Saturday, October 31, 2009

Kodak Combines W-RGB Filter with CCD

DCViews: The first public demonstration of CCD-based Kodak Truesense technology integration will be shown at Kodak's exhibit at the Vision 2009 trade show held November 3 - 5 in Stuttgart, Germany. The demonstration is based on the currently available 1080p format KAI-02150 Image Sensor combined with TRUESENSE Color Filter Pattern adding panchromatic, or “clear“ pixels to the red, green, and blue elements that form the image sensor array. Kodak claims a 2x to 4x increase in light sensitivity (from one to two photographic stops) compared to a standard Bayer color sensor.

Update: Here is the official Kodak PR.

Friday, October 30, 2009

Omnivision Improves Its VGA Sensor

Yahoo: Omnivision introduced its OV7675 VGA sensor to deliver optimized performance at a competitive cost for the high-volume OEMs developing mobile handsets for Indian and Chinese markets.

The 1/9-inch OV7675 uses 2.5um OmniPixel3-HS pixel with sensitivity of 1800 mV/lux-sec, significantly reduced noise and improved color reproduction. The OV7675 SOC sensor offers the full functionality of a complete VGA (640x480) camera, and is capable of operating at 30fps in full resolution. Its small form factor offers easy design integration for 6.5 mm x 6.5 mm drop-in replacement of VGA modules in existing handset designs.

The OV7675 comes in OmniVision's lead-free CSP3 packaging and is currently shipping in high volume quantities.

Panavision Sues Omnivision, Aptina, Canon

I missed this news item at the time, so thanks to R.C. for bringing this to my attention:

From OVTI 10-Q form, July 2009:

On March 6, 2009, Panavision Imaging, LLC, or Panavision, filed a complaint against [OVTI] alleging patent infringement in the District Court for the Central District of California. The case is entitled Panavision Imaging, LLC v. OmniVision Technologies, Inc., Canon U.S.A., Inc., Micron Technology, Inc. and Aptina Imaging Corporation, Case No. CV09-1577. In its complaint, Panavision asserts that we make, have made, use, sell and/or import products that infringe U.S. Patent Nos. 6,818,877 (“Pre-charging a Wide Analog Bus for CMOS Image Sensors”), 6,633,029 (“Video Bus for High Speed Multi-resolution Imagers and Method Thereof”) and 7,057,150 (“Solid State Imager with Reduced Number of Transistors per Pixel”). The complaint seeks unspecified monetary damages, fees and expenses and injunctive relief against us. We expect to vigorously defend ourselves against Panavision’s allegations. The Court has scheduled a claim construction hearing for December 10, 2009.

Thursday, October 29, 2009

Intertech-Pira Image Sensors 2009 Presentations On-Line

Here used to be a link to the presentations from the recent Intertech-Pira Image Sensors 2009 conference (San Diego Oct. 13-16). A lot of interesting stuff!

[Update: The links have been removed at request from the originator. This link was supposed to be only for attendees of the conference, not for the general public.]

On the technical side, Don Scansen shows Chipworks reverse engineering picture of Omnivision BSI packaging scheme:

There is one more Chipworks BSI sensor picture, apparently of Sony ClearVid camcorder sensor:

Don also presents many other observations, some of them quite paradoxical. One of the graphs shows that Microsoft has filed about 10 patent applications on BSI technology. Personally, I have not seen any of them.

Another interesting presentation comes from DxO Labs on camera module self calibration. It presents very nice features but no word is said about the price to implement them in silicon - an important omission. Also, Gr/Gb imbalance section excessively downplays the alternative approaches to the problem.

On the marketing side, there is an interesting presentation by Sanyo on digital camera market, product trends and adoption rates. Toshiba presentation mainly talks about the market consolidation trend. The Strategies Unlimited presentation disappointed me a bit, as its many data labels from the charts are removed - looks like a teaser for the paid reports.

Tuesday, October 27, 2009

Toshiba Announces Its First BSI Sensor

Toshiba announces its first BSI sensor. The 1/2.3-inch 14.6MP sensor is targeted to DSC and camera phones with video capability. The sensor is based on 1.4um pixels. Its video frame rate is 60fps at 1080p and 720p formats.

BSI approach is said to boost "light sensitivity and absorption by 40% compared to existing Toshiba products". Given already excellent QE numbers of Toshiba FSI sensors, I wonder how they can be improved by so much.

Sampling of the new sensor will begin in December and mass production will follow from the third quarter of 2010 (July—September). The new sensor will be produced at Toshiba's Oita Operations 300mm wafer lines deploying 65nm process technology. Initial production will be at a volume of 500,000 sensors a month.

Toshiba will continue to develop BSI products as a mainstream technology. CMOS image sensors are a focus product of Toshiba's System LSI business. Until now, their main application has been in mobile phones. With the new BSI sensors, Toshiba plans to expand their applications to include digital cameras.

Camera Module Calibration

Embedded: Toshiba team talks about importance of having a convenient GUI interface to camera module calibration. It appears that the article is supposed to promote Toshiba's ImaTuning tool, even though it's not called by name.

Nanya Enters CIS Production?

Digitimes reports that Nanya has been in talks with Aptina, Himax, Omnivision and Pixart for the possible release of foundry orders. Nanya is looking to maximize the utilization of its two 8-inch fabs by manufacturing image sensors.

Digitimes reminds that Powerchip and ProMOS have both stepped into CIS production at 8-inch fabs, both not very successful. ProMOS has been rumored to cease its image sensor activities, while Powerchip reportedly produces a small fraction of Omnivision products.

Omnivision Sensors in Tight Supply

Digitimes: OmniVision has notified its customers to expect limited supply of its image sensors due to strong iPhone 3GS demand for the end-year holiday season, according to Digitimes' sources. Tight supply is not expected to ease until late November 2009.

Apple has increased Q4 orders for the iPhone 3GS to its Taiwan-based manufacturing partners Foxconn and Primax by 17-20%, noted the sources. Notebook vendors are gearing up to introduce new models following the launch of Windows 7 which will also tighten the supply of CIS, the sources added.

Wednesday, October 21, 2009

Canesta Raises $16M, Quanta Joins Investment Round

Yahoo, VentureBeat: Canesta announced it has raised $16M in additional capitalization. Joining returning investors Carlyle Growth Partners, Hotung Venture Group, and Venrock are two new strategic investors – Quanta Computer, the world’s largest manufacturer of notebook computers, and SMSC, a mixed-signal ICs vendor.

“The emergence of 3-D ‘natural’ interfaces in PCs – such as ‘touchless’ gesture controls – as well as other immersive applications, has been inevitable,” commented Jim Spare president and CEO of Canesta. “The implied endorsement and strong interest in 3-D imaging technology by the world’s largest personal computer ODM will encourage other players to adopt this game-changing technology,” said Spare.

In February this year, Hitachi and Canesta demonstrated a television with 3-D interface and natural “gesture-based” controls (ISW link).

Some of the applications made possible by Canesta’s 3-D electronic perception chips and technology include:
  • Hands-free gesture control (for new user experiences that are more convenient and “fun”);
  • Robust and accurate facial recognition for user identification and security;
  • Background substitution for video conferencing and “virtual sets” (to enable every consumer to have video conferencing and content-creation capabilities that are currently only available to professionals with large budgets);
  • Avatar control for 3-D virtual communities (increasing ease-of-use and improving accessibility to a wider audience);
  • Augmented reality;
  • Immersive personalized advertising;
For several video examples, see Canesta videos. The news log about Canesta can be found here.

Update: SMSC has invested $2M in Canesta. SMSC and Canesta have agreed to collaborate in the definition of 3-D camera solutions whereby SMSC can combine its connectivity technologies and systems expertise with Canesta 3-D sensor devices and software to accelerate the adoption of this technology. The two companies intend to create economic, integrated solutions to add 3-dimensional depth-perception to automotive, PC, TV, consumer electronics and other communications devices.

Tuesday, October 20, 2009

Tessera CSP Offers Proven Reliability

Electronic Design published Tessera team article describing reliability challenges with many TSV processes, while proposing Tessera CSP solution as a proven alternative.

CMOSIS Uses Tower Process for its First Standard Product

Yahoo: CMOSIS disclosed that it uses Tower 0.18um process for its first off-the-shelf sensor CMV2000. CMOSIS selected Tower’s 0.18um process for its aggressive layout rules, allowing a novel pixel structure featuring a pixel size of 5.5µm x 5.5µm which is needed to stay close to a 2/3" optical format with 2.2M pixel resolution and form factor. The CMV2000 combines a pipelined global shutter operation that allows true correlated double sampling and a full-frame rate of 300fps.

CMOSIS estimates its target machine vision sensor market at ~$200M (CCD and CMOS combined).

Saturday, October 17, 2009

Thursday, October 15, 2009

More about Nobel CCD Controversy

IEEE Spectrum put a special page with collection of articles and podcast about the controversy around the CCD invention in Bell Labs.

Wednesday, October 14, 2009

HDcctv Alliance Got SMPTE License

Yahoo: HDcctv Alliance receives the security industry's first license to SMPTE HD-SDI Standards, bringing the HDTV digital technology for video delivery over coaxial cable to the HDcctv Specification. The HDcctv Alliance members include CSST, Gennum, Stretch, AltaSens, blueCaps, Comart, COP UK, EverFocus, theHDcctv.com, OmniVision, Ovii, and Pixim.

ST Announces 5MP EDoF Sensor Family

Yahoo: ST unveiled its new 5MP CMOS image sensors family based on 1.4um pixels. The 1/4-inch sensors are based on 65/45nm process and said to utilize ST's unique pixel-isolation techniques to maximize the sensor's SNR.

The sensors allow 5MP cameras within dimensions of 6.5 x 6.5mm and a low module height of typically 5mm to enable ultra-low-profile handset designs. From the announcement it's not clear if EDoF and ISP are integrated onto the sensor die or they are separate. It appears that ST prepares a large product family with different types of integration and feature sets, including EDoF and AF VCM drivers.

All the sensors will include single-line or dual-line 1GHz Camera Serial Interface (CSI-2), as defined by the Mobile Industry Processor Interface (MIPI) alliance, and the 650Mbit/s Compact Camera Port (CCP2) interface defined by the Standard Mobile Imaging Architecture (SMIA) group.

ST will offer standard-die and Through-Silicon Via (TSV) packaging in the 5MP family. The standard packaging will support cameras combining Chip-On-Board (COB) sensor connection and discrete optical components. TSV enables emerging technologies such as Wafer-Level Camera (WLC).

Engineering samples of the first of these sensors are available now, and the roadmap will continue with new product introductions throughout the remainder of 2009.

Reportedly these ST's 45/65nm based FSI pixels deliver better QE and SNR10 performance than BSI. This presents a problem for Omnivision to justify the added BSI complexity.

Tuesday, October 13, 2009

Stanford HDR Symposium Materials

Stanford University held Symposium and Workshop on High Dynamic Range Imaging in September 2009. The selected presentations form this event are available on-line.

Thanks to S.S. for sharing the link!

Does CCD Idea Really Belong to Smith and Boyle?

IEEE Spectrum Tech Talk blog has two posts questioning Smith’s and Boyle's ownership of the CCD idea (Post #1, Post #2 with comments). The posts suggest that the idea originally came from their Bell Labs collegue Eugene Gordon.

Thanks to E.F. for sending me the link.

Update: As said in comments, there is another post saying that the CCD idea originator is Michael F. Tompsett.

Fast Pixel Shutter Imaging Unveiled

The Engineer Online: UK Oxford University researchers presented Fast Pixel Shutter Imaging (FPSI) technology. They demonstrated the technology using a prototype built out of a digital micro-mirror array comprising a vast amount of individually moveable mirrors measuring 10um each. The mirrors can be toggled on and off and used as pixel-level optical switches. So each pixel gets individually controlled exposure.

Oxford group is working with researchers at Nottingham University on fashioning the technology into a single CMOS chip that could be integrated into small electronic devices, including mobile phones. Within six months they plan to unveil a mobile-phone-sized camera that can demonstrate FPSI.

DxO Lab Discloses Its Future Directions

In an interview with EE Times, Jérôme Meniere, chairman and CEO of DXO Labs, shrank the field of view and focused on DXO Labs' solutions for camera phones, and unveiled future research directions. DxO Labs also provides the calibration and tuning tools which allow the configuration of the ISP to be done more easily and at a lower cost.

Another marker of differentiation is DxO Labs' EDoF technology. DxO Labs said PalmPre smartphones and Nokia cell phones integrate its EDoF design.

Looking ahead, Meniere said he identifies several directions: "There are huge opportunities to help camera module manufacturers reduce their costs. The idea is to loosen some constraints that we recover and use at the image processing level. Typically, you can have a yield drop because the optic has been badly tuned. We can recover modules that otherwise would be thrown away, and here lies a huge gain."

Another direction concerns very high ISO, probably with associated noise reduction. Other possible directions were described in very general words, so I do not quote them.

DxO Labs, formerly known as DO Labs SA, was founded in 2002 as a spin-off of Vision IQ, a company in the field of computer vision. It currently employs about 100 people, including 60 percent in R&D, and has filed about 40 patent families.

Monday, October 12, 2009

KLA-Tencor Helps Toppan to Get 300mm Wafer Capability

Yahoo: KLA-Tencor announced that its 8900 defect inspection system was recently installed at the first 300mm advanced CFA fab of Toppan. Key features of the new 8900 defect inspection system include:
  • Simultaneous brightfield and darkfield optical paths to capture a wide range of defect types in a single pass, such as micro-lens deformation; resist and fall-on defects; color contamination; large stains and striations;
  • Selectable LED inspection and review illumination spectrum matched to CIS filter colors;
  • Sensitivity consistent with requirements of advanced CIS roadmaps;
  • Throughput above 110 wafers per hour at production sensitivity, for 300mm semiconductor wafers;
  • Automated binning of defects by type;
  • Automated sensor pass/fail dispositioning; and
  • Automated color review for defect verification.

Invensense Targets OIS Systems

Electronic Design: Addressing lower-cost optical image stabilization (OIS) as a major goal, InvenSense has unveiled the first digital dual-axis pitch and roll (X and Y) gyroscope for camera phones and digital still cameras. The original announcement from over a month is here. The IDG-2000 family offers the smallest such devices on the market, housed in a 4- by 4- by 0.9-mm quad flat no-lead (QFN) package and integrating 16b ADcs. The IDG-2000 measures hand jitter across a wide frequency range of 0.1 Hz to 20 Hz with noise 0.033°/s-rms.

The 20Hz upper range looks too low to me, especially for zoom equipped cameras. However, it should work fine for wide angle low light situations.

Silicon Image 18MP ISP IP Core

EETimes talks about new Silicon Image camerIC-18 ISP IP core targeted to camera phones with resolutions from 5MP to 18MP. Ron Richter, director of business development at Silicon Image, tells all the nice stories about the new ISP capabilities while saying very little about small pixel-specific artifacts. I doubt that stand-alone ISP companies can survive in the small pixel times without a deep knowledge of this kind of stuff.

Update: Here is the official Silicon Image PR.

Update #2: EETAsia: Meanwhile TI has introduced two new members of its OMAP-DM5x family of coprocessors that claim to deliver 20MP capability and and 720p HD video functionality. The new OMAP-DM515 and OMAP-DM525 coprocessors offer 8MP shot-to-shot speed of 1.4fps in high-quality mode and 2fps in burst mode.

Tessera Hits and Misses

EETimes: In 2005, realizing that it needed a new engine for growth, Tessera has embarked on a bold but little-understood strategy that could pay huge dividends--or may simply fall flat. Since 2005, it has acquired five separate companies in the imaging and optics arena (Shellcase, Digital Optics, Eyesquad, FotoNation and Dblur). Call it Tessera's big gamble: It hopes to replicate success in DRAM IP, by assembling the pieces to boost the imaging quality and functionally for camera-based handsets.

But the company has already seen a slight setback for its Imaging and Optics division. Citing a fall in capital equipment spending, the Imaging and Optics group is projected to exit the year with about $12 million in sales for 2009, significantly down from $23 million in sales last year. The group missed its forecast in other respects. At one time, the Imaging and Optics division projected that its sales would hit $100 million by 2010. The group has now pushed out those ambitious $100 million sales targets to 2011.

The article also has few interesting side stories, such as wafer level camera status at ST:

With Nokia--ST's largest camera module customer--pushing for very low-cost modules and wafer level camera technology adoption, ST also began developing wafer level optics. In light of the challenges, checks suggest that the ST's wafer level camera was not reaching the same level of image quality, causing Nokia to become less aggressive in pushing its vendors to adopt a wafer level camera approach.

iSuppli Forecasts 2B Sensors in 2012

Digitimes: iSupply predicts image sensor market to grow from 1.3B units in 2009 to more than 2B in 2012. Shipments of image sensors to the automotive market are expected to more than triple, rising from 4.2M in 2009 to 14.6M units in 2013. In 2010, more than one billion image sensors will be shipped to the handset market, iSuppli predicts.

Sunday, October 11, 2009

Omnivision BSI Backside Implant Optimizations

Omnivision BSI pixel uses P+ backside implant to passivate the backside surface. The recently published WO/2009/099491 patent application gives some details of its optimization process. The figure below shows the location of the backside P+ implant 220:


First, Omnivision optimizes the P+ layer depth, making few implants with different energies:

The sensitivity for the different depths is shown below:


Next step is the optimization of the P+ implant dosage. The three tested dosages are shown here:


The pixel sensitivity vs dosage is shown below:

One more step is optimization of the silicon layer thickness:


Saturday, October 10, 2009

Omnivision BSI Pixel Crosss-section Revealed

WIPO published Omnivision patent applications WO/2009/099778 and WO/2009/100039 on BSI photodiode profile and WO/2009/100038 on black pixels in BSI technology. While the black pixel idea is quite obvious - removing photodiode implants to make black pixel light-insensitive, a really interesting part of these applications is Omnivision BSI pixel cross-section:


One can see that the photodiode extends well underneath the transistor region. Apparently, Omnivision did a great job to achieve lag-free operation with such a complex shape of the fully depleted photodiode.

The earlier Omnivision BSI applications, such as WO/2009/099483, WO/2009/09949, WO/2009/099494, WO/2009/099491 do not mention the extended photodiode profile. By the way, the last of these has an interesting part describing optimizations of the backside implant and the sensor's substrate thickness.

More about Sony Single-Lens 3D Approach

Tech-On: Sony single-lens 3D camera is based on the concept that it is possible to achieve enough stereoscopic and depth effects with an interocular distance of several millimeters. The concept, which is called "microstereopsis," was announced in 1995.

There are two major advantages in using the new monocular video camera. First, it does not cause disagreement between the focal point of the eyes and the intersection point of the lines of sight (accommodation-vergence conflict), which is typical of normal 3D images. Second, though it requires special glasses to view 3D images, the images do not split into two even without the glasses.

Carnegie Mellon University group spent some time on microstereopsis principle research at the beginning of this decade. The approach relies on human vision perception psychology when surprisingly small disparities between left and right eyes are adequate to stimulate binocular stereopsis. So, even if the stereo base is much smaller than 6.5cm, typical for human eye-to-eye distance, a good 3D perception can be achieved.

Friday, October 09, 2009

Albert Theuwissen on CCD Nobel Prize

Albert Theuwissen publishes a nice blog post on CCDs in our personal and professional lives.

Thursday, October 08, 2009

8,188 SPADs Integrated

EETimes: Philips Research group integrated 8,188 single-photon avalanche detectors (SPAD) on a single die so that each SPAD is individually addressable. The Philips chip is a digital design with an integrated 8ps timer. It does not need a separate digital readout ASIC today's analog SPAD arrays require.

Philips engineers hope to use the chip as a block in a larger 8x8 array they plan to prototype by the end of the year. Such arrays could be used in medical PET scanners, new automotive night vision systems or high throughput DNA testers.

Gigavision Sensor Proposed

New Scientist: A team led by Edoardo Charbon, a professor from the Swiss Federal Polytechnic Institute (EPFL) in Lausanne presented their so-called "gigavision" sensor at an OMNIVIS 2009 workshop on Oct. 4 in Kyoto, Japan.

While Charbon's idea is new and has a patent pending, the principle behind it is not. It has long been known that memory chips are extremely sensitive to light. The charge stored in every cell corresponds to whether that cell is in a light or dark area. Memory cell is small, so for every pixel on one of today's sensors, the memory-based sensor could have 100 pixels. A chip the size of a 10MP camera sensor will have 100 times as many sensing cells if implemented in memory technology - hence the choice of the gigavision name.

Unlike the pixels in a conventional sensor, which record a greyscale, the cells in Charbon's memory-chip sensor are simple on-off devices: they can only store a digital 0 or 1, for which read either light or dark. To build a sensor that can record shades of grey, EPFL engineer Feng Yang, who presented the Kyoto paper, is developing a software algorithm that looks across an array of 100 pixels to estimate their overall greyscale value.

On the surface this sounds a lot like Eric Fossum's Digital Jot idea. Once EPFL patent application is published, one can see what is the difference. I'm almost sure that EPFL team is aware of the prior art and did something different.

Comments to the New Scientist article point that using of RAMs as sensors is known since 70s, so this part of the work is hardly new.

Wednesday, October 07, 2009

HH NEC Buys Grace Semiconductor

EETimes: Image sensor foundry HH NEC merges with Grace Semiconductor (buys Grace according to another article). Grace also has some image sensor history, when Cypress unsuccessfully tried to transfer its image sensor process to Grace few years ago. Since then Grace reportedly abandoned its attempts in image sensor area. On the other hand, HH NEC has entered image senors recently.

Dongbu Officially Becomes CIS IDM

EETimes: At last week's GSA Expo Dongbu has quietly devised a CMOS image sensor and LCD driver IC that is being marketed under its own brand name. So, now Dongbu goes to compete with its own customers, such as SETi and Siliconfile.

Tuesday, October 06, 2009

Kodak Announces 4MP 32fps CCD

Kodak: The 4MP KAI-04050 CCD, based on 5.5um Interline Transfer Platform, sets a new benchmark for 1" optical format devices, featuring full resolution frame rate up to 32fps. A Region of Interest (ROI) mode supports partial readout at even higher frame rates, such as 3MP in a 16:9 aspect ratio at 40fps, or standard VGA at over 90 fps.

The KAI-04050 will be available in both color and monochrome configurations. Engineering grade devices are currently available, and production availability is planned for Q4, 2009.

Kodak PluggedIn blog talks about the whole 5.5um pixel family, which includes 5 CCDs now.

CCD Inventors Willard Boyle and George Smith Got Nobel Prize

Yahoo: CCD inventors Willard S. Boyle and George E. Smith shared the 2009 Nobel Prize in physics with the fiber optics inventor Charles K. Kao. The award's 10 million kronor ($1.4 million) purse will be split between the three with Kao taking half and Boyle and Smith each getting a fourth.

In its citation, the Royal Swedish Academy of Sciences said that Boyle and Smith "invented the first successful imaging technology using a digital sensor, a CCD."

Boyle, in a phone call to the academy, said he is reminded of his work with Smith "when I go around these days and see everybody using our little digital cameras, everywhere. Although they don't use exactly our CCD, it started it all."

He added that the biggest achievement resulting from his work was when images of Mars were transmitted back to Earth using digital cameras. "We saw for the first time the surface of Mars," Boyle said. "It wouldn't have been possible without our invention."

EV Group Installs Fusion Wafer Bonders at Two Image Sensor Manufacturing Facilities in Asia

Yahoo: EV Group (EVG) announced that it has completed the installation of two automated fusion bonding systems for 300-mm wafers at a leading semiconductor foundry and at a major consumer electronics manufacturer. The GEMINI FB automated production fusion bonding systems will be employed for the production of BSI CMOS image sensors ranging from ultra-compact wafer-level cameras for mobile phones to larger form factor high-end image sensors.

Samsung Licenses Apical IP

Electronics Weekly: Samsung has licensed an image processing core from London-based IP developer Apical. Apical has developed Iridix imaging technology which provides local DR correction on per-pixel basis. The licensing agreement entitles Samsung LSI to integrate the latest version of Iridix into any of its semiconductor designs.

CMOSIS Comes Out with Off the Shelf Product

AZOM.com: Custom sensors design house CMOSIS has announced its first standard product. The 2.2MP CMV2000 imager is based on 5.5um pixel and designed to run at a frame rate of more than 300fps at full resolution. A unique feature of the CMV2000 is its novel pixel structure, which combines pipelined global shutter operation with CDS. This technique, pioneered by CMOSIS, yields a low noise level below 18 e-.

The CMV2000 integrates a high-performance 10bit column ADC on-chip. The ADC features a slow 12bit mode and multiple HDR modes.

Samples of the CMV2000 are currently released to certain CMOSIS customers. Full production will ramp up in early 2010. A 4MP version of the new sensor, named CMV4000, will be available to camera manufacturers beginning May 2010.

Monday, October 05, 2009

Welcome New Imaging Technologies - A Smart Sensor Company

It came to my attention there is a new image sensor startup in France. New Imaging Technologies (NIT) is a fabless company providing both standard products, custom design and technology consulting-licensing. NIT is a spin-off from the French National Telecommunication Institute. Lead by Professor Yang NI, the research team has worked since 1991 on CMOS smart imaging devices for a large variety of scientific, industrial, automotive, biometrics and optical communications applications. Several tens of CMOS application oriented smart sensors have been designed and delivered.

The company has been granted with a R&D contract from the French Ministry of Defense through its DGA operation. NIT will design a specific CMOS imaging sensor using its proprietary wide dynamic range technology (so called MAGIC) to be coupled with an image intensifier for night vision application.

Also, NIT won a design contract with an unnamed major Japanese firm for designing a smart imaging sensor for mass market application. The sensor is said to use a 0,18um process from a large silicon foundry.

Thanks to J.B. for letting me know about the new startup.

Fujifilm Organic CMOS Sensor

MuTaka has brought to my attention a DPNet article in Chinese on organic image sensors - see comments to Tessera post. Unfortunately, I'm unable to understand most of the article, other than seeing that Fuji is making a good progress in the organic sensors.

Potentially, multi-layered organic sensors can over-perform BSI by a big margin. Their biggest advantage is that the dye absorption could be tuned to have very high QE in each of three colors, crosstalk can be minimized and each pixel location can detect and process all three color signals. Bayer sensors, on the other hand, get only one color per pixel, effectively discarding 2/3 of the incoming signal.

Said all this there are few challenges that organic sensors should overcome on their way to the mass market:
  • Process reproducibility and reliability. So far thin film organic photodiodes experience large variation, in comparison with silicon photodiodes.
  • Large dark current, orders of magnitude higher than Si photodiodes of the same size
  • Noise, primarily kTC noise. It appears that generic 4T pixel architecture used for kTC noise suppression is not applicable for organic sensors. Other techniques for the noise reduction should be developed for them.
If these challenges are solved, multilayered organic sensors would enjoy a bright future.

Omnivision Wins Microsoft HD Webcam Design

Yahoo: Omnivision announced that its OV9712, a native 720p wide-screen HD video image sensor, was selected by Microsoft for its next generation webcam. The 1/4-inch OV9712 uses a 3-micron OmniPixel3-HS pixel architecture. The sensor supports 30fps speed.

Tessera One More Time Announces Face Reconition Technology

Yahoo: Tessera one more time announced its FotoNation FaceRecognition. Tessera acquired Fotonation at the beginning of 2008 for $29M (plus another $10M if some goals are met). By that time Fotonation already had face recognition product for sale.

I guess this announcement relates to an improvement version of it, although it's not clearly stated. The technology appears to be software based and available for a number of platforms, including COACH and ARM. No recognition speed or number of simultaneously recognized faces is mentioned, both being very important merits of the algorithm.

Thursday, October 01, 2009

Sony Goes 3D with Single Lens

Sony announced the development of a single lens 3D camera technology capable of recording 3D images of fast-moving subject matter such as sports, at 240fps. A prototype camera incorporating this technology is to be demonstrated at at "CEATEC JAPAN 2009" in Chiba city, Japan, from October 6.

In existing 3D camera systems have separate lenses for the left and right eyes. However, when operating the zoom and focus functions of such systems, a complex technology is required to ensure that each camera lens is closely coordinated, and there are no discrepancies in the optical axis, image size, and focus. The human eye is very sensitive to differences in the size and rotational movement of dual images, as well as any vertical misalignment or difference in image quality. The introduction of a single lens system is said to resolve any issues that may occur as a result of having different optical characteristics for each eye.

The optical scheme of the camera is shown below:



The main idea appears to be that the images from left and right sides of the single lens are split, so one sensor gets the image from the left half of the lens, while the other sensor gets the right half image.

It looks to me that Sony could make much more compact design on the same principle, if it used a single stereo image sensor, such as one described in a recent Kodak patent application US20090219432. Kodak relies on exotic lenticular microlens in this invention, but the same idea can be easily implemented with regular microlens, although the exact reference to this implementation escapes me.

Update: Kodak stereo sensor principle is illustrated below:

Albert Theuwissen PTC Lessons: Temperature Influence

Albert Theuwissen published another chapter in the great series of articles on Photon Transfer Curve (PTC) measurements - the influence of temperature on PTC.