Wednesday, April 10, 2019

Interview with Sony 48MP CIS Designers

Sony publishes an article "Perspectives from the creators of the image sensor “microcosm”" with interviews with the IMX586 CMOS sensor designers. Few quotes:

"With smartphone cameras getting more and more sophisticated in recent years, every company has been striving to make pixels smaller to meet the demand for more advanced cameras that are still small enough to fit in a phone. So, in order to stay ahead of the competition, we needed to develop even smaller pixels. With the IMX586, we were able to achieve a pixel size of 0.8 μm, which in turn made it possible to deliver a high resolution of 48 effective megapixels even on a compact sensor of 1/2 inch (8.0 mm diagonal).

Downsizing even 0.1 μm is, in fact, incredibly difficult... the trend of miniaturization is about to enter a turning point. That is, we will eventually reach the limit for simply making pixels smaller and face tradeoffs due to miniaturization.

...we can differentiate our product by curtailing noise so as to realize high sensitivity performance and pioneering new pixel structures and miniaturization.
In addition, at Sony, we have people nearby thinking about signal processing algorithms, and we have the manufacturing company within our Group. This proximity gives us an advantage in that it makes it easier for us to find ways to achieve overall optimization.

...for the IMX586, our algorithms played a big role in functions such as the high dynamic range (HDR) image composition, the array conversion processing for the Quad Bayer color filter array that achieves both high sensitivity and high resolution, and the phase difference detection entailed in high-speed autofocusing.

...since the pixel size of the IMX586 was a world-first at 0.8 μm, the basic development started at Nagasaki, the core manufacturing site for smartphone image sensor development. However, due to circumstances related to other product development, resources and production, we decided to develop and produce in Oita.

The team at Oita was, frankly, very surprised with that move as we did not believe that we had enough experience in image sensor development compared with other Sony technology centers, and so we never thought that we would be at the forefront of product development for such a challenging technology.

Secondly, it had only been a little while since the Oita Technology Center joined Sony Semiconductor Manufacturing, so there were many differences in development procedure and culture. For that reason, it was my mission to find a way to smoothly integrate the culture of the Oita plant with the culture of Sony Semiconductor Manufacturing. In the development of IMX586, the schedule was very tight, so there were challenges with unifying all the team members while working at the same time to meet the timeline.

The smaller the pixel, the more it becomes necessary to build the photodiodes in the depth direction of the silicon substrate. To do that, you need to use greater energy to inject impurities into the silicon.

Also, in the photolithography process, we use a thing called thick film resist. This time it was particularly difficult to address fluctuations in the imaging characteristics due to the change in shape of this thick film resist. We had to spend a lot of time improving processing reproducibility using the same equipment and uniformity in the wafer surface.


  1. So, they still use a no self-align approach...

    1. Self alignment of the pinned photodiode (PPD) relative to the (poly-)gate is possible in case of relatively flat PPDs. But the self aligned approach might be difficult, or even impossible, in case of in-depth integrated PPDs for these very small pixels. These in-depth integrated PPDs need high implant energies. Thus the dopants will pierce through the relatively thin polygate (<200nm), which makes self alignment difficult or impossible.


All comments are moderated to avoid spam and personal attacks.