In general, smartphone auto focus (AF) uses the "phase difference detection" method*1. In this method, a part of the pixels for imaging on the image sensor is used as the phase difference detection pixels, so it was necessary to interpolate the image information to the position of the phase difference detection pixels at the time of imaging. Too many phase detection pixels make the interpolation difficult and deteriorate image quality. Therefore, there is a limit to the arrangement density of phase difference detection pixels, This was a limitation of AF performance.
*1) A method to detect an optimal focus point based on the information from several incident light beams of different angles.
On the other hand, with the "all-pixel auto focus" type image sensor, all of the imaging pixels can also be used as phase difference detection pixels. By dramatically increasing the density of acquiring phase difference information, stable AF operation is realized even in scenes where it was difficult to focus. In addition to being able to focus on the subject at any position on the screen, AF operates stably even in dark environments because it collects a lot of phase difference information. In addition, the interpolation processing is not necessary when using it as an image signal, and the image quality does not deteriorate.
Reference video : It focuses accurately without hesitation even in dark scenes.
Reference video : The focus performance is improved for a subject with unclear outline.
Development and provision of multiple methods to realize all-pixel AF
Sony Semiconductor Solutions Corporation develops and provides multiple methods to realize all-pixel autofocus for various image sensor types.
(1) Dual PD method : the basic type sensor that offers all-pixel AF
(2) 2 x 2 OCL method : adopted for medium sized sensors with miniaturized, high-resolution pixels
(3) Octa PD method : adopted for large sized sensors to achieve high sensitivity and high resolution
Dual PD method
The dual PD (photo diode) takes two adjacent photo diodes to form one pixel. They share one on-chip lens (OCL*2), so they catch different incident lights, which enables to detect phase differences. The signals from both PDs are combined to be used for image data, which cancels the impact of phase difference detection.
The Dual PD method is used as a default system for all-pixel AF.
*2) Condenser lens to collect light
2 x 2 OCL method
The 2 x 2 OCL method is based on the Quad Bayer sensor, in which four pixels of the same color use one OCL. Like the Octa PD method, this method, benefits from both the Quad Bayer functionality and all-pixel AF. Unlike the Dual PD method, it does not require the signals from two PDs to be combined for image signals, which makes it easier to achieve higher resolution. The method, is therefore more suitable for medium sized sensors.
Octa PD method
The Octa PD method is a hybrid between Quad Bayer sensor and Dual PD method. It retains the features of the Quad Bayer arrangement, such as high sensitivity and HDR compatibility, while achieving the excellent AF performance through the all-pixel AF functionality, making it more suitable for large sized sensors. Its unique feature is in the HDR mode, that phase differences are detected in all pixels across the three exposure levels (long, medium, and short), enabling quick AF operation even on a low-luminance target.
Achieves both of HDR and all-pixel AF
*) SDR（Standard Dynamic Range）
For inquiries about Sony Semiconductor Solutions Group and products / solutions, specifications, quotation / purchase requests, etc., please contact us using the Inquiry form from the button below.