The charge dumping or "discharge" regions are said to improve the color separation:
Thanks to AA for the pointer!
Friday, January 18, 2013
Sony Applies for Another Layered Sensor Patent
SonyAlpha Rumors noticed yet another Sony patent application talking about a Foveon-like layered sensor, the US20130009263. The main proposal is to add photoelectron dumping regions 24-1 in between the photodiodes 23B, 23G, 23R:
The charge dumping or "discharge" regions are said to improve the color separation:
Thanks to AA for the pointer!
The charge dumping or "discharge" regions are said to improve the color separation:
Thanks to AA for the pointer!
Subscribe to:
Post Comments (Atom)
Dick Merrill would have said see! ... miss him.
ReplyDeleteI was wondering, does vertacolor make sense for BSI? What do you think? (Vertacolor was Foveon's internal name before marketing said to use X3).
I guess the consensus with these BTJ schemes is that it is hard to make pinned photodiodes with fully depleted reset and in-pixel charge transfer? And that's why noise is a problem?
And the broad spectral curves don't help noise much either...
ReplyDeleteIt could work with BSI but aside from reversal, it would be tricky to get the blue layer (so to speak) close to the back surface.
ReplyDeleteYes complete charge transfer is tricky. In-pixel CDS for 3T operation becomes more possible with stacked 3D structures but pixel pitch remains an issue.
Color performance of X3 is not easy to achieve and generally takes a lot of processing to get acceptable color. The overlaps are terribly large (likely the motivation for the subject patent) so even with CDS, YSNR10 is not so good. Basically, subtracting two large noisy signals give a small signal with even more noise. This is the issue with negative elements in the color correction matrix. Lastly, for sub-diffraction limit pixels, who cares about spatial sampling color issues? It is all blurred anyway.
For sure 3-layer devices are tantalizing. I bent my pickaxe a couple years ago on a consulting job on this topic - even with 2 layers and filters there is trouble. It is all about YSNR10. Hard to beat Bayer (or RGBnW).