A Near Infra-Red Hyper Spectral Imager

Fred Sigernes^1,2, Marie Bøe Henriksen ², Sivert Bakken ², Joseph Garrett ², Roger Birkeland ², Torbjørn Skauli ³ and Tor Arne Johansen ²

¹ The University Centre in Svalbard (UNIS), Norway
² Norwegian University of Science and Technology (NTNU), Trondheim, Norway
³ University of Oslo, Oslo, Norway

First, the source illumination in the NIR region can be identified in Figure 1. The target wavelength region 700 - 1100 nm is less intens compared to the visible part of the spectrum and dominated by two atmospheric absorption bands (O₂ and H₂O).

The design is inspired [1] by the pushbroom HSI v6 onboard the HYPSO-1 satellite [2-4]. The equation for a head on illumination of a transmitting grating is essential for the construction

n λ = a sin β,

where n is the spectral order, λ is the wavelength, a the groove spacing, and β the diffracted angle.

The optical diagram is shown in Figure 2. The center design wavelength is set to 900 nm. A 300 lines/mm blazed grating is the key element. Blaze angle is 31.7°. The efficiency is above 55% for the wavelength region 700 - 1100 nm. The effective aperture may be set to D₀ = D₁ = 18 mm which corresponds to ~ F/2.8 The input slit width is fixed with no magnification or demagnification of the height (h = 10 mm). Field Of View (FOV) along slit equals 11.4°. A slit width of w = 50 µm will result in a first order (n = 1) Full Width Half Maximum (FWHM) spectral bandpass of 3.33 nm.

All components are off-the-shelf. Three 50 mm focal length NIR objectives from the company Edmund Optics (EO) is used in combination with a standard blazed transmitting grating from Thorlabs. The detector is a Black Silicon CMOS sensor from the company SiOnyx, LLC.

The front lens focuses light from infinity onto the entrance slit plane. Light through the slit is collimated by the collimator lens to produce a parallel light beam that will illuminate the grating. These two lens objectives are chosen to be identical to preserve image quality through the system. The same mechanical solution as presented in [3] is used. See Figure 3.

The above parts are sufficient to create a prototype hyper spectral imager. All parts may now be embedded into a 3D printed design or a metal housing according to the angles and positions defined in the optical diagram. Note that a right-angle prism can be used between the collimator and the grating to make the design more compact in size. See Figure 4 below.

A 3D printed press fit solution is shown in Figure 5. A 3D printer from the company Prusa Research model Prusa i3 MK3S+ was used with a black PETG (Polyethylene Terephthalate modified with Glycol) filament. The material is known for its temperature resistance, tenacity, and flexibility. According to Prusa Research it does not shrink or warp, making it suitable for large print models.

The camera sensor evaluation kit (XRB-1350-PDK) was tested with a Schneider Kreuznach F/1.4 CCTV 17 mm focal length lens. A front cut-off filter HOYA R72 was used to cut light below 740 nm. The camera was mounted under a dome at the observatory to a motorized pan-tilt unit from the company Directed Perception (PTU-D47-70). This setup enables us to quickly point the camera to any sky target of interest using the SvalPoint tracking system.

Snapshot image
A snapshot image of a low intensity (1-2 kR) red colored post noon auroral arc was identified as soon as we turned the system on. See image below. Default automatic mode of camera was used. The gain was 3200 with an exposure time of just 11 msec running at 90 FPS.

Movie
A 60 second video timelaps is shown here.

Note that the video is scaled down 4 times to reduce storage and the real duration is only 16.7s. Nevertheless, the above auroral experiment indicates that the sensor is low light sensitive and a promising candidate for our hyperspectral imager.

The assembled instrument was turned on and pointed towards a white cardboard illuminated by a roof mounted Osram fluorescent gas tube lamp. The recorded snapshot spectrogram is shown in the below figure with identified emission lines.

Note that due to lack of a spectral line source in NIR, a UV-NIR cutoff filter was mounted to the front lens (EO #89795). The latter to was done to identified well know emission line in the second order of the visible spectrum without the NIR part contributing from the first order. With well know emission lines identified in the second order, the corresponding first order NIR wavelengths become 2 times higher. This procure can be applied since the sensor is highly sensitive in the visible part of the spectrum [6].

The instrument was wavelength calibrated using a hydrogen gas discharge spectral tube and the UV/NIR cut-off filter (EO #89795). The first order Hα line and the second order Hβ line was identified, and the spectral range is calculated to be from the visible red into NIR (618 - 1243 nm).

λ ≈ 617.805 + 0.611025 × p,

where p ∊ [0..1023] is pixel value. Assuming Full Width Half Maximum (FWHM) spectral bandpass of 3.3 nm, the number of pixels per spectral bin becomes 5.4. A narrow spectral line may therefore be sampled by 12 pixels. This corresponds to 85 images generated from of the spectral data cube.

A red colored filter is needed to cut the second order visible part of the spectrum. A mounted filter (EO #46-545) should be installed on the front lens. This must be done prior to sensitivity calibration.

The spectral imager was mounted to a Syrp Genie Mini II rotary table. A 30° horizontal sweep angle and period of 15 s were used. The target scenario is out an office window of UNIS towards East. The camera head operated at 90 FPS in automatic mode.

1000 frames were sampled by the Windows PC uTest64 program in is just 11 s. Each frame was rotated and accumulated into a spectral movie by ffmpeg. The latter procedure was used to make the hyper spectral data cube compatible with our Play Spectrogram software, which generated 3 spectral images at 700, 800 and 900 nm with image bandpass of 3.3 nm. The color composite is shown above.

Note that the analog gain was constant low at 100 during the sweep. The gain range is 100-3200. The exposure time varied from 6 - 6.5 msec. The effect of the automatic mode of the camera is seen in slight intensity changes in horizontal direction. Note that the image is resized vertically by 50% and auto-level color corrected by the program paint.net.

Item	Part / links	Description	Qty	Cost $
1	EO VIS-NIR 50 mm	50mm C VIS-NIR Series Fixed Focal Length Objective *	3	1785
2	EO 2nd order filter	M30.5 x 0.5 mounted Red filter	1	54
3	Thorlabs SM1A10	Adapter ring SM1 - C- mount internal	2	45
4	Thorlabs SM1M10	SM1 lens tube 1 inch long with internal threads	1	17
5	Thorlabs S50LK	Fixed high precision mounted slit	1	122
6	Thorlabs Spacer Rings	Thorlabs C-mount 0.25-2mm space ring kit	1	121
7	Thorlabs GTI25-03A-NIR	(25 x 25) mm 2 Blazed Trans. grating (300 grooves/mm)	1	118
8	Thorlabs right-angle prism	N-BK7 Right-Angle Prism, Uncoated, L = 25 mm	1	65
9	Sionyx RD board	Black Silicon sensor (12.3 x 9.9) mm2	1	790
10	3D printer material	PRUSA Jet Black PETG filament	1	27
		Total	13	3144

Table 1. Detailed part list NIR HSI v7. * Possible lens candidate: KOWA 50 mm.

Item	Component	Property
1	EO VIS-NIR 50 mm objective	Transmission
2	Kowa VIS-NIR 50 mm objective	Transmission
3	Thorlabs right-angle prism	Transmission
4	EO 2nd order filters	Transmission
5	Thorlabs transmission grating	Efficiency
6	Sionyx black sillicon sensor	Responsivity

Table 2. Optical component transmission, efficiency and responsivity [6]. Data adopted from manufactories webpages.

A pushbroom Near Infra-Red Hyper Spectral Imager design (NIR HSI v7) is described using off-the-shelf components. The spectral range is approximately 600 - 1200 nm with a spectral resolution less than 4 nm. Spatial resolution and senstivity is expected to be comparable to the HSI v6 on board the HYPSOP-1 satellite. Total part cost of prototype is estimated to be less than 4k USD.

1. Fred Sigernes, Mikko Syrjäsuo, Rune Storvold, João Fortuna, Mariusz Eivind Grøtte, and Tor Arne Johansen, Do it yourself hyperspectral imager for handheld to airborne operations, Opt. Express 26, 6021-6035 (2018), https://doi.org/10.1364/OE.26.006021
2. M. E. Grøtte, R. Birkeland, E. Honore-Livermore, S. Bakken, J. L. Garrett, E. F. Prentice, F. Sigernes, M. Orlandic, J. T. Gravdahl, T. A. Johansen, Ocean Color Hyperspectral Remote Sensing with High Resolution and Low Latency - the HYPSO-1 CubeSat Mission, IEEE Trans. Geoscience and Remote Sensing, Vol. 60, pp. 1-19 (2022), https://doi.org/10.1109/TGRS.2021.3080175
3. M. Henriksen, E. Prentice, C. van Hazendonk, F. Sigernes, and T. Johansen, Do-it-yourself VIS/NIR pushbroom hyperspectral imager with C-mount optics, Opt. Continuum 1, 427-441 (2022), https://doi.org/10.1364/OPTCON.450693
4. S. Bakken, M. B. Henriksen, R. Birkeland, D. D. Langer, A. E. Oudijk, S. Berg, Y. Pursley, J. L Garrett, F. Gran-Jansen, E. Honore- Livermore, M. E. Grøtte, B. A. Kristiansen, M. Orlandic, P. Gader, A. J. Sørensen, F. Sigernes, G. Johnsen and T. A. Johansen, HYPSO-1 CubeSat: First Images and In-Orbit Characterization, Remote sensing, 15(3), 755 (2023), https://www.mdpi.com/2072-4292/15/3/755
5. Post by Harron, All you need to know about Solar Radiation, http://synergyfiles.com/2016/05/solar-radiation/
6. Mark Crawford, Black silicon is ready to revolutionize photoelectronics, SPIE, The international society for optics and photonics, 08 December 2008.