Dynamic Multispectral Imaging Using the Vertical Overflow Drain Structure

Abstract

Multispectral imaging enables discrimination of spectra beyond the 3-D spectral space of human vision. Most multispectral systems are complex and bulky, and simpler monolithic designs usually require several pixels in order to generate one multispectral data point. Here, we use the vertical overflow drain (VOD) structure to enable monolithic multispectral imaging in a single pixel. We show that by controlling the substrate bias voltage it is possible to change the effective depth of the photodiode, thus enabling dynamic tuning of the pixel's spectral response. A small voltage change (<;5 V) is shown experimentally to reduce the red light quantum efficiency by ~ 40 %, while the blue light quantum efficiency is reduced by <;10%. Using this effect, we demonstrate an ability to discriminate between different monochromatic illumination sources with 20-nm spectral resolution in a single pixel. In addition, we present an RGB image taken using an off-the-shelf charge coupled devices with no color filters by relying solely on the VOD mechanism. Finally, we present process and device simulations suggesting that this mechanism can be implemented in any pinned photodiode pixel with a VOD barrier, including in CMOS image sensors fabricated on n-type starting material.