2.1

Optical design and image sensor curving

For low volume high end applications (e.g. space, scientific, astronomical applications, and more), SILINA is a service company to support the development of state-of-the-art optical instruments using curved image sensors. We can support on the optical design by considering the capability to curve the image sensor selected for the application / instrument.

SILINA does not develop new CIS. SILINA curves already existing CIS. We work closely with the instrument designers, manufacturers, integrators, but also with CMOS image sensor manufacturers. We can support the various players who discuss early stage mission phases to assess how a curved image sensor can improve the characteristics and boost the overall performance of the optical system. The curvature step is just an additional step to a full manufacturing process of an image sensor. The electronics control board and the package remain the same as for a flat image sensor.

2.2

Location and facilities

SILINA’s technological development and production site is based close to Aix-en-Provence, in France. We operates in a fully equipped 700m² ISO 3 to 5 cleanroom dedicated to back-end semiconductor. We have access to a private area in which we can operate in total confidentiality and autonomy. We have been developing our own EMVA compliant and fully automated electro-optical test-beds (image 1(a)). We also have the capability to perform environmental tests: thermal cycling, thermal shock, vacuum, and humidity (image 1(b)).

Figure 1.

CIS characterization test beds and environmental test equipment

2.3

The manufacturing process

The manufacturing process to obtain curved image sensors is decomposed by several steps. The first one consists of receiving CIS as wafer or bare die. Then, we perform a back grinding step : the wafers / sensors are thinned down to make them flexible. On wafers, we perform a dicing step in order to individualized the dies. At that step, the dies are ready to be curved. The process developed by SILINA enables to reach different shapes : spherical, aspherical, toroidal and custom shape based on instrument’s needs. We can reach concave and convex shape but also extremely flat shape. The curving process can be applied to any CIS format from 1/3” to full frame and more (no upper limit), either monochrome or color sensors, with or without micro-lenses. The process does not depend on the pixel pitch. We have already demonstrated the capabilities on front side (FSI) and back side (BSI) illuminated CIS and we are currently working on back side stacked image sensors.

2.4

Packaging capabilities

Once the sensor is curved, we can perform the final packaging steps: a wire bonding step and the bonding into the ceramic package. A glass lid on top of it can be added if needed.

Two characterization steps occur after curving and packaging : the shape analysis and the electro-optical performance characterization. The shape of the sensors is characterized with a confocal profilometer. We record the 3D shape after curving with a resolution that can reach the sub micrometer. From the 3D map, we can extract, among other parameters, the real radius of curvature and the deviations to the targeted shape. Regarding the electro-optical performance, we develop our own EMVA compliant fully automated test bed. Additional capabilities are currently integrated to perform MTF or angular response characterization, as well as a monochromator and an active thermal control. In order to assess the good operating of curved image sensors in their life condition, we perform reliability test and characterization at the end of the process.

3.1

Test conditions

After having demonstrated the performance of CMOS image sensors after curving,⁵ reliability tests are performed for the very first time on that technology. The main objective is to guarantee the good performance of curved CIS in real environmental conditions. These tests represent an important step toward the industrialization of the technology.

The standards and operating conditions are displayed in Table 1. Test conditions have been inspired from microelectronic consumers standards JESD22-A108D and JESD22-A103D. Normally, a duration of 1000 hours is required to comply with the requirement, but we decided to consider 100 hours per test since it is enough to detect early failure modes. For these tests, temporary glass lids are applied on top of the sensors and packages are made hermetic, as displayed in Figure 2. The glass lid is removed before every sample characterization step associated to the shape metrology and the electro-optical performance.

Table 1.

Description of the different reliability test conditions with the corresponding standards

Figure 2.

Picture of curved image sensor samples inside the thermal chamber

Three different environmental tests have been performed:

A CTS CS-70/200-5 climatic chamber is used for uHAST and TC tests and a CTS T-40/25 thermal chamber for HTSL tests.

3.2

Samples description

In this paper, only concave spherical shapes have been considered, with two different radii applied on a 15.7x18.7mm CMOS FSI image sensors. The sensors are packaged into a ceramic package (see Figure 3). Results are compared to classical flat samples that undergone the packaging process. Sensor distribution per test is displayed in Table 2.

Figure 3.

Pictures of packaged flat image sensor (left) and a curved image sensor(right)

Table 2.

Samples description and distribution

3.3

Measurement of the shape error

To determine the impact of the different reliability tests on the sensors, a metrology inspection of the front side is performed, in order to identify potential mechanical failure such as cracks, wire bonding pads corrosion or delamination. The characterizations occur in the 48 hours following the end of the test.

After the curving step, the shape of image sensors is measured with a confocal profilometer. A 3D map of the sensor surface is extracted (Part a, Figure 4). The theoretical shape is subtracted to the 3D map in order to observe the deviations to the target shape (Part b and c, Figure 4). Technical efforts are focused on the minimization of these residuals in order to make curved image sensor technology accessible to fast optics. Early 2022, we had a maximum deviation below 20 um peak to valley (PTV). Today we have divided it by 2 and we have recently achieved 4 um PTV.

Figure 4.

Surface map of a CMOS sensor curved with a radius of 250mm. a) Original Map, b) Residuals to the theoretical shape after curving, c) Residuals to the theoretical shape after curving and HTSL test

The shape characterization is repeated after each reliability test in order to quantify the stability of the curved sensor shape. Table 3 describes the maximum deviation (Sz) seen on flat and curved sensor surfaces, before and after reliability test. Flat sensors are considered as the reference. We observe deviations Sz from the ideal flat shape up to 6 times higher than the ones of curved sensors from the ideal spherical shape. This means that the curving process we have been developing reaches better shape accuracy than traditional packaging techniques for flat image sensors. The Sz difference between before and after the reliability test does not exceed 4 um, with a median of 2 um all tests combined. The radius of curvature of the sensors does not change after the tests. The first three reliability tests in uHAST, HTSL and TC for 100 hours show no critical failure.

Table 3.

Peak to valley deviation to the theoretical shape of curved sensors before and after reliability tests

4.1

Before reliability test

Figure 5 shows the results of the characterization after the curving process. The first part of the EO characterization is done in dark condition. We measure the Dark Current (DC) and the Dark Signal Non Uniformity (DSNU) per radius of curvature (respectively Figure 5(a) and 5(b)). The DC of the curved image sensors are similar to the flat ones with a mean of around 33 e-/s/pix. The flat sensors have a DSNU of around 9.7e- while the curved image sensors have a DSNU of around 9.3e-. The results of the dark characterization are stable from one curvature to the others, both concave and convex. The second part of the characterization consists in measuring under homogeneous flat illumination at 530 nm. Figure 5(c) represents the photo-response nonuniformity (PRNU), Figure 5(d) the Quantum efficiency (QE), and Figure 5(e) the signal to noise ratio. Flat image sensors have a median PRNU of 0.58%. Curved image sensors have a median PRNU value of 0.61%. Despite one single sample at 0.91%, we can observe that the performance are similar compared to flat image sensors and stable between samples. To be more precise, additional measurements must be performed on a large quantity of samples.

Figure 5.

Electro-Optical performance before reliability tests

4.2

After reliability test

The same measurements have been performed on curved sensors post-reliability tests. Figure 6 shows the results of the characterization for the DC, DSNU, PRNU, QE and the SNR. As seen previously, curved and flat sensors with the same thickness respond the same way. Here, the results of curved and flat sensors have been grouped per reliability test and “T0” represents the performance before the tests. Curved sensors are functional after all tests and with the same performances as before. DSNU, PRNU, QE and SNR are stable.

Figure 6.

Electro-Optical performance after reliability tests

We note on Figure 6(a) an increase of around 10e- for the DC post HTSL and TC, but the dispersion of the data covers the initial performances. During these measurement, the camera has not been thermally monitored and controlled. We do not know the temperature of the sensors during these EO measurements, so it is difficult to conclude for now. Additional measurements must be done in order to conclude on this effects. Moreover, the next upgrade of the test bed consists in measuring the temperature of the image sensor in real time during the EO measurements.