As part of its Earth Observation programs, the European Space Agency (ESA; Paris, France) is preparing the Living Planet Earth program. This new mission will include the Surface Processes and Ecosystems Changes Through Response Analysis (SPECTRA) system, which will use a hyperspectral imaging spectrometer and a bi-spectral thermal imager to observe the Earth's surface.1 The scientific objective of the SPECTRA mission is to describe, understand, and model the role of terrestrial vegetation in the global carbon cycle and its response to climate variability under the increasing pressure of human activity. The mission will generate detailed observations that will offer researchers new opportunities to derive sufficiently accurate representations of the terrestrial biosphere for use in global Earth systems models. Planning for the Future
To prepare such future spaceborne imaging spectrometers, a joint Swiss/Belgian ESA initiative yielded a project tasked with building an airborne imaging spectrometer that could represent a precursor mission. The project includes the development of the Airborne Prism Experiment (APEX), a dispersive push-broom imaging spectrometer that will contribute to preparation, calibration, validation, simulation, and application development for the SPECTRA mission, as well as to the understanding of land processes and interactions at a local and regional scale in support of global applications.2
The APEX spectrometer consists of a collimator that directs the light transmitted by an entrance slit toward a dichroic beamsplitter that separates the beam into the two spectrometer channels: visible/near IR (VNIR; 380 to 1000 nm) and shortwave IR (SWIR; 930 to 2500 nm).3 Two prisms, one for each waveband, disperse the beam into channels.
In the case of the SWIR wavelength range, the ESA could not identify off-the-shelf detectors able to meet the requirements of these missions.4 In 2000 the ESA commissioned Sofradir (Châtenay-Malabry, France) to launch the development of an appropriate SWIR detector.
The resultant detector is based on space-qualified mercury cadmium telluride (MCT/HgCdTe) technology.5 The technology offers the possibility of a SWIR focal plane array (FPA) operating at focal plane temperatures near 200K and above, depending on the cut-off wavelength; these operating temperatures are compatible with passive cooling systems, which are less expensive and more reliable than space-qualified Stirling coolers. In addition, MCT technology allows the detector cut-off wavelength to be customized for the application, which is not the case for other materials candidates like indium gallium arsenide, in particular for cut-off wavelengths greater than 2 µm.
Figure 1. A hyperspectral instrument images a scene over multiple, closely spaced spectral channels. The size of the cross-track swath combines with the pixel number/size to define spatial resolution.
Hyperspectral instruments provide images of an observed scene with a high number of spectral channels (typically more than 100) and with a high spectral resolution (typically 10 to 15 nm) in the considered waveband (see oemagazine, March 2004). These instruments incorporate 2-D detectors with spectral responses that fit the waveband of the instruments. One dimension of the detector (the x-axis) represents the spatial resolution of the system, which is driven by the relationship between the number/size of pixels and the swath, or the size of the cross-track scan. Generally, this dimension is larger than the orthogonal dimension (the y-axis), which represents the spectral resolution, with each pixel corresponding to a spectral channel (see figure 1).
The ESA derived the specifications for its SWIR detector by beginning with the Land Surface Processes and Interactions Mission (LSPIM) requirements and taking into account the APEX and SPECTRA needs. Based on this analysis, Sofradir developed a SWIR FPA of 1000 x 256 pixels with a pitch of 30 µm. This detector is used by the APEX mission and will be available for future hyperspectral missions like SPECTRA. The MCT Advantage
Figure 2. 1000 x 256 pixel detector features a 30-µm pitch.
The Earth sensing detector is based on two monolithic substrates (MCT detection module and readout circuit) hybridized by indium bumps (see figure 2). MCT material is produced by liquid-phase epitaxy on cadmium zinc telluride (CdZnTe), which is the standard Sofradir process. Specific and robust hybridization technology was used, which is based on a hot indium bump reflow process. During thermal testing of the FPA, the structure survived temperature cycling between 293K and 100K more than 1000 times without any degradation of the indium bump connection or the electro-optical performance.
The structure developed for this detector could be extended up to 4.5 cm x 1.5 cm, based on readout circuit possibilities. Such a detector would feature 1500 x 500 pixels at a 30-µm pitch or even more pixels with a reduced pitch size.
One of the main advantages of MCT material is the design flexibility offered in adapting the operating temperature and the cut-off wavelength of the detector to the application. The operating temperature can be increased to relax the cooling constraints or decreased to improve the performance as the application requires. The design performed has been validated for operation between 140K and 200K.
Starting from the typical spectral response of the detector, we can tune the detector cut-on and cut-off wavelengths precisely using a specific cold filter adapted for the required waveband. The cut-on wavelength of the detector is 0.78 µm and is driven by the bandgap of the CdZnTe substrate (see figure 3). This cut-on wavelength could be extended to the visible spectral region by removing the detector substrate after hybridization. In order to satisfy the spectral response requirements, the cut-off wavelength for the APEX detector has been tuned to 2.5 µm at 175K.
Figure 3. 1000 x 256 SWIR detector spectral response remains high across the waveband of interest.
For high-performance applications in the SWIR waveband, the detector dark current is one of the most critical parameters. In terms of dark current, our MCT technology performance is compatible with a full-performance detector operating temperature of at least 175K. Up to at least 175K, the average detector dark current remains below 0.1 pA (with a 900 µm2 pixel), which enables a detector compatibility with passive cooling systems for space applications. In the frame of the APEX development, the electro-optical performance of the detector has been fully validated (see table).6 We optimized the number of outputs as well as the maximum output rate in order to design a versatile FPA, which can answer several mission requirements with different conditions of operation. In addition to these characteristics, hyperspectral applications require specific functions concerning the control of the detector, including a function allowing the user to select the spectral lines to be output in order to deselect the lines corresponding to spectral bands not of interest, and a function to select the number of detector outputs in order to optimize the frame rate.
As presented in the table, we implemented two readout circuit gains (0.4 Me- and 1.6 Me-) to optimize the hyperspectral performance. From one spectral channel to another, the level of signal could be very different; the design allows the gains applied to be different between these channels. For each channel, the gain is selected to provide the best compromise between the maximum charge to be stored and the signal-to-noise ratio. A good figure of merit Fm is the ratio between the maximum storable charge (Qmax) and the detector noise:
where BRoic is the silicon circuit readout noise (rms) and BPD is the noise of the MCT photodiode (rms). When the level of signal is high, the best tradeoff is to maximize Qmax; because BPD is larger than BRoic, the lower gain is the best choice. When the level of signal is low, the best tradeoff is to minimize BRoic using a low Qmax; because BPD remains low, the higher gain is the best choice.
The measured linearity of the detector in function of the input flux shows good performance even at very low flux levels: The device achieves better than 99.9% linearity over a dynamic range from 5 to 95%. This is due in particular to the implementation of a capacitance transimpedance amplifier (CTIA) input stage.
Not surprisingly, detector readout noise varies as a function of the different gains available (0.4 Me- and 1.6 Me-). With the higher gain level (0.4 Me-), the mean readout noise of the detector is 547 µV, which corresponds to 103 e- noise, with a standard deviation of 92 µV (16.8%). With the lower gain level (1.6 Me-), the mean readout noise of the detector is 535 µV, which corresponds to 435 e- noise, with a standard deviation of 91 µV (17%). These values are perfectly compatible with the predicted performance and are low for the maximal charge the readout circuit can store and for the use of a CTIA with a very high linearity. Regarding defective pixels, if our criteria for defective pixels consists of those that are saturated or that have a response lower than 50% of the average response, the typical number of defects found in our detectors is fewer than 100 out of 256,000 pixels.
For the APEX application, we integrated the detector in a Dewar and cooler assembly that enables the customer to have an integrated and autonomous product. Flight models will be delivered at the end of 2004.
The Earth-sensing IR detector meets the needs of hyperspectral spectrometers and space applications very well. The device is versatile and is able to meet directly, or with minor adaptations, many different needs. Furthermore, this detector also can be an excellent answer for other high-performance applications. oe
J.Y. Labandibar et al., Proc. SPIE 5234, p. 232 (2003).
M. Schaepman, K. Itte, and M. Ras, Geographica Helvetica, Vol. 58, p. 120 (2003).
M. Schaepman et al., Proc. SPIE 5234, p. 202 (2003).
T. Dartois et al., Proc. SPIE 4130, p. 547 (2000).
P. Tribolet et al., Journal of Crystal Growth 184/185, p. 1262 (1998).
P. Chorier and P. Tribolet, Proc. SPIE 4540, p. 328 (2001).
Seeing in a New Light
It was during his year-long military service after college that Philippe Tribolet first got involved in the development of IR detectors. Tribolet graduated in 1982 from Ecole Supérieure de Mécanique et d'Electricité (College of Mechanics and Electricity; Paris, France) as an engineer specializing in electronics and software. The French military assigned him to work as a researcher at the French government's Infrared Laboratory in Grenoble, and when his tour of duty was finished he stayed on because he found the field interesting. His research centered on photovoltaic IR detectors. Working at the laboratory also allowed him to stay close to the small town of Tullins where he was born and where he enjoyed trekking through the local mountains.
After three years in the laboratory, Tribolet joined the newly formed IR detector company Sofradir (Châtenay-Malabry, France), which was founded in 1986. "It was an exciting period and a very tough job," he says.
Tribolet's military experience came in handy at Sofradir, which makes IR systems for military, aerospace, and commercial applications. For instance, he became manager of a program to develop an IR detector for the U.S. Army's Line-of-Sight Anti-tank Weapon, which earned him an award from the U.S. Army in 1994.
Tribolet worked his way up through Sofradir, becoming chief technical officer in 2001. In this position, he manages 40 people in different technical fields and oversees research and development projects for both French and international customers, from the U.S. Army to the European Space Agency. Along the way, he's published more than 40 research papers in the field.
"I enjoy the infrared field," he says, "And I believe that the number of infrared applications will continue to increase exponentially, including for daily life."
Philippe Tribolet is technical director/chief technical officer of Sofradir, Châtenay-Malabry, France.