2.1

General instrument design description

The ALTIUS instrument (currently under development by OIP NV) is meant to be integrated on top of a PROBA P200 platform developed by QinetiQ Space. The ALTIUS mission is planned for a nominal timeline of three years with a possibility of extension for two years. It will be launched as a co-passenger using the Vega C launcher.

Due to the limited allocations in volume and mass, the payload had to be developed under severe volume constrains while still ensuring the allocation of three different sub-instruments (also called channels), imaging respectively in the UV range between 250nm and 355nm, VIS range between 440nm and 675nm and NIR between 600 and 1020nm. The three channels are optically and electrically fully independent from each other. The three channels are stacked on top of each other on the same mechanical assembly, called Optical unit. Each channel is electrically controlled by an independent control electronics assembly, each three assembly being positioned along the optical unit as shown in Figure 1.

Figure 1.

ALTIUS instrument optical unit and channels control units

The design of each of the three channels had to consider multiple observation modes required by the scientific community for the ALTIUS instrument as bright limb observations, solar and stellar occultation while ensuring for them the adequate radiometric, spectral and geometric performances. The main requirements on spatial performances and spectral sampling which shall be achieved by the payload are indicated in the Table 1 below.

Table 1.

Main parameters guiding the optical design of the three channels

Based on an extensive feasibility study performed in the payload development phase B, backed up by breadboarding activities, spectral filters allowing a direct 2D spatial imaging in the relevant wavelength domain were selected for each of the channels. As such, the UV channel is based on Fabry-Perot technology; while the VIS and NIR channels integrate an Acousto-Optical Tunable filters (AOTFs) which has been identified as the best candidate based on the scientific requirements.

From a general standpoint, each channel is based on a reflective optical design organized over a unique layout:

Considering technical and programmatic optimization of the payload development, the design of each channel is strongly focused on reusing the same subassemblies. Reusability has been leading to use of a quasi-identical optical design between the VIS and NIR channels. This is also for instance the case for the focal plane assembly, which is identically based on the CIS115 sensor from Te2V [1] on each channel. A detailed description of the payload and technologies onboard was already reported in [2] and the reader can also refer to [3] presented in this same conference and showing an overview of the current development status and challenges.

2.2

VIS and NIR channels opto-mechanical design

The VIS and NIR channels of the ALTIUS are almost identical from an optical design point of view. The design is presented in Figure 2.

Figure 2.

VIS channel optical design concept

In these two channels, the front end optics (FEO) is a telecentric objective composed of three mirrors integrated on a single mechanical structure. Its role is to match the AOTF aperture and acceptance angle while correcting the chromatic aberrations induced by the AOTF crystal. The FEO is generating an intermediate image of the scene few mm behind the AOTF crystal.

The AOTF is the spectral filter allowing a versatile selection of the wavelength to be imaged by the camera. The detailed AOTF acousto-optics interaction theoretical baseline is very well known and documented, for instance in [3]. From a high level standpoint, an acoustic wave, generated through a vibrating piezo actuator, propagates in a large high purity birefringent (TeO2 for most VIS/NIR applications) crystal and defines a phase grating whose pitch is defined by the frequency of the wave. This phase grating interacts with the incoming beam generating a diffraction effect between a first diffracted order, containing the waveband of interest and a zero order which contains the undiffracted light. The central wavelength of the first diffracted order is defined by the frequency of the selected acoustic wave. The light from the first diffracted order intermediate image is then collected by a second TMA, called Back End Optics (BEO) which is a relay towards the detector. The magnification of the BEO is then determined by the dimensions of the image projected onto the sensor chip.

Table 2.

VIS and NIR channels first order parameters defining the optical design

One of the first challenge designing using AOTF technology is to filter out the narrow waveband first diffracted order from the very intense zero order containing the rest of the continuum from the incoming scene light. Fortunately, the acousto-optic interaction in a birefringent material presents some properties to ease the design:

Figure 3.

Zero order blocking method illustrated using a beam stop (red zero order, blue first diffracted order)

The combination of these techniques is allowing a significant rejection of the zero order itself. Nevertheless this approach cannot suppress possible reflections and ghosts from the zero order on the refractive elements to find their way towards the detector.

2.3

UV channel opto-mechanical design

In the UV channel, the optical design is quite different from the VIS and NIR channels. The spectral filter is based on a stack of four active Fabry-Perot interferometers. Each Fabry-Perot used for the ALTIUS mission is based on a dielectric-coated fused silica mirror pair connected together using piezo actuators with air gaps ranging between 400nm and 1000nm. Coating design of each mirror pair is tailored to ensure that the four FPI can select a narrow wavelength band in a continuous way between 250nm and 355nm. The main driver for the optical design is the limited étendue of the FPI, driven by the limitation of the mirrors clear aperture where the required parallelism could be guaranteed to achieve the required spectral performance. A maximum clear aperture of 12 mm has been identified as feasible, thus requiring for the stellar occultation SNR achievement to design a magnifier system and a line of sight changing mechanism. The optical layout is then split in two optical paths, one for bright limb and one stellar occultation as shown in Figure 4.

Figure 4.

UV channel optical design

Considering the FPI stack to be mostly transparent in the visible range, the visible spectral rejection is done thanks to the use of a low pass filter coating applied on each mirror combined with a bandpass filter applied on a fused silica substrate. The main parameters of the UV optical design are identified in Table 3 below:

Table 3.

UV channels first order parameters defining the optical design

3.1

Scenes and requirements

The scenes to be observed for the ALTIUS mission are defined by the scientists based on the modelling of the atmospheric transfer function using MODTRAN. These scenes, in particular in the bright limb period of the orbit are characterized by an extremely strong dynamic range from the bottom of the atmosphere to the top of the 100km scene observed. This distribution is shown very clearly in Figure 5, which contains all required wavelength of observation for the instrument.

Figure 5.

Irradiance of the scene modelled by the scientists as a requirement for the ALTIUS radiometric performances

Considering the above-mentioned scenes the scientists have been defining rather challenging radiometric requirements in terms of signal to noise ratio for the instrument.

Figure 6.

Required signal to noise ratio as a function of the images altitude for the scene specified in Figure 5

Based on the required collected signal they have also defined a stray light requirement specified in terms of stray light rejection based on a ratio between the stray light and the available signal. This requirement is basically allowing the stray light to be a maximum of 3x the noise from the detection chain.

Figure 7.

Stray light to signal ratio required as a function of the observation altitude

3.2

Specific sensitivities and way forward

Considering the stray light requirement and the irradiance distribution for the scene is already representing a significant challenge for the ALTIUS instrument concept as the stray light rejection shall occur in field. In such configuration there is little parameters to play with to optimize the optical design for in field stray light reduction; the scattering and the ghosting.

The scattering effect induced by each optical element plays an important role in the end to end achievable performances. The first optimization parameter in this matter is the optimization of the number of optical surfaces allowing still the design of each channel to fit within the allocated volume. The second optimization parameter is the selected technology for the manufacturing of the mirrors and refractive elements. Refractive optical elements are based on polished fused silica substrate. For reflective optical elements, selected approach is based on the reduction of the micro roughness of each unique optical surface. Considering the all-aluminum approach for the ALTIUS design, it has been decided to go for polished nickel plated mirrors which would satisfy the target of reach a surface roughness ≤ 1 nm RMS in an extended frequency range from 1.5mm^-1 to 2000mm^-1.

To support the stray light analyses, achievable performances for polished nickel plated mirrors have been validated through prototypes development. The bidirectional scattering distribution function (BSDF) has been measured at 633nm on samples and an ABg function has been fitted to the measurement, shown in Figure 8. For the mirrors, two models, one for small angle scattering (PV-EVO SAS) and one for large angle scattering (PV-EVO LAS) are combined for the roughness scattering simulation. The parameters of these models are given in Table 4.

Figure 8.

Example of BSDF measured on polished nickel plated mirrors together with the fitted BSDF ant the signature of the measurement equipment.

Table 4.

Parameters of the ABg models for the fitted BSDF of polished nickel plated mirrors

Other ABg models are defined for the scattering of the refractive surfaces, one for the UV channel and one for the VIS channel. A unique model is used for the reflective and refractive surfaces as well as the detector for the scattering from particulate contamination [4]. The Mie scattering model for particulate contamination with a slope of 0.383 and a cleanliness level of 361 corresponding to an obscuration of 600ppm has been simulated with ABg functions, see Figure 9.

Figure 9.

ABg curves used to simulate particulate contamination

Optimization of ghosts has been performed by scrutinizing the respective orientation of all surfaces which would contribute, namely the detectors, FPIs, bandpass filters, AOTFs and wire grid polarizers. Their respective orientations have been selected such that, if there is ghosting occurring it shall preferably lead to a ghosts which originates from the same altitude within the scene, to avoid cross contamination and ease the deconvolution methodology on ground. The design for each channel has therefore been optimized considering simple best engineering practices, namely:

Beside the infield stray light another challenge to be tackled for the ALTIUS is the near out of field stray light. Indeed the ALTIUS instrument shall be managing an irradiance profile coming from the earth, just below the field of view which could range between 1000 and 500 times the irradiance from the brightest part of the scene. This extremely bright near out of field elements could have a significant impact onto the stray light within the field.

3.3

Benchmarking needs

A comparison between the results obtained for the stray light values using two different optical software suites is considered necessary and of great added value for the ALTIUS stray light analysis. In the study, the softwares OpticStudio ® and FRED® are used.

The goal of the study is to assess the stray light performance of the UV and VIS channels. For that, the PSFs for different wavelengths and fields are generated and compared using a mathematical criterion that quantitatively measures how different they are. Based on the results from this criterion and also in the overall distribution of the stray light irradiance, their differences and the validity of the models will be assessed. Additionally, the stray light irradiance distribution for the Bright Limb scene is also calculated for different wavelengths and the pixel stray light values compared to the requirements.

The UV and VIS optical and mechanical parts are modelled in both softwares, together with the sources, coatings, and scatter models. For the purpose of the study and in order to keep computation times reasonable, a simplified detector of 501x501 pixels with 21 microns pitch is used. Details and particularities on the modelling of the different parts are described here below.

4.1

Scattering model

The initial finding comparing FRED and OpticStudio is coming from the way the scattering effects can be modelled. In OS, the only built-in scattering model to model scattering from polished optical surface is the ABg model.

β₀ and β are the projections of the specular and scattered ray vectors down to the surface.

The ABg model in OpticStudio has the limitation that the B parameter cannot scale below 1E-12. With the PV-EVO SAS function this limit is reached for wavelengths below 545nm. To overcome this the PV-EVO SAS function was modified slightly. The ABg models used to simulate surface roughness are given in Table 5. The full mirror scattering function, SAS + LAS, together with the PV-EVO function is shown in Figure 10 on the left.

Figure 10.

Comparison of PV-EVO BSDF and BSDF used in OpticStudio for mirror roughness (left) and comparison between ABg and Harvey-Shack model (right)

Table 5.

Parameters of the ABg models used in the OpticStudio model to simulate surface roughness

As the original stray light model was built in OpticStudio, scattering ABg functions derived from the roughness measurements of the different optical element presented in §3.2 were used.

Initially and in order to maintain consistency between the two models the ABg approach was also kept for the integration in FRED. Nevertheless initial investigation on the model consistency in FRED revealed that the obtained Total Integrated Scattering (TIS) computed with FRED appears not correct for models presenting a very sharp BSDF which is the case for the SAS mirror roughness scattering function. As an example, the scattering model for the ALTIUS mirrors prototypes has a TIS value obtained with OS as well as analytically of 3.2% at 300 nm, but the TIS computed with FRED for the same ABg model is 0.35%. This was attributed to the too high necessary sampling to perform the integration which is not supported in FRED.

Therefore it was necessary to come back with FRED to a Harvey-Shack (HS) model.

The HS parameters have been obtained using the following conversion equations between ABg and HS model:

Based on the conversion corresponding to the same ABg model the analysis of the TIS showed a value closer to the expected value but still slightly different (2.4%).

The ABg and HS parameters are given for a reference wavelength of 633 nm and a scaling with wavelength is performed to obtain the parameters for 300 and 350 nm for the UV design and 650 nm for the VIS design.

OpticStudio automatically scales the ABg function according to the following scaling laws:

For the Harvey-Shack functions this translates to:

The models parameters are presented in Table 6. OpticStudio uses the ABg functions while in FRED the mirror SAS and the fused silica are replaced by Harvey-Shack functions. The ABg and Harvey-Shack models present small differences but their impact has been checked on the resulting PSFs and they were showed to be not significant for the final results.

Table 6.

ABg & Harvey-Shack models used for the scattering of the mirrors

Note that the backward particulate contamination (PC Backward in the table below) is not included in FRED because FRED does not allow negative g values.

4.2

Conservation of energy

Modelling the surface roughness and the particulate contamination requires that several scattering functions are applied to the surface. For the mirrors, for instance, four scattering functions must be added: SAS, LAS, the PC backward and the PC forward that is reflected on the mirror. In OpticStudio, ABg files can be added in two ways: relative and absolute. With the relative option the combined scattering function is the weighted average of the ABg functions, with the absolute option, the combined scattering function is the weighted sum. Obviously the ABg functions need to be added with the absolute option.

A simple test was performed on a 100% reflective flat surface with the combined ABg functions and it was found that the total power received by a hemispherical polar detector was larger than the power of the source. This is only the case with the absolute option. With the relative option the energy is conserved. As a workaround, scaled ABg functions were created with A’ = nA where n is the number of ABg files to be added and the scaled functions were added with the relative option.

4.3

Ray trace settings

Both FRED and OpticStudio have settings for dropping rays like minimum absolute and relative power and maximum number of intersections. According to the OpticStudio manual, the maximum number of generations of child rays split off from the parent ray is controlled by the maximum intersections per ray. This has been set arbitrarily to the maximum of 4000. FRED has a setting that has no equivalent in OpticStudio: ancestry level. This parameter directly limits the number of generations for specular and scattered rays. Considering the computation intensive aspect of the stray light analysis setting a number of ancestries optimizes the computation time as allowing to trace less rays in the source in FRED compared to OpticStudio. It also permits to reconstruct final results as a stack of different layers of paths while for OpticStudio the approach is more brute force. Nevertheless, this approach also requires the optical designer to know a priori what he is looking for to observe and induce the risk to “miss” significant effects in a first instance, risk prevented in the OpticStudio approach.

4.4

Modelling of the Components

The model was first built in OpticStudio. The mechanics is simulated with native OpticStudio objects. These objects were exported as a STEP file and imported in FRED. For this study the non-optical surfaces were defined as fully absorbing.

Coating on optical surfaces are defined with a general sampled coating using reflectance or transmittance values at the wavelengths of interest as a function of several incidence angle and for both polarization states s and p.

Figure 11.

Imported mechanical element (Left: OpticStudio; Right: FRED) of the UV channel, close up of FPI’s shown at the bottom.

The UV channel is relatively easy to model both in OpticStudio and FRED. The only unconventional components are the FPI’s. The FPI’s are modelled as two fused silica rectangular volumes. In OpticStudio they are in contact. FRED doesn’t support two surfaces being positioned in the exact same position therefor there is an air space of 1μm between them. The transmission and reflection properties of the FPI are defined by means of a coating on the front face of the second plate. The front and back face of the first plate and the back face of the second plate all have a coating with 100% transmission. It shall be noted that in both models there is no integration nor consideration about the full spectral transfer function of the FPI units and or their respective constructive interference patterns. Therefore the analysis both in FRED and OpticStudio is excluding the spectral stray light which could originate from the spectral leakages due to imperfect coating design.

The VIS channel has two unconventional components that require some tricks to model: the wire grid polarizers and the AOTF.

The wire grid polarizers are double polarizers with the wire grids on the internal surfaces. They are modelled as two fused silica disk, 1mm thick, with a 0.1mm air space. Wire grid polarizers mainly reflect the rejected polarization. OpticStudio has a type of surface that is ideally suited to model this type of behavior: the Dual Brightness Enhancement Film (BEF) Surface. The Dual BEF Surface is a rectangular plane for which reflection and transmission in X and Y can be defined. FRED does not have this type of surface or coating therefor different models had to be made for X- and Y-polarized light.

In OpticStudio two Dual BEF Surfaces are placed inside the air space at 1μm from the fused silica disks. The outer surfaces of the disks have an anti-reflection coating with 99% transmission and 1% reflection, the inner surfaces have a coating with 100% transmission. The Dual BEF Surface has: reflection X 12.832%, reflection Y 99.998%, transmission X 87.168% and transmission Y 0.002%.

In FRED the polarizers are modelled with ordinary coatings on the faces of the fused silica disks and additional ideal polarizers and rotators. There are different models for the 1^st order with an X-polarized source, the 0 order with an X-polarized source and the 0 order with a Y-polarized source. In first order with an X-polarized light source, everything is straight forward and the transmission and reflection values are the same as for the Dual BEF surface except for a small correction for the second polarizer that is rotated 3° round its normal to simulate alignment uncertainty.

To determine the transmission and polarization properties of each polarizer part in the zero order, an on-axis polarized point source was traced in OpticStudio and the power in front of, in the middle and behind each dual polarizer was recorded for X- and Y-polarized light separately. The results are given in Table 7.

Table 7.

Flux in front, in the middle and behind each polarizers for a polarized source and a polarized detector, traced in OpticStudio with the Dual BEF models.

In the zero order with an X-polarized light source, the first polarizer is the same as for the first order. After travelling through the AOTF a small part of the light is Y-polarized. The first part of the second polarizer will reflect almost all the X-polarized light but transmits most of the Y-polarized light. To simulate this a rotator is placed between the two parts of the second polarizer. With a Y-polarized source, a very small part of the light is X-polarized when it reaches the first polarizer. This part is transmitted by the first polarizer. After the first polarizer the light is mainly X-polarized and thus a rotator is placed after the second polarizing face. The complete model used in FRED is given in Table 8.

Table 8.

Model of polarizers in FRED

The AOTF is modelled similarly in OpticStudio and FRED. The AOTF is modelled in different parts. The AOTF material is birefringent TeO2. The acoustic wave acts like a grating and is modelled as such. The groove spacing is chosen so that the diffracted ray has the correct angle after exiting the crystal. For a grating to function properly in both FRED and OpticStudio it needs to be immersed on both sides in non-birefringent material. In the used diffraction order, light entering the crystal is polarized perpendicular to the plane of diffraction, X-direction, and exits the crystal polarized parallel to the plane of diffraction, Y-direction. To simulate this a polarization rotator in the form of a Jones matrix is added near the grating. Because the grating normal is at a large angle θ with the incident rays, the Jones matrix doesn’t have the conventional form. It was empirically derived that the Jones matrix has to be defined as follows:

There are two main differences between the AOTF models in FRED and OpticStudio:

The zero order is modelled in OpticStudio by setting the grating order to zero, changing the coating on the AOTF efficiency object to 100% and changing the Jones matrix to A = D =1 and B = C = 0. In FRED, the grating and the Jones matrix are removed.

Figure 12.

AOTF in OpticStudio

4.5

Modelling of the Scene

The characteristics of the sources are given by the geometry of the scene and by the optical characteristics of the instrument. ALTIUS will be observing the limb of the atmosphere from a 725km orbit, which gives an object distance of about 3000 kilometers.

The bright limb scene is modelled in OpticStudio with a set of extended sources of the type Source Two Angle. The infield that covers about 100km x 100km of the bright limb is modelled with 21 sources, 2.1° wide in the X-direction and 0.09572° wide in the Y-direction. The near out-of-field is modelled with a further 21 sources on both sides of the infield sources and one wide source at the bottom. Together the sources cover a field of view of ±6° in the horizontal direction and +1.005°/-6° in the vertical direction. The details of the sources are given in Table 9.

The spectral radiance L of the bright limb source is given in photons/s/sr/cm²/μm. The radiant power Φ in W of the sources is calculated as:

with:

Table 9.

Angular dimensions and orientation of the sources used in OpticStudio to model the bright limb scene.

In FRED the bright limb is modelled with randomly distributed point sources. The infield is modelled with 8000 point sources. The distribution is shown in Figure 13. The same method is used for both left and right near out-of-field sources i.e. 8000 randomly distributed point sources for both sides. As the covered field is larger than the infield, the near out-of-field has a smaller sampling than the infield.

Figure 13.

The distribution of the UV infield sources over the field of view in FRED

The radiant power of the sources is calculated as for the OpticStudio sources but as a point source doesn’t have a solid angle, the solid angle of one detector pixel was used.

The approach of discrete random modelling was selected as a matter of convenience but further investigations have proven that the same approach than in OpticStudio can be used in FRED to simulate the scene as a continuous scene.

5.1

PSFs

Before performing a full stray light analysis, first some PSF’s of point sources were compared.

The detector in the analysis has 501x501 pixels of 21μm. This corresponds to the field of view of the VIS channel, 2°x2°. In the UV channel it corresponds to 5.7°x5.7° or almost 3 times the infield FOV.

The X- and Y- cross section of the PSF at the central field point at 350nm and at 650nm in first order are shown in Figure 14. The PSF at 300nm is very similar to the PSF at 350nm. The OpticStudio and FRED results are quite similar. In the UV, the PSF’s deviate slightly close to the specular peak, while in the VIS there is a small deviation at the edge of the detector. These differences are not explained by the differences between the ABg and Harvey-Shack models.

Figure 14.

Cross Section in X (left) and Y (right) of the PSF at the central field point at 350nm (top) and 650nm (bottom)

In the VIS channel in zero order the differences are much larger. Figure 15 shows the log10 of the PSF in W/mm² at 650nm in the zero order in FRED (left) and OpticStudio (right). From top to bottom an X-polarized source at the center field, a Y-polarized source at the center field, an X-polarized source at the corner field and a Y-polarized source at the corner field are shown. These are not really PSF’s in the traditional sense because direct light from the zero order is blocked so all light reaching the detector has either scattered or ghosted.

Figure 15.

The log10 of the PSF at 650nm in W/mm² in the zero order in FRED (left) and OpticStudio (right) from top to bottom: X-polarized source center field, Y-polarized source center field, X-polarized source corner field, Y-polarized source corner field

In FRED the energy distribution are similar for X and Y-polarized sources, only the irradiance levels are different. In OpticStudio there are large differences depending on the polarization of the source. Other differences are:

5.2

UV channel Stray Light

The stray light in ALTIUS is calculated as the stray light to signal ratio. Signal is the irradiance on the detector when a simulation is performed without scattering or ghosting. Infield stray light is found by subtracting the signal from the irradiance on the detector when a simulation is done allowing scattering and ghosting with the infield sources only. Near out-of-field (NOoF) stray light is the irradiance on the detector when a simulation is done allowing scattering and ghosting with the out-of-field sources only.

The stray light to signal ratio is shown in Figure 16 for 300nm and 350nm, for both FRED and OpticStudio, as a function of the tangent altitude. For each tangent altitude, the average over the horizontal field of view and over a 5km vertical band is taken. A negative stray light to signal ratio originates when more light scatters away from the bright lower part of the scene than light scatters back in from other parts of the scene. The stray light to signal ratios are not very different between FRED and OpticStudio. The main difference is that the ratio increases much more rapidly at higher tangent altitudes in OpticStudio. Further study is needed to explain the differences.

Figure 16.

Stray light to signal ratio as a function of tangent altitude, for 300nm (left) and 350nm (right)

5.3

VIS channel Stray Light

The stray light calculations for the VIS channel are currently in progress. The stray light to signal ratio is calculated following the same conceptual approach as for the UV, with the additional complexity of the stray light created by the 0-th order.