Image sensor blog from Japan puts together few words about Sony presentation at IEDM 2009 titled "0.9um Pitch Pixel CMOS Image Sensor Design Methodology".
As far as I was able to understand, Sony built 0.9um, 1.12um, 1.4um and 1.75um pixels using 0.18um FEOL and 65um copper BEOL process. All the pixels are FSI, rather than BSI. The pixels design was based on simulation results optimizing microlens shape, stack hight and metal opening.
1.1um pixel performance was considered OK. 0.9um pixel QE was unacceptably low. The article also mentions SNR10 being of order of 100Lux, not clear for what pixel size.
The published IEDM paper makes no reference to SNR.
ReplyDeleteIt looks like that the blogger from Japan got the SNR10 figure from his private communication. At least this is my interpretation of the Google translation of the article:
ReplyDeleteμ lens shape aperture metal, stack height and other optic simulation using the parameters and guidelines, and then actually create the device (0.9, 1.12, 1.4 and 1.75um pixels. all FSI. FEOL the 0.18um process, The BEOL Cu 65nm process) confirmed the validity of the method, and the contents of (1.1um pixels is OK. 0.9um pixels is serving out the terms of QE).
I have also discussion of other people, BEOL using the 65nm process, and, stack height is low enough to be of decent quality at FSI (SNR10 ~ 100Lux) 1.1 ~ 1.2nm resolution devices that can be received the impression (dark current, read noise and consideration, etc.)
Babylon on-line translation comes out a little bit more legible:
I perform optics simulation with parameters such as μ lens shape, metal opening, stackheight and make a guideline, and really make a device0.9, 1.12, 1.4 extendsA 1.75um pixel. All FSI. 0. FEOL18um process, BEOL are the との contents that Cu65nm process) identified the effectiveness of the techniqueThe 1.1um pixel is OK'd. The 0.9um pixel doesn't meet a condition of QE;). If discussed it with the other companies, but stackheight is low enough with a 65nm process in BEOL again; of a picture moderate even as for FSI (SNR10?100Lux)1.1?The 1.2nm pixel device had a possible impression (consider dark current, readnoise).