`sensor die and sensors packaged at
`the wafer level and protected by a
`cover glass
`
`G. Humpston, A. Grinman, O. Jackl, M. Ebel
`
`G. Humpston, A. Grinman, O. Jackl, M. Ebel, "Optical performance of bare
`image sensor die and sensors packaged at the wafer level and protected by a
`cover glass," Proc. SPIE 6897, Optoelectronic Integrated Circuits X, 68970U
`(3 March 2008); doi: 10.1117/12.764716
`Event: Integrated Optoelectronic Devices 2008, 2008, San Jose, California,
`United States
`
`Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 07 Jul 2019 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
`
`PROCEEDINGS OF SPIE
`
`SPIEDigitalLibrary.org/conference-proceedings-of-spie
`
`Apple v. Corephotonics
`IPR2019-00030
`Exhibit 2019 Page 1 of 12
`
`
`
`Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 07 Jul 2019
`Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
`
`Optoelectronic Integrated Circuits X, edited by Louay A. Eldada, El-Hang Lee
`Proc. of SPIE Vol. 6897, 68970U, (2008) · 0277-786X/08/$18 · doi: 10.1117/12.764716
`
`Proc. of SPIE Vol. 6897 68970U-1
`
`
`
`
`
`
`
`
`
`Optical Performance of Bare Image Sensor Die and Sensors Packaged
`at the Wafer Level and Protected by a Cover Glass
`
`G Humpston
`
`Tessera U.K.
`Aylesbury, Buckinghamshire, HP22 4AZ, United Kingdom
`Email: ghumpston@tessera.com Tel: +44 1296-658-299
`
`A Grinman
`
`Tessera Israel
`Manhat Technology Park, Bldg.4, P.O.Box: 48328, Jerusalem, Israel, 91483
`Email: agrinman@tessera.com Tel: +972 2-679-8890
`
`O Jackl and M Ebel
`
`Schott AG
`Hüttenstr. 1, 31073 Grünenplan, Germany
`Email: oliver.jackl@schott.com, maida.ebel@schott.com Tel: + 49 5187-771 0
`
`
`ABSTRACT
`
`
`The yield of solid state camera modules declines with increasing imager resolution. The easiest means of compensating
`for this trend is to package the imager die before they are assembled into camera modules. The packaging is preferably
`accomplished at the wafer level. The package cover is a sheet of glass that forms part of the optical train of the solid
`state camera. This paper discusses the properties required of the glass and describes a computer model that was
`constructed to provide quantitative insight into its effect on the optical performance of image sensors. The study
`corroborates practical measurements that the cover glass has no significant effect on the low light sensitivity of a typical
`camera module. The cover glass does introduce some reflection losses and minor image aberrations, but these can be
`managed through the combination of attention to the design of the total optical path and making the cover glass as thin as
`possible.
`
`Key words: Solid state imager, wafer-level package, cover glass, optical model
`
`
`
`INTRODUCTION
`
`
`Solid state image sensors are manufactured in vast quantities. Typically each year more than 1 billion image sensors are
`produced, which primarily find application in portable electronics products such as camera ‘phones, digital still cameras
`and increasingly, laptop computers. Predictions are that this market will continue growing for some years as cell phones
`with multiple cameras become the norm and automotive driver aids enter the video age, which could entail 10 or more
`cameras being installed on each vehicle.
`
`
`
`
`
`Apple v. Corephotonics
`IPR2019-00030
`Exhibit 2019 Page 2 of 12
`
`
`
`Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 07 Jul 2019
`Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
`
`Proc. of SPIE Vol. 6897 68970U-2
`
`
`
`PACKAGING OF SOLID STATE IMAGERS
`
`
`In common with most other semiconductor devices, solid state image sensors require some form of enclosure in order to
`ensure their longevity. Traditionally for imagers this was achieved using chip-on-board (COB) assembly processes. In
`this scheme the imager is attached to a substrate and wire bonds form interconnections between bond pads on the die and
`lands on the substrate. The substrate forms the base of the enclosure. Over the die is then placed a lens turret, which
`houses the optical train of the camera. The lens turret is sealed to the substrate so the lower optical surface in
`combination with the sidewalls of the turret form an enclosure for the die. This is illustrated in Figure 1.
`
`This structure has two principal drawbacks. The first is that the cost of assembly is incremental for each die packaged.
`A second limitation is that the imager die is totally unprotected until the final assembly operation when the lens turret is
`attached. It is therefore not surprising that more than 90% of defects in camera modules are related to contamination by
`particles1. The short-term solution is to invest in clean rooms and operator training. However, many solid state camera
`module manufacturers are already operating Class 10 environments, or better, so there is not a great deal of scope for
`further improvement at affordable cost.
`
`
`
`
`Camera module assembled using chip-on-board processes. The lens turret housing forms a seal over the exposed die.
`Drawing not to scale. Source: Tessera
`
`
`Fig 1
`
`
`
`
`WAFER LEVEL PACKAGING
`
`
`Wafer-level packaging (WLP) is an alternative approach where the die are packaged while still in wafer form and the
`wafer is then singulated to free individually packaged die. WLP has the advantage that the costs of packaging are shared
`among the good die on the wafer, greatly reducing packaging costs per die. A wafer-level cavity package for an image
`sensor is achieved simply by applying a picture frame of adhesive around each die, attaching a glass wafer and then
`sawing the assembly to yield individual die, each with a cover over the delicate image sensor area. This process is
`illustrated schematically in Figure 2.
`
`
`
`
`
`Apple v. Corephotonics
`IPR2019-00030
`Exhibit 2019 Page 3 of 12
`
`
`
`Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 07 Jul 2019
`Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
`
`'4 4.
`
`'4
`
`* *
`
`I.
`
`I'
`
`Proc. of SPIE Vol. 6897 68970U-3
`
`
`
`
`
`
`
`Formation of a wafer-level cavity package. Left - the device wafer containing five die. Middle - application of the seal
`material to form a picture frame around the perimeter of each die. Right – attachment of a lid material to seal the cavity over
`each die. Singulation frees the packaged die from the wafer, an example of which is shown in Figure 3. Source: Tessera
`
`
`
`
`Fig 2
`
`
`
`
`
`Fig 3
`
`
`Image sensor packaged at the wafer level2 and provided with a ball grid array interface to simplify and cheapen attachment to
`a printed circuit board. Source: Tessera
`
`
`
`Wafer-level packaging provides two benefits that have great value for image sensors. Firstly, the dies are protected from
`the very first step of the assembly process so that yield loss from contamination is minimized. The second benefit is that
`it is possible to provide the packaged die with a ball grid array interface. This permits the camera module to be soldered
`to the main printed circuit board of the product at the same time as all the other semiconductor and passive components.
`It is predicted that by 2010 more than half of all image sensors will be protected by wafer-level packages.
`
`
`
`
`Apple v. Corephotonics
`IPR2019-00030
`Exhibit 2019 Page 4 of 12
`
`
`
`Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 07 Jul 2019
`Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
`
`Proc. of SPIE Vol. 6897 68970U-4
`
`
`The cover glass used for the wafer-level package must fulfil several criteria. These are:
`
`
`COVER GLASS
`
`• Optically transparent. The glass must allow incident radiation of the visible spectrum (400-700nm) to reach the
`image sensor. The through thickness transparency of the cover glass needs to be as high as possible since any
`attenuation has a direct bearing on the low light sensitivity of the camera.
`
`• Reasonably closely matched in thermal expansivity to silicon over the temperature range -40°C to +260°C. To
`benefit from the economics of wafer-level processing, the cover glass is bonded to the silicon die while both are
`in wafer form. This means that the expansivity of the glass must be closely matched to silicon above room
`temperature since joining of the glass to silicon entails an elevated temperature process.
`
`• Free of active species that could potentially leach out of the glass and degrade the delicate image sensor. In this
`regard, the elements normally found in glass that are particularly detrimental are the alkalis. Over time these
`can leach out of the glass and poison the semiconductor junctions.
`
`• Be available in wafer form, to match the diameter and features of SEMI standard silicon wafers (e.g. flats,
`notches). Similarly, the glass wafers must be manufacturable to exacting specifications in terms of defects,
`since most types of blemish in the glass will be visible as artefacts in the image captured by the sensor. A
`typical scratch/dig specification for image sensor cover glass is 50/05. In addition, the geometric properties such
`as thickness and total thickness variation (TTV) need to be well controlled.
`
`• Cost attractiveness. The packaging market is driven by challenging price roadmaps, thus high costs for materials
`and components are not acceptable. A major cost driver in the production of glass wafers is the polishing step.
`Choosing a hot forming method for the raw glass production that offers the desired thickness directly from the
`tank saves the expensive polishing step.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`COVER GLASS SELECTION
`
`
`There are very few glass types that meet the requirements set out above. The one selected for this application has the
`designation SCHOTT AF 453. Its properties and performance are proven to meet the requirements for 200mm
`manufacturing lines in mass production for several years. In order to provide a suitable glass type for 300mm production
`lines as well as single glass WLP solutions, the glass type with the designation SCHOTT AF 324 has been developed.
`
`AF 45 as well as AF 32 are melted using selected pure raw materials. This ensures the high luminous transmittance in the
`visible wavelength, as can be seen from Figure 4.
`
`
`
`
`Apple v. Corephotonics
`IPR2019-00030
`Exhibit 2019 Page 5 of 12
`
`
`
`Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 07 Jul 2019
`Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
`
`150
`
`300
`
`350
`
`400
`
`450
`
`500
`
`..
`— AF 3117.6.07 — AF 45 Wl7fl
`d0.5[
`d0.51[
`
`550
`
`600
`
`650
`
`700
`
`750
`
`000
`
`0.0025
`
`0-0:2
`
`0.0015
`
`0-0Z1
`
`0.0025
`
`0
`
`temperature rC}
`
`—AF45
`sroduced
`04-07-20
`
`—AF32
`sroduced
`07-06-12
`
`—Silcon
`Rr. 3157Y
`
`Proc. of SPIE Vol. 6897 68970U-5
`
`._____
`-0
`
`5.0
`
`100
`
`150
`
`200
`
`250
`
`30:
`
`35:
`
`400
`
`450
`
`520
`
`
`
`Fig. 4 Transmission of AF 45 and AF 32 as a function of wavelength. Source Schott AG
`
`
`The most important difference between the two glass types is the lower coefficient of thermal expansion (CTE) of AF 32
`which has a very close match to the CTE of silicon, as can be seen in Figure 5. As a result the bond between the cover
`glass and the silicon is very flat even for wafer sizes of 300mm diameter.
`
`
`
`
`
`Fig 5 Dynamic elongation in initial length ∆l/l0 with a heat-up rate of 5K/min of AF 32, AF 45,and silicon. Source
`Schott AG.
`
`
`
`Apple v. Corephotonics
`IPR2019-00030
`Exhibit 2019 Page 6 of 12
`
`
`
`Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 07 Jul 2019
`Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
`
`Proc. of SPIE Vol. 6897 68970U-6
`
`
`
`
`
`The two glass types AF 45 and AF 32 are free of alkalis, which is indicated by the AF (Alkali Free) in the glass name.
`The glass compositions do not intentionally contain any oxides of lithium, sodium or potassium.
`
`Both AF 45 and AF 32 are easy to process into the required wafer format. Table 1, below, shows an example of typical
`wafer specifications that can be achieved when choosing the above mentioned glass types.
`
`
`
`Requirement
`Size (diameter/square)
`Thickness
`TTV
`Roughness
`Cleaning
`
`Tolerance
`Specification
`± 0.1mm
`50 – 300mm
`± 10µm - ± 15µm
`0.1 – 1.1mm
`advanced: +/- 5µm
`standard: +/- 10µm
`
`standard: ≤ 1 nm
`Ultra Sonic (US) and Mega Sonic (MS) in
`clean room class 1000
`advanced: 5/1
`standard: 10/5
`wafer boxes or equivalent
`holes and cavities possible
`
`Surface defects (scratch/dig)
`Packaging
`Structuring
`
`
`Table 1 SCHOTT AF32 and AF45 wafer specifications and tolerances. Source Schott AG
`
`
`Both AF 45 and AF 32 are manufactured by the down-draw process which enables the production of many different
`thicknesses in the range of 0.1mm up to 1.1mm at a TTV of +/- 10µm as a standard. Compared to other hot forming
`technologies, down-draw tanks are relatively small thus the transition cost from one thickness to the other is not as high
`for float glasses. The many available standard thicknesses as well as the relatively low minimum order quantity for
`customized thicknesses eliminate the need for the expensive mechanical polishing process. Beside the fact that down-
`drawn glasses come in the right thickness they have a very smooth surface with a roughness of below 1nm RMS, which
`is more than adequate for a cover glass on a camera module package.
`
`
`
`
`OPTICAL PERFORMANCE
`
` A
`
` key determinant of whether a camera module is suitable for a particular application is its optical performance. A
`camera module fabricated using COB assembly and one built using an imager protected by a WLP differ in one key
`regard, namely that the component packaged at the wafer level has a cover glass as an additional element in the optical
`path. The objective of this work was to evaluate the effect of the cover glass on the optical performance of miniaturized
`camera modules, typified by the type of device commonly employed in camera ‘phones, using a combination of
`experimental techniques and numerical simulation.
`
`
`
`
`EXPERIMENTAL TECHNIQUE
`
`
`The test vehicle was a CMOS imager of 1.3 Mpixel resolution, with a pixel dimension of 3.3µm. The imager was
`married to test station optics that gave a field of view diagonal between -1.0 and +1.0 coordinates i.e. 81 degrees
`horizontal and 59 degrees vertical. The photosensitive area illumination efficiency (PAIE) was measured for each pixel
`colour (red, green and blue), resolved to 8 bits, with different cover glass thicknesses of 300, 400 or 500µm. Variation in
`PAIE between different imagers from the same wafer was also measured.
`
`
`
`
`
`
`
`Apple v. Corephotonics
`IPR2019-00030
`Exhibit 2019 Page 7 of 12
`
`
`
`Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 07 Jul 2019
`Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
`
`Proc. of SPIE Vol. 6897 68970U-7
`
`
`
`RESULTS OF THE EXPERIMENTS AND DISCUSSION
`
`
`The measured PAIE per pixel colour for different cover glass thicknesses is given in Figure 6. Two conclusions can be
`drawn from the data. As expected, the PAIE is highest in the centre of the image area and decreases toward the
`periphery, decreasing abruptly beyond the field of view. By normalising the PAIE measurements relative to the 300µm
`cover glass it can be seen that the PAIE is reduced for each colour as the cover glass increases in thickness, as can be
`seen in Table 2. The magnitude of the trend is placed in context by measuring multiple die from the same wafer, each
`with a 500µm thick cover glass. Normalising the results to the first die, the fluctuations of PAIE for other die, presented
`in Table 3, show clearly that the die-to-die performance variation on a given wafer is at least of the same magnitude, if
`not larger than the impact of the cover glass thickness.
`
`
`
`PAIE measurements for different top glass thicknesses
`
`-0.8
`
`-0.6
`
`-0.4
`
`-0.2
`
`Field of view
`0
`
`0.2
`
`0.4
`
`0.6
`
`0.8
`
`225
`
`175
`
`PAIE
`
`Red 300
`Green 300
`Blue 300
`Red 400
`Green 400
`Blue 400
`Red 500
`Green 500
`Blue 500
`
`
`Table 2
`
`
`
`
`
`
`Measured PAIE per pixel colour normalized for a cover glass thickness of 300µm. Source: Tessera.
`
`Red coloured pixels
`Green coloured pixels
`Blue coloured pixels
`
`
`Table 3
`
`Min. fluctuation die-to-die
`
`0.9%
`-2.5%
`5.1%
`
`Average fluctuation of die-
`to-die
`2.3%
`-0.6%
`9.9%
`
`Max. fluctuation of die-to-
`die
`4.6%
`1.6%
`13.6%
`
`Measured PAIE per pixel colour for multiple die from the same wafer, with 500µm thick cover glass, normalized
`to the first die measured. Source: Tessera.
`
`
`
`Fig 6 Measured PAIE per pixel colour for different cover glass thicknesses. Source: Tessera.
`
`
`
`
`
`Red coloured pixels
`Green coloured pixels
`Blue coloured pixels
`AVERAGE
`
`Glass 300 microns
`100%
`100%
`100%
`100%
`
`Glass 400 microns
`98.9%
`99.2%
`99.5%
`99.2%
`
`
`
`Glass 500 microns
`96.7%
`99.8%
`90.6%
`95.7%
`
`Apple v. Corephotonics
`IPR2019-00030
`Exhibit 2019 Page 8 of 12
`
`
`
`Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 07 Jul 2019
`Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
`
`Proc. of SPIE Vol. 6897 68970U-8
`
`
`
`
`
`The results of the experiment imply the cover glass has no significant impact on the optical performance of the image
`sensor. However the difficulty of making accurate measurements, plus the die-to-die variation encountered suggested
`numerical modelling of the optical system might be a worthwhile alternative approach.
`
`
`
`NUMERICAL SIMULATION
`
`
`The objective of the numerical simulation was to analyse the optical performance of a solid state camera module, with
`and without a cover glass over the imager, via PAIE and crosstalk calculations. To do this a computer model (ZEMAX)
`was constructed to perform the required non-sequential ray trace simulation.
`
`The imaged object was a 0.28mm x 0.28mm square surface located 100mm from the imaging lens. The source emits
`unpolarized light at 510nm. The object was a diffuse cosine source with distribution:
`
`I(θ) ≈ I0×cos(θ)Cn
`
`where Cn = 10,000 such that the illumination is hyper-Lambertian, namely a narrow lobe with rotationally symmetric
`distribution about the optical propagation axis.
`
`The lens is a 50mm double-Gauss wide field of view lens with an effective focal length of 5.8mm. The lens field of view
`is 46° and the simulations were carried out at three fields of view: 0°, 13.4° and 20°. The lens f-number (f#) was varied
`between 1.2, 2.0 and 4.0 by changing the diameter of the lens entrance pupil.
`
`The CMOS device was reduced to a resolution of 5x5 pixels to decrease the computation time. A larger array would not
`add further insight as all of the optical phenomena of interest are already present in this low resolution device. The pixel
`size is 3.3µm.
`
`On top of the CMOS imager was constructed a 5x5 micro lens array, built from toroid lenses. The lenses have a flat back
`face, 2.0µm thickness and identical radius of 2.37µm on both curvatures.
`
`Where present, the cover glass is 300µm thick and spaced 40µm from the front face of the CMOS imager to mimic the
`real life wafer-level package. The insertion of the AF45 glass plate to the system introduces a focal shift of
`0.142±0.01mm. This is expected and was accounted for in the simulation by shifting the CMOS imager to the new focal
`plane.
`
`Simulating the optical behaviour of the camera was carried out in two steps. First, a sequential ray trace was used to
`locate the image plane of the double-Gauss lens for each field angle and f-number used. The criteria used for
`optimization was minimum spot size. Next, the CMOS imager with micro lens array was added and a non-sequential ray
`trace was performed with the source object inserted 100mm in front of the lens. The elements were positioned such that
`the imaging lens was focused on the photodiode layer. A detector object was placed on the back side of the photodiode
`layer with dimensions: 0.02×0.02 mm and 128×128 pixels. A Monte Carlo ray trace with 200,000 rays was run and the
`incoherent intensity of all rays hitting the detector plane recorded. The total power and peak irradiance were then
`derived. The performance of the model was assumed adequate when the resulting image resolved the discrete 5×5 pixel
`array. Such an image indicates that the micro lens array is functioning and the illumination is concentrated on the pixels.
`
`PAIE is defined as the normalized, integrated intensity of the active pixels. Calculation of this parameter was as follows:
`
`
`a) Non-sequential calculation of the intensity image using Monte-Carlo simulation.
`b) The diffraction spot of the system was calculated from the point spread function for each F-number (1.2, 2.0 and
`4.0).
`
`
`
`Apple v. Corephotonics
`IPR2019-00030
`Exhibit 2019 Page 9 of 12
`
`
`
`Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 07 Jul 2019
`Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
`
`Proc. of SPIE Vol. 6897 68970U-9
`
`
`
`
`
`
`c) Both intensity image and diffraction spot were convoluted.
`d) A binary 2-D grid corresponding to the pixel dimension (pitch, size) and fill factor (40%) was generated.
`e) The grid and convoluted image were multiplied.
`The PAIE is the ratio of the integrated image from step (e) and the convoluted image, step (d).
`
`The pixel crosstalk is calculated using the following scheme using diffraction-limited paraxial optics instead of the
`double-Gauss lens. This is in order to eliminate aberrations due to the imaging lens:
`
`
`a) The diffraction spot is calculated via the point spread function of a single micro lens + CMOS layers.
`b) An intensity image is simulated with a Monte-Carlo run. However, unlike the previous case, a point source was
`used and the beam was focused on a single pixel in the micro lens array.
`c) Both images were convoluted and multiplied by the 2D grid as in the PAIE routine.
`d) The intensity of each of the central, illuminated pixel and the nearest 8 pixels is summed.
`e) The XT is defined using the following two methods:
`• The ratio of the highest neighboring pixel and the illuminated pixel (XTMAX).
`• The ratio of all 8 neighboring pixels and the illuminated pixel (XTTOT).
`
`RESULTS OF THE NUMERICAL SIMULATION AND DISCUSSION
`
`
`Table 4 gives the total power and peak irradiance calculated at the detector plane following non-sequential tracing of
`200,000 rays for a COB imager and one in a wafer level package. Table 5 gives the calculated PAIE and XTMAX and
`XTTOT for the same two package styles.
`
`
`
`Peak Intensity (W/cm2)
`
`7.59E+05
`7.16E+05
`5.78E+05
`7.02E+05
`6.48E+05
`5.09E+05
`
`7.6%
`9.5%
`11.9%
`
`Package
`
`Field
`
`Total Power (W)
`
`COB
`COB
`COB
`WLP with Schott AF 45
`WLP with Schott AF 45
`WLP with Schott AF 45
`
`COB vs. WLP
`COB vs. WLP
`COB vs. WLP
`
`
`
`0.253
`0.00
`0.164
`0.67
`0.091
`1.00
`0.223
`0.00
`0.143
`0.67
`0.081
`1.00
`Comparative results
`0.00
`11.9%
`0.67
`12.7%
`1.00
`11.1%
`
`
`Table 4 Total power and peak irradiance at the detector plane (200,000 rays). Source: Tessera.
`
`
`
`
`Apple v. Corephotonics
`IPR2019-00030
`Exhibit 2019 Page 10 of 12
`
`
`
`Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 07 Jul 2019
`Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
`
`Proc. of SPIE Vol. 6897 68970U-10
`
`Package
`COB
`COB
`COB
`COB
`COB
`COB
`COB
`COB
`COB
`
`field
`0.00
`0.67
`1.00
`0.00
`0.67
`1.00
`0.00
`0.67
`1.00
`
`0.00
`0.67
`1.00
`0.00
`0.67
`1.00
`0.00
`0.67
`1.00
`
`F#
`1.20
`1.20
`1.20
`2.00
`2.00
`2.00
`4.00
`4.00
`4.00
`
`1.20
`1.20
`1.20
`2.00
`2.00
`2.00
`4.00
`4.00
`4.00
`
`PAIE
`0.654
`0.589
`0.600
`0.673
`0.659
`0.666
`0.674
`0.660
`0.666
`
`0.662
`0.595
`0.579
`0.683
`0.622
`0.592
`0.683
`0.638
`0.607
`
`XTMAX
`[%]
`2.22
`
`
`2.05
`
`
`2.05
`
`
`
`3.15
`
`
`2.38
`
`
`2.32
`
`
`
`XTTOT
`[%]
`7.90
`
`
`7.33
`
`
`7.33
`
`
`
`10.80
`
`
`8.46
`
`
`8.25
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Table 5 PAIE and XT for camera modules with and without a cover glass over the image sensor die. Source: Tessera.
`
`Under ideal conditions, the only difference between the two package configurations should be reflection and absorption
`losses due to the cover glass. These effects will be most noticeable at large field angles. The AF45 glass has an index of
`1.53 at 0.5µm and thus a reflection loss of about 9% is expected. The measured differences in the total power between
`amount to approximately 12% (Table 4) and are within the experimental error of these simulations, as some light is lost
`due to scattering from surfaces. However, as shown in Table 6, the difference in PAIE values is small, amounting to
`about 3% when averaging over all field angles and F-numbers. This value is also small compared with the measurements
`reported above for die-to-die variations within a single wafer.
`
`
`WLP with Schott AF 45
`WLP with Schott AF 45
`WLP with Schott AF 45
`WLP with Schott AF 45
`WLP with Schott AF 45
`WLP with Schott AF 45
`WLP with Schott AF 45
`WLP with Schott AF 45
`WLP with Schott AF 45
`
`
`
`
`
`
`mean
`Std dev
`
`COB
`0.649
`0.032
`
`WLP
`0.629
`0.040
`
`Overall PAIE comparison between imagers with and without cover glass. Source. Tessera.
`
`
`Table 6
`
`It should be noted that PAIE represents the efficiency of the system and therefore the reflection losses shown in Table 4
`are cancelled out in the PAIE calculation. Crosstalk values are slightly lower for the COB imager since the AF45 cover
`glass adds some noise to the system. This could be compensated for by thinning the glass. Both methods of XT
`calculation result in similar ratios for the two cases. Crosstalk is generally more sensitive to added surfaces in the
`vicinity of the surface plane than to surfaces further out, implying that a taller cavity package is preferable whenever
`permitted.
`
`
`
`
`Apple v. Corephotonics
`IPR2019-00030
`Exhibit 2019 Page 11 of 12
`
`
`
`Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 07 Jul 2019
`Terms of Use: https://www.spiedigitallibrary.org/terms-of-use
`
`Proc. of SPIE Vol. 6897 68970U-11
`
`
`
`
`
`CONCLUSIONS
`
`
`Packaging of image sensors at the wafer level is becoming essential as resolutions increase and pixel dimensions
`decrease. While this is technically tractable and economically favourable a packaged image sensor must have a cover
`glass as an additional component in the optical train. The cover glass has to meet very exacting specifications but
`suitable products are commercially available nevertheless. The effect of the cover glass on the optical performance of the
`camera has been measured experimentally and analysed by numerical simulation. The cover glass has negligible effect
`on the low light sensitivity of the camera module because its transmittance is so high. As expected, the cover glass
`introduces reflection losses and some minor image aberrations. The reflection losses can be managed by designing the
`optical system to function at low chief ray angles while the aberrations are minimized by making the cover glass as thin
`as possible.
`
`
`
`REFERENCES
`
`Chowdhury, A, "Camera Module Assembly and Test Challenges", Semiconductor International, January 2006
`
`Tessera, SHELLCASE® RT image sensor package
`
`Schott AG, SCHOTT AF 45TM aluminoborosilicate glass technical specification TE - AF45TM and physical and
`chemical specification PCP – AF 45TM
`
`Schott AG, SCHOTT AF 32TM aluminoborosilicate glass technical specification TE – AF 32TM and physical and
`chemical specification PCP – AF 32TM
`
`
`
` 1
`
`
`
` 2
`
`
`
` 3
`
`
`
` 4
`
`
`
`
`
`Apple v. Corephotonics
`IPR2019-00030
`Exhibit 2019 Page 12 of 12
`
`