Tuesday, April 08, 2014

Omnivision Proposes BSI Sensor with No Color Filter

Omnivision patent application US20140084135 "Backside-Illuminated Photosensor Array With White, Yellow ad Red-Sensitive Elements" by Gang Chen, Duli Mao, Hsin-Chih Tai proposes to use different PD depths in BSI sensor to get pixels sensitive to white (PD extends all the way to the backside), Yellow, and Red (shallow PD). I'm not sure what are the advantages of this approach, but it's a simple and nice idea:


  1. Foveon x3 style colour filtering but without the fill factor advantage. Is there some advantage to the BSI process that can create more accurate spectral characteristics for the filtering?

  2. I think to work as intended there is at least one missing layer in this structure. You need a sort of VOD = vertical overflow drain to make this work right. So, starting with 109, the pinning layer, then 124 the n- SW "absorber" layer, then you need a p-layer, then an n-layer to form the VOD hump. This is so that carriers absorbed near the backside are repelled from diffusing and being collected in the absorber layer. I read this a few times and it seems this hump is neither described nor drawn. Without it, the QE of each pixel is going to be similar.

    So, aside from seeming not to work as intended, the proposed device has all the color separation problems of X3 without the advantage of no aliasing.


All comments are moderated to avoid spam.