PTB >>Wafer Level Camera Technology

Film cameras were traditionally manufactured by discrete assembly, which means each component was fabricated as an individual item, tested if necessary, and then assembled into the final working product. The advent of solid state imaging did little to change this approach. The film was merely replaced by a light-sensitive electronic component. The only significant change was that the mechanical shutter was rendered obsolete and its action generated electronically within the imager die. Film and solid state cameras co-existed happily until 2001 when the first mobile phone incorporating a camera was introduced. Solid state cameras are now found in cars, trucks, toys, laptops, net books, machine tools, security systems, etc. In less than 10 years, the number of solid state cameras produced annually has exploded from tens of thousands to billions.

By far the largest market for solid state cameras remains mobile platforms, principally cell phones and compact digital still cameras. These markets are extremely fashion-conscious and the current vogue is for extreme thinness. While the discrete approach to manufacturing camera modules has thus far been able to deliver smaller and cheaper, that may not be possible in the future. The laws of physics limit the dimensions to which the camera module can be shrunk, while there is a finite amount of cost that can be squeezed out of a supply chain that manufactures a plethora of discrete parts and assembles them into a product. The solution to both of these problems is to switch to wafer level manufacturing. A wafer level camera has three principal components: the image sensor, the housing for the image sensor, and the optical train.

Image Sensors
Image sensors used for wafer-level cameras differ from conventional camera sensors in one important regard - pixel dimensions. Camera module height is strongly influenced by pixel size. In modern camera phones, sensors with 1.4μm pixels are common, and most imager companies show roadmaps out to at least 0.9 μm. One consequence of using imagers with very tiny pixels is that camera module yield can suffer badly from particle contamination. Even a Class 10 clean room environment, working to specification, contains particles that are large enough to cause black pixels if they get lodged on the sensor surface. The solution to this problem is to enclose the sensor die in a protective housing as the very first step of camera module manufacture. Discrete packaging would be prohibitively expensive. However, with approximately 3,000 die on an imager wafer, a wafer level approach results in a package cost of a few cents per die.

Figure 2. Imager die housed in a wafer-level package that uses a low cost through silicon via technology to connect the die bond pads to the package ball grid array interface. (Source: Tessera)

Wafer Level Packaging
Wafer-level packaging of image sensors is conducted as follows (Figure 1). First, a picture frame seal of adhesive is placed around the optically active area on each die. Next a cover glass is attached. Finally the wafer is diced to free individually packaged die. The wafer-level package provides two very important benefits to camera phones. Obviously, the package provides protection from the ambient environment (humidity, salt etc.) and any dirt that does land on the cover glass is removed from the focal plane, causing no defect in the image. Second, the perfectly flat glass lid of the package is an ideal substrate for attaching the camera optics. The problem with the wafer-level package structure is that the bond pads on the front face of the die are trapped beneath the cover glass and are inaccessible.

The traditional and commercially successful solution is to use SHELLCASE interconnects, that effectively warp the bond pads round the edge of the package to lands on the underside. To date more than 1 billion imagers have been manufactured using this approach. More recently, the technology has been developed to manufacture through silicon vias (TSVs) to connect the die bond pads to the package lands. An example is shown in Figure 2. Wafer level packages for imagers are provided with a ball grid array interface, making it possible to attach the package to a printed circuit board at the same time as all the other surface mount components required to build a compact camera or cell phone.

Figure 3. Lens arrays can be manufactured at the wafer scale with circular base lenses (l) or square base lenses (r), which can have the same spacing and pitch. Square base lenses would be expensive to make by conventional manufacturing techniques, but wafer-scale manufacturing methods are insensitive to lens shape. (Source: Tessera)

Wafer Level Optics
In its simplest form, wafer-level lens manufacturing involves making wafersized masters, one for each optical surface of each lens. Each master contains a closely packed array of cavities. One method is to create the lens shapes in photoresist using grey-scale processing to obtain 3D profiles. The photoresist is then cured and coated with a thin layer of electroless nickel applied to boost the durability. A polymer is cast over the nickel to provide mechanical support, and the photo resist is then removed. A glass plate, or wafer, is spin-coated with liquid polymer. The master is then pressed into the liquid film, which flows to fill the cavities.

The polymer is cured to fix the lens shape enabling the master to be released and reused. Not only is the cycle time fast, but the key economic advantage of wafer-scale lens manufacture is that thousands of lenses can be made simultaneously on a single 200- mm diameter wafer, with the materials and process cost being divided among the good parts on the wafer. Favorable economics isn't the only reason for the adoption of wafer-level lens manufacture. This process makes it possible to manufacture optical components that have functionality far beyond what polished glass or injection-molded plastic lenses can achieve. For example, using wafer level processing it is no more difficult to make axially symmetric lenses than asymmetric lenses.

Figure 4. Exploded and perspective view of an optical train for a field-widened pin hole imaging system. It comprises a half ball lens, aperture disc and plano-convex lens and almost 180 degrees field of view with considerable depth of field. The substrate plate forms part of the plano-convex lens. Note the two lenses have different diameters and can be made of different materials. This type of optical component is expensive, if not impossible, to manufacture using discrete components. (Source: Tessera)

Close-packed lens arrays are optically more efficient when the lenses are rounded squares, rather than circles. As can be seen by comparing the two cases in Figure 3, the fill factor of the square-base lenses is proportionately higher. Another benefit of wafer level manufacture is that the two optical surfaces of each lens are made using totally independent processes. This means their shapes have no physical connection and they can even be different sizes. More interesting still is that the two optical surfaces can be produced from different materials possessing different optical properties.

Because the master only comes in contact with liquid polymer it does not have to be rigid like a mold for an injectionmolded plastic lens. Instead, the master can be compliant, or slightly rubbery. This makes it possible to design optical surfaces with re-entrant features and other unusual profiles. One of the limits on wafer level lenses is achievable lens sag, which, for commercial optical polymers and mastering

techniques, is around 200 μm. Higher values allow a camera module to maintain the same form factor but grow progressively in resolution. By developing proprietary materials and processes at least one company is demonstrating lenses with >500um sag and >50 degree slopes on pre-production samples for 8 and 12 MP (megapixel ) wafer level cameras. At the other extreme, a wafer level lens for a VGA camera with 60 degree field of view and F-number of 2.8 can be built that is only 0.5mm thick.

The optical train of a camera module consists of more than just lenses. To function correctly it must also contain, apertures, infrared filter, pupils, and anti-reflection coatings to name a few. All of these components can be realised with semiconductor processing techniques at the wafer level. The net result is a completely integrated optical component that has all the functionality necessary for a solid state camera. Figure 4 shows an innovative example of the type of integrated optical component that can be realized using wafer-scale lens manufacturing techniques. Two very important metrics in manufacturing optical assemblies are alignment and focal length. Because the integrated optical component is made using semiconductor-based processes and equipment it can be mated with an imager die without the need for manual adjustment of focus, resulting in significant savings in the cost of test.

Figure 5. Solid state camera manufactured at the wafer level comprising an integrated optical component and an imager die packaged using a SHELLCASE wafer-level packaging technology. The camera module measures 2.5mm high and has a BGA interface, enabling it to be attached to a circuit board using standard lead-free solder reflow. (Source: Tessera)

Wafer Level Camera
Manufacturing a wafer-level camera module requires mating a wafer-level optical component and a wafer-level packaged imager. The resulting product is depicted in Figure 5. Critical to this operation is achieving precise alignment in plane and rotation, together with accurate spacing and planarity in the vertical direction. By making use of the novel lens shapes made possible by wafer-level lens manufacture, together with desensitising positional errors when translating from wafer alignment to component alignment, the necessary tolerances can be achieved with commercially available equipment.

Indeed, it is possible to design the camera module so that it can be assembled without requiring live adjustment of focus. Ideally, the components of each camera module would all be fabricated on wafers of identical diameter, the wafers stacked and bonded, then the assembly diced to produce individual camera modules. Currently, this is not the optimum approach since the diameter of the optical components and the imager die do not match; the semiconductor die is larger than its optically active area. Instead the optical wafers are combined and individual lens stacks bonded to each die. This is only possible because the lid of the wafer-level package is perfectly flat and level and spaced exactly from the imager die.

Through judicious choice of materials, it is possible for a wafer level camera to survive the thermal excursion of the lead-free solder reflow cycle. Con - ventional camera modules are unable to withstand these temperatures and must be interfaced to the phone by a flexible lead and connector. This arrangement is expensive and mechanically not robust, with failure of this interconnect being one of the leading causes of camera phone returns. A reflowable wafer level camera can be attached and interconnected to the main printed circuit board of the handset using the ball grid array interface on the underside of the wafer level imager package at the same time as other components using a standard surface mount process.

Wafer level cameras clearly address the issues of cost and form factor, but as yet do not have any significant impact on performance. Performance improvements are likely to follow from the incorporation of technologies like Smart Optics, which give rise to optical innovations such as continuous depth of field, ultra fast lenses and optical zoom with no moving parts. According to market analysts, this could eventually result in reflowable wafer level cameras taking more than 30% market share by 2012.

This article was written by Giles Humpston, Director of Research, and Yehudit Dagan, Vice President Program Integration, Tessera (San Jose, CA). For more information, contact Ms. Dagan at ydagan@tessera.com, or visit http://info.hotims.com/22920-200.