ebrary.net
Open in
urlscan Pro
5.45.72.163
Public Scan
URL:
https://ebrary.net/205103/engineering/chip_stepped_fabry_p_rot_filters
Submission: On September 03 via manual from IL — Scanned from NL
Submission: On September 03 via manual from IL — Scanned from NL
Form analysis
1 forms found in the DOMhttps://ebrary.net/2/search
<form action="https://ebrary.net/2/search" id="cse-search-box" accept-charset="UTF-8">
<div><input type="hidden" name="cx" value="partner-pub-9237573666471195:3011315463"><input type="hidden" name="cof" value="FORID:10"><input type="hidden" name="ie" value="UTF-8"><input type="text" name="q" placeholder=""
style="background: url("https://www.google.com/cse/static/images/1x/en/branding.png") left 9px top 50% no-repeat rgb(255, 255, 255);"></div><input name="siteurl" value="ebrary.net/205103/engineering/chip_stepped_fabry_p_rot_filters"
type="hidden"><input name="ref" value="" type="hidden"><input name="ss" value="" type="hidden">
</form>
Text Content
< Prev CONTENTS Source Next > ON-CHIP STEPPED FABRY-PÉROT FILTERS Fabry-Perot interferometer-based filters have been adopted to transmit the selected spectral bands of interest. It is named after Charles Fabry and Alfred Perot, who developed the filter in 1899 (Fabry and Perot 1899). A Fabry-Perot filter consists of two tightly spaced parallel mirror surfaces, i.e., an etalon and an interferometer, between which incoming light reflects and interferes (Figure 1.10). This interference allows some wavelengths to pass, while others are filtered out. Its transmission spectrum, as a function of wavelength, exhibits peaks of large transmission corresponding to resonances of the etalon. By carefully controlling the distance between the reflecting mirror surfaces, a Fabry-Perot filter can be designed to transmit only a narrow-wavelength band wanted. A practical solution for fast, compact, user-friendly and low-cost hyperspectral imaging sensors can be achieved by monolithically integrating optical interference filters on top of complementary metal-oxide-semiconductor (CMOS) based image sensors. These integrated filters can be made by traditional semiconductor fabrication tools, enabling mass production and. therefore, low-cost manufacturing. In order to integrate Fabry-Perot filters on top of a CMOS detector array, the design of the filters requires the selection of CMOS-compatible materials and suitable fabrication steps. However, the process steps and materials used for post-processing the filter structures must be carefully selected. They are crucial for the quality of the Fabry-Perot filters but may not damage the underlying CMOS detector arrays. Geelen et al. (2013) reported that Fabry-Perot filters in a wavelength range of VNIR (400-1000 nm) have been processed directly on top of the wafer that contains silicon-based CMOS image sensor chips. Figure 1.11 shows the Fabry-Perot filters for a line scan imager. Each row of a CMOS detector array is integrated with a row of same band-pass Fabry-Perot filters. Each Fabry-Perot filter has a fixed distance between the two reflecting mirror surfaces for a particular band-pass. The use of monolithically integrated Fabry-Perot filters on top of a detector array is an innovative approach to combining spectroscopy with imaging technology for hyperspectral sensors. This can result in low cost, compactness, and high speed. As a natural evolution of this innovative technology, the wavelength range of the CMOS-based Fabry-Perot filters has been extended to shortwave infrared (SWIR): 1000-1700 nm, together with InGaAs-based detector arrays in 2018 (Imec 2018). FIGURE 1.11 Fabry-Perot filters in a wavelength range of visible and near-infrared (400-1000 nm) integrated on top of silicon-based CMOS image sensor chips at the wafer level. This new integration approach is unique and has attractive advantages: First, combining the production of filters and detector arrays into one CMOS-compatible process leads to an overall simplification and cost reduction, and enables massive production. Second, the monolithic integration induces less cross-talk between neighboring bands and reduces stray light in the system. This also has a positive effect on the system’s sensitivity and speed. The upper part of Figure 1.12 shows a silicon-based CMOS detector array of format 1088 rows by 2014 pixels working in a wavelength range of VNIR. The same band-pass Fabry-Perot filters are integrated on top of the detector pixels in a particular row' of the detector array w'hen the detector array was fabricated at the wafer level. The spectral band-pass of the filters is controlled by the distance between the two parallel mirror surfaces (i.e., the thickness). The same band-pass Fabry- Perot filters can be deployed on multiple row's of the detector array. In the figure, the same band-pass filters (having the same thickness) are integrated on top of every six adjacent rows of pixels. These multiple row's of pixels have the same thickness (or height) of deposited filters and look like a step of a stair. This is why this kind of Fabry-Perot filters on top of a detector array is also referred to as on-chip stepped filters (OCSF), as illustrated in the low'er part of Figure 1.12. The configuration of the OCSF is a series of filter rows disposed one behind the other in the along-track direction that allow's to create a pushbroom hyperspectral imager same as in a LVF- based hyperspectral imager. Most importantly, this configuration provides the choice to sense objects on the ground using multiple detector rows (such as 6 rows) of the same spectral band-pass to boost the SNR using the on-chip time-delayed integration (TDI) or off-chip data binning in the spectral dimension. The VNIR detector array with the OCSF shown in Figure 1.12 has 181 spectral bands spanning a wavelength range from 450 nm to 960 nm with a spectral sampling interval of 5 nm. The FWHM of each spectral band is about 15 nm. The width of a step is 6 detector row's (these rows are deposited the same Fabry-Perot filters). The OCSF can be adjusted to the center pixel to an accuracy of about ±1 nm. The instrumental line shape is a Lorentzian function. The FWHM of the line shape is approximately between 7 nm and 20 nm with an accuracy of about ±1 nm. The OCSF showm in Figure 1.12 covering the spectral range from 450 nm to 960 nm, this spectral range has been extended to 420-1000 nm. Typical transmittance efficiencies are between 80% and 95% when not considering the detector quantum efficiency. Figure 1.13 illustrates the concept of data acquisition of spectral components using an OCSF- based hyperspectral sensor versus using a conventional dispersive element based hyperspectral sensor. In the case of the conventional dispersive element based hyperspectral sensor (assuming working in the pushbroom mode), all spectral components (A,, A2, ..., A„) of the ground sampling cells w'ithin a cross-track line (jc,) are acquired simultaneously at the moment T = t,. All spectral FIGURE 1.12 On-chip stepped filters covering a wavelength range of visible and near-infrared (400-1000 nm), each step having 6 rows with the same spectral band-pass filters. components (X,, X2, ..., X„) of the ground sampling cells within the next cross-track line (x2) are acquired simultaneously at the moment T = t2 when the satellite flies in the along-track direction, and so on, until a sufficient amount of cross-track lines are acquired to generate a datacube. In the case of the OCSF-based hyperspectral sensor, at the moment T = t,, the ground sampling cells in a total of n cross-track lines are acquired, but only for one corresponding spectral component, i.e., the ground sampling cells in cross-track line x, are acquired for spectral component X,, in cross-track line x, for spectral component X2, ..., and in cross-track line x„ for spectral component X„. In order to generate a datacube, the ground sampling cells in each cross-track line need to be scanned by all the filter steps and acquired. Let’s use cross-track line x, as an example to see how it works. At the moment T = t,, the ground sampling cells in cross-track line x, have been acquired with spectral component A.,. At the moment T = t2, with the satellite flight in the along- track direction, the detector array with filter step X2 chipped in over cross-track line x„ the ground sampling cells in cross-track line x, are acquired with spectral component X2. At the moment T = r„ the ground sampling cells in cross-track line x, are acquired with spectral component X2, and so on, until at the moment T = the ground sampling cells in cross-track line x, are acquired with spectral component Xn. After /; moments, the ground sampling cells in each cross-track line are acquired with all spectral components, and thus a datacube can be constructed. The main difference between an OCSF-based hyperspectral sensor and a dispersive element based hyperspectral sensor is that at one moment the former acquires a 2D image of a scene containing n cross-track lines, each of which obtained only one spectral component, whereas the later acquires the ground sampling cells in only one cross-track line, but obtained all the spectral components simultaneously. Figure 1.14 shows two COTS hyperspectral cameras that were fabricated by Interuniversitair Micro-Electronica Centrum (IMEC) using an OCSF. These hyperspectral sensors consist of a foreoptics and a silicon-based CMOS detector array with integrated OCSF, plus imaging controlling electronics. There are no spectrometers in the hyperspectral cameras. That is why their volume and mass are all small. The volume is about 6 cm x 6 cm x 8 cm and the mass is around 0.6 kg. The hyperspectral camera shown in the lower part of Figure 1.14 uses IMEC’s first-generation OCSF covering a spectral range 600-1000 nm with about 100 spectral bands and 5-nm bandwidth. The hyperspectral camera shown in the upper part of Figure 1.14 uses IMEC’s second-generation OCSF covering a spectral range 470-960 nm with about 150 spectral bands. The author of this book worked with his federal government partners at the Department of National Defence (DND) and the members of Canadian industry teams and conducted studies on the OCSF-based hyperspectral sensors. An OCSF-based hyperspectral imaging breadboard was built using an IMEC hyperspectral camera. The breadboard hyperspectral sensor was characterized in an optical laboratory. Hyperspectral images were collected using a translation stage setup to acquire images and analyze spatial alignment and spectral fidelity of the reconstructed cubes. The characterization of the breadboard was carried out on various targets with interested or known spectra to verify the capability of the OCSF in identifying the spectrum of materials. Five rare earth doped spectralon targets were used because their spectra are well known and often used as standards to validate spectrometers. Figure 1.15 shows the rare earth holmium (WA17A- 1622) doped spectralon target used in the tests. Figure 1.16 presents spectra of the rare earth holmium doped spectralon acquired by using a field portable spectrometer manufactured by the Analytical Spectral Devices (ASD) and the Fabry-Perot filter based hyperspectral camera with the second-generation OCSF (covering a wavelength range from 470 nm to 960 nm). In the figure, the spectrum acquired by the OCSF-based hyperspectral camera is the mean of multiple singlepixel spectra selected manually in a region of interest. Two spectra of the standard deviation upper and lower bound (+std, -std) are also shown. It can be seen from the figure that the spectrum of the rare earth holmium doped spectralon acquired by the OCSF-based hyperspectral camera is quite similar to that acquired by the ASD spectrometer, especially for the absorption features at around 540 nm, 640 nm, 760 nm, and 880 nm. The narrow absorption features at 650 nm and 665 nm are not well captured by an OCSF-based hyperspectral camera. This is understandable because the FWHM of the Fabry-Perot filters integrated on top of the silicon-based CMOS detector array is around 15 nm, even though their spectral sampling interval is 5 nm in the wavelength range from 450 nm to 960 nm. Following the success of the laboratory tests, hyperspectral images were acquired outdoors using the breadboard in parallel with a conventional dispersive element based hyperspectral instrument. In-scene targets included white and dark panels, rare earth targets, an orange cone, and people. The reconstructed datacubes from the OCSF-based hyperspectral sensor were compared to those from both the reference ASD spectrometer and the dispersive element based hyperspectral sensor to evaluate the spectral output of the OCSF-based hyperspectral sensor. FIGURE 1.16 Spectra of the rare earth holmium doped spectralon acquired by an ASD spectrometer and an on-chip stepped filter based hyperspectral camera. In order to demonstrate the capability of an OCSF-based hyperspectral sensor, the breadboard hyperspectral camera underwent two airplane flight campaigns together with the conventional dispersive element based hyperspectral instrument. Figure 1.17 shows an example of the airplane flight campaigns. The airborne setup consisted of a platform for airborne measurement with the dispersive element based hyperspectral instrument (HySpex) in which the OCSF- based hyperspectral breadboard was integrated. A scene near a baseball field in Saint-Henri, Quebec, Canada was selected. Ten regions of interest (ROIs) were identified in the scene, as listed at the bottom of Figure 1.17. The airplane flew at an altitude of 10,000 feet at a speed of about 100 knots. In the figure, the spectra of the material in ROI 10 acquired by both OCSF- based hyperspectral breadboard and the dispersive element based hyperspectral instrument are shown. These are the spectra of at-sensor radiance after processing of spatial alignment and spectral calibration. It can be seen that they are quite similar and this technology shows great promise. The team continues to work on the processing methodology to improve the spectral performance. ELECTRONICALLY TUNABLE FILTERS An ETF-based hyperspectral sensor uses a filter that is mounted in front of a monochrome camera by electronically tuning its spectral transmission (i.e., band-pass) to produce a stack of image slices at a sequence of wavelengths. An ETF is a device whose spectral transmission can be electronically controlled by applying voltage, acoustic signal, etc. (Gat 2000). The advantage of an ETF-based imaging spectrometer is that an entire 2D spatial image of a spectral band is formed instantly when the filter tunes to a particular band-pass wavelength. Unlike dispersing element based and LVF/OCSF-based hyperspectral sensors, there is no need to observe multiple crosstrack lines by the satellite flight motion to obtain the second spatial dimension or to accumulate the spectral dimension. This advantage is at the cost of the additional time required to tune the filter to cover the whole wavelength range before the satellite moves its FOV to observe the next FOV on the ground. FIGURE 1.17 An example of the airplane flight campaigns to demonstrate the capability of an OCSF-based hyperspectral sensor. There are typically three classes of ETFs: * 1. Liquid crystal tunable filter (LCTF), which uses electronically controlled liquid crystal elements to transmit a desired wavelength of light and block others. A LCTF has high image quality and is relatively easy to be integrated into an optical system. A disadvantage of a LCTF is its lower peak transmission values in comparison with conventional fixed-wavelength optical filters due to the use of multiple polarizing elements. A LCTF is also effected by temperature. Higher temperature can decrease the transition time for the molecules of the liquid crystal material to align themselves and for the filter to tune to a particular wavelength. Lower temperature can increase the viscosity of the liquid crystal material, and thus increase the tuning time of the filter from one wavelength to another. * 2. Acousto-optic tunable filter (AOTF), which is based on the principle of diffraction. An acousto-optic modulator, also called a Bragg cell, uses the acousto-optic effect to diffract and shift the frequency of light. Compared with a LCTF, an AOTF has faster tuning speed (microseconds versus milliseconds) and a wider wavelength range. A disadvantage of an AOTF is its relative poor imaging quality due to the acousto-optic effect of sound waves to diffract and shift the frequency of light. * 3. Interferometer-based filters. For example, a Fabry-Perot filter is an interferometer-based filter that has been described in Section 1.2.2.2. Abdlaty et al. (2018) reported a comparison of performance between an AOTF-based and LCTF- based hyperspectral imagers in medical applications for the purpose of highlighting the leverage points of the two type of filters to facilitate their selection in hyperspectral imager design. In their experiments, three parameters were examined: spectral resolution, out-of-band suppression (spectral cross-talk), and image quality in the sense of spatial resolution. The experimental results demonstrated that AOTF-based hyperspectral imager showed superiority in spectral resolution, out-of-band suppression and random switching speed between wavelengths, whereas LCTF-based hyperspectral imager had better performance in terms of the spatial image resolution, both horizontal and vertical, and high definition quality. They concluded that an efficient design of a hyperspectral imager is application-dependent. For medical applications, for instance, if the tissue of interest required more spectral information for undefined optical properties, or contains close spectral features, AOTF might be the better option. Otherwise, LCTF is more convenient and simpler to use, especially if the tissue chromophore’s spatial mapping is needed. The Visible and Near-infrared Imaging Spectrometer (VNIS) aboard Chinese Chang’E-3 lunar spacecraft is an AOTF-based hyperspectral imager. Chang'E-3 achieved lunar orbit on December 6, 2013 and landed on December 14,2013, becoming the first spacecraft to soft-land on the moon since the Soviet Union’s Luna 24 in 1976. The VNIS is described in detail in Section 2.15. At the time of this book being written, to the best knowledge of the author there was no space- borne hyperspectral imager based on a LCTF. Figure 1.18 shows a concept drawing of a Fabry-Perot filter based hyperspectral sensor. The Fabry-Perot filter is mounted in front of the imaging optics and lets pass only light that is at the resonance condition X = 2d (first-order). Each plate separation generates a 2D image at a wavelength with spectral extent given by the FWHM of the transmission response. The way a Fabry-Perot filter is used here is different from that in the on-chip Fabry-Perot filter based hyperspectral sensor. Fabry-Perot filters described in Section 1.2.2.2 are deposited on the top of a CMOS detector array at the wafer fabrication level with fixed cavity for a fixed band-pass, whereas Fabry-Perot filters used here are tunable for band-pass by electronically controlling the width of the cavity. At the time of this book being written there were at least two spaceborne hyperspectral imagers that are based on the technology of electronically tunable Fabry-Perot filters. The Greenhouse Gas Satellite-Demonstrator (GHGSat-D) microsatellite, which was launched in June 2016 (Germain et al. 2016), uses an electronically tunable Fabry-Perot filter operating in a wavelength region between 1600 nm and 1700 nm with a spectral resolution on the order of 0.1 nm. This wavelength range and spectral resolution were selected for the presence of spectral features for greenhouse gases methane and carbon dioxide, as well as relatively little interference from other FIGURE 1.18 A concept drawing of an electronically tunable Fabry-Perot filter based hyperspectral sensor. atmospheric species, H20 in particular. This miniaturized ETF-based hyperspectral imager has a mass of 5.4 kg and a volume of 36 cm x 26 cm x 180 cm. This spaceborne hyperspectral imager is described in Section 2.17. Another spaceborne hyperspectral imager is called Aalto-1 Spectral Imager (AaSI) onboard the Aalto-1 3U CubeSat. A tunable Fabry-Perot filter is electronically controlled in a closed capacitive feedback loop by three different piezo actuators to cover a spectral range from 500nm to 900 nm with a spectral resolution of 6-20 nm. For detailed description of this spaceborne hyperspectral imager, refer to Section 2.18. Source < Prev CONTENTS Source Next > * ON-CHIP FABRY-PÉROT FILTERS The IMEC in Belgium reported an innovative snapshot hyperspectral imaging technology (Imec 2013). The key concept of their snapshot hyperspectral imager is the use of Fabry-Perot filters that are directly post-processed at the wafer level on top of a CMOS detector array. A snapshot hyperspectral imager... (Hyperspectral Satellites and System Design) * FILTER BUBBLES AND DIGITAL ECHO CHAMBERS Judith Möller Are filter bubbles and echo chambers two names for the same phenomenon? Filter bubbles and echo chambers are often named as key drivers of political polarisation and societal fragmentation (Pariser 2011; Sunstein 2001). Both concepts are based on the notion that people are excluded... (The Routledge Companion to Media Disinformation and Populism) * ARE FILTER BUBBLES AND ECHO CHAMBERS TWO NAMES FOR THE SAME PHENOMENON? Filter bubbles and echo chambers are often named as key drivers of political polarisation and societal fragmentation (Pariser 2011; Sunstein 2001). Both concepts are based on the notion that people are excluded from information that is different from what they already believe. Very often, they refer... (The Routledge Companion to Media Disinformation and Populism) * EMPIRICAL INQUIRY INTO FILTER BUBBLES AND ECHO CHAMBERS According to a range of empirical studies, users of online information seek out diverse information. This is associated with less rather than more political polarisation. Empirical evidence stemming from large-scale panel studies demonstrates that social networks of users online are often quite diverse,... (The Routledge Companion to Media Disinformation and Populism) * SPECTRAL INTERROGATION BASED ON TUNABLE FILTERS One of the most successful techniques for spectral interrogation of FFP sensors is based on the use of a tunable narrow bandpass filter [13], as shown in Figures 5.8 and 5.9. The commonly used filter is a fiber-pigtailed and piezo-electrical transducer (PZT)-driven tunable FP filter (TFPF) [14]. In Figure... (Fiber-optic Fabry-Perot sensors an introduction) * TUNABLE OPTICAL FILTERS There are different types of tunable optical filters and fixed-tuned optical filters used in optical network in tunable transceiver. The viability of many local WDM networks depends on the speed and range of tunable filters [54]. Filter Characteristics Tunable optical filters [55] are analyzed... (Fundamentals of Optical Networks and Components) * FILTER BUBBLES AND DIGITAL ECHO CHAMBERS Judith Möller Are filter bubbles and echo chambers two names for the same phenomenon? Filter bubbles and echo chambers are often named as key drivers of political polarisation and societal fragmentation (Pariser 2011; Sunstein 2001). Both concepts are based on the notion that people are excluded... (The Routledge Companion to Media Disinformation and Populism) * ARE FILTER BUBBLES AND ECHO CHAMBERS TWO NAMES FOR THE SAME PHENOMENON? Filter bubbles and echo chambers are often named as key drivers of political polarisation and societal fragmentation (Pariser 2011; Sunstein 2001). Both concepts are based on the notion that people are excluded from information that is different from what they already believe. Very often, they refer... (The Routledge Companion to Media Disinformation and Populism) Home Log in / Register Accounting Business & Finance Communication Computer Science Economics Education Engineering Environment Geography Health History Language & Literature Law Management Marketing Mathematics Others Political science Philosophy Psychology Religion Sociology Travel Home Engineering * Academic library - free online college e textbooks - info{at}ebrary.net - © 2014 - 2023