College of Engineering -- University of Wisconsin-Madison
ELECTRICAL & COMPUTER ENGINEERING
n response to needs arising from new technologies ranging from medicinal therapy to high-speed data communications, novel, high-performance aluminum-free diode lasers developed by Assistant Professor Luke Mawst (pictured) and Professor Dan Botez soon may replace their aluminum-based counterparts.
In current Al-based technology, the aluminum content in a laser's active region determines its wavelength; lasers that achieve shorter wavelengths require more aluminum. However, because aluminum-containing compounds in the active region are highly reactive, impurities such as oxygen can deteriorate the laser's reliability. Eliminating aluminum from the laser's active region improves its reliability, and removing it from the semiconductor layers around the active region makes it easy to fabricate complex high-performance diode lasers.
In an ongoing project funded by NSF and industrial grants, Mawst and Botez are growing InGaAsP-based materials on GaAs substrates using metalorganic chemical vapor deposition (MOCVD), and fabricating lasers using microelectronic processing technology. The resulting high-power prototypes are being tested in photodynamic therapies for cancer, atherosclerotic cardiovascular disease and age-related macular degeneration; and also to spin-polarize gases such as xenon or helium for use in a new form of magnetic resonance imaging (MRI), which has demonstrated image resolution of a million times greater than conventional MRI.
Mawst and Botez, of the Reed Center for Photonics, are collaborating with Princeton and Duke universities to develop and implement the Al-free laser sources, and Mawst also is developing Al-free vertical-cavity surface-emitting lasers for laser printing, optical recording and sources for data communications via optical fibers. Mawst, Botez and graduate student Tom Earles are also starting a spin-off company called Alfa Light based on the technology.
Integrating electronics into automotive accessories
Traditionally, automotive accessory functions such as steering and braking operate hydraulically, pneumatically or mechanically, but Professor Thomas Jahns would like to see them powered electrically. Although these "drive-by-wire" vehicles of the future will feature improved performance, fuel efficiency and reduced emissions, they may require as much as five times more electrical power than is generated in current internal combustion engine vehicles.
With colleagues from MIT, Jahns is developing a combined starter/alternator system that uses an ac permanent-magnet machine mounted in place of the flywheel between the engine and the transmission. The system will both start the engine and generate up to six kilowatts of electrical power--roughly five times that of today's high-end vehicles--and Jahns is supervising research to develop a digital control system to operate the magnet machine from standstill to maximum engine speed.
However, this drive-by-wire technology can be costly and vulnerable to failures in the severe under-the-hood environment. In related research through the Center for Power Electronics Systems, a new National Science Foundation engineering research center, Jahns hopes to reduce those barriers by developing integrated power-electronic modules that not only process electric power for automotive accessories, but pave the way for affordable and reliable power electronics in everything from air conditioners to washing machines.
Making microprocessors more efficient
New developments in microchip technology, hardware and software are the impetus for researchers in the tiny world of microprocessor microarchitecture to make big strides, and Professors James Smith and Gurindar Sohi are working to keep microprocessors up to speed.
They're investigating methods for increasing simultaneous instruction execution, or instruction-level parallelism (ILP). In their research they examine microprocessor designs that automatically find as many simultaneous operations as possible, and determine ways to enhance those operations by predicting the outcomes of certain operations in advance and initiating new operations based on the predictions.
To carry out ILP studies, Smith and Sohi rely heavily on real-program simulations on the computer, which they study to characterize accurately the types of simultaneous operations and predictabilities that "naturally" occur. To help, they've developed detailed processor models and conducted experimental simulations involving many design choices and programs. Smith and Sohi also are focusing on high-performance processors that can overcome increasing on-chip communication delays by localizing processor functions and distributing ILP operations. Their developments translate into increased performance not only in personal computing, but also in areas like entertainment, environmental monitoring, manufacturing systems and electronic commerce.
Copyright 1999 University System Board of Regents
Content by firstname.lastname@example.org