Hyperspectral cameras divide the light spectrum into many frequency bands with the aim of associating each pixel in the image of a scene with a specific portion of the spectrum. The goal is to find spectral “signatures” that can help locate objects and reveal processes. For example, a known spectral signature for oil helps mineralogists find new oil fields. These cameras are used in astronomy, agriculture, biomedical imaging, mineralogy, physics, surveillance and other areas. At present they are complex, bulky, heavy and expensive due to their optical components.
Imec researchers will discuss a wafer-scale CMOS-compatible approach that may lead to compact, high-performance, scalable and mass-producible versions. They monolithically built and integrated optical filter arrays atop CMOS image sensors using standard semiconductor fabrication techniques. The arrays separate light into different wavelengths by means of interferometers with varying cavity lengths. They built three different filter arrangements that each showed well-defined narrowband frequency response: a staircase-like structure; a tiled layout with filters laid out in large squares atop groups of pixels; and a mosaic layout with filters arranged onto individual pixels at a 5.5µm pitch. These arrangements show that the technique can be used to create customized filter layouts to match an application’s requirements.
In the images above, from left to right, (a) shows 200mm image sensor wafers with different hyperspectral filter arrangements; (b) is a conceptual representation of a tiled hyperspectral snapshot imager and (c) is this sensor configuration in a package; (d) is a conceptual representation of the mosaic hyperspectral snapshot imager and (e) is this mosaic image sensor in a package; and (f) is a series of microscope images of mosaic snapshot imagers with different filter configurations, showing that different arrangements are possible. Each pixel within a cell is a different spectral filter.
(Paper #10.5, “A CMOS-Compatible, Integrated Approach to Hyper- and Multispectral Imaging,” A. Lambrechts et al, Imec)