A.

Spectrometer concepts

The classical imaging spectrometers mentioned in the introduction, are based on the pushbroom principle in which both dimensions of a detector array give: (1) a spatial dimension (one sample line) and (2) a spectral dimension (one spectrum per spatial sample). The motion of the vehicle or a scanning mirror usually recovers the second spatial dimension. The optical principle of such an instrument is given in fig.1 (layout 1). The slit determines the field of view of the instrument. The collimator and the dispersive system make images of the slit in all the wavelengths spread over the spectral direction, and theses images are focused on the detector pixels using the objective.

Fig. 1.

Spectrometer concepts.

A LVF is a narrow-band transmission filter with a peak wavelength varying linearly along one dimension of the filter. Coupled to an array detector, it forms a simple and compact imaging spectrometer as drawn in fig.1 (layout 2) and described in [2]. Besides the compactness, another advantage of this system is the reliable spectral registration, due to the rigid connection between the filter and the detector array, which allows getting rid of an internal calibration device. This concept is also interesting for infrared spectroscopy as it reduces the thermal background. Indeed a pixel sees only its useful narrow wavelength band, whereas in a classical spectrometer, a pixel sees the background in the whole spectral range. The main drawback of the system is that at a given time the pixels see different lines of sight. Image resampling can correct this misregistration, well noticed in [3], but needs very precise geometry information.

Adding a dispersive system in front of the objective, with the same dispersion law than the LVF, allows the pixels to point the same line of sight. In this system drawn in fig.1 (layout 3), the dispersive optics work in a collimated beam, thus in the best conditions. The alignment between the two optical groups {dispersive system + objective} and {filter + detector} is not stringent because a decenter creates an offset of the line of sight that can be calibrated.

B.

Choice of the dispersive system

We have basically the choice between a diffraction grating and a prism. Considering its linear dispersive law and mass criterion, the diffraction grating would be the preferred solution. However the absence of field stop at the entrance of the instrument prevent from using a grating. Indeed, for a wavelength λ₀ diffracted in the direction β₀ in the first order of the grating, there are several angles of incidence (thus fields) for which rays at λ₀ are diffracted in the same direction β₀. This is due to the existence of multiple diffraction orders of the grating and there is no solution to avoid them.

A prism is therefore chosen as dispersive system. The main concern of the prism is its non-linear dispersive law. This will be discussed later in the paper. A typical optical design of a wedge filter imaging spectrometer is illustrated in fig.2. The optical elements are limited to a prism and the four lenses of the objective. The folding mirror is added to compact the instrument volume.

Fig. 2.

Optical design of a wedge filter imaging spectrometer.

C.

Linear variable filter

Off-the-shelf LVF are available, for example at JDS Uniphase (fig.3 left). They supplied several units of LVF working in the [620-1080nm] range. The substrates dimensions are 15.0 x 5.5 x 1.5mm³. The local bandwidth is of the order of 1% of the central wavelength and the peak transmission is higher than 50%. Fresnel Institut in Marseille characterized the LVF (fig.3 right) to verify the spectral resolution and the wavelength gradient (40nm/mm).

Fig. 3.

Left: Linear variable filters from JDSU (Courtesy JDSU). Right: Characterization of a linear variable filter by Fresnel Institute.

Other spectral bands are available up to 14μm. The spectral band of a filter is limited due to the absence of blocking filter.

A.

Numerical simulations

A Mathcad tool runs analytical simulations of the system, and in parallel, an IDL program models the instrument. Both models compute the total field of view (FoV) of the instrument, the instantaneous field of view (IFoV) of each pixel and the spectral resolution (R) of each pixel. The entrance parameters are the spectral band, the optical data (prism material, angles, focal lengths, apertures), and the filter characteristics (dimensions, spectral resolution, wavelength gradient). Two other parameters are adjustable:

The principle of the Mathcad model (fig.4) is to take the origin on the detector surface and describe the optical path until the entrance of the prism to retrieve the desired parameters (FoV, IFoV and R). Conversely, the IDL program samples the object spatially and spectrally and computes the detector response.

Fig. 4.

Principle of Mathcad simulation.

B.

Instrument behavior

Due to the deviation of the prism and filter dispersion laws, each pixel of the detector matrix points a line of sight, θ₂ in fig.4, slightly different from its neighbors. The FoV is defined as the variation of θ₂ along the spectral dimension of the detector. The Mathcad tool models this effect (fig.5 left) and allows estimating the best relative position of the filter and the spectrum dispersed by the prism, i.e. the position minimizing the FoV. For a position where the dispersion laws are fitted, the FoV is about 12mrad (0.7°). A way to reduce the deviation of the dispersion laws is to linearize the prism dispersion by coupling two prisms. However the improvement is not so interesting compared to the increase of the mass. Another solution would be to design non-linear variable filters adapted to the prism dispersion law.

Fig. 5.

Left: FoV versus relative position of the filter and prism spectrum. The optical axis is defined as the direction of the rays at 850nm dispersed by the prism. The x-axis represents the wavelength transmitted by the filter on the optical axis. Right: Optical parameters used for the simulations.

Unless the filter is directly deposited on the sensitive area of the detector, which is not the case here, the distance between the filter and the detector is not null. Indeed, coupled to the aperture of the beam, the surface intercepted on the filter increases with the distance between the filter and the detector.

This results in an increase of the IFoV (fig.6 left) and a decrease of the spectral resolution (fig.6 right). Thus, the distance between the filter and detector should be minimized. For a distance of 100μm, the expected IFoV is about 2.2mrad @650nm and 0.9mrad @1050nm.

Fig. 6.

IFoV and spectral resolution versus distance between filter and CCD.

The spectral resolution of the instrument is defined as λ/Δλ, where Δλ is the spectral band contributing to 90% of the energy registered by a pixel. As the filter tranmission profile is gaussian, Δλ is equal to 1.4 times the full width at half maximum of the gaussian function (FWHM). For a filter to detector distance of 100μm, the expected spectral resolution is equal is about 68. Note that the filter resolution, defined as λ/FWHM, is equal to 100, thus 1.4 times higher than the spectrometer resolution. This is just a consequence of the different definitions.

Adapting the dispersion laws of the filter and prism, as described above, would allow reaching IFoV of about 1mrad. The IFoV would still vary with wavelength, though to a lower extent, because it is proportional to the spectral band locally transmitted by the filter FWHM (R=λ/FWHM is a constant for the filter, thus FWHM is varying). A minimum spectral resolution of 150 could be easily achieved with better resolution of variable filters (typically 200).

C.

Example of instrument

LESIA has proposed a “classical” infrared imaging spectrometer for the Binary Asteroid in-situ Explorer (BASiX) Mission [4], based on prims dispersion. BASiX will make the first quantitative measurements of a near-Earth asteroid’s strength and seismic properties, and be the first to explore a binary asteroid system’s mass morphology (target 1996 FG₃). A near infrared imaging spectrometer is classical and unavoidable instrument for remote characterization. This instrument covers the [0.4-4.0μm] band, with a spectral resolution of 200, a mass of 7.5kg and an optical head volume of 400 x 300 x 128mm3 (fig.7 left).

Fig. 7

Left: Opto-mechnical design of a classical imaging spectrometer. Right: Opto-mechanical design of a wedge filter imaging spectrometer.

Focusing on the main scientific goals, and thus reducing the spectral band ([2.7-3.8μm] typically), a wedge imaging filter spectrometer would comply with the scientific requirements with a mass in the 1.5kg range and dimensions within 160 x 230 x 54 mm³ (fig.7 right).

A.

General description

The objectives of the spectrometer breadboard (fig.8) are to integrate a LVF in a camera, align the assembly with respect to a prism and characterize the performances in terms of spectral resolution and field of view.

Fig. 8.

Spectrometer breadboard.

Locating a slit at a collimator focus creates the artificial object at infinite distance. A monochromator (spectral resolution better than 2nm) and its output optics illuminate homogeneously the slit. The slit is mounted on a translation stage to explore the field of view. The off-the-shelf prism (N-SF11 glass) works at its minimum deviation. An objective focuses the dispersed beam onto the detector.

B.

The camera assembly

In order to reduce the cost, we chose to supply a commercial camera and integrate the filter close to the detector in our laboratory. The camera must comply with several requirements: (1) no cover glass for the CCD, (2) no window at the entrance of the camera, (3) a CCD area and detector package large enough to locate the filter (15.0mm length) at 100μm distance from the CCD sensitive area. The camera Insight IN1400 from SPOT Imaging Solutions (fig.9 top left) complies with all these requirements. The Kodak CCD KAI-4021-M, with 2048 x 2048 pixels (7.4μm pitch), leads to a 15.2 x 15.2mm² sensitive area. SPOT kindly accepted to deliver a CCD without cover glass and to remove the camera window.

Fig. 9.

Camera assembly. Top left: front view of the camera with unprotected CCD. Top right: mechanical interface between filter and camera. Bottom: intregration of the filter assemby in the camera housing.

The LESIA design department developed a custom mechanical mount to maintain the filter close to the CCD sensitive area (fig.9 top right). The filter is glued on a mount that can be translated from 100μm to 2mm from the CCD, with a ±10μm precision. The filter mount assembly uses the F-mount of the camera as mechanical interface (fig.9 bottom).

C.

Metrology

The distance between the filter and the CCD is an important parameter to monitor, not only because the filter must not touch the CCD sensitive area, but also because it directly impacts the spectrometer performances. Thus we need to have at our disposal a non-contact distance sensor with a precision better than a few microns. We chose the confocal-chromatic sensor and controller from Micro-Epsilon. The confocal chromatic measuring principle works by focusing polychromatic white light onto the target surface using a multi-lens optical system. The lenses are arranged in such a way that the white light is dispersed into a monochromatic light by controlled chromatic deviation (aberration). A certain deviation (specific distance) is assigned to each wavelength by a factory calibration. Only the wavelength that is exactly focused on the target surface or material is used for the measurement. This light reflected from the target surface is passed through a confocal aperture onto a spectrometer, which detects and processes the spectral changes.

The filter mount was adapted to point successively the detector surface and the filter first face (fig.10). The filter thickness was previously measured with several methods, and in particular using the confocal-chromatic sensor that measures transparent elements thicknesses (knowing the refraction index of the material).

Fig. 10.

Distance measurement between the LVF and detector.

The distance between the filter rear face (where the filter function is located) and the CCD sensitive area is measured with a precision better than ±3μm.