VLSI IC design

VLSI Physical Design and Image Sensors

Program leader: Edoardo Charbon
Researchers: Claudio Favi, Matthew Fishburn, Mohammad Karami, Shingo Mandai, Yuki Maruyama, Juan Mata Pavia, Mauro Scandiuzzo, Chokalingam Veerappan, Hyung-June Yoon

Mission

Our research mission is to model and develop hardware/software systems based on quantum devices. Particular emphasis is on high-speed 2D/3D optical sensing, biomedical imaging, embedded and reconfigurable processing architectures, single-photon avalanche devices (SPAD), and design optimization techniques.

Teaching

Our teaching mission is to equip engineers and scientists with a multi-disciplinary knowledge on state-of-the-art hardware/software systems.
ET 4054 Methods and algorithms for system design

Research

• Ultra-fast IC Design
• Fundamentals of Single-photon Detection
• Single-Photon Imaging
• Biomedical Imaging
• DNA Detection
• 3D Imaging
• In-vivo Brain Imaging
• Optically synchronied instruction set extensions

Projects

MEGAFRAME: time-correlated systems for biophotonics and medical imaging

MEGAFRAME is a FP6 FET project (2006-2010) aimed at the research of low-cost techniques for single photon counting. MEGAFRAME yielded a family of imagers implemented in 130nm CMOS technology capable of on-pixel single-photon time-of-arrival detection with 55-119 picosecond resolution. The chips operate at 1 megaframe per second in time-correlated single-photon counting (TCSPC) and time-uncorrelated photono counting (TUPC) mode.

MEGAFRAME's long-term goal is to bring single photon imaging technology to an advanced low-cost deep-submicron CMOS platform, so that massive arrays of single photon detectors can coexist with and interface to large networks of parallel digital processing units on the same chip. The target fields are biology, physics, and medical imaging. However, any disciplines requiring time-resolved ultra-fast optical sensors are prime candidates.

SPADnet: fully networked, digital components for photon-starved biomedical imaging systems

SPADnet is an FP7 STREP (2010-2014) to follow onto MEGAFRAME footsteps. SPADnet aims to develop a new generation of smart, CMOS-based large area networked image sensors for photon-starved biomedical applications, build ring-assembly modules for Positron Emission Tomography (PET) imaging, and carry out performance tests in a PET system evaluation testbed.

Single-Photon detection fundamentals

Single-photon sensors have existed for decades, but only recently CMOS implementations of a class of such detectors, known as single-photon avalanche diodes (SPADs), have emerged. In this research we are looking at the fundamentals of SPADs from several points of view, from noise to sensitivity, from dynamic range to image quality. The study is done in the context of future large array implementations currently being developed.

Deep-submicron single-photon detection

Miniaturized single-photon detectors may represent the future of large imaging systems with multifunctionality emebedded on chip. In this project we propose to use deep-submicron CMOS processes to achieve ultra-fast miniarturized devices that can be scaled up in order to achieve megapixel arrays. We study the propoerties of devices implemented in technologies that are optimized for speed and the trade-offs possible in these technologies.

Time-resolved brain imaging

This project deals with imaging algorithms and sensors to enable 3D imaging of the brain activity in young infants to monitor potential lesions in the very early stages of development. The technique is based on the use of near infrared diffused illumination in combination with SPAD based imagers capable of detecting the response of a cluster of neurons using picosecond time resolution.
This project is in collaboration with the Unversity Hospital Zurich.

Endoscopic Gamma/positron camera

Cancer detection, especially at an early stage, has become critical in successfully treating several forms of tumor. Surgeants are however often blind in the O.R. due to thesize and relatively low contrast characterizing tumor cells. In this project we propose to develop fluoresecence cameras capable of detecting tumoral cells through fluorophore enhancement. The technique should enable safer radioactivity-free detection of super-small structures and micro-metastatic fibers.
This project is in collaboration with Dr. Christian Mester, CHUV Unversity Hospital, and Forimtech.

Three-dimensional vision sensors

Three-dimensional sensing has become one of the most active areas of research in Computer Vision recently, with the introduction of more and more advanced 3D interpretation and rendering algorithms. One of the weak points of the field is the lack of low-cost high-precision 3D cameras. With the recent introduction of true photon counting detectors integrated in CMOS technology, the architectural paradigm in 3D sensing has shifted from analog to digital methods. In this research activity we investigate new, more compact, pixel-based, architectures for optical rangefinding and 3D imaging for a number of applications in medicine, entertainment, and industrial vision systems.

Label-free DNA detection

Gene analysis has already been applied for tailor-made healthcare. It allows us to prevent serious diseases. Additionally, doctors can provide the most suitable treatment tailored to each patient based on his/her genetic information. This project offers biochemists the full potential of label-free CMOS DNA sensors, while enabling low-cost and highly accurate detection in a fraction of the time required by conventionalt sensiors.

Optically synchronized instruction set extensions in multiprocessors

Quantum optical detectors will generate large amounts of data on-demand. The volume of data will depend on the structure of the picture and/or its depth map, illumination, etc. To deal with such as sparse data we propose a globally asynchronous locally synchronous (GALS) approach based on instruction set extensions (ISEs) synchronized optically on chip.

Gigavision: a fully digital, high dynamic range vision concept

In this project we propose a new concept to the design of imagers based on fully digital pixels. Grey levels are achieved by measuring the density of pixels that had detected at least a photon. Due to the simplicity of theier design, pixels may be much smaller than their analog counterparts, smaller than the diffraction-limited spot, and thus the lens acts as a low-pass filter and pixels behave very similarly to Silver grains in conventional photosensitive emulsions. The advantage is the near-perfect logarithmic response without excessive` pixel-to-pixel variability.

BSIspad: back-illuminated SPAD technology

In this project we propose to use backside illumination in combination with single-photon imaging so as to improve a number of performance measures. The technique, in combination with capacitive/inductive coupling, enables to combine several technologies in one package thus taking advantage of benefits of some 3D integration processes.

|  17 June 2010   |