Battery Independence: Reducing the Dependence on Batteries in Wearable Computing Through Energy Harvesting Techniques

Abstract

This paper discusses one of the most troublesome and frustrating issues for mobile devices users, battery life. Once a battery is utilized in a system, battery charging must be incorporated seamlessly so that the workflow of the user is uninterrupted. In today's fast-paced society, people don't have the time to stop what they are doing to charge their mobile devices and wait for them to complete. A secondary known issue concerning batteries is that their effective capacity diminishes over time and use. In turn the process of battery charging would become more frequent and disruptive to the user. The need for an effective means to continually charge the battery through smart energy harvesting techniques will be investigated in the following paper. Utilizing the findings from other articles, we were able to generate easy-to-use formulas to help estimate the power across multiple real-world use cases (running, walking, sleeping, sun, shade, etc.) Testing the formulas across multiple use cases, only two of the cases generated more power in our application than what the system utilized. This excess power would be utilized to keep the battery's charge topped off. In the use cases where we didn't generate enough power, the battery would need to supplement the system.

Full text

Battery Independence: Reducing the Dependence on
Batteries in Wearable Computing Through Energy
Harvesting Techniques
David Pasko, Michael Mrazik and Khaled Elleithy
Computer Science and Engineering Department
University of Bridgeport
Bridgeport, Connecticut
DPasko@my.bridgeport.edu, MMrazik@my.bridgeport.edu, elleithy@bridgeport.edu
Abstract—This paper discusses one of the most troublesome
and frustrating issues for mobile devices users, battery life. Once
a battery is utilized in a system, battery charging must be
incorporated seamlessly so that the workflow of the user is
uninterrupted. In today’s fast-paced society, people don’t have
the time to stop what they are doing to charge their mobile devices
and wait for them to complete. A secondary known issue
concerning batteries is that their effective capacity diminishes over
time and use. In turn the process of battery charging would
become more frequent and disruptive to the user. The need for an
effective means to continually charge the battery through smart
energy harvesting techniques will be investigated in the following
paper. Utilizing the findings from other articles, we were able to
generate easy-to-use formulas to help estimate the power across
multiple real-world use cases (running, walking, sleeping, sun,
shade, etc.) Testing the formulas across multiple use cases, only
two of the cases generated more power in our application than
what the system utilized. This excess power would be utilized to
keep the battery’s charge topped off. In the use cases where we
didn’t generate enough power, the battery would need to
supplement the system.
Keywords—energy harvesting; low power; wearable computing
I. INTRODUCTION
In 2015, there was an estimated 15.4 billion dollars of sales
revenue worldwide on wearable devices [1]. Many of these
devices have one common theme, their batteries require re-
charging. This often requires that the wearable devices be
removed from the user (i.e. removal of a smart watch in order to
plug it in and charge it). This process of battery charging could
become cumbersome and impede the user experience. As a
battery is used and ages, its effective capacity diminishes so this
cumbersome charging process would increase in frequency. In
this paper, the need for continual unobtrusive charging is
presented.
The primary problem with utilizing batteries in wearable
devices is the re-charge time. According to Apple, it takes
approximately two hours to charge the Apple Watch up to 100%
capacity [2]. If a user were to be relying on this device (text
messages, heart rate monitor, sleep tracking, etc.) the two hour
charge may be unacceptable downtime. Continuing to use the
Apple Watch as an example, the battery can last up to eighteen
hours (based on a specific use case) [2]. This means that a
battery’s cycle is roughly twenty hours (two hours of charge plus
eighteen hours of discharge) with the charge time being 10% of
the total cycle. Batteries also degrade over time and use so their
effective capacity diminishes over time. As an example, a user
with an Apple Watch may not get the same eighteen hour run
time after owning the watch for two years so the 10% charge
portion of the battery cycle would increase (as the battery
capacity decreases). Figure 1 from Battery University illustrates
how battery capacity degrades over time which would cause the
user to require recharging more frequently [3].
Figure 1: Battery Degradation over time [3]
A benefit of being a wearable device is that the physical and
environmental conditions are usually changing. This could be
in the form of a user walking or running, stepping outside into
the sun, or temperature differentials between the user’s skin and
the external environment. In each of these scenarios there is
energy available to be harvested and utilized for the powering of
the wearable device. While having a single source of harvestable
energy is not often feasible for all use cases, having multiple
sources allows a system to have redundancies to account for
circumstances where an energy source is no longer available (i.e.
the user is in low-light conditions where solar is not
feasible). Optimizing the system to run off the highest energy-
rich source is an effective means to deliver the most power to
the system and the battery charging. The following sections will
978-1-5386-1104-3/17/$31.00 ©2017 IEEE 37

discuss in detail the methods on how to harvest energy from the
environment so that a wearable device no longer needs to be
charged by the user.
II. RELATED WORK
In the article [4], the authors show an example of energy
harvesting that can be examined and improved upon for future
development. The harvesting form revolves around the normal
movement of the human hand. This motion is intriguing to us
as it fits into the realm of normal motion that can be experienced
by a wearable device located on the wrist.
Analyzing the results of [4], it is interesting to see that the
repetitive and normal movement of a human hand results in
voltages and power levels that are acceptable as inputs into
rectification and boost circuitry typically used in the embedded
design industry. Through the experiment it can be seen that
with the hand movement model the peak to peak voltages of 30-
40V were observed. After rectification and conversion, these
voltages were a steady 2.5-3.5V DC. The success of the circuit
was also the ability of the authors to power either an LED
circuit or a small LCD which are both potential devices used in
wearable devices. This will be discussed further in Section III.
The authors of [5] present an experimental prototype to
study the viability of harvesting solar and thermoelectric energy
in the context of wearable devices. The experiment was setup
as follows for the energy harvesters: 1) There were three types
of solar cells utilized (one rigid and two flexible) and one
thermoelectric generator (TEG). 2) The system contained a
power management unit (PMU), 3) Experimental system was
tested under several environments in two places on the body
(upper arm and wrist). 4) Thermoelectric Energy Harvesting
was performed indoor and outdoor while sitting still, walking
at variable rates, and running at variable rates. 5) Solar Energy
Harvesting was performed in an indoor office setting (complete
fluorescent lighting), indoor with a window (natural light as
well as supplementary fluorescent lighting), and outdoor on a
sunny day at 13°C (55°F).
The results of the experiment showed that the highest
power generated from the Thermoelectric Generator was
811uW on the upper arm and 1.65mW on the wrist. These
measurements were taken while running outdoors. According
to the authors, this can mostly be attributed to the fact that there
is more air flow (wind) and more body movement outdoors [5].
Table 1 presents full results from [5].
Table 1: Full TEG Results from Article [5]
The results of the experiment also showed that the highest
power generated from the Solar Energy Harvesting was
113mW and these measurements were taken on the rigid cell
outdoors. Table 2 has full results from [5].
Table 2: Full Solar Results from Article [5]
Although the data shows the rigid cell performing the best
outdoors, the flexible cells produced more power than rigid in
all of the indoor environments.
The test setup of this experiment was fairly straight
forward and provided valuable test results including the TEG
placement and the rigid vs flexible solar cells.
Similar to the work presented in [5], the authors of [7]
utilize solar and thermoelectric generators equipped on a jacket
to charge a battery pack. The experiment was setup as follows
for the energy harvesters: 1) There were sixteen solar cells
partitioned into four zones for the chest, back, right shoulder,
and left shoulder. In addition to the solar cells, there were a
total of twelve thermoelectric generators (TEGs) configured as
six in series (to increase voltage) and two sets of the six in
parallel (6s2p). 2) The system also incorporates power
management controllers (PMCs) for the four sets of solar cells
and one set of TEGs. The PMC component utilized is a
benchmark part that has built in low power step up voltage
conversion, battery management, and maximum power point
tracking (MPPT). 3) All five of the PMCs are connected in
parallel to output to two AAA sized NiMH batteries connected
Power
(
mW
)
ѐT(°C) Power
(
mW
)
ѐT(°C )
Si tt ing 0.164 5.
6
0.23
8
5.9
Wal kin g 2.0 mp h 0.224 7.2 0.267 7.4
Wal kin g 3.0 mp h 0.238 7.3 0.29
6
7.6
Wal kin g 4.0 mp h 0.319 7.
4
0.335 7. 7
Running 5.5 mph 0.382 7.
8
0.465 8. 0
Running 6.5 mph 0.400 8.0 0.55
8
8.1
Running 7.5 mph 0.431 8.2 0.612 8.4
Running 8.5 mph 0.468 8.3 0.673 8.7
Si tt ing 0.743 9. 1 0.92
8
9.7
Wal kin g 0.602 10.0 1.030 10.
4
Running 0.811 12.9 1.650 13.1
WristUpper Arm
Indoo
r
Outdoo
r
Thermoelectric
Energy Harvesting
on the Body
38

in series. 4) The prototype jacket was tested outdoors under
various weather conditions including sunny, cloudy, and partly
cloudy days in July, August, and January. 5) The power levels
were measured instantaneously when the prototype equipped
person was both stationary and in motion.
The results of the experiment concluded that the harvested
power with all of the solar cells (PV – photovoltaic referenced
below) in full sun was 710mW [7]. Figure 2 below shows that
the average power decreases as the number of solar cells in the
shade increases.
Figure 2: Article [7] Average Power vs Cells Shaded
In addition to the above data, measurements were taken on
the solar cell groups individually with the wearer remaining
stationary and facing away from the sun. The back and left
shoulder were under full sun and generated around 190mW and
170mW respectively [7]. For the same test, the right shoulder
and the chest were in the most shade so they both generated less
than 50mW. The author(s) stated that “the power of the
shoulder sections is sensitive to their positions relative to the
sun due to placement of PV cells around a round shape” [7].
This statement is their justification for why one shoulder had a
much higher power output than the other. Additional testing
was performed while the wearer was walking during full sun
and during cloudy weather. Figures 3 and 4 below illustrate the
important test results from [7]:
Figure 3: Power Levels showing Shading on Body
Figure 4: Power Levels showing Walking in Full Sun
It is clear that the power output during motion is not
consistent (spikes up and down) and that walking around in full
sun allows for much higher power output than walking during
cloudy weather. The author(s) attribute the spikes and drops
when walking due to more active partial shading and possible
movement of the solar cells [7].
Additionally to the solar harvesting results, the power of
the TEGs under light movement was negligible compared
against the solar harvesting. In full sun, the maximum TEG
power was 1.25uW and dropped to 0.5uW in the shade.
The author(s) made a note in their conclusion that “the key
improvement for our thermal energy harvesting system is to
increase the temperature gradient of TEGs” [7]. We believe
one of the major fall backs of this design/experiment is the
utilization of the jacket for TEGs. To get the maximum
temperature gradient between the skin and the ambient air you
require good thermal conductivity and we believe the jacket
materials (layers, linings, etc.) increased the thermal resistance
which compromised the data of this experiment.
III. METHODS AND MATHMATICAL MODEL
Our proposed solution seeks to remove the need for battery
charging using the concepts of redundancy and priority-based
input selection. These energy harvesting techniques would
supplement the system and charge the battery and therefore
require less charging and prolong the life of the battery.
Figure 5 presents a block diagram of our proposed system
architecture. The functional blocks that make up the overall
architecture are described in greater detail in each sub-section.
Figure 5: System Architecture Block Diagram
In this segment, we will aim to utilize the data and results from
the “Related Works” section to generate our mathematical
models and calculate our hypothesis.
A. Piezoelectric Input
As discussed in [4] we aim to capture energy in the motion
surrounding the human hand and wrist. This energy was higher
than that of the motion of the elbow during walking/running so
we believe this is the appropriate starting point for our
experiment to run the system and charge the battery. Figure 6
shows the Piezoelectric Input components from the
architecture drawing:
39

Figure 6: Piezoelectric Input Components
Article [4] utilizes a PZT sheet consisting of three layers: steel
sheet, PZT layer, and the conductive adhesive tape. Table 3
from article [4] has been converted into inches in order to
remain consistent with all of the other numbers presented in this
paper.
Layer Length Width Thickness
Component Inch Inch Inch
Steel 2.638 0.709 0.008
PZT 2.047 0.669 0.008
Conductive Adhesive 1.929 0.591 0.004
Table 3: Article [4] dimensions converted to inches
Considering that the largest contributing factor to the overall
size is the steel base, we will use that in all of our size
calculations proceeding this point. Based on the article, a piezo
beam constructed with the above sizes created a voltage
waveform when a “hand-like” type of motion was applied
similar to the movement of running. This voltage waveform is
displayed in Figure 7:
Figure 7: Voltage Waveform
Further breaking down the voltage waveform provided we can
obtain the following information:
- There is a 40V peak to peak range
- 2.638in / 40V = 0.0659in/V for length calculation.
- 0.709in / 40V = 0.0177in/V for width calculation.
These calculations as well as the rest of the components from
the table above have been calculated the same way. If we were
to keep the size and voltage balanced for the application and
choose 5V (10V peak to peak) as the maximum input voltage
(because 5V is a common voltage seen in embedded systems),
then we would get Table 4:
Layer Length Width Thickness
Component inch Inch inch
Steel 0.6594 0.1772 0.0020
PZT 0.5118 0.1673 0.0020
Conductive
Adhesive 0.4823 0.1476 0.0010
Table 4: Calculations of Table 3 in inches/Volt
Aside from the physical component sizes, another portion
of effective piezoelectric operation is that it must flex to
generate the voltages required. According to the experiment in
the article [4], the flex must roughly 30 degrees. With the total
length of 0.6594” (from table above) we calculated a maximum
vertical travel of 0.381” (Tan (30) = x / 0.6594”). Because this
distance needs be traveled upwards and downwards then the
total distance is 0.762”. This dimension will dictate how much
physical space is needed internally to the wearable device to
generate the needed voltages.
The power measurements recorded in the paper [4] were
on the order of 1-10uW. Although this may seem like an
insignificant source of pure energy, it does allow some extra
energy to supplement a battery in high motion situations. Since
our device has been miniaturized to fit into a wearable that
could be worn on the wrist, we have lost some of that potential
power. If we were to make the assumption that a 40V input
gives you 10uW as specified in the paper [4], then we could
expect ¼ of that since we have decreased the overall size by
four (40V peak to peak down to 10V peak to peak). This
equates to a maximum estimated power generation of 2.5uW
when there is a 10V peak to peak input. Using Ohms Law and
calculating the current for 2.5uW @ 10V, this would be
0.25uA. For the purpose of our calculations and because article
[4] wasn’t actually tested on a human (simulation only), we will
make the assumption that walking creates ½ the power that
running does. Although neither of these actions would create a
significant source of power to be provided to the system, this
small current may help trickle charge the battery.
Based on the data and discussion above, we can come up
with the following constants to utilize in our scenarios later on:
ܲ݅݁ݖ݋ܧ݈݁ܿݐݎ݅ܿܰ݋ݐܯ݋ݒ݅݊݃ ൌ Ͳݑܹ
ܲ݅݁ݖ݋ܧ݈݁ܿݐݎܹ݈݅ܿܽ݇݅݊݃ ൌ ͳǤʹͷݑܹ
ܲ݅݁ݖ݋ܧ݈݁ܿݐݎܴ݅ܿݑ݊݊݅݊݃ ൌ ʹǤͷݑܹ
B. Solar Input
As discussed in [7] we aim to capture energy from the sun
through a photo cell attached to a human wrist. Since a human
normally performs activities aside from sleep with some
amount of light present, we aim to capture this energy during
periods of low energy output from either the piezoelectric or
thermoelectric inputs. Figure 8 shows the Solar Input
components from the architecture drawing:
40

Figure 8: Solar Input Components
There was a useful table from article [5] shown previously in
the “Related Works” section. This table summarized the power
(units are mW) found across multiple indoor and outdoor
conditions. Our particular model will focus on rigid
photovoltaic cells so we will therefore only utilize the results
from the “rigid cell” utilized in article [5]. According to the
article, the rigid cell used was 2.65” x 1.05” in size (converted
from the millimeters actually specified in the article). From the
table previously mentioned, we calculated the mW per square
inch by dividing each value in the table by the square inches of
the cell (2.65”x1.05”=2.783in²). Performing this calculation to
each of the items in the table gives us Table 5:
Indoor Power (mW)
per in²
Fluorescent only - perpendicular to
light 0.140
Fluorescent only - facing light 0.200
Room w/ Window, facing Window 3.026
Room w/ Window, facing away
from window 0.431
Outdoor
On a sunny day 40.611
Table 5: Article [5] Converted to mW/ in²
From Table 5, we can come up with the following formulas to
utilize in our use cases later on:
ܵ݋݈ܽݎܲ݋ݓ݁ݎܫ݊݀݋݋ݎܨ݈ݑ݋ݎ݁ݏܿ݁݊ݐܱ݈݊ݕܰ݋ݐܷ݊݀݁ݎܮ݄݅݃ݐݏ
ൌ൤
ሺܵ݋݈ܽݎܥ݈݈݁
௜௡
మ
ሻכͲǤͳͶͲܹ݉
݅݊
ଶ
൨
ܵ݋݈ܽݎܲ݋ݓ݁ݎܫ݊݀݋݋ݎܨ݈ݑ݋ݎ݁ݏܿ݁݊ݐܱ݈݊ݕܨܽܿ݅݊݃ܮ݄݅݃ݐݏ
ൌ൤
ሺܵ݋݈ܽݎܥ݈݈݁
௜௡
మ
ሻכͲǤʹͲͲܹ݉
݅݊
ଶ
൨
ܵ݋݈ܽݎܲ݋ݓ݁ݎܫ݊݀݋݋ݎܰ݋ݐܨܹܽܿ݅݊݃݅݊݀݋ݓ
ൌ൤
ሺܵ݋݈ܽݎܥ݈݈݁
௜௡
మ
ሻכͲǤͶ͵ͳܹ݉
݅݊
ଶ
൨
ܵ݋݈ܽݎܲ݋ݓ݁ݎܫ݊݀݋݋ݎܨܹܽܿ݅݊݃݅݊݀݋ݓ
ൌ൤
ሺܵ݋݈ܽݎܥ݈݈݁
௜௡
మ
ሻכ͵ǤͲʹ͸ܹ݉
݅݊
ଶ
൨
**Note, we will not utilize the outdoor calculation as the
outdoor characteristics from article [7] are better defined.
Comparatively, article [7] described in the “Related
Work” section offered additional information that is useful in
performing the similar calculations shown above. Article [7]
illustrated the power generated by each zone (each zone
contained four photovoltaic cells). As done with article [5],
calculations were performed to get these numbers down to the
mW/in² unit. In order to do so, we also had to divide each of
the measurements in the table by four since they were for four
photovoltaic cells connected in parallel. Each cell measured
1.22” x 1.22” in size and so was therefore 1.488 in² [7]. The
calculated results are in Table 6:
Outdoor Shading
Power (mW)
(per in²)
Min Max Average
Fully Shaded 0.840 5.039 2.939
Shaded 4.199 16.797 10.498
Full Sun 25.195 33.593 29.394
Table 6: Article [7] Converted to mW/ in²
From Table 6, we can come up with the following formulas to
utilize in our scenarios later on:
ܵ݋݈ܽݎܲ݋ݓ݁ݎܱݑݐ݀݋݋ݎܨݑ݈݈݄ܵܽ݀݁
ൌ൤
ሺܵ݋݈ܽݎܥ݈݈݁
௜௡
మ
ሻכʹǤͻ͵ͻܹ݉
݅݊
ଶ
൨
ܵ݋݈ܽݎܲ݋ݓ݁ݎܱݑݐ݀݋݋ݎ݄ܵܽ݀݁݀
ൌ൤
ሺܵ݋݈ܽݎܥ݈݈݁
௜௡
మ
ሻכͳͲǤͶͻͺܹ݉
݅݊
ଶ
൨
ܵ݋݈ܽݎܲ݋ݓ݁ݎܱݑݐ݀݋݋ݎܨݑ݈݈ܵݑ݊
ൌ൤
ሺܵ݋݈ܽݎܥ݈݈݁
௜௡
మ
ሻכʹͻǤ͵ͻͶܹ݉
݅݊
ଶ
൨
As a reference, for these calculations we utilized the
average power found in the experiment. Additionally, the solar
cells utilized in the system from article [7] were rated for peak
voltages of 1.2V and open circuit voltage of 2.2V. We will
utilize these voltages in Section E: DC/DC Converter.
C. Thermoelectric Input
As discussed in [7] the thermal energy emanating from the
human skin can be a valuable source of electrical energy if a
temperature differential is present. Since in most cases the
human skin will have either a positive or negative temperature
differential from ambient air, we should be able to gather
energy whenever the user has the wearable device on. In
periods when the temperature of the air is the same as the
human skin or if the user does not have the wearable device on,
this input will be deselected by the priority-based voltage
combiner.
Figure 9 shows the Thermoelectric Input components from
the architecture drawing:
Figure 9: Thermoelectric Input Components
As illustrated in the “Related Works” section, we believe that
there were flaws to the way the TEG model was constructed in
article [7] so we will only focus on the results from article [5].
Article [5] states in the introduction that commercial TEGs can
provide 20-40uW/cm² (converts to 0.129-0.258mW/in²) on the
body, which results from a 5-10°C temperature difference
across the TEG. Unfortunately, this same article doesn’t
41

actually specify the dimensions of the TEGs used in their
experiment. Instead, they only tell the reader that it is a
Micropelt product from Germany [5]. Therefore, we will be
unable to utilize the mW measurements found in that article but
can still make use of the ǻT(°C) values measured across
multiple actions (running, walking, etc.)
As a baseline, we are going to use the data from a Micropelt
MPG-D655 Thin Film Thermogenerator [12]. Assuming that
both of the TEG plates are the same dimensions as the “top”
side of the device, the plates would be 0.011in² (calculated from
2845x2424um). Figure 10 are from the Micropelt datasheet
illustrate the open circuit voltage and the matched output
power:
Figure 10: Micropelt Datasheet Graphs
Article [5] published measurements for sitting, walking, and
running both indoors and outdoors with the TEG placed on both
the upper arm and the wrist. There are differences in the table
provided (in “Related Work” section) between the indoor and
outdoor walking and running speeds. The outdoor speeds were
defined as “normal” while the indoor speeds had more
specification. For the sake of simplification, we have averaged
the indoor walking and running speeds.
Below is a table containing the average ǻT(°C) findings
from article [5], the open circuit voltage determined from the
ǻT(°C) on the graph above, and the matched power (mW)
calculated from the datasheet [12] based on the ǻT(°C) values
in the graph above. We are most interested in the results from
the wrist calculations for our design so we will not present the
data for the upper arm measurements.
Action
Wrist
ǻT(°C) Power
(mW)
Volts
(V)
Indoor
Sitting 5.90 0.25 0.47
Average
Walking 7.57 0.45 0.61
Average
Running 8.30 0.55 0.66
Outdoor
Sitting 9.70 0.70 0.78
Walking 10.40 0.76 0.83
Running 13.10 1.15 1.05
Table 7: Average ǻT(°C) findings from article [5]
From this table, we have calculated the power per square inch
by dividing each of the mW measurements by 0.011in² (as
calculated previously). Table 8 shows mW/in² calculated for
each of the actions (walking, running, and sitting) based on the
ǻT (°C) rise from article [5] for the wrist.
Action
Wrist
ǻT(°C) Power (mW)
per in²
Indoor
Sitting 5.90 23.50
Average
Walking 7.57 42.29
Average
Running 8.30 51.69
Outdoor
Sitting 9.70 65.79
Walking 10.40 71.43
Running 13.10 108.08
Table 8: Calculated mW/in²
Based on the data calculated in Table 8, we can come up with
the following formulas to utilize in our scenarios later on:
݄ܶ݁ݎ݉݋ܧ݈݁ܿݐݎ݅ܿܫ݊݀݋݋ݎܵ݅ݐݐܹ݅݊݃ݎ݅ݏݐ
ൌ൤
ሺܶܧܩ݌݈ܽݐ݁
௜௡
మ
ሻכʹ͵ǤͷͲܹ݉
݅݊
ଶ
൨
݄ܶ݁ݎ݉݋ܧ݈݁ܿݐݎ݅ܿܫ݊݀݋݋ݎܹ݈ܹܽ݇݅݊݃ݎ݅ݏݐ
ൌ൤
ሺܶܧܩ݌݈ܽݐ݁
௜௡
మ
ሻכͶʹǤʹͻܹ݉
݅݊
ଶ
൨
݄ܶ݁ݎ݉݋ܧ݈݁ܿݐݎ݅ܿܫ݊݀݋݋ݎܴݑܹ݊݊݅݊݃ݎ݅ݏݐ
ൌ൤
ሺܶܧܩ݌݈ܽݐ݁
௜௡
మ
ሻכͷͳǤ͸ͻܹ݉
݅݊
ଶ
൨
݄ܶ݁ݎ݉݋ܧ݈݁ܿݐݎܱ݅ܿݑݐ݀݋݋ݎܵ݅ݐݐܹ݅݊݃ݎ݅ݏݐ
ൌ൤
ሺܶܧܩ݌݈ܽݐ݁
௜௡
మ
ሻכ͸ͷǤ͹ͻܹ݉
݅݊
ଶ
൨
݄ܶ݁ݎ݉݋ܧ݈݁ܿݐݎܱ݅ܿݑݐ݀݋݋ݎܹ݈ܹܽ݇݅݊݃ݎ݅ݏݐ
ൌ൤
ሺܶܧܩ݌݈ܽݐ݁
௜௡
మ
ሻכ͹ͳǤͶ͵ܹ݉
݅݊
ଶ
൨
݄ܶ݁ݎ݉݋ܧ݈݁ܿݐݎܱ݅ܿݑݐ݀݋݋ݎܴݑܹ݊݊݅݊݃ݎ݅ݏݐ
ൌ൤
ሺܶܧܩ݌݈ܽݐ݁
௜௡
మ
ሻכͳͲͺǤͲͺܹ݉
݅݊
ଶ
൨
42

As stated, two of the tables utilized in this section have a
column for voltage. This voltage is based on the ǻT (°C) that
is calculated using a formula in the datasheet [12]. This same
value can also be viewed on the graphs from the datasheet that
were copied in this section. These voltages will be necessary
in Section E: DC/DC Converter.
D. Rectification and Regulation
The next important part of the piezoelectric circuit is the
rectification and regulation circuitry. As described, the output
of our piezoelectric circuit is a sine wave with a peak to peak
voltage of 10V. If we were to use a full-wave rectifying bridge
this would allow both the positive and negative halves of the
wave to be transferred to the output up to 5V. Figure 11 is an
example of a full-wave rectifying bridge per [13].
Figure 11: Full-wave rectifying bridge example
In the above circuit, if we were to use schottky diodes we would
have less voltage drops and therefore less loss. Assuming a
single diode drop of 0.15V, two combined would give 0.3V.
This would affect the overall system efficiency in the following
manner:
- 5V – 0.3V = 4.7V output from the rectification circuit.
- 0.3V / 5V * 100 = 6% loss. Therefore, we
automatically would have a 6% system loss.
Finally, even though we have designed the circuit
mechanicals to only allow for a deflection that would provide a
10V peak to peak sine wave, we should still add in protection
in the event that the voltage ever exceeded this. That being said,
we would utilize a Zener diode (possibly 9V) on the output of
the rectification circuit. This would allow the output to never
go above 9V and therefore protect the next component in line
from over voltage failures.
E. DC/DC Converter
As with the piezoelectric input described previously, both
the solar and temperature inputs need to have a voltage
conversion so that they are useful for an embedded system to
operate. As discussed in [7], the open circuit voltage
characteristics of these sensors call for a slightly different
approach when handling the voltage conversion. For this we
have chosen to implement an integrated circuit by Texas
Instruments named BQ25504 [11]. This circuit will need to be
implemented twice as they are only single input and we will
need one for the solar input and another for the TEG.
To estimate the power losses in a system utilizing solar cells
and/or TEGs, we will refer to the BQ25504 [11] efficiency
graph in Figure 12.
Figure 12: BQ25504 Efficiency Graph
As previously determined in the sections on the Solar and TEG
inputs, the minimum input voltage that would be seen is around
0.45V and that would be on the TEG placed on the upper arm
while the user is sitting. Based on the two graphs above and the
input current, the efficiency of the DC/DC conversion would be
roughly 65-85%. To be safe, we will utilize a standard
efficiency of 70% in our power system calculations.
F. Priority-Based Voltage Combiner
The problem with having multiple inputs into the system is
that some logic is required as to which input should be passed
through to the output. To do this we have selected to use the
LTC4417 [10] by Linear Tech. This specific part allows a
system to automatically transition between input sources to feed
them to the output. We chose this part because it is completely
hardware controlled and therefore doesn’t rely on software to
operate. It operates on the basis that the three inputs are ranked
in priority order so input one is higher priority than input two
and three. By detecting which input source is currently present
to be enabled, the output will be triggered accordingly. Another
important feature about this part is that the current output status
is represented in the form of status pins. Three status pins can
tell the system which input is currently sourcing the output.
With this information the system can choose to behave
differently if needed. This method of selective operation was
successfully used in article [6]. By knowing the input power
budget for an input power source the system could turn
peripherals on or off as warranted by the power available.
Additionally, there could be circuitry added to each of the
inputs that could qualify the power levels and switch the
priorities based on the best input to the system. For the sake of
our calculations, we are making the assumption that this
circuitry is in place so that the highest power output get passed
to the battery/system.
43

G. Wearable Device
For the purposes of our system, this is simply the end load
that will be powered by the combination of the battery and
energy harvesting inputs. This can be any wearable device but
for our application we mainly focused on devices similar to
smart watches that are worn on the wrist. This maximizes the
amount of potential energy that can be harvested through the
means of motion, temperature, and solar energy. In terms of
the electrical requirements for powering the system, we have
extracted the data from the popular Apple Watch. Based on the
battery size specified online [15], it can be seen that the watch
operates off of a 205mAh pack. With an estimated runtime of
18 hours for estimation-sake [2], that equations to 205mAh / 18
= 11.38mA on average continuously throughout the day. In a
given 24 hour day cycle though that means that it is assumed
the watch is not in use for 6 hours. So if we could generate
energy in our 24 hour continuous-use cycle it would only need
to be on the order of 205mAh / 24 hours = 8.54mA. We are
also making the assumption that the battery voltage is its
nominal value of 3.76V for all calculations. In theory, if we
could transfer 8.54mA of current into the battery continuously,
it would never need to be charged manually.
IV. RESULTS
The purpose of this section is to take our proposed power
sources and calculations to real situations and see how the
system would function. The goal is to have the input sources
create enough power in all situations to supplement the battery
and prevent it from fully discharging. Our calculations will
take into account the expected inefficiencies that come with
converting raw inputs (solar, thermal, and motion) into clean
outputs required by an embedded system.
Our design is going to consist of an electronic watch with a
custom face. For visualization purposes, we have utilized the
face of a watch from FaceRepo [14] and have drawn a black
ring around the outside to look like a solar panel. Figure 13 is
a dimensional image of this watch:
Figure 13: Dimensional image of concept solar panel
The total size of the watch is 1.50” in diameter (0.750”
radius). The outside rim of the watch is a custom made solar
panel that is 0.250” thick all the way around the face.
Therefore, the surface area of the solar panel is equal to the total
surface area minus the surface area of the face. Below is this
calculation:
ܵ݋݈ܽݎܥ݈݈݁
௜௡
మ
ൌߎݎ
ଶ
ܶ݋ݐ݈ܽ െ ߎݎ
ଶ
ܨܽܿ݁
ܵ݋݈ܽݎܥ݈݈݁
௜௡
మ
ൌ ͵ǤͳͶሺͲǤ͹ͷͲ̶ሻ
ଶ
െߎሺͲǤͷͲͲ̶ሻ
ଶ
ܵ݋݈ܽݎܥ݈݈݁
௜௡
మ
ൌ ͵ǤͳͶሺͲǤ͹ͷͲ̶ሻ
ଶ
െ͵ǤͳͶሺͲǤͷͲͲ̶ሻ
ଶ
ܵ݋݈ܽݎܥ݈݈݁
௜௡
మ
ൌ ͳǤ͹͸͸݅݊
ଶ
െͲǤ͹ͺͷ݅݊
ଶ
ܵ݋݈ܽݎܥ݈݈݁
௜௡
మ
ൌ ͲǤͻͺͳ݅݊
ଶ
For the thermoelectric generator, we will use the base of
the watch (the aluminum plate that makes contact with the
skin). This dimension will be the same as the as the surface
area of the face itself. Therefore, the surface area of the TEG
will be 0.785in².
To reiterate the dimensions of the piezoelectric design,
Table 9 has the sizes used in the calculation of each use case.
Additionally, the value 2.5uW was used for the 5V rectified
input.
Layer Length Width Thickness
Component inch inch inch
Steel 0.6594 0.1772 0.0020
PZT 0.5118 0.1673 0.0020
Conductive
Adhesive 0.4823 0.1476 0.0010
Table 9: Calculated Sizes of Piezoelectric Design
Below are the use cases that we are evaluating:
Use Case 1: Sleeping at Night While Wearing Device
Solar
(mW)
Thermal
(mW)
Piezo
(uW)
System
Draw
(mW)
Power
Remaining
(mW)
0 12.92 0 31.78
-18.86
Table 10: Use Case 1 Calculations and Result
Equations Used to Obtain Numbers Above
- ܵ݋݈ܽݎܲ݋ݓ݁ݎܦܽݎ݇݊݁ݏݏ ൌ ૙࢓ࢃ
- ݄ܶ݁ݎ݉݋ܧ݈݁ܿݐݎ݅ܿܫ݊݀݋݋ݎܵ݅ݐݐܹ݅݊݃ݎ݅ݏݐ ൌ
ቂሺܶܧܩ݌݈ܽݐ݁
௜௡
మ
ሻכ
ଶଷǤହ଴௠ௐ
௜௡
మ
ቃ
- ܶܧܩ
௜௡
మ
ൌ ͲǤ͹ͺͷ݅݊
ଶ
- ݄ܶ݁ݎ݉݋ܧ݈݁ܿݐݎ݅ܿܫ݊݀݋݋ݎܵ݅ݐݐܹ݅݊݃ݎ݅ݏݐ ൌ ሺͲǤ͹ͺͷכ
ʹ͵ǤͷͲሻൌ ͳͺǤͶͷܹ݉
- ݄ܶ݁ݎ݉݋ܧ݈݁ܿݐݎ݅ܿܧ݂݂݅ܿ݅݁݊ܿݕܮ݋ݏݏ ൌ ͳͺǤͶͷܹ݉ כ
͹ͲΨ ൌ ૚૛Ǥૢ૛࢓ࢃ
- ܲ݅݁ݖ݋ܧ݈݁ܿݐݎ݅ܿܰ݋ݐܯ݋ݒ݅݊݃ ൌ ૙࢛ࢃ
- ܵݕݏݐ݁݉ܦݎܽݓ ൌ ͺǤͶͷ݉ܣ כ͵Ǥ͹͸ܸ ൌ ૜૚Ǥૠૡ࢓ࢃ
- ܴ݁݉ܽ݅݊݅݊݃ܲ݋ݓ݁ݎ ൌ ͳʹǤͻʹܹ݉ െ ͵ͳǤ͹ͺܹ݉ ൌ
െͳͺǤͺ͸ܹ݉
44

Use Case 2: Sleeping at Night While Not Wearing Device
Solar
(mW)
Thermal
(mW)
Piezo
(uW)
System
Draw
(mW)
Power
Remaining
(mW)
0 0 0 31.78
-31.78
Table 11: Use Case 2 Calculations and Result
Equations Used to Obtain Numbers Above
- ܵ݋݈ܽݎܲ݋ݓ݁ݎܦܽݎ݇݊݁ݏݏ ൌ ૙࢓ࢃ
- ݄ܶ݁ݎ݉݋ܧ݈݁ܿݐݎ݅ܿܰ݋ݐܹ݋ݎ݊ ൌ ૙࢓ࢃ
- ܲ݅݁ݖ݋ܧ݈݁ܿݐݎ݅ܿܰ݋ݐܯ݋ݒ݅݊݃ ൌ ૙࢛ࢃ
- ܵݕݏݐ݁݉ܦݎܽݓ ൌ ͺǤͶͷ݉ܣ כ͵Ǥ͹͸ܸ ൌ ૜૚Ǥૠૡ࢓ࢃ
- ܴ݁݉ܽ݅݊݅݊݃ܲ݋ݓ݁ݎ ൌ Ͳܹ݉ െ ͵ͳǤ͹ͺܹ݉ ൌ
െ͵ͳǤ͹ͺܹ݉
Use Case 3: Walking Indoors
Solar
(mW)
Thermal
(mW)
Piezo
(uW)
System
Draw
(mW)
Power
Remaining
(mW)
0.1 23.24 1.18 31.78 -8.54
Table 12: Use Case 3 Calculations and Result
Equations Used to Obtain Numbers Above
- ܵ݋݈ܽݎܲ݋ݓ݁ݎܫ݊݀݋݋ݎܨ݈ݑ݋ݎ݁ݏܿ݁݊ݐܱ݈݊ݕܰ݋ݐܷ݊݀݁ݎܮ݄݅݃ݐݏ ൌ
ቂሺܵ݋݈ܽݎܥ݈݈݁
௜௡
మ
ሻכ
଴Ǥଵସ଴௠ௐ
௜௡
మ
ቃ
- ܵ݋݈ܽݎܥ݈݈݁
௜௡
మ
ൌ ͲǤͻͺͳ݅݊
ଶ
- ܵ݋݈ܽݎܲ݋ݓ݁ݎܫ݊݀݋݋ݎܨ݈ݑ݋ݎ݁ݏܿ݁݊ݐܱ݈݊ݕܰ݋ݐܷ݊݀݁ݎܮ݄݅݃ݐݏ ൌ
(0.981 * 0.14) = 0.14mW
- ܵ݋݈ܽݎܲ݋ݓ݁ݎܧ݂݂݅ܿ݅݁݊ܿݕܮ݋ݏݏ ൌ ͲǤͳͶܹ݉ כ ͹ͲΨ ൌ
૙Ǥ૚૙࢓ࢃ
- ݄ܶ݁ݎ݉݋ܧ݈݁ܿݐݎ݅ܿܫ݊݀݋݋ݎܹ݈ܹܽ݇݅݊݃ݎ݅ݏݐ ൌ
ቂሺܶܧܩ݌݈ܽݐ݁
௜௡
మ
ሻכ
ସଶǤଶଽ௠ௐ
௜௡
మ
ቃ
- ܶܧܩ
௜௡
మ
ൌ ͲǤ͹ͺͷ݅݊
ଶ
- ݄ܶ݁ݎ݉݋ܧ݈݁ܿݐݎ݅ܿܫ݊݀݋݋ݎܹ݈ܹܽ݇݅݊݃ݎ݅ݏݐ ൌ ሺͲǤ͹ͺͷכ
ͶʹǤʹͻሻൌ ͵͵ǤʹͲܹ݉
- ݄ܶ݁ݎ݉݋ܧ݈݁ܿݐݎ݅ܿܧ݂݂݅ܿ݅݁݊ܿݕܮ݋ݏݏ ൌ ͵͵ǤʹͲܹ݉ כ
͹ͲΨ ൌ ૛૜Ǥ૛૝࢓ࢃ
- ܲ݅݁ݖ݋ܧ݈݁ܿݐݎܹ݈݅ܿܽ݇݅݊݃ ൌ ͳǤʹͷݑܹ
- ܲ݅݁ݖ݋ܧ݈݁ܿݐݎ݅ܿܧ݂݂݅ܿ݅݁݊ܿݕܮ݋ݏݏ ൌ ͳǤʹͷݑܹ כ ͻͶΨ ൌ
૚Ǥ૚ૡ࢛ࢃ
- ܵݕݏݐ݁݉ܦݎܽݓ ൌ ͺǤͶͷ݉ܣ כ͵Ǥ͹͸ܸ ൌ ૜૚Ǥૠૡ࢓ࢃ
- ܴ݁݉ܽ݅݊݅݊݃ܲ݋ݓ݁ݎ ൌ ʹ͵ǤʹͶܹ݉ െ ͵ͳǤ͹ͺܹ݉ ൌ
െͺǤͷͶܹ݉
Use Case 4: Walking Outdoors
Solar
(mW)
Thermal
(mW)
Piezo
(uW)
System
Draw
(mW)
Power
Remaining
(mW)
7.21 39.25 1.18 31.78 7.47
Table 13: Use Case 4 Calculations and Result
Equations Used to Obtain Numbers Above
- ܵ݋݈ܽݎܲ݋ݓ݁ݎܱݑݐ݀݋݋ݎ݄ܵܽ݀݁݀ ൌ ቂሺܵ݋݈ܽݎܥ݈݈݁
௜௡
మ
ሻכ
ଵ଴Ǥସଽ଼௠ௐ
௜௡
మ
ቃ
- ܵ݋݈ܽݎܥ݈݈݁
௜௡
మ
ൌ ͲǤͻͺͳ݅݊
ଶ
- ܵ݋݈ܽݎܲ݋ݓ݁ݎܱݑݐ݀݋݋ݎ݄ܵܽ݀݁݀ ൌ(0.981 * 10.498) =
10.3mW
- ܵ݋݈ܽݎܲ݋ݓ݁ݎܧ݂݂݅ܿ݅݁݊ܿݕܮ݋ݏݏ ൌ ͳͲǤ͵ܹ݉ כ
͹ͲΨ ൌ ૠǤ૛૚࢓ࢃ
- ݄ܶ݁ݎ݉݋ܧ݈݁ܿݐݎܱ݅ܿݑݐ݀݋݋ݎܹ݈ܹܽ݇݅݊݃ݎ݅ݏݐ ൌ
ቂሺܶܧܩ݌݈ܽݐ݁
௜௡
మ
ሻכ
଻ଵǤସଷ௠ௐ
௜௡
మ
ቃ
- ܶܧܩ
௜௡
మ
ൌ ͲǤ͹ͺͷ݅݊
ଶ
- ݄ܶ݁ݎ݉݋ܧ݈݁ܿݐݎܱ݅ܿݑݐ݀݋݋ݎܹ݈ܹܽ݇݅݊݃ݎ݅ݏݐ ൌ
ሺͲǤ͹ͺͷכ͹ͳǤͶ͵ሻൌ ͷ͸ǤͲ͹ܹ݉
- ݄ܶ݁ݎ݉݋ܧ݈݁ܿݐݎ݅ܿܧ݂݂݅ܿ݅݁݊ܿݕܮ݋ݏݏ ൌ ͷ͸ǤͲ͹ܹ݉ כ
͹ͲΨ ൌ ૜ૢǤ૛૞࢓ࢃ
- ܲ݅݁ݖ݋ܧ݈݁ܿݐݎܹ݈݅ܿܽ݇݅݊݃ ൌ ͳǤʹͷݑܹ
- ܲ݅݁ݖ݋ܧ݈݁ܿݐݎ݅ܿܧ݂݂݅ܿ݅݁݊ܿݕܮ݋ݏݏ ൌ ͳǤʹͷݑܹ כ
ͻͶΨ ൌ ૚Ǥ૚ૡ࢛ࢃ
- ܵݕݏݐ݁݉ܦݎܽݓ ൌ ͺǤͶͷ݉ܣ כ͵Ǥ͹͸ܸ ൌ ૜૚Ǥૠૡ࢓ࢃ
- ܴ݁݉ܽ݅݊݅݊݃ܲ݋ݓ݁ݎ ൌ ͵ͻǤʹͷܹ݉ െ ͵ͳǤ͹ͺܹ݉ ൌ
͹ǤͶ͹ܹ݉
Use Case 5: Running Outdoors
Solar
(mW)
Thermal
(mW)
Piezo
(uW)
System
Draw
(mW)
Power
Remaining
(mW)
7.21 59.39 2.35 31.78 27.61
Table 14: Use Case 5 Calculations and Result
Equations Used to Obtain Numbers Above
- ܵ݋݈ܽݎܲ݋ݓ݁ݎܱݑݐ݀݋݋ݎ݄ܵܽ݀݁݀ ൌ ቂሺܵ݋݈ܽݎܥ݈݈݁
௜௡
మ
ሻכ
ଵ଴Ǥସଽ଼௠ௐ
௜௡
మ
ቃ
- ܵ݋݈ܽݎܥ݈݈݁
௜௡
మ
ൌ ͲǤͻͺͳ݅݊
ଶ
- ܵ݋݈ܽݎܲ݋ݓ݁ݎܱݑݐ݀݋݋ݎ݄ܵܽ݀݁݀ ൌ (0.981 * 10.498) =
10.3mW
- ܵ݋݈ܽݎܲ݋ݓ݁ݎܧ݂݂݅ܿ݅݁݊ܿݕܮ݋ݏݏ ൌ ͳͲǤ͵ܹ݉ כ ͹ͲΨ ൌ
ૠǤ૛૚࢓ࢃ
45

- ݄ܶ݁ݎ݉݋ܧ݈݁ܿݐݎܱ݅ܿݑݐ݀݋݋ݎܴݑܹ݊݊݅݊݃ݎ݅ݏݐ ൌ
ቂሺܶܧܩ݌݈ܽݐ݁
௜௡
మ
ሻכ
ଵ଴଼Ǥ଴଼௠ௐ
௜௡
మ
ቃ
- ܶܧܩ
௜௡
మ
ൌ ͲǤ͹ͺͷ݅݊
ଶ
- ݄ܶ݁ݎ݉݋ܧ݈݁ܿݐݎܱ݅ܿݑݐ݀݋݋ݎܴݑܹ݊݊݅݊݃ݎ݅ݏݐ ൌ ሺͲǤ͹ͺͷכ
ͳͲͺǤͲͺሻൌ ͺͶǤͺͶܹ݉
- ݄ܶ݁ݎ݉݋ܧ݈݁ܿݐݎ݅ܿܧ݂݂݅ܿ݅݁݊ܿݕܮ݋ݏݏ ൌ ͺͶǤͺͶܹ݉ כ
͹ͲΨ ൌ ૞ૢǤ૜ૢ࢓ࢃ
- ܲ݅݁ݖ݋ܧ݈݁ܿݐݎܴ݅ܿݑ݊݊݅݊݃ ൌ ʹǤͷݑܹ
- ܲ݅݁ݖ݋ܧ݈݁ܿݐݎ݅ܿܧ݂݂݅ܿ݅݁݊ܿݕܮ݋ݏݏ ൌ ʹǤͷݑܹ כ ͻͶΨ ൌ
૛Ǥ૜૞࢛ࢃ
- ܵݕݏݐ݁݉ܦݎܽݓ ൌ ͺǤͶͷ݉ܣ כ͵Ǥ͹͸ܸ ൌ ૜૚Ǥૠૡ࢓ࢃ
- ܴ݁݉ܽ݅݊݅݊݃ܲ݋ݓ݁ݎ ൌ ͷͻǤ͵ͻܹ݉ െ ͵ͳǤ͹ͺܹ݉ ൌ
ʹ͹Ǥ͸ͳܹ݉
V. CONCLUSION
After performing all of the use cases above and testing the
different methods of operation, we can begin to see how
effective this application would be in real-world situations. In
the six different use cases presented, only two of them showed
results that demonstrated energy going back into the battery
beyond just powering the system. In these two cases the battery
would be charged while being worn by the end user. In the
other four use cases, a significant amount of energy exited the
battery to help power the system. While this proposed
architecture may not function in the full capacity we set out for,
it could help increase the time between battery charges by
helping offset some of the system load in most cases. The
results of this paper show that there are many potential areas for
improvement for the possible application. The need still exists
for a solution that solves the problem of recharging battery
powered devices and this proposed architecture is a step in the
right direction. Therefore, below is a list of the potential
improvements for the system moving forward:
- Look into other potential applications where the average
system load is smaller than 8.54mA. This would give us
more headroom to power the system.
- Investigate alternative power generation methods beyond
the ones included in this paper. While the solar and
thermoelectric methods produced energy densities that
allowed significant system power to be extracted, the
piezoelectric method was insignificant and should be
replaced by another method.
- Examine a true voltage/power combiner instead of the
priority-based controller currently proposed. The method
of choosing the highest priority device that is currently
connected throws away precious energy from the other
sources that may be present but are not producing the same
quantity of power. Utilizing the total power of all sources
would help the system become fully independent.
- Produce a physical prototype of this device. All
calculations included are theoretical and draw from the
experience of others. In order to prove our assumptions
and calculations are correct, it would be beneficial to create
a prototype to compare against.
REFERENCES
[1] Consumer Technology Association: ABI Research (2015). “Wearable
sales revenue worldwide by category from 2015 to 2021 (in billion
U.S.dollars) [Online]. Available:
https://www.statista.com/statistics/641865/wearables-sales-by-category-
worldwide/
[2] Apple: General Battery Information (2017). “General Battery
Information: Charge Time” [Online]. Available:
http://www.apple.com/watch/battery.html
[3] Battery University, Cadex Electronics (2017). “BU-808: How to Prolong
Lithium-based Batteries” [Online]. Available:
http://batteryuniversity.com/learn/article/how_to_prolong_lithium_base
d_batteries
[4] Q. Cheng, Z. Peng, J. Lin, S. Li, and F. Wang, “Energy Harvesting from
Human Motion for Wearable Devices,” 10
th
IEEE-NEMS 2015, Xi’an,
China, pp. 409-412, April 2015.
[5] T. Voss, V. Subbian, and F. Beyette, Jr., “Feasibility of Energy
Harvesting Techniques for Wearable Medical Devices,” 2014 36
th
Annual International Conference of the IEEE Engineering in Medicine
and Biology Society, Chicago, IL, 204, pp 626-629.
[6] M. Alhawari, T. Tekeste, B. Mohammad, H. Saleh, and M. Ismail,
“Power Management Unit for Multi-Source Energy Harvesting in
Wearable Electronics,” 2016 IEEE 59
th
International Midwest
Symposium on Circuits and Systems (MWSCAS), Abu Dhabi, UAE, pp
1-4, October 2016.
[7] Q. Brogan, T. O’Connor, and D. Sam Ha, “Solar and Thermal Energy
Harvesting with a Wearable Jacket,” 2014 IEEE International
Symposium on Circuits and Systems (ISCAS), Melbourne VIC, pp. 1412-
1415, June 2014.
[9] Linear Technology: LTC3588-1 Datasheet (2015). “LTC3588-1
Nanopower Energy Harvesting Power Supply Datasheet” [Online].
Available: http://cds.linear.com/docs/en/datasheet/35881fc.pdf
[10] Linear Technology: LTC4417 Datasheet (2012). “LTC4417 – Prioritized
PowerPath Controller Datasheet” [Online]. Available:
http://cds.linear.com/docs/en/datasheet/4417f.pdf
[11] Texas Instruments: bq25504 Datasheet (2015). “bq25504 Ultra Low-
Power Boost Converter With Battery Management for Energy Harvester
Applications” [Online]. Available:
http://www.ti.com/lit/ds/symlink/bq25504.pdf
[12] Micropelt: MPG-D655 Datasheet (2015). “MPG-D655 Thin Film
Thermogenerator Preliminary Datasheet” [Online]. Available:
http://micropelt.com/downloads/datasheet_mpg_d655.pdf
[13] ElectronicsTutorials: Full Wave Rectifier (2017). “Full Wave Rectifier”
[Online]. Available: http://www.electronics-
tutorials.ws/diode/diode_6.html
[14] FaceRepo: Downloadable Smart Watch Faces (2017). [Online].
Available: https://assets.facerepo.com/app/image/preview/black-pixel-
158a4670068.png?v=1590b2d1d0f
[15] Wikipedia: Apple Watch (2017). “Apple Watch” [Online]. Available:
https://en.wikipedia.org/wiki/Apple_Watch
46