Sony announced the new chip architecture at ISSCC this week, as the latest in its long line of innovations in sensor design. The addition of on-board DRAM will help solve at least three major issues with current smartphone cameras: rolling shutter artifacts, slow-motion videos, and multi-image noise reduction artifacts.
A great band-aid for rolling shutter headaches
One problem with low-cost imagers, like those used in smartphones, is that they don’t have a mechanical shutter. Their images are read out one piece at a time while the sensor is still actively recording data (called a rolling shutter). That means that objects moving at high speed appear distorted, like the locomotive in this illustrative example from Sony:
On the left, the typical 1/30s readout time smears the profile of the moving locomotive. On the right, while still technically a rolling shutter, the new chip’s 1/120s readout time greatly reduces the potential for distortion.
While the additional DRAM layer looks pretty simple in this block diagram, it changes the physics of the chip, and introduces additional opportunity for electrical noise. So far, the chip isn’t in production, so it isn’t possible to benchmark its real-world performance. But Sony claims it has been able to solve the new noise problems.
Support for super-slow-motion video and special effects
The DRAM also allows the sensor to capture 1000fps video at 1080p HD resolution. This makes for some very impressive super-slow-motion potential. By mixing high-frame-rate video with standard frame rate the chip also makes it possible to do some cool special effects right in a smartphone, as you can see in this Sony example video:
<iframe width="320" height="241" src="https://www.youtube.com/embed/dFdU-JjypWs" frameborder="0" allowfullscreen></iframe>
Expect record-setting image quality results from this sensor
Less obvious, but perhaps most importantly, the 30fps readout of full-resolution 19.3MP images is perfect for the sophisticated multi-image noise reduction that’s increasingly-common in high-end smartphones. For example, Google’s HDR+ silently combines between 3 and 9 frames to create a single higher-dynamic-range, lower-noise, final output image. That process can introduce artifacts if objects in the scene, or the camera, are moving. By having the frames captured and read out more quickly, those artifacts can be further reduced.
