Band 2 Filter (sapphire)

Band 4 Filter (Si)

,—Dichroic Beam Splitter BS2 (Si)

-e

The Si:As arrays for WISE were developed specifically for this program, building on the experience gained from other programs such as the Spitzer Space Telescope’s 128x128 pixel Si:As BIB detector arrays.7 The relatively high background that WISE will see due to zodiacal emission implies that relatively relaxed requirements on dark current and read noise are acceptable. The Si:As array temperature of 7.8 K has been set to optimize dark current performance and linearity. The arrays are each divided into four output quadrants which are read out from the corners to the center at ~240 kHz. It is possible to anneal the Si:As arrays to 20 K using a set of redundant heaters.

2.1. Data Collection Method WISE will survey the entire sky in a single operating mode. WISE will employ a sample-up-the-ramp (SUTR) data collection scheme that is synchronized Band 3 Filter (ZnSe) J with the scan mirror scan pattern. The instrument will Dichroic Beam Splitter BS3 (Si) collect nine samples, each 1.1 seconds in duration, up the ramp for each frame and will digitize the samples. Dichroic Beam Splitter BS1 (lnSb) A saturation bit will be set if the first sample triggers — Band 1 Filter (sapphire) an adjustable threshold value. The individual samples’ weightings can be varied for each of the four Figure 2: Beam splitter assembly arrays if necessary. A slope is fit to the samples, reducing the data rate by a factor of ~9 prior to downlink. For the HgCdTe arrays, an anomalous first sample effect was found, resulting in the first coefficient being set to zero weight. All nine samples will be used for the Si:As arrays. The relative simplicity of the WISE data collection scheme and single nominal operating mode significantly reduces the complexity of the test program necessary for the WISE detector arrays. As part of this test program, we have evaluated


the usual detector performance parameters such as dark current, read noise, responsivity, response to radiation, latent image behavior, etc. The results of these measurements to date are described in the following sections.

1.

-20

0

I

20

IL 40

60

60

Figure 3: Dark frames were measured at DRS using the engineering unit electronics and flight cables. Top left = W1; top right = W2; bottom left = W3; bottom right = W4. These data were collected at the WISE sampling cadence of nine 1.1 second samples. Dark frames were computed CDS. The number of high dark current pixels can be seen to increase between bands W1 and W2; this is caused by the increasing cutoff wavelength of the HgCdTe material, although both arrays easily meet their dark current requirements. The banding offsets in bands W3 and W4 Si:As arrays can been seen.

2.2. Dark Current


0

111111111 111111111 111111111

N

0 0

x—

0

(0

x—

liii'''

0 x— 0

111111111

-x

DNk Curreni Sign (elections)

—

WISE does not have a shutter in its payload. However, prior to ejection of the aperture cover approximately two weeks after launch, bands W1 and W2 should be quite dark, although bands W3 and W4 will be saturated due to the ~90 K cover temperature. This will allow us to collect measurements of the dark current for W1 and W2 on-orbit. For W3 and W4, however, we must rely on ground-based measurements alone to determine the dark offset maps since the zodiacal background will always be significantly higher than the dark current. In order to both assess the performance of the flight arrays and to ensure that we have the data products necessary to complete the WISE survey, the dark currents of all four flight arrays and flight spares have been measured at DRS using engineering unit electronics and flight cables (see Figure 3). Several data sets have been collected, including measurements of the dark current with varied operating temperatures and detector bias voltages. We have also assessed dark current stability by collecting four hours of dark frames taken in the WISE standard data collection cadence. Measurements of the dark current are underway now at SDL using the flight analog and digital electronics and cables. Analysis of the effects of non-uniform dark current on point source photometry results in a requirement on dark current stability. Although we can tolerate relatively high dark currents due to the zodiacal backgrounds, particularly in bands W3 and W4, changes to the dark current structure can adversely affect photometric accuracy. In order to have 600 e-/sec. It was determined that there are burst noise in the HAWAII-1RG mutliplexer. ~6700 pixels with dark current exceeding 5 e/sec for the W1 array and ~10,200 pixels for the W2 array. Due to the ~1.5% inter-pixel capacitance noted in the HgCdTe arrays, it is also necessary to reject any pixel that is immediately adjacent to one with dark current >300 e-/sec. Figure 4: HgCdTe pixels with extremely high dark current (such as this one on the W1 array) exhibit non-linear behavior.

IL?

TliI1

J


Table 2 summarizes the variations shown in dark current and read noise with temperature and detector bias voltage for the HgCdTe arrays. These measurements were made at DRS prior to delivery to the payload with test electronics; they will be repeated at the payload level using flight electronics and cables. In all cases, the data were collected with the WISE integration time (8.8 sec per frame) and sampling cadence (nine samples up the ramp). Sixteen frames were collected at each test condition. Because we know from the 1000 sec ramp measurements that the average dark current on the HgCdTe arrays is extremely low (~0.013 e-/sec), the total dark current signal collected during these short exposures (0.13 e-) is much lower than the noise (~16 e- CDS) for a single pixel. We therefore fit a Gaussian to the histogram of all 1024x1024 pixels in the array; its mean is taken to be the mean dark current signal. We can also see how many pixels will fail to meet our dark current specification of 5 e-/sec as a function of temperature and bias. Band

Temperature (K)

Bias (V)

Mean Dark Current (e-/sec)

W1 W1 W1 W1 W1 W2 W2 W2 W2 W2

32 30 34 32 32 32 34 31 32 32

0.250 0.250 0.250 0.200 0.300 0.250 0.250 0.250 0.200 0.300

0.05 1.66 0.00 -0.08 0.19 -0.38 -0.59 -0.61 -0.75 -0.44

# of Pixels with Dark Current >5 e/sec 30,918 160,945 31,484 27,175 31,955 58,232 60,107 59,647 41,259 99,901

Mean CDS Read Noise (e-) 15.8 16.0 16.1 15.8 15.8 15.4 15.6 15.5 15.7 16.2

Table 2: Dark current and read noise were measured at the nominal, upper, and lower limits of the temperature and detector bias voltage ranges.

2.2.1.1. RTS noise/signal HgCdTe arrays are known to suffer from an effect known as random telegraph signal (RTS; also known as popcorn noise or burst noise).8 Burst noise originates in the HAWAII-1RG multiplexer, and it is characterized by sudden current level shifts in either the positive or negative directions, and the magnitude of the shift can be variable. The cause of burst noise is thought to be due to the tunneling of charge trapped in defects through the barrier layers of the multiplexer, resulting in a change in output signal voltage. It can be attributed to the multiplexer because the phenomenon can sometimes be observed in the reference pixels, which are completely unconnected to the detector material. Identifying pixels that are strongly affected by burst noise is important for WISE since burst noise results in an unstable dark current signal. Figure 5 shows an example of a pixel in the W1 array that was found to have significant burst noise. We have completed a preliminary analysis of the burst noise in the WISE arrays by examining collections of 1000 sec duration dark frames and looking for pixels with high standard deviations (after excluding pixels with high dark current). We eliminated the possibility of mistaking cosmic rays for burst noise by implementing an algorithm to search the individual ramps for sudden increases in signal on a single ramp only. We found that ~1% of the pixels in the W1 array were affected by burst noise that resulted in a dark current instability larger than the DId < 0.3 e-/sec requirement. However, it will be necessary to compare this to dark data taken at the 8.8 sec WISE frame time before making a final determination of which pixels should be considered unusable. We are still evaluating the number of pixels affected by RTS in the W2 array. 2.2.2. Si:As Arrays The dark current requirements for both Si:As arrays are relaxed (3 e-/sec from the pre-dose level; this fraction dropped to 8% following a thermal anneal to 20 K. None of the pixels which were formerly good had dark currents >100 e-/sec after the dose. While this is a relatively large number of pixels, it is hoped that since the 2 krad(Si) represents the expected lifetime dose of the WISE mission, the actual day-to-day degradation of the dark current will be more gradual. The WISE radiation environment will be dominated by passages through the South Atlantic Anomaly (SAA) several times a day on average. To simulate this, the Ames group applied a “SAA-equivalent” dose of 0.5 rad(Si) over one minute prior to dosing the array with the full 2 krad(Si). The resulting dark frames collected after the 0.5 rad(Si) dose show that only ~0.9% of all formerly good pixels showed changes in dark current >3 e-/sec. We conclude from these tests and analyses that it will be necessary to track each pixel through time to look for evidence of systematic changes to the dark current; however, this is implemented as a standard part of the WISE data reduction pipeline. Band

Temperature (K)

Bias (V)

Mean Dark Current (e-/sec)

# of Pixels with Dark Current >100 e-/sec

Mean CDS Read Noise (e-)

W3 W3 W3 W3 W3 W4 W4 W4 W4 W4

7.8 7.3 8.3 7.8 7.8 7.8 7.3 8.3 7.8 7.8

2.0 2.0 2.0 1.7 2.3 2.0 2.0 2.0 1.7 2.3

11.6 9.5 16.5 10.2 13.4 14.0 10.9 26.7 10.0 19.1

65 56 80 53 83 609 587 744 423 583

103.5 105.4 101.1 103.6 103.7 101.6 103.6 99.3 100.2 102.2

Table 3: Dark current and read noise were measured at the nominal, upper, and lower limits of the temperature and detector bias voltage ranges.

2.3. Read Noise Although the WISE arrays will be read out in SUTR mode, noise specifications were given in correlated double sample (CDS) mode for simplicity during testing. The WISE SUTR algorithm will reduce the CDS noise by ~20% on average. The CDS read noise for each pixel was computed by taking the standard deviation of a set of 42 dark frames collected at the WISE sampling cadence; the slope of each frame was computed by subtracting the first sample from the last in the ramp (using the second and last samples for the HgCdTe arrays, which have an anomalous first sample). The CDS noise measured for the HgCdTe arrays at DRS was ~16 e-. More recent measurements at SDL using the integrated flight focal plane electronics box and flight cables suggest that the CDS noise for both HgCdTe arrays will be ~19 e-. For both Si:As arrays, the CDS read noise is ~103 e-. 2.4. Responsivity We have conducted preliminary assessments of the responsivity of all four flight arrays plus the flight spares at DRS. Pixel-to-pixel relative responsivity is measured by uniformly illuminating the arrays. As the light sources used to illuminate the arrays were not completely uniform on large spatial scales, we were only able to obtain lower limits on the responsivity. These flat field measurements will be repeated during tests at SDL with a test chamber capable of creating a much more uniform illumination pattern with a blackbody source. Nonetheless, the results in hand to date


from the testing at DRS indicate that all flight and flight spare arrays meet their relative response uniformity and pixel operability requirements. The ground-based flat field measurements derived during payload testing will be used to initialize the WISE data processing pipeline. Every effort will be made to ensure that the illumination during flat field testing at SDL will be as uniform as possible. These ground-based flat field measurements should be adequate, at least on small spatial scales. It is possible to use the sky itself as an illumination source, particularly for bands W3 and W4, where the backgrounds due to zodiacal emission are high. For bands W1 and W2, we will need to combine hundreds of frames to measure the responsivity due to reduced emission from the ~300 K zodiacal dust at these wavelengths. These on-orbit measurements can be combined with the ground-based flats. Representative flat fields for all four flight arrays are shown in Figure 6. Flats were produced using six different illumination levels ranging from ~9% to ~50% of the full well depth. The median flat field values were found to be stable and invariant to these changes in illumination to within 0.4% for the W1 array; 0.7% for the W2 flight array; 0.2% for the W3 array; and 0.3% for the W4 array. These values were calculated by finding the mean of a Gaussian fit to the histogram of all pixels in the active areas. The response of the Si:As flat field to radiation was measured by the Ames group during proton testing at the U. C. Davis cyclotron.5 Following the 2 krad(Si) dose, the median pixel’s response decreased by 13%; however, thermally annealing the array returned the median response to within 4% of its original pre-radiation value. This indicates that corrections to the flat fields in bands W3 and W4 may be necessary following SAA passages; however, since the 2 krad(Si) dose applied was much larger than the anticipated 0.5 rad(Si) dose for each SAA passage, the effect on the flat fields may be much more gradual. 2.5. Bad Pixels The data processing pipeline written by the WISE Science Data Center requires bad pixel masks for all four arrays. While these masks will be updated on orbit by searching for anomalous behaviors in pixel coordinates, the pipeline must be initialized with masks computed from ground-based data. Our initial criteria for bad pixels are as follows, although these may change as the payload testing progresses: 1) Anomalous dark current (>5 e-/sec for HgCdTe arrays and >100 e-/sec for Si:As arrays or 38 e- CDS for HgCdTe arrays and >200 e- CDS for Si:As arrays). 3) Low relative responsivity (0.3 e-/sec for HgCdTe and >3 e-/sec for Si:As arrays). 5) For HgCdTe only: All four nearest neighbors of pixels with Id > 300 e-/sec (these pixels’ signals will likely be excessively high due to inter-pixel capacitance). Table 4 summarizes the net results when these criteria are applied to our flight array acceptance test data. Note that the total bad pixel count is not simply the sum of all individual bad pixels, as some pixels are considered bad for multiple reasons. We have demonstrated that all four of our flight arrays have exceeded our requirement of >90% pixel operability. Band W1 (136) W2 (147) W3 (26) W4 (37)

Low responsivity 2,478 3,719 2,035 2,079

Bad dark

Bad noise

Total bad pixels

10,393 11,669 64 674

6,222 10,180 124 382

17,563 67,890 3,213 2.762

Bad (%) 1.8 6.8 0.3 0.3

pixels

Table 4: Preliminary bad pixel counts were developed using a number of different criteria.

Simulation software has been developed by the WSDC to model the effects of a particular bad pixel mask on the survey performance. This tool helped to optimize the choice of flight arrays by examining the average sky coverage that results when a particular bad pixel mask is imposed on a simulation of the WISE survey pattern.


2.6. Quantum Efficiency Quantum efficiency for both HgCdTe and Si:As arrays was determined by DRS using process evaluation chips (PECs) drawn from the same wafers as the flight arrays. The four arrays were each given anti-reflection (A/R) coatings to enhance their response in the respective bandpasses. The measured net quantum efficiencies exceed the requirements (>70% for HgCdTe and >60% for Si:As) for all four arrays. Since there is only one flight spare of each array type, it was necessary to choose between coating types for the spares. The HgCdTe flight spare was given the same coating as band W2, which will result in a 5% net loss of throughput if this array must be used for W1. We chose a W4 coating for the Si:As flight spare array to minimize throughput losses to band W4 as band W3 currently has more sensitivity margin to spare.

I

-1

Figure 6: Representative flat fields were computed for all four flight arrays; top left = W1; top right = W2; bottom left = W3; bottom right = W4. The large-scale variations observed in all four flat fields are the result of nonuniform illumination rather than real non-uniform responsivity in the arrays.


Dark Current (e-/sec)

2.7. Banding During dark measurements of the Si:As arrays, a banded appearance to the dark frames was noticed (Figure 3). This banding appears as an offset change that starts and stops on small clusters of defective pixels with high dark currents. Initial tests have shown that the banding appears to be stable to within 1-2 e-/sec from one exposure to the next over a period of several hours (Figure 7). If the pattern is fixed, it should subtract out from the on-sky images. The offsets are believed to result from capacitive coupling between some unused circuitry in the multiplexer and the output. It may be possible to subtract out the banding pattern by using the arrays’ reference pixels as described below. 14

12 ——

A

6

I

I

I

I

fr:

k

4

I

501

I

I

I

I

I

I

1001

I

I

I

I

I

I

1501

Frame Number Figure 7: The banding pattern observed on the WISE W3 array appears to be stable to within 1-2 e-/sec over a period of many hours, well within our stability requirements.

2.8. Reference Pixels Both types of hybrids were designed with four columns and four rows of reference pixels around the outermost edges of the hybrids. The reference pixels are simply ‘pixels’ that lack indium bumps to connect the detector material to the multiplexer unit cells beneath them. The signals from these unconnected pixels are therefore representative of the arrays in the absence of illumination, the main difference being that their capacitance is lower without the detector material. For the Si:As arrays, the reference pixels’ noise is approximately 80 e- (CDS), compared to the 103 e- found in the active area; for the HgCdTe arrays, the reference pixels have ~7-10% lower noise than the active area.

We have made a preliminary assessment of how the data processing pipeline should make use of the reference pixels. In the case of the HgCdTe arrays, we find that using the 4x1024/16 = 256 reference pixels at the top and bottom of each of the 16 outputs results in an increase in effective read noise, so they should not be used in general. For the Si:As arrays, we can use the row by row reference pixels to attempt to remove the “banding” offsets. In general, the offset between the reference pixels in a given band region and the adjacent active area is constant with time. Each quadrant must be considered separately (see Figure 3). Once the offsets are known between the reference pixels and the banding regions, they can be removed. However, since the banding offsets start and stop on defect clusters in the array, which occur at random locations, this method may have to be adjusted on these rows. It does not appear to be helpful to use the reference pixels along the top and bottom of the Si:As arrays. 2.9. Latent Images Preliminary tests of the latent image behavior of the Si:As arrays have been performed using both flight and non-flight FPAs. These initial tests have shown that the Si:As arrays will have latent images caused by bright sources; we are in the process of characterizing the phenomenon. A more detailed set of tests are planned at the payload level of integration. 2.9.1. Annealing Thermal annealing has been shown to be an effective means of removing latent images and the effects of radiation damage from Si:As BIB detectors. We have the capability to thermally anneal the Si:As arrays by heating them to 20 K for approximately two minutes. This can be done approximately twice per day without significant loss to WISE’s cryogenic lifetime. Once the arrays are heated to 20 K, we must wait until the arrays reach thermal equilibrium to resume data collection. We plan to execute the anneals during downlinks to minimize time lost from the survey. Thermal models have shown that it will take ~20 minutes for the array temperature to return to 7.8 K at the beginning of the mission when the cryogen tanks are full; this time will increase slightly as the hydrogen tanks deplete over the six


month mission lifetime. Tests on a non-flight Si:As array at NASA-Ames have shown that annealing can reduce the effects of radiation on both the dark current and the responsivity as described above. We are currently evaluating the effects of annealing on latent images. 2.10. Radiation Effects WISE will orbit the Earth at an altitude of ~525 km. The primary radiation source in this environment will be protons encountered during passage through the South Atlantic Anomaly, which will occur several times per day on average. The WISE HgCdTe arrays are smaller counterparts of the mosaic of 2048x2048 HgCdTe arrays planned for the James Webb Space Telescope. The JWST program has conducted extensive radiation tests of these arrays to 5 krad(Si), well in excess of the expected WISE total lifetime dose of 2 krad(Si) over the six month nominal mission.9 McKelvey et al. (2005) report that no significant changes to the read noise were observed after this dose, and the median pixel responsivity changed by no more than a few percent. The output voltage of the HAWAII-2RG source-follower was shown to remain stable up to doses as high as 50 krad(Si). Approximately 10% of the irradiated array’s pixels were defined as activated following dosage; however, JWST’s dark current requirements (~