Contributed by Ray Fontaine

The 6^th annual Image Sensors conference was held in London, UK on March 19 to 21. Having attended in 2010, it was nice to return this year and find an ever increasing scope of technical content and diversity of attendees. The best aspect of the conference is the variety of end-use applications covered including automotive, broadcast and cinema video, consumer, machine-vision, medical, etc. It is a good reminder that imaging technology is involved in nearly every aspect of our lives. Unfortunately, like many technical conferences the Image Sensor 2013 Conference (ironically) doesn’t allow pictures to be taken. Therefore, articles like these lack a certain multimedia panache.

Highlights from the oral presentations included research in the area of single-photon avalanche diode (SPAD) arrays and time-of-flight (ToF) imaging systems. Advanced driver assistance systems (ADAS) are emerging as a reality and will continue to be developed as regulatory bodies worldwide push for safer cars. The practical use of ToF systems in consumer electronics was debated, although a new product from STMicroelectronics offers reflectivity independent short range distance measurement and could potentially disrupt the current IR-based proximity detector market for smartphones. Multi-aperture imaging has a lot of positive attributes, including improved low light sensitivity and high dynamic range (HDR) performance, depth mapping, and potential for thinner camera modules. These offerings come with the expense of greater computational requirements and increased camera module footprint. Two memorable (and entertaining!) keynotes on video systems and color reproduction included good reminders of the nature of the human vision system and human perception.

Rather than attempt to regurgitate all of the papers and presentations here, I’ll focus on two papers that should be of interest to our teardown and technology blog readership (who tend to gravitate towards small pixel mobile and digital camera applications). First up on Wednesday was Howard Rhodes from OmniVision who started with a review of OmniVision’s small pixel mobile device roadmap. New for 2013 is a move to TSMC 55 nm technology for third generation OmniBSI with a 1.1 µm pixel size. It is worth noting that while OmniVision reported advanced design rules for OmniBSI-2 in the pixel array only, third generation OmniBSI (presumably with OmniBSI-3 branding) will get sub-65 nm  generation design rules for the pixel array and logic region. Looking forward to 2014, Howard and his team are developing fourth generation OmniBSI to deliver sub-micron pixels (SMP) fabricated using 45 nm design rules. Beyond that, Howard is bullish on continued pixel shrink and notes that, while a lot of advanced technologies are up and running in R&D labs, it is ultimately pull from end clients that dictates when devices hit the streets.

As OmniVision marches down the small pixel roadmap, it always has its 1.75 µm pixel generation as a reference for each new platform. Third generation OmniBSI technology deployed in a 1.75 µm pixel shows continued improvements in sensitivity and reduction of crosstalk. So how do they do it? The domain of materials science offers a few knobs to turn and OmniVision has been working on optimizing its color filter and on-chip microlens technology. Chiefly among these innovations is the transition from Bayer patterned RGB to a proprietary Clear Process Technology. While there is no standard for RGBC patterning, OmniVision has chosen to use 1/2 clear, 1/4 green, 1/8 red, and 1/8 blue. Coupled with a custom image signal processor dubbed the L6, the prototype RGBC sensor delivers a 2x sensitivity improvement (as compared to Bayer RGB), and a 6 dB improvement in SNR in low light (<10 lux). The total system solution also shows improved high dynamic range (HDR) video performance. Adding in clear pixels represents a win for low light performance with a cost of minor color aliasing (for everyone, not just OmniVision).  OmniVision is continuously improving its system, and really, it will be the OEMs that will decide what they want.

I wouldn’t want to minimize the engineering and manufacturing challenges facing the small pixel guys, but it appears there are no fundamental barriers to innovation for the next little while. It’s not hard to predict that the small pixel mobile crowd is excited about stacked CIS and RGBC technologies to enable near term performance improvements. I suspect we’ll be hearing more details and performance metrics from Howard in June at this year’s International Image Sensors Workshop.

The second presentation I wanted to mention was given by Junichi Nakamura of Aptina Japan. While conventional digital compact cameras have been losing market share to smartphone cameras, it was mentioned elsewhere at the conference that there still is a market there and hopefully sensor and camera improvements can rejuvenate sales. On a personal note, I agree with this thinking. I intentionally travelled to London with only my Samsung Galaxy S III smartphone, leaving behind a bulky Nikon DSLR.  While relatively impressed by the performance of my smartphone camera, the release of bridge cameras and compact mirrorless interchangeable lens cameras (MILCs) make sense as gap fillers. Regardless of whether it is a growth or declining market, it is not controversial to say there is strong demand for high frame rate video capability in this space.

Nakamura san gave a nice overview of high frame rate imaging in DSCs. The development of the AR1011HS represented a balance between sensor resolution and full resolution video frame rate. The resulting 1 inch format sensor features 3.4 µm generation pixels, a 10.8 Mp resolution, and full resolution video frame rate of 80 fps. Aptina suggests that at 80 fps, there is almost no issue with the electronic rolling shutter effect, enabling cameras without mechanical shutters if desired. The sensor uses a 13 bit column parallel successive approximation ADC (SA-ADC). This Aptina sensor uses the dual conversion gain (DCG) mode pixels branded as DR-Pix. Nakamura san showed a table comparing the new DCG sensor to previously published metrics from other Aptina sensors. In low conversion gain mode, the AR1011HS has a full well capacity (FWC) of 25 ke-, and in high conversion gain mode it has a FWC of 9.1 ke-. In high conversion gain mode, the noise floor is 1.6 e-. Nakamura san gave performance comparisons to other recently published work and indicated more results will be presented at IISW 2013.

There truly are several memorable moments from IS2013, and I regret not drawing attention to the other presentations. They were of high quality, but I’m sure the conference organizers would prefer you attend next year or purchase a copy of the proceedings!