A.

Instrument Overview

The MISTIGRI architecture briefly presented hereafter is the result of technical studies and trade-off led by CNES. For details, the reader can refer to paper [2]. The multispectral pushbroom imager, sensitive to the 8.6, 9.1, 10.3 and 11.5 microns regions, provides surface radiometric temperature and emissivity.

The Focal Plane is based on a single uncooled microbolometer VGA array of 640 (cross-track) x 480 (along-track) with a pixel pitch of 25 microns developed by French company ULIS. The Time Delay Integration (TDI) algorithm is applied to improve the NETD performance by summing around 30 independent spatial measurements of each line on the ground.

B.

Focal Plane Overview

The multispectral filter assembly is depicted figure 3. The filter is manufactured by assembling four linear sliced filter elements purple in this picture, so-called strips, each corresponding to one spectral band. The strips are bonded together side-to-side. These strips associated to one spectral channel are cut from one germanium plate over-coated with the required bandpass optical filter using a two dimensional accuracy precision process. The size of each stripe is approximately 15(l) x 3.5(w) with a thickness of 1mm.

Figure 3.

LWIR multispectral filter assembly

Next the filter assembly will be integrated close to the bolometer window, i.e. around hundred microns, as seen figure 4. The filter assembly will be first assembled into a structural housing with two lateral bezels and then bonded on the bolometer package thanks to specific alignment techniques. Small labyrinthine vents between bolometer window and filter will allow pressure equalization during lifetime.

Figure 4.

LWIR focal plane prototype

Figure 5.

Ray tracing simulating the propagation of light

During this development activity, market off-the-shelf LWIR narrow bandpass filters were purchased. We selected two different coating suppliers offering appropriate performances, i.e. bandwidth and rejection, and providing solutions in applications where a severe environment is anticipated.

C.

Ghost image analysis

The goal of this analysis was to predict quantitatively the amount and distribution of stray light that reaches the focal plane. The analysis was carried out with FRED® software.

For this study, since the filter assembly is in close proximity of the detector and with the relatively low f-number of the uncooled imaging system, the stray light is mainly caused by ghost image artifacts. The critical aspect was the optical crosstalk between neighboring bands.

Ghost images are produced by multiple-reflection (even number of times) of light on optical element surfaces. The magnitude of a ghost image depends upon the surface reflectivity (mainly the product of two reflectances: detector and bandpass filter) and the relative distances between optical surfaces which induce the out-of-focus and the lateral shift of the ghost image. The lateral shift is sensitive to misalignment, particularly to the tilt of the filter assembly parts.

Based on the modeling, the crosstalk level should not exceed 0.15% on the useful areas.

A.

Filter Assembly Level

First the VNIR flowchart and the manufacturing processes were re-assessed since machining operations are more challenging on the germanium substrate which is more friable than glass. In addition the gluing process was adjusted and revalidated on representative samples.

We also validated the opacity of epoxy used between strips i.e. transmission less than 0.1% in LWIR.

Two representative prototypes of four-color LWIR assembled arrays were produced showing the good reproducibility of this technology.

During the manufacturing stage we found that strips are very sensitive and any physical contact will cause damage. Thus the delicate nature of the LWIR strips requires that special procedures be followed for handling and cleaning. By practicing two different coating manufacturers we also found that the supplier origin of one of the products is less fragile than the other one.

The current state-of-the-art in term of positioning and sizing features is given in the table 1 with the comparison with VNIR domain.

TABLE I.

POSITIONANING AND SIZING FEATURES

(a)

with VNIR strip dimensions: (L x W x H) 100x1x1.5 mm

(b)

R&T results before industrialization activities

Additional samples and witnesses were used during environmental test campaign.

One prototype was subjected to laboratory environmental durability tests that consisted of a combination of slow temperature change thermal cycles and damp heat.

The sequence included:

Before and after each environmental test the co-planarity, the tilting and the center-to-center spacing between strips were measured to check the stability of the filter assembly design.

No significant defects in the useful areas were detected after visual inspection with automated optical inspection system.

And no positioning changes between strips, i.e. less than 2 microns, were measured by means of the use of three-coordinate measuring machine (figure 6).

Figure 6.

Filer assembly metrology tools

B.

Focal Plane Level

One prototype was produced see view on figure 4.

This focal plane assembly was then subjected to laboratory environmental durability tests that consisted of a combination of slow temperature change thermal cycles and damp heat.

The sequence, same as on filer assembly alone, included:

In addition in order to establish any changes in the behavior of the prototype, radiometric tests were realized thanks to the following equipments available at Sodern premises:

Figure 7.

CALIPSO& IASI Imagers test set-up

Figure 8.

Bolometer video acquisition chain

The radiometric evaluation was accomplished with two extended area differential blackbodies with temperatures between 30°C and 100°C.

Corrections of bolometer drift and non-uniformity as offset correction for each pixel in the array have been considered with our electronics. The first stage of the analog front-end electronics provides signal amplification and differential subtraction of the pixel-to-pixel offset. Digitization is performed on 14 bits, for an effective resolution of 12 bits. On micro-bolometers, the fixed pattern noise (FPN) caused by pixel-to-pixel offsets can typically reach 1/3rd of the detector electrical dynamic range. The subtraction function is driven by a coarse 4-bit digital to analog converter working at the video data rate. The correction of pixel-to-pixel offsets reduces by 16 the signal amplitude variations induced by FPN, and allows using a high video gain in order to minimize the quantization noise when observing low contrast scenes as shown figure 9.

Figure 9.

Flat field image before coarse FPN correction (right) and after correction (left). Nine pixels are significantly outside the normal distribution and were indicated as bad pixels.

This uncooled focal plane enables the user to view a scene as depicted figure 10. Dead zones between neighboring strips bands including bond line width limits the field of view.

Figure 10.

LWIR focal plane prototype