Distinguished Lectures Series
The ECE Distinguished Lectures Series brings world-class researchers to the University to share their research and discoveries.
Random Ideas about Biological Networks
Why does the functioning of biological systems seem miraculous? One reason is that we do not know how to design systems that do what cells do, namely molecular computing. In contrast, we know how to design highly complex information systems. The fundamental reason for the successful evolution of information systems is the development of mathematical abstractions that enable efficient and robust design processes. In particular, Claude Shannon in his classical 1938 Master Thesis demonstrated that all Boolean functions can be computed by relay circuits, leading to the development of digital logic and resulting in computer chips with over a billion transistors. Motivated by the challenge of analyzing stochastic gene regulatory networks, we generalize the notion of logic design to probabilistic logic design. Specifically, we consider relay circuits where deterministic switches are replaced by probabilistic switches and present efficient algorithms for synthesizing networks that compute probability distributions.
Integrated Power Conversion – The Switched Capacitor Approach
Technical merits and challenges of the switched capacitor approach to dc-dc power conversion are discussed. A detailed analysis enabling a strategic comparison among switched-capacitor converter topologies and also enabling comparison of switched-capacitor topologies with conventional magnetic topologies is outlined. The analysis framework allows a quantitative comparison of the various popular power conversion circuits in terms of their utilization of switch technology and also their utilization of energy storage devices (eg. capacitors, inductors). Roughly, the analysis views a power converter as an ideal transformer with a given or adjustable conversion ratio, with losses modeled with a series output resistance for load-dependent losses, and with a parallel resistive impedance to capture frequency-dependent and static leakage losses. Significantly, the analysis shows that for a wide range of conversion applications, switched capacitor converters outperform the conventional buck, boost, and transformer-based converters with respect to component utilization.
Since switched capacitor converters contain no magnetic devices, they are well suited to integration in a range of CMOS processes. Further, since devices can be effectively stacked, extended voltage operation can be realized with low voltage processes. However, switched capacitor converters also present a number of challenges in voltage regulation and in ripple performance. Design strategies to meet these challenges are outlined. Data developed from on-going experimental work will be discussed.
Preparation & properties of multi-component nanocrystal superlattices & nanocrystal based devices
Christopher B. Murray
Semiconductors synthesized in, and processed from, the solution phase offer new avenues to achieving large-area, low-cost electronic and optical devices. Areas of application include flexible transistors for controlling displays; highly spectrally-tuned light emitters and sensors; and new solar materials. The synthesis of colloidal nanocrystals with controlled crystal shape, structure and surface passivation provides a rich family of nanoscale building blocks for the assembly of new solid thin films and novel devices. The tunability of the electronic, magnetic, and optical properties of the nanocrystals has lead to them being compared to a set of “artificial atoms”. This talk will provide key insights into the development of “best practices” in preparation, isolation and characterization of semiconducting quantum dots, nanocrystal phospors and magnetic nanoparticles. A very brief discussion of the organization of monodisperse nanocrystals in to single component superlattices that retain and enhance many of the desirable mesoscopic properties of individual nanocrystals will transition into a discussion of multicomponent assembly. The potential to design new materials expands dramatically with the creation binary nanocrystal superlattices BNSLs. I will show how we synthesized differently sized PbS, PbSe, CoPt3, Fe2O3, Au, Ag and Pd nanocrystals and then these nanoscale building blocks into a rich array of multi-functional nanocomposites (metamaterials). Binary superlattices with AB, AB2, AB3, AB4, AB5, AB6 and AB13 stoichiometry and with cubic, hexagonal, tetragonal and orthorhombic packing symmetries have been grown. The opportunity to optimize materials for applications in solution processable photovoltaic systems and phosphor based luminescent concentrators will be highlighted. We have also identified a novel method to direct superlattice formation by control of nanoparticle charging. Although modular nano-assembly approach has already been extended to a wide range of nanoparticle systems, we are confident that we have produced only a tiny fraction of the materials that will soon accessible. Recent progress in the extensions to the formation of quasicrystalline colloidal phases will be shared. Progress toward large area (~1cm2 ) processing and integration and device fabrication.
Encapsulating Designer Knowledge: Improving Digital and Mixed Signal Design
For the past 40+ years society could count on the scaling of silicon technology to make information technology faster, lower power, and cheaper. Today we face two huge challenges. The first is that while silicon technology continues to scale, physical limitations mean that the energy and performance gains are modest. The second is through scaling we are now building incredibly complex systems, which commensurate design costs. The net result is a paradox of needing to build specialized chips to continue to scale performance, and having technology that is amazingly capable, but no one can afford the large (>$20M) upfront design costs to use it.
In this talk I will review my group’s research to reduce this large design cost by embedding designer knowledge in the designs created. For analog and mixed signal design, this involves making analog design more like digital design, with real reusable cells, and functional and electrical rules checking. For digital design this involves creating flexible building blocks and using them to create flexible chip designs that can generate the specialized instance desired. I will provide some examples from my group indicating the promise of this approach – including improving the energy efficiency of a H.264 encoder by 200x.
Efficient Light and Solar Energy Generation Through Nanoscale Control of Organic Materials
In this talk we will look at several promising approaches to creating nanoscale morphology in small molecular weight thin films. Moving from the conventional vacuum deposited CuPc/C60 based system, we will examine a number of different routes to creating nanostructures based on new materials such as squaraines, carbon nanotubes and subphtalocyanine. Combinations of solution and vapor phase growth will be discussed. Routes to demonstrating very high efficiency single junction and tandem architectures with these materials and growth combinations are considered. In addition, we will consider how morphological control can also be used to generate high efficiency light emission from organic coherent emitters.
Resonant Circuit Oscillators: How they Work, why they are Imperfect
Every electronic system that processes signals must have references built into it that define a scale for amplitude and time. A local oscillator serves as the time reference, and except in some rudimentary uses, it is based on a resonator. The principles for the oscillator are the same, whether it uses a quartz crystal, an inductor-capacitor tank, or a machined cavity as the resonator.
In this talk I will describe the electronic circuit that develops oscillatory power from these resonators, and show in simple but accurate terms how oscillation starts and settles into steady-state. This necessarily involves a discussion of nonlinearity. Fluctuations in the circuit’s currents and voltages introduce what is termed phase noise and jitter, and I will show that these can be predicted without too much fuss.
The presentation is suited to EE undergraduates and graduate students who know basic signal theory and simple transistor circuits.
Next-generation Ultra-Low-Power System Design
Anantha P. Chandrakasan
Next-generation handheld devices and wireless sensors for health and environmental monitoring, will require dramatic reduction in energy consumption. The ultimate goal is to power these devices using energy-harvesting techniques such as vibration-to-electric conversion, or through body heat. A system-level approach must be used to optimize such devices. Relevant considerations include ultra-low-voltage digital circuit operation, application-specific digital and mixed-signal architectures, extreme parallelism, computation vs. communication trade-off, and integrated energy-processing circuits. The use of analog-assisted digital circuits (such as embedded switched-capacitor power management, and offset compensation in sense amplifiers) will be critical in dealing with device variability and low-voltage operation. Efficient energy-processing circuits (for generation, buffering, and conversion) is critical in many applications. Several system examples will be shown, covering portable biomedical and multimedia devices.
ABC: An academic “industrial-strength” verification tool
About 6 years ago, ABC was started as a replacement for our synthesis tool SIS. As part of this development, we needed to put together an equivalence checker for assuring that synthesis algorithms had been implemented correctly. Some of the new synthesis algorithms in ABC involve scalable sequential operations, such as retiming and sequential SAT sweeping, which yield impressive results. A companion scalable sequential equivalence checker (SEC) was needed and developed. The resulting SEC engine was more general and is a powerful property as well. The verification part of ABC was packaged and submitted it to the model checking competitions of CAV’08, at which it won in 2 of 3 categories and at CAV’10 it won again in 2/3 categories.
Working on both synthesis and verification has allowed us to borrow ideas from the best of both worlds and to observe the growing synergy between the two areas. This synergy works in two directions; the ability to synthesize large problems and show impressive gains spurs on development of equivalently scalable SEC methods and the ability to scalably verify sequential equivalence problems spurs on the development, use, and acceptance of some sequential synthesis methods. In ABC, similar concepts and algorithms are used in both synthesis and verification, e.g. use of AIGs, rewriting, use of SAT, sequential SAT sweeping, retiming, and interpolation.
In this talk, we will provide an overview of ABC, discuss some ways in which synthesis and verification ideas have influenced each other, and illustrate how various algorithms have been assembled and employed to create a powerful model checking engine that can rival some commercial offerings.