Monday, September 30, 2013

Teledyne DALSA Announces Full-Frame 32MP CCD Family

Teledyne DALSA releases a new FTF7046 family of full frame image sensors with 32MP. Using the latest 5.2um pixel, the company says it is able to preserve DR (>72dB linear), QE, and FF (>90%) in the FTF7046 even as they reduced pixel size to deliver 32MP in a 35mm optical format. The FTF7046 image sensors are available in color (RGB) and monochrome models. The sensor's frame rate is 1.2fps with 2 x 25MHz parallel outputs.

IHS: iPhone 5s Camera Gets Cheaper

Electronics360: Andrew Rassweiler, senior director of Teardown Analysis and Cost Benchmarking Service at IHS Electronics & Media, presented cost analysis of the new iPhone 5s and 5c, with both rear and front cameras taking roughly 6-7% of the BOM:

Preliminary Cost Estimate for iPhone 5s
Preliminary Cost Estimate for iPhone 5s
Preliminary Cost Estimate for iPhone 5c
Preliminary Cost Estimate for iPhone 5c

This is somewhat lower than IHS' last year estimation of $16.40 for the rear camera of iPhone 4s:

iPhone 4s Cost Analysis

Now I wonder how IHS came to the conclusion that iPhone 4S' 8MP 1.4um-pixel BSI camera module is cheaper than 5S' set of front and rear cameras 1.2MP+8MP, where the rear one has 1.5um pixel stacked sensor? Where do they find the savings? My guess is that they just estimated the chip areas, and, naturally, the stacked sensor came out smaller.

Sunday, September 29, 2013

Sony Playstation 4 to Support Gesture Control

IGN, Forbes: At Tokyo Game Show, Sony confirmed the PlayStation 4 will support gesture control via the PlayStation Camera, similar to Xbox One’s Kinect. But while Kinect will come bundled with the console, Playstation 4 users will need to buy the optional gesture recognizing camera separately.

Saturday, September 28, 2013

STMicro Develops B&W+R Sensor for Automotive

EETimes: STMicroelectronics develops B&W automotive image sensors for ADAS. An interesting feature is that it's able to single-out the red colored details, while everything else is B&W. When asked why red, ST BU director Martin Duncan said, "Red indicates a lot of important information - such as red traffic light, red road signs, brake lights, and tail lights."

OmniBSI-2 Video

Omnivision publishes a Youtube OmniBSI-2 video:



Friday, September 27, 2013

Nokia Publishes More Lumia 1020 Camera Details

The official Nokia Conversations blog published few more 41MP Lumia 1020 camera details coming from Eero Salmelin, Nokia’s Head of Imaging:

"First we have the optics system with its six element lens, which... is the highest performing optics ever been put into a Nokia device."

"In Nokia’s OIS solution the whole lens system is resting on ball bearings. The special gyro components are very accurate and fast in detecting how the phone is moving and the lens is moved by the motors in the opposite direction very quickly to compensate for the handshake."

"The camera module is the casing that packages everything together so that it slots neatly inside the Lumia 1020. It measures 25mm by 17mm and contains well over 130 individual components. Every single camera component is tested, calibrated and the quality is verified. We make sure that each module performs extremely well."

Lumia 1020 camera module

Fujifilm and imec Develop New Photoresist for Organic Image Sensors

Fujifilm and imec have developed a new photoresist for organic semiconductors that enables the realization of submicron patterns and does not hurt the light sensitivity of the organic photodetectors. For technical verification, Fujifilm and imec developed organic photo detectors (OPD) using the new photolithography technology, and tested their performance. Organic semiconductor materials were patterned to produce OPD composed of fine light receiving elements down to 200μm×200μm size (for some reason marked as 300um x 300um on the graph below). Generally, patterning of organic semiconductor materials degrades the property of converting light into electricity, but the OPD developed in this case were patterned without degradation:

Organic semiconductor materials were patterned to form OPD
of 300μm×300μm size, and photoelectric conversion property
with irradiation of artificial sunlight (AM1.5G) (red dotted line)
was examined. When compared with photoelectric conversion
property without patterning organic semiconductor materials (red
solid line), it did not show degradation due to patterning.
Similarly, there was no difference between with patterning (black
dotted line) and without patterning (black solid line) for
unirradiated dark current.

Toshiba Announces Stereo 3D Depth-Enabled Camera Module

Business Wire: Toshiba announces what it says the industry’s first dual camera module for smartphones, tablets and mobile devices. TCM9518MD incorporates twin 1/4 inch optical format 5MP camera modules (5MP x 2 arrays) and simultaneously outputs recorded images and depth data. Images captured with the module can subsequently be manipulated to change the depth of field and point of focus. Samples will be available by January 2014, with mass production scheduled for April 2014.

The dedicated companion LSI of the “TCM9518MD” measures and appends depth data to objects in the image. These data can be used for a wide variety of applications, including focus and defocus, and even to extract and erase objects from the picture. Used in combination with customers’ applications, the module supports the creation of new functions.

The companion LSI generates 13MP images by upscaling images taken by the twin 5MP cameras, realizing a lower module height than that of conventional 13MP camera modules. The TCM9518MD brings high resolution and computational camera functionalities to the CMOS image sensor market for smartphones, tablets and mobile devices.

The pixel size is 1.4um and the camera module measures 8.0 x 12.0 x 4.65 mm according to the PR (or 18 (W) × 12 (D) × 4.65 (H) mm according to the company's product page):

Thursday, September 26, 2013

LensVector Partners with Hosiden

Marketwired: LensVector and Hosiden announce an agreement for Hosiden to be LensVector's manufacturing partner. Production is scheduled to start early in 2014 in Hosiden's Shiga, Japan LCD facility. The agreement expands LensVector's manufacturing capabilities to serve the market for smartphones and other mobile devices. Hosiden will be using their LCD production technology, developed over a 40-year history of LCD production, to manufacture autofocus camera components to meet the quality and volume needs of the mobile phone market.

"Partnering with a world class manufacturer like Hosiden is an important step in establishing LensVector as the premier provider of solid-state autofocus lenses," said Howard Earhart, CEO, LensVector. "LensVector's manufacturing strategy is to merge high quality, low cost mature LCD panel manufacturing with our proprietary thin film process. This combination will provide very high capacity from existing Fabs yielding arrays of thousands of lens elements per panel."

Wednesday, September 25, 2013

Why There is No Camera Startups

Wired publishes an article by Marc Barros, the co-founder and former CEO of Contour, an action hands-free camera company. Marc discusses difficulties of creating a hardware company from scratch using camera startup as an example:

"Take cameras, for example. To make an amazing product you need: (1) a quality lens, (2) the latest image sensor, and (3) a powerful processor.

The best lenses are made in Japan (often by the camera makers themselves), so access to these components begins with $500K up front in engineering services and a guaranteed minimum order well into the thousands. Meanwhile image sensor companies are quickly being consolidated — so if a purchasing company isn’t a big name, it can’t even get access to the good stuff. For processors at least, the U.S. companies who created them are willing to provide access (to their true roadmap, SDK documentation, and engineering services), because they understand the importance of helping entrepreneurs build a product that maximizes their platform.

Overall, however, this lack of components means a hardware startup has to build volume with a crappy camera before they can make a really good one.

And of course, the established hardware players know their advantage in components is a massive barrier to entry. A few of them, like Sony and Samsung, are willing to sell other companies the same components used in their products as long as they don’t directly compete with them. Other companies, like Canon, build their own components to get ahead of the competition.
"

Tuesday, September 24, 2013

TSMC Develops Hardmask for Deep Implants

TSMC is going to present its recent developments in hardmask for deep implants in small pixel sensors at NMDC, Taiwan on Oct 6-9, 2013. The paper is titled "Nanotechnology Development for CMOS Image Sensor Applications" by C.C. Wang, T.H. Hsu, S.F. Ting, C.Y. Chen, K.C. Huang, J.C. Liu, S. G. Wuu. TSMC has developed a nice hardmask process for pixel size of order of 1um:

imec Presents its Image Sensor Technology

imec publishes its 2013 presentation on image sensor process and plans, presented on imec technology forum 3 weeks ago in Taiwan. Few slides from the doc:

Samsung Announces ISOCELL Pixel Technology

Business Wire: Samsung announces its new pixel technology, ISOCELL. "This new technology substantially increases light sensitivity and effectively controls the absorption of electrons [I'd guess they meant collection of electrons], resulting in higher color fidelity even in poor lighting conditions."

"Previous sensor technology developments focused on improving the light absorption of each pixel, and have progressed pixel technology from FSI (Front Side Illumination) to BSI (Back Side Illumination) which places photodiode at the top to maximize photoelectric efficiency. While being very effective at the time, this BSI technology also faced limitations in improving image quality as pixel sizes continued to decrease.

Building on these past advances and continuing the push toward higher quality image sensors for mobile devices, Samsung has developed ISOCELL the next generation of pixel technology, which is patent pending. ISOCELL technology forms a physical barrier between neighboring pixels – isolating the pixel. This isolation enables more photons to be collected from the micro-lens and absorbed into the correct pixel’s photodiode minimizing undesired electrical crosstalk between pixels and allowing expanded full well capacity (FWC).

Compared to conventional BSI pixels, the ISOCELL pixels decrease the crosstalk by approximately 30 percent which results in higher color fidelity to reproduce the original color with sharpness and richness, and increase the full well capacity (FWC) by 30 percent which leads to greater dynamic range.

Additionally, an imager designed with ISOCELL can feature a 20 percent wider chief ray angle (CRA), reducing the height of the camera module. This makes it suitable for slim and small form factor mobile devices with challenging low z-height requirements.

As the first Samsung image sensor to adopt this new technology, the S5K4H5YB 8Megapixel imager utilizes a 1.12um ISOCELL pixel and has a 1/4inch optical format. The S5K4H5YB is currently sampling to customers with mass production scheduled for Q4 2013.

According to market research firm Techno System Research, in 2014, approximately 66 percent of smartphones will feature image sensors with 8Mp or higher resolution.
"

Samsung Tomorrow, the official Samsung blog, publishes ISOCELL vs BSI comparison picture (click to expand):

Monday, September 23, 2013

IR on Aliasing Artifacts

Imaging Resource posted an interesting note on the recent trend to remove optical low pass filters over the sensors:

"There's been a strong move in the camera industry lately to remove low-pass filters (aka anti-aliasing filters or LPFs) from cameras, in pursuit of greater image sharpness... At IR, we feel strongly that eliminating low-pass filters is a bad idea, and a mistake for the industry. While the vast majority of natural subjects aren't subject to aliasing and moiré issues, many man-made objects have the sort of regular patterns that trigger the problem."

Imaging Resource supports its claim by real-world examples shot by one of the recent cameras with no optical anti-aliasing filter:

Example of color aliasing, caused by the fine thread patterns in the model's outfit.
Example of luminance moiré in the form of the swirly lines
in what should be diagonal louvers on the building's front.

Sunday, September 22, 2013

Pelican Imaging Publishes Full Resolution Original Images

Pelican Imaging publishes its oncoming paper at SIGGRAPH Asia 2013, including many high resolution images straight out of their reference system:

PiCam: An Ultra-Thin High Performance Monolithic Camera Array
Kartik Venkataraman, Dan Lelescu, Jacques Duparré, Andrew McMahon, Gabriel Molina, Priyam Chatterjee, Robert Mullis, and Shree Nayar

The full resolution pictures on Pelican's site are 3264 x 2448 and over 30MB in size, too large to post here. A down-scaled version of 1000 x 750 and full resolution crop of one of the images are below:


There are many more original images on the Pelican site, including ones taken at 30-50lux/15fps and 5-10lux/7fps ones. It's quite rare that an image sensor manufacturer posts the official full-resolution samples, including ones shot in the difficult lighting conditions.

An official press release is expected tomorrow. Thanks to LG for giving me a heads-up!

Update: The official PR is published on PRNewswire now.

Mike Tompsett's Presentation

IISW 2013 page publishes the presentation slides of Mike Tompsett, the inventor of the first CCD imager. The presentation slides talk about the early solid state imager history and more.


Thanks to EF for letting me know!

Saturday, September 21, 2013

Aptina Clarity+ Patent Application

Aptina's Clarity+ RB-WW color filter processing appears to be described in US20130242148 patent application "Imaging Systems with Clear Filter Pixels" by Marko Mlinar and Brian Keelan:

Aptina Applies for a Gated-Clock ISP Patent

Aptina's patent application US20130235239 "Power Saving Techniques for Image Sensors" by Vickie Wu describes a gated clock arrangement in image processing pipeline:

Sony Publishes September New Product Page

Sony publishes an English version of its new product page. The Japanese-language page was published earlier and has an identical content.

Friday, September 20, 2013

iFixIt: iPhone 5S Sensor Probably Made by Sony

iFixit team teared down the new Apple iPhone 5S. The conclusion regarding the main camera is:
  • We uncover the iSight camera.
  • The back of the iSight camera is labeled DNL333 41WGRF 4W61W.
  • According to our good friend Jim Morrison, Vice President of the Technology Analysis Group at Chipworks, "the DNL markings are consistent with the markings on the camera modules housing the Sony IMX145 we saw in the iPhone 4s and on the iPhone 5. The marks on the side of the module are different, but our industry insiders tell us this is Sony's again"
  • As Apple has stated the pixel pitch on this camera is 1.5 µ, this sensor should not be the IMX145, but a newer variant.


Update: Chipworks reverse engineering confirmed Sony sensor inside iPhone 5s camera. It's a stacked sensor:

Superpix Publishes Flyer of its DSLR Sensor

Superpix has published a flyer of its 12MP DSLR sensor, said to be the first China-made DSLR product and the first Chinese sensor beyond 10MP:

Thursday, September 19, 2013

STMicro Promotes its ToF Proximity Sensor

ST published a promotional video of its VL6180 proximity sensor, the first SPAD device from a major semiconductor manufacturer:

IHS Predicts Fast Growth in Automotive Gesture Recognition Systems

IHS forecasts that the global market for automotive proximity and gesture recognition systems to control in-car infotainment with a simple wave of their hand will grow to more than 38M units in 2023, up from about 700,000 in 2013. Nearly 40% of all new automobiles sold worldwide in 2023 will come with some degree of proximity or gesture recognition, according to a new IHS Automotive report entitled "Emerging Technologies: New Human-Machine Interface Trends."

Gesture recognition is defined as the use of cameras or sensors to track and convert a user’s movements into inputs for the infotainment system without any physical touch input. This would include things like waving a hand to the left or right to change radio presets or go to the next song in a playlist, or turning the hand clockwise or counterclockwise to raise or lower the volume.

A*STAR IME and nanoX Imaging to Develop MEMS X-ray Imager

Singapore A*STAR's Institute of Microelectronics (IME) and Japan-Israel-USA startup nanoX Imaging are to jointly develop a MEMS X-ray imager based on field emission detection. "We are confident that that the partnership with IME will be long-standing. In this first collaboration, we look forward to present to the medical industry an ideal solution of an imaging system that will offer high performance at low radiation risks," said Hitoshi Masuya, CEO of nanoX Imaging.

OMRON to Launch Imaging System for Interpretation of People's Intentions and Conditions

OMRON announces its Human Vision Component (HVC), a new image sensing component that can be embedded in a variety of devices and equipment. The HVC integrates ten algorithms from OKAO Vision and a camera module to a system capable of recognizing the intentions and conditions of people through face recognition, estimation of facial expressions, age estimation, gaze estimation, hand detection, and other functions. Launch of the standard-type HVC is scheduled for December. From next year onward, OMRON plans to introduce a succession of application-specific HVC products, which will incorporate optimized hardware and optical systems to enable customized functions that meet the requirements of each specific industry.

Examples applications include air conditioners capable of automatically adjusting the temperature to the level that people feel most comfortable; lighting products with efficient automatic control depending on the presence of people, their movements, or other conditions; automatic vending machines that suggest a product that matches the preference of a specific consumer; home appliances controllable via hand movements; robots that can react to the facial expressions of users; cars that prevent drowsy or inattentive driving by recognizing signs of fatigue in the driver's face; and more.

The ten functions to recognize the intentions and conditions of people are:
  1. face detection
  2. human body detection
  3. gender estimation
  4. age estimation
  5. gaze estimation
  6. facial pose estimation
  7. face recognition
  8. estimation of facial expressions (satisfied, unsatisfied, five different expressions: happiness, surprise, anger, sadness, and neutral)
  9. hand detection
  10. blink estimation

Wednesday, September 18, 2013

Aptina Engineers Shot by Aptina Sensors in 448 Megapixels

The Duke Imaging and Spectroscopy Program (DISP) posts a Gigapan snapshot of a of Aptina staff with David Brady (inventor of multi-scale cameras) at Aptina's headquarters in San Jose, taken with 32 Aptina 14MP sensors using the multi-scale camera architecture.

Thanks to SM for the letting me know!

Adimec on Smaller Pixels in Security

Adimec blog's recent post discusses the benefits and issues of using smartphone-style small pixels in 24/7 security camera applications. The pros and cons of small pixels are:

Pros:
  • Lower costs and weight through smaller sensors, optics, and cameras
  • Increased resolution with same sensor size and optics
  • Sensitivity in dim light is acceptable
  • Color reproduction can be good enough

Cons:
  • Give up global shutter (get rolling shutter)
  • Slower frame rates
  • More noise
  • Lower full well capacity
  • Lower MTF
  • Optical/Electrical crosstalk
  • Lower quality available optics

Thanks to GA for the link!

Truesense Publishes Detailed Spec of its First CMOS Sensor

Truesense Imaging has published a detailed 51-page spec of its first CMOS sensor KAC-12040, including QE, angular response and pretty much everything else one can think of. I believe Truesense's spec together with ON Semi's are the most detailed image sensor specs that are openly available on the net:

Monday, September 16, 2013

Truesense Announces Production Start of Fast CMOS and CCD Sensors

Truesense Imaging announces that its 12MP/70fps, 4.7um pixel KAC-12040 CMOS sensor, and 7.4um pixel based 1080p60 KAI-02170 and 4MP/28fps KAI-04070 Interline CCDs have now been released for production. "Through their features, functionality, and performance, these new products demonstrate the complementary nature of CCD and CMOS devices, and show the importance of having both technologies available in a full-featured product portfolio," said Chris McNiffe, CEO of Truesense Imaging.

The company has published a brief spec of its first CMOS sensor based on 5T GS pixels and said that it's the first of the future family:

Sunday, September 15, 2013

Canon Captures Firefies 30fps Video at its Full-Frame DSLR Sensor

Canon announces that it was able to capture a video of Yaeyama-hime fireflies flying in darkness on the high-sensitivity 35 mm full-frame CMOS sensor with 19um pixels developed by the company for video capture and announced in March 2013. No artificial lighting was used during shooting, which took place after sunset amid the island's mountains. Despite an exceptionally dark shooting environment of less than 0.01 lux, a level in which the naked eye would have difficulty discerning surrounding objects, the CMOS sensor at 30fps was able to capture not only the color of the light emitted by the fireflies, each of which measures only a few millimeters in length, and their movements, but also the surrounding vegetation in which the species lives.

Yaeyama-hime fireflies amid jungle vegetation
(Photomontage created from video footage)

A Youtube copy of the video from Canon site shows the footage:



Saturday, September 14, 2013

IISW 2013 Papers On-Line

IISW 2013 papers are open for free on-line access now. Thanks to EF for the link and for the work!

Thursday, September 12, 2013

PC Magazine Reviews Image Sensor Technology

PC Magazine published a surprisingly big and complete, albeit popular, review of image sensors history, principles, and technology. The 10-page article spans from film to CCD and CMOS sensors, from doping types to pn-junctions to process and then to packaging. Overall, it gives quite a complete picture for beginners.

Oops, the article is dated by 2001, quite old.

Movidius on Future of Imaging

Wired: Movidius CEO Remi El-Ouazzane publishes an article on future technologies in mobile imaging. The main points are:
  • Always-On Mobile Computer Vision Systems Coming Soon
  • Advanced camera topologies and novel image sensors; specifically lightfield cameras and array cameras.
  • Depth-enabled applications: enable next-generation applications such as Bokeh effects; background substitution in video conferencing; computational refocusing; improved gesture recognition; and image relighting using depth.
  • HDR video. A key element of HDR still imaging and video is to compress the dynamic range using tone mapping.
  • Stereo 3D imaging.
  • Simultaneous localization and mapping (SLAM): Mobile applications take advantage of the SLAM capability to map and navigate in complex unknown indoor environments.
  • Eulerian video magnification to reveal hidden information. All manner of micro changes in video streams to be magnified and observed, from the flow of blood as it fills the face to the movement of babies.
  • Sensor metadata: For example, multiple sensors (microphone, GPS, accelerometer, compass, etc.) can augment the cameras while network can supply rich semantic information

Thanks to DM for the link!

Intel on PerC

Fox News interviews Intel Israel president Mooly Eden on Perceptual Computing initiative:

IntoPix Presents TICO Light Video Compresion

Belgium-based IntoPix presents TICO - "visually lossless light compression specifically designed for the industry. This revolutionary technology is extremely tiny in hardware (FPGA, ASIC) , fast and powerful in software (CPU), robust for real-time operation with no latency." TICO is said to be "a smart upgrade path to manage higher resolutions (4K, 8K…) and frame rates while assuring visual quality, keeping power and bandwidth at a reasonable budget and reducing significantly the complexity and cost of the system." It features:
  • Visually Lossless quality between 2:1 and 4:1.
  • Persistent and Robust: Indistinguishable image loss over multiple generations
  • Latency: Few Microseconds – very few line of pixels ( selectable from 1 to x)
  • Small complexity and compact codec: easy to implement in low-cost FPGA or ASIC. It uses little internal memory and no external memory.
  • Powerful , Real-time or faster than real-time in CPU
  • Compatible with different resolutions, from mobile to 4K/8K UHDTV, via multiple usual transport schemes.
  • Designed to be a standard for the industry-wide support: TICO compression technology is available on multiple software and hardware technologies.

Wednesday, September 11, 2013

SiOnyx Shows 1mlux Imaging

Vision System Design: SiOnyx has shown its IR sensors at this year's SPIE Defense and Security Symposium in Baltimore, MD:

Comparison at below 1 mLux for a Best in Class CCD (left)
and SiOnyx XQE-1310 1.3MP 10um pixel imager (right).

According to Dr. Martin Pralle, VP of Business Development at SiOnyx, all the XQE sensors feature a 72dB native DR and an on-chip HDR feature that allows up to 120dB DR to be achieved. To date, the company has not yet openly published the spectral characteristics of these devices but, Pralle says. All Pralle would say was that at approximately 1064nm, the devices exhibited a QE approximately ten times that of standard CMOS image sensors. For night-vision applications, the US Army recently tested the SiOnyx XQE-1310 sensor in its labs and confirmed imaging at 1 mLux (1×10-4 Ft-candles).

Apple and Sony Smartphone Cameras

Apple announces the newest iPhone 5C and 5S smartphones. The 5S model features an improved 8MP camera with 1.5um pixels and brighter lens with F2.2 aperture. The sensor is said to be 15% larger in area. A combination of the brighter lens and the larger sensor provides 33% improvement in sensitivity. The 1.5um pixel size is quite unique. I can't recall any standard product with that pixel size. The closest size of 1.54um was used in 2-year old Canon Ixus-120IS camera. It appears that Apple sensor has been custom designed for the company.


In another publication, Sony Mobile explains its claim that its 20.7MP Xperia Z1 camera is best in class:

Tuesday, September 10, 2013

Pelican Imaging to Present its PiCam at SIGRAPH Asia 2013

Pelican Imaging publishes a Youtube video with its PiCam presentation, to be presented by the company's co-founder and CTO Kartik Venkataraman at SIGGRAPH Asia in Nov 2013 in Hong Kong:



While we are at Pelican, its lead investor in the recent round, Nokia Growth Partners, published a statement that it's not acquired by Microsoft and stays with Nokia, together with its two latest imaging investments: Pelican Imaging and Invisage.

Toshiba Image Sensors in Comics

Toshiba CIS web site promotes its CIS technology in comics now:


See more comics on HDR, BSI, and camera module on Toshiba web site.

Omnivision to Use BDA FastSPICE

Business Wire: Berkeley Design Automation announces that Omnivision has selected the company’s Analog FastSPICE (AFS) Platform for block-level characterization and full-circuit verification of their single-chip image sensors for consumer and commercial applications.

"We have stringent circuit verification requirements to achieve the high resolutions, low light levels, and resulting image quality of our CMOS image sensors," said T.J. Dai, director of analog and layout at OmniVision. "We selected the AFS Platform because it provides the accuracy and speed that enables us to perform full-spectrum device noise analysis to characterize the quality of our single-chip CMOS image sensor designs."

Bekeley DA's white paper on CMOS sensors simulations shows a transient noise simulation of a comparator, apparently in ramp ADC:

Comparator Output - Eye Diagram Window = 5 periods and Jitter Histogram

Monday, September 09, 2013

Gain-Exposure Drift Mystery Explained

Albert Theuwissen has selected the right answer on the gain-exposure mystery. The reason for the drift was LED light source variation in time.

TowerJazz CIS Update

TowerJazz Sept 2013 newsletter has few words on the image sensor business by Russell Ellwanger, the CEO:

"Our CMOS image sensor (CIS) business continues to gain market share in both X-ray and high-end camera segments. We are focusing on several initiatives including ramping to production a volume gesture control product; completing the IS11 platform, ramping to production our two major wins in dental and medical X-ray sensors, and completing the development of two mask stitching projects."

Soitec Licenses BSI-Related Patents to TSMC

PRNewswire: Soitec licenses some of its IP portfolio related to BSI technology to TSMC. Soitec says that the licensed IP covers "some of the key process steps of Soitec's Smart Stacking generic technology". BSI is said to be the key enabling technology in the race to develop small-pixel, high-quality image sensors used in consumer products such as digital cameras, smart phones and other portable electronics.

"We are very pleased to strengthen our relationship with TSMC. This licensing agreement demonstrates the value of Soitec's extensive and generic patent portfolio in manufacturing advanced devices such as BSI image sensors," said Paul Boudre, Soitec COO. "We look forward to continuing to leverage our IP portfolio by providing our partners with access to our technologies."

Sunday, September 08, 2013

World's Largest 320Giga-Pixel Panorama

This Youtube video shows an amazing level of details found in world's largest resolution 320Gigapixel panorama of London:

Saturday, September 07, 2013

Albert Theuwissen's Courses in 2013-2014

Albert Theuwissen pas published the schedule of courses for the next year:

  • Advanced Course on Image Sensor Technology
    October 21-23, 2013
    Barcelona, Spain
  • Basic Introduction to CMOS Image Sensors
    November 6-7, 2013
    Munich, Germany
  • Hands-On Measurements and Evaluation of Digital Cameras
    November 11-12, 2013
    Amsterdam, the Netherlands
  • Digital Imaging : Image Capturing, Image Sensors, technologies and Applications
    November 25-29, 2013
    Dresden, Germany
  • Basic Introduction to CMOS Image Sensors
    March 26-27, 2014
    Munich, Germany
  • Advanced Course on Image Sensor Technology
    April 7-9, 2014
    Dresden, Germany
  • Hands-On Measurements and Evaluation of Digital Cameras
    April 10-11, 2014
    Amsterdam, the Netherlands
  • Digital Camera Systems
    September 15-17, 2014
    Amsterdam, the Netherlands
  • Digital Imaging : Image Capturing, Image Sensors, technologies and Applications
    October 20-24, 2014
    Barcelona, Spain
  • Basic Introduction to CMOS Image Sensors
    November 6-7, 2014
    Munich, Germany

Yole Reviews Uncooled IR Imaging Market

Yole Developpement publishes "Infrared imaging: Uncooled infrared imaging technology & market trends report". The market experienced a strong downturn in 2012: though total shipments slightly increased by + 1%, overall revenue fell by 12% to $1,824M, mainly due to the weakness of the military market. A shift toward commercial higher volume markets is predicted in 2013-2018:


The major manufacturers (FLIR, DRS, Raytheon, ULIS, GWIC) have now moved to 8” production lines from 6” to reduce wafer cost. Several have even outsourced their production to foundries to further reduce production cost. These two elements are preliminary signs of a steep microbolometer cost reduction that will open up cost driven applications such as smartphones.


The major trend is to move to wafer level manufacturing of both imager and optics to maximize benefits of high volume manufacturing. Secondly, a race toward shrinking pixel pitch is on-going. All microbolometer manufacturers focus on developing a 12µm pixel, improving sensitivity enough to reach the 50mK target.

Friday, September 06, 2013

Fujitsu Announces Milbeaut ISP in SuVolta Process

Fujitsu announces the 7th generation of its Milbeaut ISP, featuring SuVolta DDC low power process. Other than the new process, the ISP features a newly developed circuit for correcting focus drop-off at the edges, an improved distortion correction and purple fringing countermeasures, improved noise filter, has a processing speed of 12fps at 24MP (said to be twice the speed of the previous generation), supports JPEG-HDR, a JPEG-compatible HDR format from Dolby Labs, and can do H.264 encoding/decoding at 1080p30 or 1080i60 rate.

Getting back to SuVolta process, DDC stands for Deeply Depleted Channel transistors and claims to achieve power saving and/or speed-up in more-less standard bulk CMOS techniology:


A 4.5-min long video explains SuVolta's basic claims:


The basic claim is a reduction of transistor variability by about a factor of 2:


In theory, these ideas could bring a new life to current-mode image sensors that traditionally suffer from high PRNU due to Gm variations. The regular 3.3V transistors have stddev Gm variations of 1-2% per sq.um area, depending of the fab, process, and device type. Reducing it down to 0.5-1% per sq.um could make current mode pixel more competitive.

For those who have more time, there is also a 3-part lecture in Stanford University introducing DDC technology and comparing it with FinFET and other approaches (part 1, part 2, part 3):