25-08-2017, 09:32 PM
1454733315-MAIN.pdf (Size: 1.22 MB / Downloads: 6)
Introduction
Since the 1990’s, mobile computing has transformed its penetration from niche markets and early prototypes to ubiquity.
Personal Digital Assistants (PDAs) evolved from GRiD’s PalmPad and Apple’s Newton in 1993 to the Palm, Handspring,
and Microsoft-based models that support the multi-billion dollar industry today. While BellSouth/IBM’s Simon may have
been the only mobile phone to offer e-mail connectivity in 1994, almost every modern mobile phone provides data services
today. Portable digital music players have replaced cassette and CD-based systems, and these “MP3 players” are evolving
into portable repositories for music videos, movies, photos, and personal information such as e-mail. Laptops, which were
massive and inconvenient briefcase devices in the late 1980’s, now outsell desktops. Yet all these devices still have a common,
difficult problem to overcome: power.
This chapter will review trends in mobile computing over the past decade and describe how batteries affect design tradeoffs
for mobile device manufacturers. This analysis leads to an interesting question: is there an alternative to batteries? Although
the answer has many components that range from power management through energy storage [142], the bulk of this chapter
will overview the history and state-of-the-art in harvesting power from the user to support body-worn mobile electronics.
2 Technology Trends in Mobile Computing
Mobile phone companies often sell more batteries than phones to consumers. The phones sold to users include a rechargeable
battery so that the device is immediately useful, but a certain number of consumers are expected to own more than one battery
during the life of their phone. The same can probably be said for laptops and camcorders. Yet, there is little incentive for
consumers to buy new batteries except for when they fail or when the consumer feels the need for a larger battery. Unlike
other areas of mobile computing that benefit from exponential improvements in performance, battery energy density (as
measured by joules per kilogram or joules per cubic centimeter) changes slowly so that there is little pressure for consumers
to upgrade.
2.1 Battery Energy Density as a Lagging Trend
As Figure 1 shows, battery energy is one of the most laggard trends in mobile computing. Figure 1 shows the progression
of technology in the last 13 years for laptop computers, a technology now mostly mature. In general, the laptop technology
represented in the graphs would, if repackaged in a body worn device, weigh seven pounds or less and could be used while
standing on a street corner in a major United States city. While some mobile computers existed prior to 1990, most weighed
over 10 pounds or did not include hard drives. In addition, commercial wireless data networks in the United States were not
openly available before 1990 or required amateur radio licenses to operate.
The graph depicts increases in performance as multiples of the state of the technology from 1990 (e.g. the amount of
RAM available in a laptop increased by 256X from 1990 to 2003). Due to the exponential nature of the improvements, the
y-axis in Figure 1 is on a logarithmic scale.
The laptop specifications shown were determined by examining advertisements in the December issues of popular computing
magazines (e.g. Byte, PC Computing, etc.) for each year. The numbers used reflect a composite from the highest–end
machines available at that time. An example of a high–end machine from 1990 (the base value of 1 in the graph) would
be a 16MHz 80386 with 8 megabytes (MB) of RAM and 40MB of hard drive space using a nickel-cadmium battery and
communicating at 4800 baud over the ARDIS network. Processor performance is compared in terms of Intel’s
R
index as derived from www.cpuscorecard.com; RAM and disk storage are compared by size; wireless networks are compared
by maximum bits per second of data transfer; and battery energy density is determined by the type of technology used
(nickel cadmium, nickel metal hydride, or lithium ion) and the progression these technologies made in increasing the joules
stored per kilogram (J/kg). The wireless connectivity graph represents the first author’s pursuit of the commercial city-wide
networks available in the United States (cellular standards; not emerging 802.11 “hotspots”).
While disk storage density has increased over 4000X since 1990, the lowly battery has only increased a factor of three in
energy density. New materials, along with nano and micro fabrication technologies, have recently enabled “micro fuel cells”
[145] aimed at recharging handhelds like cell phones with power plants the size of a small candy bar [177], and eventually
powering wireless sensor nodes with fuel cells on a chip [163, 102, 120]. Although the technology is rapidly advancing
[60], laptop-sized plants (e.g., 30-50 W-hr), have tended to be in an awkward place for fuel cells - too big to directly power
with micro cells, but small enough that the overhead in mass needed to handle the standard fuel cell chemistry is significant
(not to mention safety factors associated with the fuel, high expense of the platinum membrane, etc.). Nonetheless, several
companies have announced prototypes designed for laptops [32], which should make it to market over the next couple of
years and gradually improve. More exotic emerging power technologies tend to have characteristics that force them into
niche applications - e.g., radioactive batteries [81] can last for decades, but provide very little current, while devices that
actually burn fuel [3], such as microturbines [63] and microengines [72], have potential issues with safety and byproducts
like exhaust, heat, noise, or thrust.
The lesson to mobile device designers is clear: specify the battery or power source first, then design the mobile device’s
electronics around it. Battery technology is the least likely element to change in the 12 month development cycle and may be the most limiting factor in the design with respect to size, weight, and cost.
Trading Storage and Processing for Wireless Connectivity
Wireless connectivity is also a conundrum for mobile designers. While the designer can control the CPU, RAM, disk, and
battery in his device, wireless connectivity is often provided by another party. In the extreme case, a wireless provider may
go out of business and significantly impact the quality of service that can be expected. Such a situation is reflected in Figure
1 where the removal of the Metricom network reduced the maximum available throughput from 128,000 bits per second to
19,200 in several major U.S. markets.
Even on a minute-by-minute basis, a wireless connection may or may not be available at any given moment. The device
designer must either cache information for the user or refuse service when the network is not available [107]. Thus, many
devices, such as wireless PDAs, have non-volatile RAM or disks so that the user can work “off-line.” Using mass storage
strategically can save significantly on battery consumption, as both receiving and transmitting data from cellular and 802.11
networks require substantial power [174]. More specifically, the power needed for transmitting is proportional to the distance
to the fourth power [47]. Given the exponential trends in disk density above, it may soon be a viable power-saving strategy to
cache a good fraction of static Internet content for a mobile web surfer instead of connecting over power-hungry and potentially
expensive wireless networks! One can imagine a system that examines the user’s e-mail, web history, and downloads
and, based on this data, continuously updates the user’s mobile cache while the device has wired (or low power) connectivity.
An interesting illustration of this point is to compare the power required to retrieve information from modern flash memory
with the power required to transmit a request of that information from a remote source. Suppose that we have the option of
storing information on a cellular phone in the form of a flash disk or sending a wireless request for the information to
the network. Reading a bit from modern flash memory requires approximately 10 pJ or 1 x
J/bit [13]. However,
transmitting a single bit at 0.6W from a mobile phone at an aggressive 1 Mbps rate would require 6 x
J. Thus, for every
bit transmitted in the wireless request for information, the same amount of energy could be used to read 60,000 bits from
a flash drive. This calculation is conservative as it ignores the inefficiencies in the radio, the overhead generally associated
with transmission error checking, and the amount of power that would be required to receive, process, and store the response
from the network. Thus, a mobile device designer should always consider how much information can be stored or cached on
the device itself as opposed to depending on wireless services.
The sensor network community is very concerned with a similar tradeoff - i.e., how much data to process locally at a
sensor node vs. how much data to wirelessly transmit. As it takes between 100 and 10000 times more power to transmit
one bit across even a short range than to execute a single processor instruction (depending on the implementation) [153], it’s
often advantageous to analyze and/or compress the node’s data before broadcasting [151]. In order to reduce the node’s power
requirements down to the point where ambient energy harvesting is practical, researchers are pursuing joint optimization of
the processor hardware, radio circuitry, and network protocols [132]. Although we do not explicitly consider the amount of
power required to receive the information, this is often not negligible, especially in short-range networks, where it can take
more power to receive and decode a bit than to transmit one [153].
3 Power from Incident Radiation
3.1 Catching the Ambience
With so many RF transmitters of various sorts distributed throughout today’s urban environments, one might consider background
RF as a potential power reservoir for mobile devices. Electronic systems that harvest energy from ambient radiation
sources, however, tend to be extremely power-limited and generally require a large collection area or need to be located very
close to the radiating source. A classic example can be found in old-fashioned crystal radio kits [106] that draw their power
directly from AM radio stations, which play audibly through high-impedance headphones without needing a local source
of energy. The size of the required antenna, however, can be prohibitive for wearable applications unless the bearer is very
close to a transmitter, and access to a good ground is usually required. Even so, the received power is very limited in a
standard crystal radio, where set builders typically see received powers on the order of 10’s of W, approaching a millliwatt
for proximate stations. An interesting adaptation of a crystal radio set is described in US Patent# 2,813,242 [56], where a
resonant tank circuit tuned to a strong, nearby station provides enough power to run a single-transistor radio with a small
loudspeaker that can be tuned to other stations. An analysis of RF power scavenging at higher frequencies by Yeatman [194]
crudely approximates the power density produced by a receiving antenna as , where is the radiation resistance of
3
free space (377 Ohms). An electric field (E) of 10 V/m thus yields 26
at the antenna. Field strengths of even a
few volts per meter are rare in habitated environments, however, except when very close to a powerful transmitter [123]. In
a related note, power can also be extracted from the earth, across a large ground loop, tapping the AC potential difference
between grounds at different locations. A harvest of 1.4 mW has been reported using a pair of grounds separated by 50 feet
[169].
An example of ambient RF power harvesting in the mobile sphere at higher frequency is found in aftermarket modules
that flash LED’s when your cell phone rings. Several of these designs are batteryless, but need to be extremely close (or
right against) the cell phone’s antenna to work, as they draw their energy through near-field capacitive or inductive coupling.
Perhaps another mobility example, much further afield, comes from the strange, scattered and usually anecdotal reports of
people receiving strong, nearby radio broadcasts from spontaneous detectors formed by loose fillings in their teeth [86, 40],
and the passive, implantable receiver design that this has inspired [152, 7].
Higher up in the electromagnetic spectrum, it’s not uncommon to see very low power consumer items, such as simple
calculators, run off photovoltaics with ambient illumination. The energy conversion efficiency of easily available and relatively
inexpensive crystalline silicon solar cell modules (without going to IC-grade silicon or stacked junction structures)
is generally below 20%, and closer to 10% for flexible amorphous silicon panels [26]. Accordingly, mobile applications,
which generally imply limited surface area, tend to be constrained, especially in scenarios without strong and consistent sunlight
(standard solar cells produce roughly 100 mW/
in bright sun and 100
in a typically illuminated office).
Nonetheless, products like solar battery chargers for cell phones that purport to produce up to 2 Watts of power [117] and
PDA’s that run off a panel of solar cells lining their case [164] currently exist, and researchers continually strive to refine
solar cell materials [92, 26] and technologies [27] to increase efficiency [78], as well as explore unusual form factors, such
as flexible photovoltaic fibers [97], that promise to be more amenable to wearable implementations.
3.2 Get on the Beam
Rather than relying on the limited energy that can be scavenged from ambient radiation, other approaches actively beam
power from a transmitter to remote devices. The wireless transfer of power originates with Heinrich Hertz who, ushering
the dawn of radio in the late 1800’s, induced sympathetic sparks across a gap interrupting a resonantly tuned ring placed
several yards away from a transmitting antenna that was directed with a parabolic reflector [178]. The dream of wirelessly
broadcasting power to an urban area dates back to the turn of the 20’th century and Nicola Tesla [180], who experimented
with grandiose concepts of global resonance and gigantic step-up coils that radiated strong, 150 kHz electromagnetic fields
able to illuminate gas-filled light bulbs attached to a local antenna and ground at large distances [50]. Wireless power research
continued with the work of H.V. Noble [38], who in the early 30’s at the Westinghouse Laboratory, demonstrated the transfer
of several hundred watts between 100 kHz antennas separated by 25 feet, leading to public demonstrations of this technology
at the Chicago World’s Fair in 1933. The development of radar [39], hence powerful microwave transmitters, enabled further
work in directed energy transmission, a highlight of which was the wireless powering of a small helicopter by William
C. Brown in 1964 [38]. Microwave-to-DC converters, termed “rectannas” can be extremely efficient; efficiencies of over
90% have been produced in laboratory experiments and 30 kilowatts have been transferred across more than a mile at 84%
efficiency [38]. This has led to proposals for beaming massive amounts of power to earth from solar collectors in space [75]
and remotely beaming propulsion to interstellar probes from an earth-orbiting 10 gigawatt transmitter [69].
Closer to home, FCC and safety regulations (e.g., IEEE/ANSI C95.1) along with public perception [70] have restricted
the beaming of any significant amount of power in the proximity of people. Nonetheless, researchers have experimented with
microwave transmission of power in domestic environments, transferring several mW across meters to sensors for ubiquitous
and wearable computing applications [19]. At much lower power levels, short-range wireless power transmission is now commonplace
in passive Radio Frequency Identification (RFID) systems [65], which derive their energy inductively, capacitively,
or radiatively from the tag reader. As most RFID chips talk back to the reader by dynamically changing their impedance
or reflection coefficient, they require minimal power, generally between 1 and 100 W, depending on their implementation
and operating frequency (lower-frequency, magnetically-coupled tags consume less power). Today, people commonly carry
RFID transponders, most often for keyless entry systems. Simple resonant RF tags that change their tuned frequency or Q
as a function of a local or environmental parameter have been used as passive sensors in several applications [155, 99]. Examples
include LC (inductive-capacitive) tags for wireless displacement and pressure sensors in human-computer interfaces
[143], measuring tire inflation with pressure-varying backscatter from crystal bulk resonators [24], tracking tire strain with
surface-acoustic wave (SAW) devices [149], and proposed studies for using such SAW sensors as implantable blood pressure
monitors [129].
4
The reverse, where people carry the reader to interrogate tags in the environment, is not as feasible, since the readers
tend to be power hungry and large (e.g., several orders of magnitude more massive than the tags). Researchers, however,
in wearable and ubiquitous computing have adapted reader circuits to identify tagged objects when handled with readerintegrated
gloves [165, 146] or put into coil-lined pockets [94], and small, single-chip readers are now becoming available by
companies like EM Microelectronic and Innovision Research & Technology for very short-range, lower-power applications