Driving toward millivolt electronics
Researchers at the University of California—Berkeley are developing devices that will yield a millionfold reduction in power for future generations of electronic systems.
Pallab Chatterjee, Contributing Technical Editor -- EDN, December 1, 2011
Thanks to new behaviors and the characteristics of materials at
small geometries, nanotechnology has the potential to introduce
great change to the electronics arena. The University
of California—Berkeley’s Center for E3S (Energy Efficient
Electronics Science) is working to develop fundamental
devices that will result in a millionfold reduction in power
for future generations of electronic systems.The overall goal of the center, which receives its funding from the National Science Foundation, is to develop high-performance devices and circuits that require low power to run. Current electronics nominally operate at 1.25 to 5V. The E3S Center is researching devices that can operate at millivolts. The more-than-1000-mV reduction in required power would result in a power-reduction factor of 1 million, assuming that power is proportional to voltage squared.
Eli Yablonovitch, PhD,
leads the research group, which
focuses on nanoelectronics,
nanomechanics, nanophotonics,
nanomagnetics, and system
integration (Figure 1). The
nanoelectronics program focuses
on eliminating the voltage mismatch
between devices, allowing
circuits to operate on just a few
millivolts of supply power and
still have a large enough noise
margin. The group is investigating
switches—from the level of
modulating thermionic-emission
current over a barrier to the level
of modulating the tunneling
current through a barrier. The
challenges include modeling
and tuning tunneling probability
and identifying and manufacturing
materials exhibiting the
desired band-edge-energy alignment.Tsu-Jae King Liu, PhD, leads the nanomechanics research team in addressing the off-state leakage in CMOS switches, which sets a lower limit in energy per operation. Mechanical switches have zero off-state current. The mechanical switches have a speed limitation of reaching saturation at the speed of sound at approximately 340m/sec in air. When the switches use a diamond-substrate architecture, they have a saturation point close to that of an electron, which is comparable to that of a CMOS switch. The researchers are addressing the complexity of manufacturing the devices and creating higher-density circuits.
Ming C Wu, PhD, is leading the nanophotonics research group, which focuses on circuit and data communication using light rather than electrons. Photons can consume less energy for longer links, and recent demonstrations have shown optical interconnects achieving 10−12J of energy per bit. The challenge includes the creation and identification of fundamentally new approaches for transmitters and detectors to get the operating energy to the goal of communicating with one optical bit using 10−17J of energy.
Jeff Bokor, PhD, leads the nanomagnetics
researchers, who are investigating
nanomagnetics devices for
logic functions rather than memories.
Nanomagnetic logic allows a spin
degree of freedom with low energy
dissipation—even less than the
Landauer Limit—on each transition.
(The Landauer Limit is
the minimum possible amount of
energy necessary to change 1 bit
of information.) Theoretically, a
room-temperature circuit or memory
operating at the Landauer
Limit could change at a rate of 1
Gbps and expend only 2.85 trillionths
of 1W of power. Scientists
have recently challenged this
principle, and the circuits in this
research program are using energy
below this threshold.The last area of focus is the development of systems that use these devices and techniques to allow for the realization of known common circuit functions. These circuits would use models that only nanotechnology can realize.
Pallab Chatterjee is on the IEEE Nanotechnology Council.
Talkback
-
Assuming that power is proportional to voltage squared also assumes that all digital circuitry are purely resistive, which I believe is very far from the truth
In digital circuitry power is mainly consumed during switching going from high to low and from low to high to overcome the Gate capacitances.
H Mohan - 2011-7-12 17:43:33 PST -
... and even a "1000 mV" reduction from 1.25 V to 0.25 V would mean a reduction of 96%, way far from the stated factor of 1 million
Roberto Mariani - 2011-2-12 01:39:32 PST -
Dear author,
stating that
"The more-than-1000-mV reduction in required power would result in a power-reduction factor of 1 million, assuming that power is proportional to voltage squared."
is an awful mistake: it would be right only if it was a 1000-fold reduction, but moving (for example) from 1 kV down to 999 V would reduce power consumption by no more than 0.2%... and that was a 1000 mV reduction.
Roberto Mariani - 2011-2-12 01:33:30 PST -
The whole world is analog, digital is just one special case. The problem that comes with lower voltages in the digital realm is reduced noise immunity, and increased ingress of cross talk. The challenge remains is that it still takes power to change states in the digital mode, and as the source voltage is lowered the effect of random resistances becomes larger. In addition the effects of static discharges is a larger problem because of the lower voltage thresholds. It would be much more efficient to make the code smaller and more efficient, instead of running huge files of bloatware. The thing that puzzles me most is why nobody else has made this point.
William Ketel - 2011-1-12 15:24:18 PST -
Call me old-fashioned, but when exactly did "electronics" come to mean digital electronics. This article caught my attention because, at first glance, it seemed to indicate that some breakthrough discovery had been made to reduce random noise in analog electronics, making wide dynamic ranges available with low supply voltages. But, instead, it's the same game of making DIGITAL signals smaller, thereby making higher bit rates easy. It's especially easy if you neglect the analog problems that come with trying to move such tiny signals around in the real world. I'm tired of this constant usage that, in effect, marginalizes analog electronics by assuming that everyone knows that electronics means digital!
Bill Whitlock - 2011-1-12 14:30:52 PST
