Spotlight on CMOS Image Sensors

Article By : David Haynes

Sensors and MEMS systems affect a wide range of systems in the commercial, industrial, and automotive markets...

When we talk about technology, much of our focus tends to be on the major drivers of electronics: CMOS logic and memory. While those may account for the bulk of the devices, there are many other specialty technologies required to create systems that are useful to humans. Many of those technologies affect how we interact with our electronics. Examples include:

  • Sensors, including CMOS image sensors (CIS) and micro-electromechanical systems (MEMS)
  • Radio-frequency (RF) circuits for sending and receiving wireless signals
  • Power electronics, built with devices like MOSFETs, insulated-gate bipolar transistors (IGBTs), and those fashioned using bipolar-CMOS-DMOS (BCD) technology
  • Optical devices, including displays and photonic components

These technologies affect a wide range of systems in the commercial, industrial, and automotive markets. Of significant recent interest have been Internet of things (IoT) devices and cellular technology (5G in particular) both of which make extensive use of specialty technologies. Together, these markets may account for as much as 30% of global IC demand by 2023. (Source: IC Insights, McClean and OSD Reports, 2019)

Cameras in the the spotlight
Of the sensors, CIS devices have become particularly important lately. Within the consumer market, smartphones are making use of an increasing number of cameras. Once a “nice to have” in a phone, their prominence has risen as a primary benefit featured in marketing messages. In 2017, Apple promoted the camera in roughly 10% of its iPhone X messaging time; two years later, it spent as much as 49% of its time touting the iPhone 11 cameras over other phone features. (Source: CNBC)

Future automotive designs are also making use of multiple cameras of different types (in addition to radar and/or lidar) to help with advanced driver-assist systems (ADAS) and autonomy. Those cameras will effectively surround the car, eliminating blind spots and providing better situational awareness for the driver and for the autonomy systems.

The CIS market is expected to see significant growth over the coming years, with a unit compound annual growth rate (CAGR) of 6.6% through 2024. The highest-growth segments of the CIS market will see growth significantly higher than this, with a consumer CAGR of 24.6%, a security CAGR of 17.1%, and an automotive CAGR of 14.1%. (Source: Yole)

The CIS category includes cameras that record visible light as well as those that operate in the infrared (IR) or near-infrared (NIR) regimes. Visible-light cameras provide the video streams that we expect to use for identifying objects and the overall surrounding environment. Those in the infrared realm have become increasingly useful for facial recognition using structured light in a manner that doesn’t require illuminating the face with a visible beam.

 

Building the best CIS chips
Early CIS chips received their illumination from the top, or front, of the chip. This was the most obvious way to get photons through a thin layer of silicon to the sensing devices. The downside was the fact that metallization and other chip features would obscure part of that light — only the light that could enter around those obstacles would be captured.

Newer CIS devices are illuminated on the backside, where there are no such obstacles. This means, of course, that the back side of the wafer must be thinned down to capture as many photons as possible without them being scattered and absorbed by a thick wafer. Increased integration is also being achieved by stacking dice to combine the image sensor with memory and other logic functions in a single package. Because the back side of the CIS die is illuminated, the front side can be bonded to other wafers without interfering with the light being sensed.

Doing this, however, requires advanced techniques in order to achieve the most efficient light capture. Two specific examples stand out: deep-trench isolation (DTI) and through-silicon vias (TSVs).

DTI allows pixel circuits to be more effectively isolated from each other. As photons enter a pixel, they can be scattered and move around — potentially drifting from the pixel into which they entered to a neighboring pixel. When high resolution is desired, this creates a fuzzing effect as pixels bleed into each other. DTI effectively builds a wall between pixels, keeping photons contained and yielding a sharper image.

Die stacking relies on TSVs. The front side of a die has all of the metal interconnect and any traditional pads, so when bringing the front sides of two dice together, those signals can be bonded together to make the connection. But the back sides of the dice don’t have those signals. So, when stacking a back side to either a front side or another back side, there must be some way to get the signals from the front side of the die to the back of the die so that they can be connected. TSVs are deep metal “pipes” drilled through the silicon that perform that function.

DTI and TSVs are non-trivial technologies, requiring care and precision in order for them to be effective. They are technologies that Lam is particularly good at implementing, and Lam expects significant growth in the use of the tools necessary to manufacture them as the CIS market grows. While CIS and other specialty technologies may not get the constant attention that mainstream technologies attract, they are no less essential for the effectiveness of the systems we will all see and interact with in the coming years.

Leave a comment