New version of MIPI CSI-2 looks (literally) beyond mobile phones

July 20, 2017

Haran.Thanigasalam-July 20, 2017

When the first camera phones hit the market in 2000, few people could have imagined all of the applications those devices would enable: not just the obvious snaps and selfies, but also electronic check deposits, biometric authentication, and telemedicine, to name just a few.

A major reason why these and other applications proliferated is because mobile camera sensors and their adjunct systems have improved dramatically over the past 17 years. The latest milestone is version 2.0 of the MIPI Alliance's Camera Serial Interface (CSI-2) specification for connecting a mobile device's camera to its host processor.

Version 1.0 is the world's most widely used interface for imaging and video applications in mobile devices. Version 2.0 builds on that success by giving system designers a variety of new features and capabilities, which in turn enable new market opportunities.


CSI-2 Imaging Conduit Solutions (Click Here to see a larger image. Source: MIPI Alliance)

For example, MIPI CSI-2 v2.0 is designed for use not only in smartphones, but also Internet of Things (IoT) devices, wearables, medical devices, augmented and virtual reality (AR/VR) gear, drones, and vehicles. That expansion reflects the MIPI Alliance Membership: Many of these companies started in the mobile phone market, but are now expanding into applications such as IoT, automotive, drones, and perception and contextual awareness. Their evolution is driving MIPI CSI-2's evolution. For example, automakers and their suppliers can use MIPI CSI-2 v2.0 to give drivers gesture-based controls so their eyes and attention can stay focused on the road instead of fumbling with knobs, buttons, and sliders.

Systems designers have two physical layer options for implementing CSI-2 v2.0: MIPI C-PHY and MIPI D-PHY. Performance is lane-scalable, with support for up to 24 Gbps using a three-lane (nine-wire) C-PHY interface, or 18 Gbps using four-lane (10-wire) D-PHY interface.

Both PHY specs have been updated to support CSI-2 v2.0, enabling backwards compatibility. This also reduces cost and time-to-market for users because they can leverage their existing product designs as they migrate to CSI-2 v2.0.

Keep your eyes on the road
MIPI CSI-2 v1.3 supports a wide variety of resolutions, including 1080p, 4K, and 8K, in both single- and multi-camera implementations. CSI-2 v2.0 enhances those capabilities by adding support for RAW-16 and RAW-20 color depth, which significantly improves intra-scene high-dynamic range (HDR) and signal-to-noise ratio (SNR).

These capabilities are a particularly good fit for automotive applications such as Advanced Driver Assistance Systems (ADAS). For example, CSI-2 v2.0 enables those applications to maintain high performance even when the environment changes suddenly and dramatically, such as when a car emerges from a dimly lit tunnel into bright daylight.


Automotive system diagram (Click Here to see a larger image. Source: MIPI Alliance)

The HDR and SNR enhancements also benefit the Simultaneous Localization and Mapping (SLAM) used by drones for vision-based navigation. When a drone is subject to big changes in lighting conditions, for example, the HDR and SNR enhancements help it quickly adapt to ensure consistent performance.

Another key addition in CSI-2 v2.0 is Latency Reduction and Transport Efficiency (LRTE), which has three major benefits in automotive applications:

  1. It increases the number of supported image sensors by as much as 40 percent without requiring costly Serializer/Deserializer (SerDes) or PHY changes.
  2. It facilitates optimal transport while preserving system integrity for multiple vision applications.

  3. It substantially reduces frame transport latency for real-time perception and decision-making needs.

A third key enhancement for automotive applications is the addition of scrambling. This reduces Power Spectral Density (PSD) emissions, thus minimizing interference when sensitive receivers are installed in the vehicle.

Many automakers are adding collision-avoidance features both to maximize safety and help differentiate their vehicles in the eyes of buyers who seek those kinds of advanced safeguards. The more things that a collision-avoidance system can sense, the better able it is to protect the vehicle's occupants.

CSI-2 v2.0 supports up to 32 virtual channels (up from four virtual channels in the previous version), which can carry data from the vehicle's image sensors to its collision-avoidance system. The additional channels also help systems designers accommodate multi-exposure, multi-range sensor fusion for ADAS applications. Just as important, all 32 channels support multiple data types, including raw, compressed, and meta.

Finally, CSI-2 v2.0 adds support for Differential Pulse Code Modulation (DPCM) 12-10-12 compression qualified over varying degrees of image set intensity, contrast, and Modular Transfer Function (MTF) to ensure superior Image Quality (IQ). This is particularly useful for vision applications sensitive to compression artifacts such as automotive and drone navigation.

New capabilities atop a proven foundation
The MIPI Alliance was founded in 2003 to provide something that the marketplace desperately needed: a standardized camera interface specification for mobile devices. But MIPI Alliance members went a few steps further by ensuring that CSI-2 also provides the high bandwidth necessary for high-resolution images, the low power consumption that's key for battery-operated devices such as IoT sensors and wearables, and very low electromagnetic interference (EMI) for mission-critical applications such as health care.

That forward-looking design is a major reason why CSI-2 v2.0 is ideal for more than just smartphone applications. Another reason is that CSI-2 is mature and proven, giving systems designers the confidence that it can be the foundation for IoT devices, drones, automotive, and other applications outside the traditional mobile space.

Click Here for a deeper dive into CSI-2 v2.0's features and capabilities. The MIPI Alliance Camera Working Group already is hard at work developing the next round of camera, imaging, and vision enhancements. Click Here to learn more about MIPI membership and to become engaged.

Haran Thanigasalam is chair of the MIPI Alliance Camera Working Group and senior platform architect at Intel Corporation.

Loading comments...