Wednesday, August 24, 2016

Microsoft Talks About Hololens Vision Processing

EETimes: Microsoft says that HoloLens processing unit (HPU) fuses input from five cameras, a depth sensor and motion sensor, compacting and sending it to the Intel SoC. It also recognizes gestures and maps environments including multiple rooms. The TSMC 28nm HPU packs 24 Tensilica DSP cores and 8MB cache into a 12x12mm package with 65M transistors. A GByte of LPDDR3 is included in the HPU’s package.

HPU die

Mobileye and Delphi Partnership to Invest Hundreds of Million of Dollars into Self Driving Technology

SeekingAlpha transcript on Mobileye and Delphi partnership announcement has a statement that the two companies invest significant funds to develop self-driving car technology:

"When you think about what is needed to bring the Level 4/5 autonomy to series production, there is sensing – interpreting sensing on one hand, building an environmental model where all the moving objects and obstacles and all the path and symmetric meaning, but there is another component to it, which is being able to merge into traffic in a way that mimics human driving behavior. And there is machine intelligence to be ported into this. And there is a synergy between Delphi’s core IT in that area and Mobileye's core IT in that area and together we can bring a new class of machine intelligence into this project. you can imagine just given the level of technology required and the amount of integration, on a combined basis [our investment] is hundreds of millions of dollars.

Tuesday, August 23, 2016

TowerJazz and TPSCo Announce Stacked Deep PD Technology

GlobeNewsWire: TowerJazz and TowerJazz Panasonic Semiconductor Co. (TPSCo) announce a new state of the art CIS process based on stacked deep PD, allowing customers to achieve very high NIR sensitivity and realize extremely low cross-talk while keeping low dark current characteristics, using small pixels and high resolution.

This solution targets 3D gesture recognition and gesture control for the consumer, security, automotive and industrial sensors markets. NIR is becoming more and more popular in 3D gesture recognition applications and in automotive active vision applications for better visibility in harsh weather conditions. These ToF applications are using a NIR light source and ToF, creating a 3D image.

Current solutions generally use a thick epi on p-type substrate to achieve high sensitivity, but this creates high cross talk (low resolution) and high dark current values. The novel pixel structure developed by TowerJazz and TPSCo has a stacked deep photodiode, providing both high sensitivity and low cross talk at NIR. This allows very low dark current values, especially at elevated temperatures, required in the automotive market.

The tremendously fast growth of 3D gesture application in the consumer market such as PC and mobile as well as in the automotive area will allow us to attract many customers with this technology that is the best the market has to offer,” said Avi Strum, SVP and GM, CMOS Image Sensor Business Unit, TowerJazz.

The process was developed on TPSCo’s 65nm CIS technology on 300mm wafers in its Uozu, Japan fab and is already in production for leading edge automotive and security sensors. It will also be available for new designs in TPSCo’s 110nm fab in Arai, Japan and in TowerJazz’s 180nm fab in Migdal Haemek, Israel.

Monday, August 22, 2016

2016 Harvest Imaging Forum

Agenda of 2016 Harvest Imaging Forum has been published. The Forum is devoted to "Robustness of CMOS Technology and Circuitry outside the Imaging Core : integrity, variability, reliability." The 2016 Harvest Imaging forum is split into two parts, divided over two days:
  1. As all CMOS robustness topics are related to the basic CMOS devices and their operation, an in-depth knowledge of the most important fundamentals of CMOS physics, CMOS device and circuit operation, fabrication and design are necessary to ease the understanding of the robustness topics. For that reason time the first part of the forum will concentrate on the topics that have to do with CMOS physics, devices, circuits, fabrication and design, such as:

    CMOS device physics including the basic MOS device operation of nMOS and pMOS transistors, transistor current expression, the MOS diode and the MOS capacitor, the temperature dependence of the devices, the effect of the continuous scaling of CMOS technology and its problems, such as mobility reductions and leakage mechanisms,

    CMOS process technology including the basic CMOS process flow, advanced planar and FinFET technologies,

    CMOS circuit design, including basic logic gates, cell libraries, design flow and terminology.

  2. The robustness of advanced CMOS integrated circuits. The second part of the forum includes a lot of CMOS problems that can show up as artefacts in the final captured image. Most of the imaging engineers are familiar with the effects on a display or hard-copy, but what can be root cause of the image quality problems? Topics that will be discussed in the forum are:

    Signal integrity issues such as cross-talk, signal propagation, interference between ICs, current peaks, supply noise, substrate and ground bounce, on-chip decoupling capacitors and design consequences,

    Variability issues including difference between random variations and systematic variations, causes of process parameter spread, proximity effects, random dopant fluctuations, transistor matching and design consequences

    Reliability issues and topics such as electro-migration, latch-up, hot-carrier effects, NBTI, soft-errors (by cosmic rays and alpha-particles), electro-static discharge, etc.

The 2016 Harvest Imaging Forum will include a copy of: “Nanometer CMOS ICs, from Basics to ASICs” (Springer 2016) and “Bits on Chips” (Springer, 2016), as well as a hard copy of all sheets presented.

Hikvision Secures $6b Credit Lines

China-based surveillance camera maker Hikvision secures a credit facility of RMB ¥20b (more than USD $3b) with Export-Import Bank of China. In November 2015, Hikvision secured another USD $3b line of credit with China Development Bank.

So large credit lines from China state-owned banks reportedly raise some concerns in the industry that Hikvision gets an unfair advantage over its competitors.

The company's 2015 revenues were USD $3.88b, representing a YoY growth rate of 47%. Hikvision also has liquid funds available in the amount of RMB ¥11.8b (more than USD $1.78b). Hikvision is the world’s largest provider of video surveillance products and solutions for the fifth consecutive year, and the No. 1 global provider of IP cameras, according to IHS Research.

Sunday, August 21, 2016

High Speed Image Sensor Applications

Tokyo University Ishikawa Watanabe Lab publishes a couple of Youtube video exploring high speed image sensor applications. "High-speed 3D Sensing with Three-view Geometry Using a Segmented Pattern" demos 1000fps 3D camera:

"High-Speed Image Rotator for Blur-Canceling Roll Camera" demos rotation-compensating camera:

Saturday, August 20, 2016

e2v Onyx 10um Pixel Demo

e2v publishes a Youtube video showing night vision capabilities of its NIR 1.3MP Onyx sensor with 10um pixels:

Friday, August 19, 2016

Canon Proposes "Teardrop" Microlens

Canon patent application US20160233259 "Solid state image sensor, method of manufacturing solid state image sensor, and image capturing system" by Yasuhiro Sekine proposes a "teardrop" shaped microlens in order to reduce lens shading on the periphery of the pixel array:

A Race to Self-Driving Taxi Has Begun

Bloomberg, IEEE Spectrum: Starting later this month, Uber will allow customers in downtown Pittsburgh to summon self-driving cars from their phones, crossing an important milestone that no automotive or technology company has yet achieved.

The minute it was clear to us that our friends in [Google] Mountain View were going to be getting in the ride-sharing space, we needed to make sure there is an alternative [self-driving car],” says Uber CEO Travis Kalanick. “Because if there is not, we’re not going to have any business.” Developing an autonomous vehicle, he adds, “is basically existential for us.

Uber’s modified Volvo XC90 SUV.

Thursday, August 18, 2016

Ford Self-Driving Car Plans

Ford announces the steps to mass produce a fully autonomous driving car in 2021. To get there, the company is investing in or collaborating with four startups to enhance its autonomous vehicle development, doubling its Silicon Valley team and more than doubling its Palo Alto campus.

On the imaging side, Ford announcing four key investments and collaborations in advanced algorithms, 3D mapping, LiDAR, and radar and camera sensors:

  • Velodyne: Ford has invested in Velodyne, the Silicon Valley-based company dealing with light detection and ranging (LiDAR) sensors. The aim is to quickly mass-produce a more affordable automotive LiDAR sensor. Ford has a longstanding relationship with Velodyne, and was among the first to use LiDAR for both high-resolution mapping and autonomous driving beginning more than 10 years ago
  • SAIPS: Ford has acquired the Israel-based computer vision and machine learning company to further strengthen its expertise in artificial intelligence and enhance computer vision. SAIPS has developed algorithmic solutions in image and video processing, deep learning, signal processing and classification. This expertise will help Ford autonomous vehicles learn and adapt to the surroundings of their environment
  • Nirenberg Neuroscience LLC: Ford has an exclusive licensing agreement with Nirenberg Neuroscience, a machine vision company founded by neuroscientist Dr. Sheila Nirenberg, who cracked the neural code the eye uses to transmit visual information to the brain. This has led to a powerful machine vision platform for performing navigation, object recognition, facial recognition and other functions, with many potential applications. For example, it is already being applied by Dr. Nirenberg to develop a device for restoring sight to patients with degenerative diseases of the retina. Ford’s partnership with Nirenberg Neuroscience will help bring humanlike intelligence to the machine learning modules of its autonomous vehicle virtual driver system
  • Civil Maps: Ford has invested in Berkeley, California-based Civil Maps to further develop high-resolution 3D mapping capabilities. Civil Maps has pioneered an innovative 3D mapping technique that is scalable and more efficient than existing processes. This provides Ford another way to develop high-resolution 3D maps of autonomous vehicle environments