Professional Documents
Culture Documents
A Thesis
Presented to
The Academic Faculty
by
Christopher R. Valenta
In Partial Fulfillment
of the Requirements for the Degree
Doctor of Philosophy in the
School of Electrical and Computer Engineering
Approved by:
iii
ACKNOWLEDGEMENTS
The completion of this thesis would not have been possible without the love, support, and
friendship of so many people. I am truly in debt to all of you and without your help, I
would not be where I am today. My parents, John and Debra Valenta, have continually
supported my endeavors. Whether it has been traveling around the world or pursuing the
crazy idea of spending ten years in college to get a PhD, their love and support have always
Dr. Greg Durgin decided to take a risk and bring a curious, first year graduate student
from a small, engineering college in Indiana into his lab. His mentorship, teaching ability,
and sheer brilliance enabled that young student to develop into a real researcher. Greg, you
are easily the best advisor a graduate student could ask for.
My PhD committee, Dr. Gisele Bennet, Dr. Andrew Peterson, Dr. Gabriel Rincón-
Mora, and Dr. Zoya Popović - Thank you all for the time you spent reviewing my proposal
and dissertation. Your constructive comments helped to transform how I envision and
attack a problem.
My former lab mates, Rujou Liu, Lorne Liechty, Joshua Griffin, Ryan Pirkl, Albert
Lu, Patrick Graf, Greg Koo, and Matthew Trotter, thank you for all of your guidance and
mentorship. Much of the knowledge i’ve gained through graduate school has been because
of your time and effort. This thesis surely would not have been possible without you.
My current lab mates, Blake Marshall, Bashir Akbar, Marcin Morys, Raj Bhattacharjea,
and Francesco Amato, thank you all for the meaningful discussions surrounding my thesis
work and letting me simply bounce ideas off you all. Although you may not realize it,
you have all helped to solve many a problem I encountered while trying to come up with
solutions in this thesis. As for the future, I came into graduate school hoping to get out as
fast as I could. However, now that I find myself leaving, I realize all the things I will miss
about the intellectual and academic freedoms associated with being a graduate student.
iv
Use this time wisely and explore things that interest you. You may find yourself quickly in
Finally, to all of the students and professors at the Politecnico di Torino, Technische
Universität München, and the University of Osaka, thank you all for your kindness and
hospitality during my time spent at your respective universities. While my time spent as
a graduate student was full of intellectual growth in engineering, the cultural experiences I
had and friends I made abroad have made these past years some of the most enlightening
I’d also like to acknowledge all of the companies, organizations, and governments that
funded the path that led to my degree. Without your financial support, I doubt I would have
ever decided to attend graduate school and earn a PhD. I hope all of you continue to fund
academic research and enable more young scientists and engineers to embark on the many
challenges of writing a thesis. In no particular order, many thanks to the Georgia Institute of
Technology; the National Science Foundation; Southern States, LLC.; the U.S. Department
v
TABLE OF CONTENTS
DEDICATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
ACKNOWLEDGEMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv
LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x
LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi
SUMMARY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxiv
I INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Historical Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Wireless Power Transfer . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.2 RFID . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2 Current Trends . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3 Research Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
vi
2.6.4 ASIC/More efficient semiconductor/packaging . . . . . . . . . . . . 26
2.6.5 Optimized energy harvesting circuit design . . . . . . . . . . . . . . 26
2.6.6 Improved excitation waveform . . . . . . . . . . . . . . . . . . . . . 27
vii
4.3 POW Energy-harvesting circuit considerations . . . . . . . . . . . . . . . . 89
4.3.1 Rectenna under POW excitation . . . . . . . . . . . . . . . . . . . 91
4.4 POW Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.4.1 Subcarrier Spacing & Phase . . . . . . . . . . . . . . . . . . . . . . 93
4.4.2 Subcarrier Number . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
4.5 Designing energy harvesters for use with POWs . . . . . . . . . . . . . . . 96
4.5.1 Guidelines for Using Rectennas Under POW Excitation . . . . . . 96
4.5.2 Design Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . 98
4.6 POW Challenges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.6.1 Communication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.6.2 Transmitting and receiving . . . . . . . . . . . . . . . . . . . . . . . 100
4.6.3 Channel effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
4.6.4 Regulatory Spectral Considerations . . . . . . . . . . . . . . . . . . 101
4.6.5 Safety . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
VI CONCLUSIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
6.1 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
6.2 Original Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
6.3 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
viii
6.4 Publications and Presentations . . . . . . . . . . . . . . . . . . . . . . . . . 137
6.4.1 Peer-Reviewed Journal Articles . . . . . . . . . . . . . . . . . . . . 137
6.4.2 Refereed Conference Proceedings . . . . . . . . . . . . . . . . . . . 137
6.4.3 Presentations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
VITA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
ix
LIST OF TABLES
x
LIST OF FIGURES
xi
9 A standard Schottky diode model. A resistance Rs is in series with a variable
junction resistance Rj in parallel with a variable junction capacitance Cj
which change as a function of input power. . . . . . . . . . . . . . . . . . . 37
10 Impedance of single shunt rectenna using an Avago HSMS-2860 diode with
a 316 Ω load matched at 5.8 GHz and 16 dBm input power. The large range
of possible impedances makes it difficult for an energy-harvesting circuit to
work efficiently over a wide range of input powers and frequencies because of
the reflected power due to impedance mismatch [7]. . . . . . . . . . . . . . . 38
11 Parasitics in a SOT-23 package [8]. These parasitic values cause additional
impedance mismatch which can translate to additional power lost due to
reflections. Note that this package can be outfitted with one or two (shown
in dotted lines) diodes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
12 The theoretical energy-harvesting efficiency of a single, silicon-nickel Schot-
tky diode under CW excitation delivered from a 50 Ω source. This figure
illustrates the performance variation when various diode parameters (voltage
threshold VT in 12a, breakdown voltage Vbr in 12b, input frequency in 12c,
series resistance Rs in 12d, and load resistance RL in 12e) are varied. Pa-
rameters are varied about those of a standard Avago HSMS-286x diode as
shown in the enclosed table [2]. . . . . . . . . . . . . . . . . . . . . . . . . . 44
13 The maximum efficiency
of an ideal diode (VT =0 V and Vbr =∞) for various
Rs
values of r r = RL . A low-pass filter effect is due to the diode junction
capacitance and series resistance. Maximum efficiency occurs in the passband
of this filter and increases with decreasing values of the series resistance. . . 46
14 State-of-the-art RF and microwave-energy conversion efficiencies. A variety
of topologies including charge pumps and rectennas are used under a variety
of load conditions and technologies [9]. . . . . . . . . . . . . . . . . . . . . . 47
15 Various energy-harvesting circuit topologies used in RFID and SSP applica-
tions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
16 The original Dickson charge pump was a voltage multiplier which increased
a DC input signal by pumping charge across pairs of diodes and capacitors.
This pumping was controlled by an external clock signal which controlled the
rate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
17 The theoretical energy-harvesting DC output voltage for a single diode; two-,
four-, six-, and eight-stage Dickson charge pumps with 300 Ω, 600 Ω, 1200 Ω,
1800 Ω, and 2400 Ω loads, respectively. These circuits use Avago HSMS-285x
diodes under 5.8 GHz CW excitation. . . . . . . . . . . . . . . . . . . . . . 57
18 The theoretical energy-harvesting DC output power for a single diode; two-,
four-, six-, and eight-stage Dickson charge pumps with 300 Ω, 600 Ω, 1200 Ω,
1800 Ω, and 2400 Ω loads, respectively. These circuits use Avago HSMS-285x
diodes under 5.8 GHz CW excitation. . . . . . . . . . . . . . . . . . . . . . 58
xii
19 The theoretical energy-harvesting efficiency for a single diode; two-, four-, six-
, and eight-stage Dickson charge pumps with 300 Ω, 600 Ω, 1200 Ω, 1800 Ω,
and 2400 Ω loads, respectively. These circuits use Avago HSMS-285x diodes
under 5.8 GHz CW excitation. . . . . . . . . . . . . . . . . . . . . . . . . . 59
20 The theoretical energy-harvesting DC output voltage for a single diode; two-
, four-, six-, and eight-stage Dickson charge pumps all with a 300 Ω load.
These circuits use Avago HSMS-285x diodes under 5.8 GHz CW excitation. 60
21 The theoretical energy-harvesting DC output power for a single diode; two-
, four-, six-, and eight-stage Dickson charge pumps all with a 300 Ω load.
These circuits use Avago HSMS-285x diodes under 5.8 GHz CW excitation. 61
22 The theoretical energy-harvesting efficiency for a single diode; two-, four-,
six-, and eight-stage Dickson charge pumps all with a 300 Ω load. These
circuits use Avago HSMS-285x diodes under 5.8 GHz CW excitation. . . . . 61
23 The methodology for designing an energy-harvesting circuit. . . . . . . . . . 63
24 The Avago HSMS-286x, HSMS-285x, and the HSMS-282x theoretical efficien-
cies are plotted as a function of input power. These curves were evaluated
assuming an excitation frequency of 5.8 GHz and a 300 Ω load. . . . . . . . 64
25 Flowgraph for optimizing energy-harvesting circuits in Agilent ADS. . . . . 67
26 The layout of a 5.8 GHz single-shunt rectenna on a 4-layer FR-4 substrate.
The input impedance was matched to 50 Ω at an input power of 16 dBm.
Schematic elements shown were simulated as SMDs. . . . . . . . . . . . . . 69
27 The percentage of input power converted to DC (energy conversion efficiency)
and the percentage of power reflected into the source of the designed rectenna.
Note that the maximum efficiency of 70.5% at 16 dBm occurs when the
reflected power is minimized. The large discrepancy between the simulated
results and theory is due to the FR-4 substrate losses. . . . . . . . . . . . . 70
28 The theoretical and simulated output voltage of the single shunt rectenna
follow a similar trend. Much of the difference at high powers is due to sub-
strate losses in the FR-4. At lower powers, impedance mismatch causes a
much greater deviation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
29 An equal-energy power-optimized waveform (POW) and continuous wave-
form in the time and frequency domain. This 4-POW has eight equally
weighted subcarriers which combine in-phase to increase the peak-to-average
power ratio (PAPR) of the waveform to increase energy-harvesting efficiency. 74
30 A POW can cause a larger DC bias across a diode for a lower average power
than a CW. This figure shows an equal power 4-POW and a CW for a set
DC bias level. Note that since the CW does not have sufficient voltage to
turn-on the diode, its DC bias level would decrease. . . . . . . . . . . . . . 75
xiii
31 Three examples of power-optimized waveforms and their parameters. From
top to bottom, M-POW, Square POW, and Gaussian POW. Notice the dif-
ferent PAPR of each waveform along with the frequency content and shape
of their respective spectra. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
32 Harmonic content generated with a CW excitation (a) and POW excitation
(b) when used with a diode-based energy-harvesting circuit. Like a CW
excitation, DC and harmonics are generated along with the fundamental
excitation frequency. However, POWs also cause intermodulation products
at integer multiples of the POW period TPOW . A low-pass filter should help
to remove some of this content, but intermodulation products may many
times be too low to effectively filter out. . . . . . . . . . . . . . . . . . . . . 82
33 The theoretical DC output voltage for an energy-harvesting circuit with a
single Avago HSMS-2860 diode and 300 Ω load under 5.8 GHz POW excita-
tion of various PAPR/duty cycles. As the PAPR increases, the diode is able
to rectify DC voltage more easily at lower powers while at higher powers, the
DC voltage is lower because of an increased voltage drop across the diode
series resistance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
34 The theoretical DC output power for an energy-harvesting circuit with a sin-
gle Avago HSMS-2860 diode and 300 Ω load under 5.8 GHz POW excitation
of various PAPR/duty cycles. As the PAPR increases, the diode is able to
rectify DC power more easily at lower powers while at higher powers, the DC
power is lower because of an increased voltage drop across the diode series
resistance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
35 The theoretical energy-conversion efficiency for an energy-harvesting circuit
with a single Avago HSMS-2860 diode and 300 Ω load under 5.8 GHz POW
excitation of various PAPR/duty cycles. As the PAPR increases, the effi-
ciency increases at lower powers while at higher powers, the efficiency de-
creases due to power lost in the diode series resistance. . . . . . . . . . . . . 86
36 The theoretical maximum energy-conversion efficiency for an energy-harvesting
circuit with a single Avago HSMS-2860 diode and 300 Ω load under 5.8 GHz
POW excitation as a function of PAPR. As the PAPR increases, the maxi-
mum efficiency approaches zero. . . . . . . . . . . . . . . . . . . . . . . . . . 87
37 The input power for which the diode turns on for increasing PAPR in an
energy-harvesting circuit with a single Avago HSMS-2860 diode and 300 Ω
load under 5.8 GHz POW excitation. . . . . . . . . . . . . . . . . . . . . . . 88
38 The maximum energy-conversion efficiency plotted against the turn-on power
in an energy-harvesting circuit with a single Avago HSMS-2860 diode and
300 Ω load under 5.8 GHz POW excitation. The PAPR increases as the
efficiency decreases. An optimal PAPR would be found at the inflection
point. In this case, an optimal PAPR of 55 yields a turn-on power of -32 dBm
and a maximum efficiency of 30%, a 14 dB and 45% decrease, respectively. . 89
xiv
39 As the POW pulse becomes narrower (PAPR is increased), the RMS time
width decreases which reduces the amount of time the output capacitor has
to charge. Furthermore, the time spent discharging is increased. This change
results in increased voltage ripple [7]. . . . . . . . . . . . . . . . . . . . . . . 90
40 Simulated POW gain, output DC voltage, and output ripple voltage for a
5.8 GHz single shunt rectenna excited with a 1-POW with 10 MHz subcar-
rier spacing. At low powers, voltage ripple is large as the rectifier cannot
efficiently convert RF to DC and intermodulation products dominate the
output (From [7]). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
41 The POW gain of the designed rectenna was simulated with a 1-POW whose
subcarrier spacing was varied (top). It was determined that the maximum
POW gain was obtained when the subcarrier spacing was made equal to the
rectenna load time constant. The percent ripple voltage was also plotted
(bottom). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
42 DC output voltage and DC output power of the single-shunt rectenna with
increasing numbers of equal amplitude subcarriers with 31.6 MHz spacings.
Larger number of subcarriers allow the diode to operate more efficiently at
low power as evident from the larger voltages and powers between -30 dBm
and -8 dBm compared to CW. . . . . . . . . . . . . . . . . . . . . . . . . . 96
43 POW gain of M-POW as the number of subcarriers is increased. Subcarrier
spacing was set to be 31.6 MHz as determined from the previous section.
Note that the maximum gain occurs where the energy conversion efficiency
is typically around 10%. Also note that POWs do decrease the energy con-
version efficiency at high powers. . . . . . . . . . . . . . . . . . . . . . . . . 97
44 The methodology for designing an energy-harvesting circuit and POW. . . . 100
45 The channel effects on a POW may vary the relative amplitude and phase
of each POW subcarrier. This variation will usually result in a sub-optimal
shape will may be less beneficial to the energy harvester than a CW. . . . . 101
46 Plot of the measured, simulated, and data sheet supplied I-V curves from an
Avago HSMS-2682 series diode in a SOT-23 package. . . . . . . . . . . . . . 104
47 Log plot of the measured, simulated, and data sheet supplied I-V curves from
an Avago HSMS-2682 series diode in a SOT-23 package. . . . . . . . . . . . 105
48 Simulated S-parameters from interdigitated capacitor used in the prototyped
rectennas. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
49 Physical layout for the input interdigitated capacitor used for the prototyped
rectenna circuits. The length of the IDC is 740 mil and the maximum width
is 689 mil. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
50 Block diagram of the energy-harvesting circuit measurement system for CW
excitations. A circulator provides a means of measuring reflected power to
determine the matching of the circuit while a coupler allows the input power
level to be monitored. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
xv
51 Block diagram of calibrated section of the charge-pump measurement system.
By measuring the losses of the various signal paths, an accurate value of the
input and reflected power can be rectified. . . . . . . . . . . . . . . . . . . . 111
52 Measured insertion loss SBA of the coupler-circulator sub-circuit shown in
Figure 51. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
53 Measured insertion loss SCA of the coupler-circulator sub-circuit shown in
Figure 51. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
54 Measured insertion loss SDC of the coupler-circulator sub-circuit shown in
Figure 51. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
55 Block diagram of the system used to generate POWs and characterize their
performance with energy harvesters. This system requires additional atten-
tion to ensure the amplifiers operate linearly so as to not distort the POW. 114
56 Photograph of the experimental setup for measuring energy-harvester effi-
ciency under POW excitation. . . . . . . . . . . . . . . . . . . . . . . . . . . 115
57 Annotated photograph of the high-power rectenna. Note the rectenna was
placed on standard engineering graph paper with a grid spacing of 0.25 inches
or 6.35 mm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
58 Measured return loss of the high-power rectenna at a power of -10 dBm. . . 117
59 Measured output power and energy-harvester efficiency for the fabricated
high-power rectenna under CW, 1-POW, 2-POW, and 3-POW excitations
with a center frequency of 6.06 GHz and subcarrier spacing of 10 MHz. . . 118
60 Theoretical, simulated, and measured output power and energy-harvester
efficiency for the fabricated high-power rectenna under CW excitation with
a center frequency of 6.06 GHz. . . . . . . . . . . . . . . . . . . . . . . . . . 119
61 Theoretical, simulated, and measured output power and energy-harvester
efficiency for the fabricated high-power rectenna under a 1-POW excitation
with a center frequency of 6.06 GHz and subcarrier spacing of 10 MHz. . . 120
62 Theoretical, simulated, and measured output power and energy-harvester
efficiency for the fabricated high-power rectenna under a 2-POW excitation
with a center frequency of 6.06 GHz and subcarrier spacing of 10 MHz. . . 121
63 Theoretical, simulated, and measured output power and energy-harvester
efficiency for the fabricated high-power rectenna under a 3-POW excitation
with a center frequency of 6.06 GHz and subcarrier spacing of 10 MHz. . . 122
64 Example measured outputs from the prototyped high-power rectenna for dif-
ferent POWs. Note how the ripple increases with increasing PAPR. . . . . 123
65 Annotated photograph of the low-power rectenna. Note the rectenna was
placed on standard engineering graph paper with a grid spacing of 0.25 inches
or 6.35 mm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
66 Measured return loss of the low-power rectenna at a power of -10 dBm. . . 125
xvi
67 Measured output power and energy-harvester efficiency for the fabricated
low-power rectenna under CW, 1-POW, 2-POW, and 3-POW excitations
with a center frequency of 5.46 GHz and subcarrier spacing of 10 MHz. . . 127
68 Theoretical, simulated, and measured output power and energy-harvester
efficiency for the fabricated low-power rectenna under CW excitation with a
center frequency of 5.46 GHz. . . . . . . . . . . . . . . . . . . . . . . . . . . 128
69 Theoretical, simulated, and measured output power and energy-harvester
efficiency for the fabricated low-power rectenna under a 1-POW excitation
with a center frequency of 5.46 GHz and subcarrier spacing of 10 MHz. . . 129
70 Theoretical, simulated, and measured output power and energy-harvester
efficiency for the fabricated low-power rectenna under a 2-POW excitation
with a center frequency of 5.46 GHz and subcarrier spacing of 10 MHz. . . 130
71 Theoretical, simulated, and measured output power and energy-harvester
efficiency for the fabricated low-power rectenna under a 3-POW excitation
with a center frequency of 5.46 GHz and subcarrier spacing of 10 MHz. . . 131
72 Measurements of high-power and low-power rectenna efficiencies plotted with
the state-of-the-art. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
73 A standard diode model with parallel output resistance. This model assumes
that all parasitic values have been tuned out with reactive elements when
possible. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
74 The voltage across the diode and the output voltage. Note how Vd is delayed
by Φ and when the voltage is greater than VT , it remains constant at this
value between −Θoff and Θoff (Adapted from [6]). . . . . . . . . . . . . . . 148
75 A 2-stage Dickson charge pump diode model with output resistance. This
model assumes that all diode parasitic values have been tuned out with re-
active elements, but includes the circuit capacitances for RF grounding. . . 153
76 The voltage across the diode and the output voltage. For a POW or signal
with intermittent transmission, there is a period of Ψ when the waveform is
‘off’ and only the rectified DC component is present. . . . . . . . . . . . . . 158
77 Theoretical and simulated DC output voltages of N-stage Dickson charge
pumps. Theoretical results were produced with the aforementioned theory.
Simulated results were computed with Agilent ADS using ideal components
and Avago HSMS-286x Schottky diodes. . . . . . . . . . . . . . . . . . . . . 164
78 Theoretical and simulated efficiencies of N-stage Dickson charge pumps. The-
oretical results were produced with the aforementioned theory. Simulated
results were computed with Agilent ADS using ideal components and Avago
HSMS-286x Schottky diodes. . . . . . . . . . . . . . . . . . . . . . . . . . . 165
xvii
79 Theoretical and simulated output powers of rectennas under various POW ex-
citations. Theoretical results were produced with the aforementioned theory
using duty-cycled sinusoids or square POWs. Simulated results were com-
puted with Agilent ADS using ideal components, Avago HSMS-286x Schottky
diodes, and excited with M-POWs. . . . . . . . . . . . . . . . . . . . . . . . 166
80 Theoretical and simulated efficiencies of rectennas under various POW exci-
tations. Theoretical results were produced with the aforementioned theory
using duty-cycled sinusoids or square POWs. Simulated results were com-
puted with Agilent ADS using ideal components and excited with M-POWs. 167
81 The normalized probability density function of a product Rayleigh distribu-
tion for increasing numbers of RFID tag antennas in an 1 × L × 1 Dyadic
Backscatter Channel with fully correlated channels. Note that as the number
of tag antennas increase, the envelope PDF becomes a Rayleigh distribution.
Note that these fades are more severe than for a simple Rayleigh. Modified
from [10]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
82 A 3-POW with modulated data. Notice how orthogonality is preserved sim-
ilar to OFDM. At the peak of each subcarrier, the side lobes from the other
subcarriers are zero. Provided that the signals are properly sampled, no self
inter-carrier interference will occur. . . . . . . . . . . . . . . . . . . . . . . . 187
83 An RF tag configured to communicate on a single POW subcarrier. The
dotted line corresponds to the frequency response of the bandpass filter. Note
how the charge pump receives all subcarriers in order to increase charge pump
efficiency. However, the RF tag will only backscatter on subcarrier f3 . . . . 188
84 Different methods for transmitting POWs. Method (a) is the most simplistic
and uses a DAC with an inefficient and expensive Class A amplifier. Method
(b) also uses a DAC but uses N parallel Class AB or C amplifiers for the POW
signal. Finally, method (c) creates, amplifies, and combines each POW at
RF reducing the need for POW upconversion. From [5]. . . . . . . . . . . . 189
85 Probability that at least one subcarrier is not in a fade for the fully correlated,
product Rayleigh fading channel for various fade margins. Note that as
the number of subcarriers increase, the probability of at least one of the
subcarriers not being in a fade goes to unity. . . . . . . . . . . . . . . . . . 192
xviii
LIST OF SYMBOLS OR ABBREVIATIONS
A area.
Bf total bandwidth of the ideal filter used for calculating the RMS bandwidth
of a square POW.
GP OW POW gain.
xix
2
k Boltzmann’s constant (1.381 × 10−23 ms2 K
kg
).
L Length.
λ free-space wavelength.
RL load resistance.
xx
Vbr Diode reverse breakdown voltage.
VtagDC M IN minimum DC voltage at the output of the energy harvester required for
the completion of a task.
Vp peak voltage.
DC direct current.
HF high frequency.
xxi
IEEE institute of electrical and electronic engineers.
LF low frequency.
RF radio-frequency.
xxii
WISP wireless identification and sensing platform.
xxiii
SUMMARY
The wireless transfer of power is the enabling technology for realizing a true internet-
related biological data, and building utility usage are just a small fraction of the myriad of
applications which are part of an ever evolving ubiquitous lifestyle. Realizing these systems
requires a means of powering their electronics sans batteries. Removing the batteries from
the billions or trillions of these envisioned devices not only reduces their size and lowers
cuits extends the range and reliability of passive sensor networks. Multi-frequency wave-
forms are one technique that assists in overcoming the energy-harvesting circuit diode volt-
age threshold which limit the energy-conversion efficiency at low RF input powers typically
circuits along with multi-frequency waveform excitation. Using this methodology, a low-
power 5.8 GHz rectenna showed an output power improvement of over 20 dB at -20 dBm
input power using a 3-POW compared to continuous waveforms. The resultant efficiency is
the highest reported efficiency for low-power 5.8 GHz energy harvesters. Additionally, new
theoretical models help to predict the maximum possible range of the next generation of
passive electronics based upon trends in the semiconductor industry. These models predict
improvements in diode turn-on power of over 20 dB using modern Schottky diodes. This
xxiv
CHAPTER I
INTRODUCTION
The idea of (WPT) has been around since the inception of electricity. In the late 19th
century, Nikola Tesla described the freedom to transfer energy between two points without
man [11].” A truly wireless device, capable of being remotely powered, not only allows the
obvious freedom of movement, but also enables devices to be more compact by removing the
necessity of a large battery. Applications could leverage this reduction in size and weight
to increase the feasibility of concepts such as paper-thin, flexible displays [12], contact
lens-based augmented reality [13], and smart dust [14], among traditional point-to-point
power transfer applications. While several methods of wireless power have been introduced
since Tesla’s work, including near-field magnetic resonance and inductive coupling, laser-
based optical power transmission, and far-field RF/microwave energy transmission, only
RF/microwave and laser-based systems are truly long-range methods. While optical power
transmission certainly has merit, its mechanisms are often ill-suited for RFID-based systems
1
Two major communities have made significant contributions towards fulfilling Tesla’s
dream of wireless power: space-based solar power (SSP) or solar powered satellite (SPS) and
working in SPS have made advancements in energy conversion at great distances and high
powers (greater than 1 Watt) to enable electricity to be shared via radio waves, namely
collecting solar energy in orbital stations around earth and beaming it to ground stations
via microwave power transfer [15, 16, 17]. Meanwhile the RFID community has focused on
the ultra-low power harvesting (usually much less than 1 Watt) necessary to provide energy
for wireless, batteryless RFID tags and backscatter sensors [18]. Future development in
WPT will certainly leverage the work of both SPS and RFID researchers and lead to the
The short-range transfer of power between induction coils had been understood since the
invention of the transformer by Nicholas Callan in 1836. However, Tesla had much more
ambitious goals of world-wide, wireless power distribution. His experiments in the 1890s
utilized relatively high frequencies (kHz) for the time and resonant coils, which enabled him
to remotely power light bulbs and other simple devices over much greater distances than
those allotted by using a simple transformers. This success came at the expense of efficiency
Tesla failed to achieve his dream of wireless power; a lack of direction antenna beams (the
relatively low frequencies of his signals made fabricating high gain antennas difficult) and
the removal of his funding from J. P. Morgan secured the end of his work [19, 20, 21, 22].
After Tesla, the idea of WPT continued to be researched, but no significant progress was
It wasn’t until 1933 at the Chicago World Fair that H.V. Noble of Westinghouse Labo-
ratory demonstrated a functional wireless power transfer system to the public. This system
was able to turn on light bulbs in a room using 100 MHz radio waves much to the amaze-
ment of the fair attendees. Noble also carried out more scientific laboratory experiments
2
using this same system to transfer hundreds of watts over a distance of 25 feet [23]. While
this demonstration marked an improvement in efficiency, it was evident that more focused
antenna beams (higher frequencies) were necessary to effectively transmit power wirelessly.
However, these improvements would have to wait for a breakthrough in high frequency radio
The First and Second World Wars resulted in significant technological advancements,
most notably in the area of high-powered microwave devices. This availability filled in a
critical, missing piece of the WPT puzzle. With the ability to generate large amounts of
microwave energy, researchers such as William C. Brown began revisiting Tesla’s idea of
WPT. In 1963, Brown, a researcher with Raytheon under a contract from the United States
Department of Defense, demonstrated the first microwave power transfer system using a
thermionic diode [23]. This system successfully drove a 100 W DC motor via a 3 GHz RF
link from 400 W of generated power. While thermionic diodes proved useful for high-power
applications, they required heater power to begin working. Furthermore, Brown found them
to be ‘unreliable and short-lived device[s].’ [23] These issues led Brown to begin investigating
other methods of rectification. Semiconductor diodes, which at the time were not widely
ever, semiconductor diodes did not have the same power handling capacity as thermionic
diodes or cyclotron wave converters (CWC). (In fact, Brown’s first semiconductors were
Brown’s lack of experience in this area led him to two Purdue University Professors,
Roscoe George and Elias Sabbaugh, working on semiconductor diodes in microwave recti-
fiers. Combining their work with Brown’s, the rectenna (rectifying antenna) was born. The
rectenna is a simple device which has a rectifier attached directly to an antenna. In this
manner, DC power could be readily obtained from RF signals incident on the antenna. As
1
Note: Although other devices have been developed which can convert RF to DC, including cyclotron
wave converters (CWC), they will not be discussed in the remainder of this thesis as they are outside the
scope and practical application towards small, ultra-low power devices. Brown discusses several alternatives
in his many reports.
3
a solution to the low-power handling capacity of the rectenna, Brown and the Purdue pro-
fessors developed a rectenna array. This array consisted of many rectennas linked together
across a common load. Thus, the DC power provided was the sum of the power from all
the rectennas. One of the first semiconductor rectenna arrays used much lower power than
Brown’s previous designs at 8 W, but still had an efficiency of 50%, much higher reliability,
and lower weight. The low-weight aspect was particularly important as the project spon-
sor, the U.S. Air Force, was interested in microwave-powered aircraft. Brown continued to
helicopter outfitted with a 2.4 GHz rectenna array that could fly 60 feet above an antenna
the design and was tasked by the U.S. Air Force to build a microwave-powered platform
for communication or surveillance. Several years later despite his success, the Air Force
decided not to continue this project, and very little progress was made from 1965 - 1970.
Peter Glaser introducing a concept known as Space Solar Power (SSP) or Solar-Power Satel-
lite (SPS) [15]. He envisioned placing massive solar panel farms in space and then beaming
the power down to earth as an effort to reduce fossil fuel consumption and meet energy
demand. Since the panels would be able to collect energy 24 hours a day and have larger
energy density available, it seemed to NASA an idea worth exploring. They awarded Brown
system. Many systems-level engineering challenges for the SSP system were addressed dur-
ing this work, most relevantly the investigation of more exotic semiconductor diodes. Much
of this work collimated in 1975 with the demonstration of a one-mile-long WPT system
tion had a DC-to-DC efficiency of 54% (RF-to-DC efficiency of 82.5% at 2.45 GHz) with a
DC output power of 30 kW [23]. After some follow on programs including flexible printed
antennas and high efficiency (90%+) rectennas, NASA ended the program in 1980 with
the conclusion that no major technical problems existed in the SSP program. However,
4
follow-on research and funding failed to materialize.
Two major activities in the 1980s in the area of microwave-power transfer came about
in Japan and in Canada. In 1983, Hiroshi Matsumoto and his team investigated high-power
microwave beam interaction in the ionosphere as part of the Microwave Ionosphere Non-
linear Interaction eXperiment (MINIX) for SSP applications [16]. Later in 1987, Canadian
scientists developed an airplane that was powered solely by microwave beam in a project
known as the Stationary High Altitude Relay Platform (SHARP). While a full-size model
was not constructed, a 1/8th scale model of the plane was built and successfully tested.
Notably, the SHARP team considered using 5.8 GHz as a potential frequency although the
test used 2.45 GHz [26]. The suggestion of a higher frequency meant that antenna sizes
could be smaller, thereby reducing the size and cost. Matsumoto’s research group in Japan
eXperiment (MILAX) in 1992. While SHARP utilized a parabolic antenna which manually
In the 1990s and 2000s, there was a large amount of work exploring higher frequencies
for WPT for SSP applications as well as lower frequencies for RFID and WSN applications.
The challenge for the SSP community has been to create the most efficient and economic
manner to transfer microwave energy. Thus, a trend to higher frequencies where smaller
aperture sizes would be possible has been a focus. Additionally, there has been work by
several researchers working on WPT for RFID applications. These applications tend to have
a significantly lower power density available which makes energy harvesting a challenge.
While there are too many projects to name individually, Naoki Shinohara [16, 27, 28] and
Kai Chang [6, 29, 30] have been among the most recognized names working on WPT focusing
on SSP.
Some of the important trends in WPT history are summarized in Table 1. There has
been a progression from high powers at low frequencies to lower powers at higher frequencies.
This change stems from the ability to more easily focus radio wave energy when at high
frequencies in addition to reducing the required antenna apertures for harvesting energy.
This trend also shows a need to generate less power because of higher efficiencies. Higher
5
efficiencies became possible due to a larger understanding of the RF-DC rectification process
and the loss mechanisms encountered. In fact, Brown’s understanding of the rectification
process led to the highest RF-to-DC efficiency ever recorded at 90.6% from his work for
NASA in 1977.
Table 1: Historical trends in WPT frequencies, powers, and efficiencies.
1.1.2 RFID
communications for low-cost tracking. One of the first instances of the use of backscatter
occurred in World War II. With the advent of RADAR, both sides were able to tell when an
aircraft was approaching. However, there was no method to identify whether that aircraft
was friendly or an enemy. The German Air Force is credited with discovering that they
could identify their aircraft by having their pilots maneuver their planes in a certain manner
which changed their radar cross-section (RCS), and therefore the response on the radar
screen [31]. This method is a precursor to how backscatter communications work and will
be discussed in Chapter 2. A few years later in 1937, the British debuted an identify friend
or foe (IFF) transponder system called the XAE, and later Mark I, as a more secure method
of identifying their aircraft. This system, when hit with a RADAR pulse, would send out a
signal which identified the aircraft. It is accredited with being the first active RFID system.
As early as 19452 , Soviet Scientist Léon Theremin had designed a passive method of
listening to conversations. This device, now known as the ‘Great Seal Bug,’ allowed the
2
The exact date is unknown since this was developed as an espionage device.
6
Soviets to spy on the American embassy in Moscow for seven years until its discovery [32].
It used a resonator which, when sound wave vibrated a membrane, modulated any sound in
the room onto an external RF signal that Soviet Spies pointed at the American Embassy.
The Americans were unable to discover this device because it had no active components
which to detect. It is accredited with being one of the first passive RFID systems despite
During the time that the Great Seal Bug was in use, Harry Stockman officially referenced
the first method of communication by reflected radio waves in a paper published in 1948 [33].
While much of the necessary technology for performing true passive, backscattered commu-
nication did not exist, Stockman nevertheless foresaw the benefits of reflected radio waves
for passive communication. It wasn’t until a few years later in 1952 that D.B. Harris filed
a U.S. Patent referencing the first relative of the modern RFID tag [34]. While Harris’s
power transfer mechanisms were not as advanced as Brown’s, the path for integration was
set.
The 1960s and 1970s represented a huge leap in the progress of RFID tags thanks to
plications such as Sensormatic and Checkpoint electronic article surveillance (EAS) devices
for retailers [35], the Ratheon Raytag [36], and Los Alamos National Laboratory nuclear
material and cattle tracking [37, 38] increased the interest in RFID technology and led to
further investment and development by the government and industry alike. The work out
of Los Alamos led to the development of passive UHF RFID which is directly linked to the
At the time of this thesis, RFID technology has been widely adapted in the industry for
uses in logistical tracking, highway tolling, building access, and transportation ticketing.
These applications have generally been limited to the original intended use of RFID, that is
identification and a replacement for bar codes. Research in this area continues as venders
wish to increase the read rate, range, and reliability of their RF tags.
7
The success of RFID has attracted the attention of the WSN community [39]. The desire
for an internet of things (IOTs), to ‘sense everything’ in current society, has been increas-
these ideas with RFID providing a particularly attractive solution because of their low cost
Most current trends in RFID research mention sensing to at least some degree. The
system design as the sensors must also be powered. Thus, researchers have begun examining
some of the work of the SSP community. While the SSP community typically focuses on
high-power (greater than 20 dBm) energy conversion, their work on semiconductors and
rectennas can be directly applied to the very low-power (less than 0 dBm) situations where
Some of the first work in this RFID-enabled sensing area has come from Zoya Popović at
the University of Colorado at Boulder [40] and Joshua Smith at the University of Washing-
ton [41]. The Intel Wireless Identification and Sensing Platform (WISP), which Smith did
most of his work with, is a passive UHF RFID with a programmable microcontroller that
can communicate via the EPC Global Generation (Gen) 2 protocol using off-the-shelf RFID
readers. This research tool has proven very effective as the Gen 2 protocol is one of the
most used and the project code, schematics, and WISP samples have been made available
to researchers and hobbyists alike. It has spawned a great deal of interest and work in the
sensors [42, 43], computation [44, 45], protocol [46], and security [44, 47] areas. Much of the
other work in the RFID space has involved other methods of energy harvesting [48, 49, 50],
moving to higher frequencies [51], novel antenna designs [52], flexible substrates [53], and
Both SSP and RFID applications have numerous advantages in moving to microwave
frequencies. Microwave antennas are substantially smaller than some of their UHF and VHF
counterparts, and this smaller size helps to reduce the antenna aperture required and to
8
focus the antenna beams more easily. Since the antenna size is the dominating component
in the size of a rectenna or RFID-enabled sensor, space can be used more efficiently. This
reduction in size (and therefore weight) can help SSP applications reduce the amount of land
required for ground-based stations, and help RFID applications minimize the profiles of their
tags and sensors. The world-wide availability of the unlicensed 5.8 GHz microwave band, an
ISM frequency band in the United States, has led research in microwave-power transfer to
focus on this frequency. In fact, several other benefits of operating at 5.8 GHz for RFID have
been discussed apart from the reduced footprint including the following: greater bandwidth
availability, reduced on-object penalty [51], spatial signaling techniques [10, 57, 58], greater
material sensing capability [59, 60], and plasma penetration ability [61, 62] among others.
Another key technological development in RFID and WPT is the power-optimized wave-
form (POW) [63, 64] or multi-sine excitation [65, 66]. A POW is a waveform whose spec-
trum is shaped to increase the efficiency of RF energy harvesting circuits, especially at low
powers. By increasing this efficiency, the range and reliability or power available to the
RF tag will also be increased. A multi-sine accomplishes this increase in efficiency via a
large peak-to-average power ratio (PAPR) which helps to overcome the energy harvesting
circuit diode-voltage threshold limitation. When incident signals are not greater than this
threshold, the circuit efficiency suffers. Trotter [64] and Boaventura [65] have independently
shown that POWs do increase the efficiency of charge-pump circuits. However, their work
has not explored in detail the limitations of this increase in energy harvesting efficiency or
This research is novel in its systems-level approach to the design of the excitation waveform
and the energy harvesting circuit. A theoretical analysis of the excitation waveform will
be derived and compared with simulated and measured results. Furthermore, an algorithm
will be discussed which will help to determine the maximum range and reliability of an RF
9
Overall, this research seeks to answer the following question:
Given a design space limited by application, regulatory standards, and physical limitations,
what are the RF-to-DC energy harvesting efficiency improvements gained by utilizing
Chapter 2 introduces some of the background on RFID tags and RFID-enabled sensors.
Chapter 3 will discuss in detail RF energy harvesting in RFID tags specifically addressing
Chapter 4 will address power-optimized waveforms (POWs), their design and implemen-
Chapter 5 will cover the measurements of the microwave power transfer systems and
discuss similarities and differences between theoretical, measured, and simulated results.
Chapter 6 will close this thesis and summarize the original contributions, directions for
future work, and mention the publications and presentations pertaining to this research.
Appendix A contains data in tabular form for the current state-of-the-art energy-harvesting
conversion efficiencies for a wide range of input powers, frequencies, loads, and topologies.
Appendix B covers step-by-step the theoretical derivations for the efficiency of an energy-
harvesting circuit with N diodes and under POW excitation. Computer simulations which
Appendix E provides details on the Agilent ADS simulation schematics and layouts used
10
CHAPTER II
• Discussion of tag power consumption and methods to improve tag range and reliability.
Radio-frequency Identification (RFID) has traditionally been utilized for logistical pur-
poses as a replacement for bar codes. The ability to simultaneously read data from hundreds
of RFID tags in a few seconds without the need for visual contact enormously decreases the
costs and increases the reliability associated with logistical tracking. As RFID technology
has matured into the market, it has found uses not only in logistical tracking and inventory,
but also for toll collection, employee identification, and ticketing, among others [32]. As
this technology has spread, economies of scale have taken over, greatly reducing the cost
of RFID hardware. Along with this price reduction, advances in semiconductor technology
sensors over traditional sensors is their low cost and ability to operate maintenance free.
These advantages are possible because of the use of backscattered radio waves.
It is important to note that the term RFID has become synonymous for a variety of
technologies. Typically, they revolve around an electronic means of identification, but the
tags. This new terminology may also involve hardware which communicates via backscatter
11
radio, the same method which semi-passive and passive RFID tags use. Therefore, the term
RFID may be used to describe not only traditional identification devices, but also new
RFID tags typically use one of two methods for communication: magnetic or induc-
tive coupling and backscatter. Magnetically coupled systems typically operate at LF and
HF at 135 kHz and 13.56 MHz, respectively. These systems use the same terminology as
backscatter RFID systems, but have a shorter range as magnetic coupling is a near-field
phenomena. Usually, this range is limited to the greatest dimension of the transmit an-
tenna [32]. Backscatter radio systems, which will be discussed henceforth, rely on far-field
and thus have a greater range. The rest of this thesis will concentrate on backscatter radio
systems.
RFID systems are comprised of two parts: the reader (also known as an interrogator or
base station) and the tag [31]. RFID systems differ from conventional radio systems in that
they use backscattered radio waves to communicate. While traditional radio systems have
active transmitters on both the mobile station and the base station, RFID systems only
have one active transmitter at the reader. This reader typically broadcasts a continuous
wave (CW) signal which travels to the mobile station (RFID tag) as shown in Figure 1. The
tag modulates its data onto this signal by changing its radar cross-section (RCS) by means
of switching its antenna load. This modulated signal then returns to the reader where it is
In the United States, RFID tags typically operate in Federal Communication Commis-
sion (FCC) designated, unlicensed Industrial, Scientific, and Medical (ISM) communication
bands at 433/915 MHz, 2.4 GHz, and more recently 5.8 GHz [31]. Table 2 lists some common
parameters for the ISM bands in the United States including maximum power transmission
and effective isotropic radiated power (EIRP). Similar frequency bands exist elsewhere in
12
Figure 1: A typical semi-active/passive RFID system communicates via backscatter radio.
A reader broadcasts a CW signal to the environment where RF tags then backscatter this
signal back to the reader. The reader then must process the received signal and demodulate
the tag data.
the world, but exact regulations tend to vary by country. Note that inductively-coupled
tags operating at 135 kHz and 13.56 MHz are absent from this list.
At the time this thesis was written, 915 MHz UHF RFID systems offer the largest range
for the lowest cost of any passive RFID system. Therefore, these systems are among the most
widely commercially adopted. However, many additional benefits at higher frequencies exist
including smaller footprints, additional bandwidth, and a larger permissible EIRP, among
others [51]. Thus, it is possible to meet and even exceed the limitations of the additional
Besides the frequency and transmit power standards which are set by regulatory agencies,
RFID systems follow a set of communication standards which set various protocols and
13
modulations. The most well-known of these standards is the EPC Class-1 Generation-2 UHF
RFID (also simply known as ‘Gen2’) [67]. There are other ‘low-power’ standards that are
designed for WSNs including IEEE 802.15.4, Bluetooth, ZigBee, and many other proprietary
standards. However, the Gen2 standard is used by the majority of UHF RFID tags, which is
the de facto standard for inventory and logistics tags. While the specifics of this standard are
outside the scope of this discussion, it is important to note that the communication protocol
affects the power consumption of the tag. For example, when two-way communication
between the tag and the reader is desired (in send-and-acknowledge schemes such as in
the EPC standard), the tag must use an ADC so it may sample and decode the incoming
reader command. Not only must the tag power an additional peripheral, it must also remain
Protocols are designed in part to avoid packet collisions, to increase security, and to
ensure maximum spectral efficiency. Removing or simplifying them may not always be a
ity must be made when the protocol is selected. In some applications, such as building
temperature monitoring, packet error rate may not be as important as low-cost sensors.
Unlike typical radio communication systems, in RFID systems there are two different radio
channels. The forward channel expresses the radio environment from the reader to the tag
while the backscatter channel describes the environment from the tag back to the reader.
While each channel environment may be modeled identically, the statistics of the channel
fading will change since the radio wave will pass through two unique channels, not necessarily
the same one twice. These channels can be modeled according to the Friis Transmission
formula. However in backscatter, each channel will include other terms not typically used
The forward channel or ‘power-up’ link budget is expressed in (1). This log-scale link
budget is called power-up because this equation models the amount of energy which is
available at the output of the tag antenna. For battery-free tags, this is the maximum
14
power available to power-up the tag.
4πrf
Pt = PT + GT + Gt − 20 log + Xf − Θ − Bf − FP , (1)
λ
where Pt is the received power at the RF tag, PT is the transmitted power by the base
station, GT and Gt are the gains of the reader transmitter and tag antennas, respectively,
rf is the forward-link distance between the RFID transmitter and the RF tag, λ is the free-
space wavelength, Xf is the polarization mismatch between the RFID transmitter and the
RF tag antennas, Θ is the RF-tag antenna on-object gain penalty, Bf is the forward-link
path blockage loss, and FP is the power-up fade margin, typically chosen as Rayleigh or
Rician.
On the backscatter link, a similar link budget is used, but several factors are counted
twice and others are added. These additional terms are largely due to the fact that the
forward and backscatter signal paths do not need to be symmetrical. The backscatter link
where PR is the received power at the RFID reader, GR is the gain of the RFID reader
antenna, rb is the backscatter-link distance between the transmitter and receiver, Xb is the
polarization mismatch between the RF tag and RFID reader receiver antennas, M is the
modulation factor, Bb is the backscatter-link path blockage loss, and Fb is the backscat-
ter fade margin, typically chosen as a product Rayleigh or product Rician. An in-depth
RFID readers typically come in two forms. In mono-static systems, the reader has a single
antenna which is used for both transmitting and receiving. It accomplishes this configura-
tion via an RF circulator. However, poor isolation in the circulator can lead to increased
interference in the receiver because of transmit leakage. Bi-static systems eliminate the
need for a circulator and better isolate the transmit and receive paths through the use of
15
two antennas. By having independent antennas, the isolation, and therefore sensitivity, can
RFID readers must demodulate backscattered signals from tags in the presence of a
variety of noise and interference sources. Received backscatter signals typically have very
low amplitudes as their path loss increases with the fourth-power of distance. In addition
to common thermal and flicker noise sources, RFID readers have the added disadvantage
transmitted energy leaking directly into the receive chain as well as signals which reflect off
of background clutter. These interferers can be many orders of magnitude greater than the
signal level of the backscatter signals. Thus, readers typically use a direct down-conversion
(homodyne) receiver architecture with AC coupling as shown in Figure 2. The use of this
architecture allows the use of a single RF oscillator to down convert and effectively eliminate
the CW power in the RF carrier, including the high-power unmodulated backscatter and
self-interference. This fully coherent receiver can also be used to make very sensitive phase
measurements. This topology also allows the reader to transmit and receive continuously
in full-duplex mode.
16
As part of the homodyne receiver architecture with AC coupling, capacitors are required
before the signal(s) can be sampled by an ADC to eliminate any DC component as a result
form a high-pass, RC filter with a time constant controlled by the value of the capacitor.
This value must be large enough to remove any strong, unmodulated signals, but small
enough to keep the data signal itself from decaying. To improve the performance of RFID
systems, which employ direct down-conversion receivers, the RF tag modulation may be
RF tags can be active, semi-passive, or passive. Active tags require an on-board battery
which powers the tag electronics and their active transmitter. They do not communicate
via backscatter and operate like a traditional radio communication system. Due to their
active transmitter, active RFID tags act like a transponder, and therefore typically have
the largest ranges and fastest data rates. However, these advantages come at the cost of
increased price, size, and complexity due to the necessity of a battery and on-board RF
hardware. Semi-passive tags also use a battery, but only to power the tag electronics when
within reader range. All communication is done via passive backscatter, which eliminates
the necessity of complex RF hardware. Passive tags use no battery and harvest all of their
energy from signals from the reader or some other source. The reader-delivered signal is
used for both backscatter communication and energy harvesting. Due to the need to harvest
their energy, passive tags typically have the shortest ranges and lowest data rates. However,
their simplicity, and thereby low cost, makes them ideal for a variety of applications.
Active and semi-passive tags are generally backscatter-link limited. This means that
their maximum range is governed by the sensitivity of the receiver. It must be able to
detect the extremely weak backscattered signal in the presence of non-linear reflections,
hand, passive RF tags are generally forward-link or power-up link limited. This descrip-
tion means that the maximum range at which a tag can operate is limited by the voltage
17
available at the tag antenna. When the voltage fails to reach a specified value, the tag does
not have sufficient energy to power-up. This failure occurs because of a limitation in the
electronic circuit utilized for energy harvesting. Unless the incident signal produces a volt-
age level greater than the threshold voltage of the diodes in the energy harvesting circuit,
current will not be able to flow, and the RF tag will not have any energy for processing or
communication.
Passive or battery-free RFID tags are an attractive option for WSNs and RFID applications
due to the fact that passive RFIDs do not utilize a battery for power storage. Instead,
passive tags must harvest energy from some other source. Options can include solar [48],
motion or vibration [49], ambient RF [50], or most commonly, an RF signal generated by the
RFID reader. RFID tags typically rely on their readers for energy as the reader presence
usually signifies a desire to read the tag, therefore a need for energy. However, in some
WSN applications, additional methods of energy harvesting could be required if the tag is
expected to be removed from the presence of a reader for a long period of time.
Passive RFID tags are unique communication devices in that they do not have complex
much simpler RF front end and no battery while allowing the tags to be small, low-cost,
and power efficient. Figure 3 shows the main components of a semi-passive/passive RFID
tag.
controlled by a piece of digital logic which stores the tag identification information, receives
and processes reader commands, interrogates sensors (if applicable), and backscatters in-
formation back to the reader. While commercial solutions utilize custom ASICs for this
digital logic, low-yield applications in the research field tend to use microcontrollers [41],
GAs) [39]. Many of these devices tend to have integrated oscillators, timers, comparators,
18
Figure 3: This block diagram shows the main components of a semi-passive/passive RFID
tag. Some form of digital logic samples sensors, receives reader commands, and controls a
load modulator to backscatter data back to a reader. An energy harvester provides power
c
for all of these devices (Reprinted from [5],
M.S. Trotter).
ADCs, DACs, and digital communication buses, thereby allowing for a single integrated
One or more antennas are used by the tag for backscatter communication, receiving
reader commands, and for energy harvesting. This/These antenna(s) can be linearly or
circularly polarized depending on their application and are often designed to be as omni-
directional as possible. This uniform coverage allows the tag to communicate with the reader
regardless of its orientation. However, if the tag position can be controlled or is known a
priori, a directional antenna with much greater gain can be used. If a single antenna or
and envelope detector which all affect the load on the antenna, an important consideration
so that the tag can communicate with the reader. It is controlled by digital logic which sends
the encoded data to the modulator. Typically, this load modulator is a simple transistor
switch which changes between two different impedances. This choice limits the tag to two-
state ASK or PSK modulation schemes. While higher-order modulation schemes have been
19
demonstrated by [71, 72], low-order modulations are typically used because of the low SNR
at the receiver. In the event that a single antenna is used, one impedance state may be the
energy harvester input impedance while the other state would be a short or open circuit.
When this occurs, the modulation index of the tag is reduced which decreases the differential
backscattered power to the reader [73]. However, a separate antenna could be used for both
the modulator and energy harvester to maximize the differential backscattered power levels.
An energy harvester and power management circuitry are responsible for collecting suf-
ficient energy to power the tag and any associated sensors. As previously mentioned, this
energy can come from a variety of sources but is typically reader-delivered RF power. Ad-
sample the received waveform from the reader. This waveform can then be decoded into a
reader command, and the tag can operate accordingly. It is important that the capacitors
in the envelope detector be sized appropriately to the reader command chip rate so that an
Associated sensors in the tag are powered by the energy harvester and interrogated by
the digital logic. These sensors may measure quantities such as temperature, pressure, or
gas concentration among others. Several researchers have also integrated sensing into the
Passive RFID tags typically rely solely on RF power from their readers for all of their
energy. Energy harvesting circuits such as rectennas or charge pumps convert RF energy
to DC energy that is able to power the tag electronics. However, these circuits suffer from
inefficiencies and behave non-linearly. Equation (3) follows from (1) and gives the DC power
where PtagDC is the DC power at the output of the energy harvester, LP C is the energy
20
harvester RF-to-DC conversion gain which is a function of the tag received RF power Pt as
defined in (4) [75, 76]. All other terms follow from (1).
PtagDC
LP C = 10 log10 (ηRF −DC ) = 10 log10 , (4)
Pt
where ηRF −DC is the RF-to-DC conversion efficiency of the energy harvesting circuit. The
largest value of LP C will be zero. This number will occur only if the energy-harvesting
In the absence of multi-path interference, as the tag moves farther from the reader, the
close distances when PtagDC ≥ Ptask , where Ptask is the power requirement to complete a
task on-board a tag, the data throughput is said to be protocol limited [39]. The tag can
continuously power all on-board sensors, receive commands from the reader, and perform
necessary communications. The only tag limitation is the amount of data it can deliver
to the reader as determined by the protocol. However, many times the power available is
insufficient for continuous operation. In this case when PtagDC < Ptask , the tag is said to
be power limited. Energy must be stored in an energy storage device, typically a capacitor,
until enough energy exists to complete a task. The tag communication rate is thereby
reduced. Under these circumstances, a capacitor or battery must charge until the stored
energy Estored is greater than or equal to the task energy Etask according to (5) and (6).
1 2 2
Estored = C(Vtag − Vdd ) (5)
2 DC
Z T0 +Ttask Z T0 +Ttask
Etask = Ptask dt = Vdd Itask dt, (6)
T0 T0
where C is the value of the energy harvesting circuit output capacitor, Vdd is the voltage
level at which the circuit logic operates, Ttask is the time required for the task, and Itask is
the time varying current draw of the logic during the task. Equation (5) describes the usable
tag energy stored for the circuit to operate. Since the logic requires a minimum voltage Vdd
to function, the voltage level must be greater than this value. Only voltage overhead, the
21
available voltage above Vdd , can be used for tag operations. Equation (6) gives the energy
level required for a tag to complete a specific task. This equation is simply the power
consumption of the tag multiplied by the time it takes to finish its task. Rearranging (5)
and solving for VtagDC gives the minimum voltage required at the output of the energy
When the tag moves further away from the reader and VtagDC M IN < VtagDC , it is said
to be voltage limited and will not power on. Some energy harvesting circuits will step up
the voltage that is received by the tag. In this case, VtagDC M IN may fail to be large enough
because received power levels are too low for the energy harvester to charge at all or power
levels are too intermittent and leakage currents deplete available charge before large enough
levels can be reached. Therefore, VtagDC M IN sets the maximum operating range for a given
Since the tag power level can vary with distance from the reader and multi-path interference,
the tag must have a mechanism to detect whether or not the current energy level is sufficient
for operation. This mechanism must be able to detect the energy level and be able to wake
up the digital logic when sufficient energy is present. Most often when waiting for energy
accumulation, the tag remains in an ultra-low power state with its logic circuitry and the
majority of its peripherals powered down. In this state, the tag waits for the energy harvester
to charge with enough energy so that it can power on and complete its task.
Voltage supervisor circuits are used to monitor the charge accumulation on the energy
harvesting circuit output capacitor by measuring the voltage on the capacitor VtagDC . When
this voltage reaches VtagDC M IN , this implementation sends the main logic a wakeup signal
to tell it that there is sufficient energy stored to complete its task. An ADC may also
be used to measure this voltage, but this method requires additional logic and time and
22
While digital logic can operate over a range of voltages, there is typically a preferred
voltage level from a power consumption point of view. Furthermore, input voltage swings
can lead to variation in on-chip oscillators and affect the device timing and stability. Thus,
voltage regulators are used to supply a steady voltage level for the RF tag. It is important to
note this as the voltage regulator together with the voltage supervisor are the circuits from
which the energy harvester load and consequently conversion efficiency are determined.
Figure 4 shows an example of the power management circuitry for a passive RF tag.
Output current from the energy harvesting circuitry builds up charge on an output capacitor
and supplies power to a voltage supervisor and voltage regulator. Simultaneously, the
voltage supervisor monitors the output capacitor voltage. The regulator provides power to
the tag logic and associated peripherals, which are powered down or in a low power state
while the output capacitor is charging. Finally, a zener diode acts as overvoltage protection
Figure 4: The power management circuitry for an RF tag must regulate and measure the
energy harvesting circuit output voltage. When this voltage reaches a predetermined level,
it will wake up the tag logic. Furthermore, this circuitry sets the load conditions for the
energy harvesting circuit which affects its RF-to-DC conversion efficiency.
The limited power harvested by a passive RF tag must be used for a variety of tasks both
tags can vary, some of the basic components which must have their power consumption
23
taken into account are listed below. Care must be taken to minimize the current draw of
• Baseband logic/Microcontroller
– CPU
– Oscillator(s)
– Analog-to-digital converter
– SPI/I2C digital communications
– Timers
• RF switch
• Voltage regulator
• Power Management
• Sensor(s)
tocols and tag control, it is important to consider the power consumption of the software
power consumption can drastically vary. Some major tag operations are listed below.
• Initialization
• Sensor interrogation
• Packet formation
• CRC calculation
• Coding/Modulation
• Backscattering
24
2.6 Increasing the Range and Reliability of RF Tags
Passive RFID tags and RFID-enabled sensors are typically forward-link limited due to the
lack of power available to the tag. When this minimal power level is not met, the tag will
fail to power up and communication cannot occur. Several items can be addressed to extend
this forward-link limit and thereby increase the range and reliability of RF tags. Many of
these will be application specific and have impacts on the system performance.
Items in the power-up link budget from (1) can be adjusted to increase the tag received
power Pt and therefore the available DC power PtagDC . If these levels are increased, the
tag maximum operating range can be extended. First, while increasing the transmit power
PT and the reader transmit antenna gain GT will increase range, the EIRP is typically
government regulated. Second, the tag antenna gain Gt can potentially be increased at the
expense of coverage. However, tag orientation may be uncontrollable. Thus, this increase
will depend on the application. Third, moving to a lower frequency will also increase the
power received by the RF tag, but the available frequencies are limited by regulating agen-
cies. The antenna sizes required to efficiently collect the RF energy at these low frequencies
may also be impractical. Furthermore, lower frequencies typically have less bandwidth
available and more congestion which may lead to communication challenges. Finally, if the
tag location is controllable, the polarization mismatch Xf , on-object gain penalty Θ, and
path blockage Bf can be reduced to zero. If it is not, then appropriate statistical models
25
must be made to ensure the tag will function within an accepted probability of error. This
analysis could be similar to the one used to estimate the power-up fade margin, which is
The duty cycle of the tag in its on state could be reduced. This change could come in the
form of lowering the number of samples taken by a sensor or lowering the number of times
a tag communicates with the reader. By reducing the need to perform calculations and/or
sample sensors, RF tag task energy Etask could be lowered. Following from (7), if Etask is
Similar to decreased duty cycle, if the communication protocol is simplified, the amount of
energy required by the tag is reduced. Complex modulations and encoding all require an
increased number of clock cycles on the digital logic. By minimizing the number of clock
cycles, the power requirements of the tag can be reduced. However, the application may not
tolerate the lack of security or high probability of error associated with eliminating some of
It is well known that ASICs offer superior efficiency and lower operating voltages compared
to discrete designs. Discrete semiconductors tend to have high losses associated with their
into improving these methods would help to decrease these losses and improve the overall
be expected to have a larger range. However, the high cost and inflexibility with an ASIC
Choosing the most efficient energy harvesting circuit design is critical to obtaining the high-
est RF-to-DC conversion efficiency possible. The previously mentioned techniques all focus
26
on obtaining the highest amount of RF energy available at the antenna output or reducing
the tag power consumption. However, this technique focuses on efficiently converting the
received power to usable DC energy. There are a variety of energy harvesting circuit designs
in literature, but few attempts at comparing them. However, there are certainly tradeoffs in
the number of stages in charge pumps, the diodes chosen, and the values of the capacitors
Changing the reader delivered waveform that powers the tag from the standard continuous
waveform (CW) to a high peak-to-average power ratio (PAPR) waveform such as a POW
or multi-sine can increase the power available to the RFID tag. POWs and multi-sines can
high instantaneous power helps to increase the RF-to-DC conversion efficiency of energy
27
CHAPTER III
Wireless power transfer (WPT) systems have attracted significant attention as elec-
tronics have become more and more integral into our daily lives. The need to power these
electronics has proliferated from the desire for an internet-of-things, the desire for cable-free
consumer electronics, and the desire for a reduction in the use of copper, among others.
While several methods of wireless power have been introduced, including magnetic reso-
nance [77], inductively coupled devices [31], and RF/microwave energy transmission [23, 50],
only RF/microwave energy transmission is a truly long-range method. As RFID and RFID-
enabled sensors aim to operate at appreciable distances, this dissertation will only focus on
RF/microwave energy transmission WPT systems have several core components which al-
low energy to flow between two points in space. These core components are illustrated in
Figure 5. First, RF/microwave power must be generated at the base station via a mag-
netron or solid state-source. This choice is usually dominated by efficiency, cost, desired
28
transmit power, and spectral clarity [16]. For most sensor-based applications, solid-state
transmitters are sufficient because of their low cost, clean spectral content, and appropriate
power generation. This transmitter connects to an antenna which may or may not have
a directional pattern. At this point, its important to note that this directivity and EIRP
After the base station-radiated energy propagates through the channel between the two
locations, the mobile node must capture it. The mobile node has an energy conversion
rectifying circuit, and low-pass filter. The band-pass filter helps to ensure that the antenna
is correctly matched to the rectifying circuit, and that harmonics are not re-radiated to
the environment. The rectifying circuit can exist in a variety of topologies which will be
discussed later in Section 3.4.. Finally, an output low-pass filter removes the fundamental
and harmonic frequencies from the output and stores charge for later device consumption.
While various energy-harvesting circuit topologies are possible to convert RF to DC, nearly
are usually able to handle relatively small amounts of power (from a SPS perspective),
and their low cost and small form factor makes them ideal for a variety of applications.
29
Schottky diodes are typically chosen because of their lower voltage threshold and junction
capacitance than PN diodes. This low threshold allows for more efficient operation at low
powers, and the low junction capacitance increases the maximum frequency at which the
diode can operate. In addition to Schottky diodes, diode connected transistors implemented
Diodes are typically modeled according to the small signal, non-linear relationship expressed
in (8)
where I(V ) is the diode current as a function of the diode voltage VD , IS is the diode
q
saturation current, and αD is the diode temperature-idealty factor where αD = nkT and q
is the charge of an electron (1.602 × 10−19 C), n is the diode idealty factor, k is Boltzmann’s
2
constant (1.381 × 10−23 ms2 K
kg
), and T is the temperature in Kelvin. These regions are
illustrated, in part, in Figure 6. This I-V curve is characterized by three major regions.
For low voltages below Vbr , the reverse breakdown voltage, the diode is said to be reverse
biased and will conduct in the reverse direction. Between Vbr and VT , the turn-on voltage,
the diode is off and only a very small amount of leakage current will flow. Above VT , the
Rectification occurs in the following way: Assuming no initial charge (Vo = 0) on the
output capacitor (not pictured), when an RF signal with a peak voltage Vp greater than VT
is across the diode, charge will flow through the diode and build up DC bias on the output
capacitor. Provided the input RF signal Vp and the diode DC bias VD,DC are larger than
the threshold voltage (Vp + VD,DC > VT ), the DC bias will continue to increase as additional
charge is transported through the diode. This case is illustrated in Figure 7c. Here, it is
obvious that the input signal level is greater than the voltage threshold, so the DC bias
will increase. Note that this DC bias is shown as negative as it is referenced to the diode
voltage VD .
30
Figure 6: A typical diode I-V curve with annotated breakdown and turn-on voltages. The
schematic on the left-hand side is assuming that all reactive elements have been tuned out
and a sufficient RF power exists across the diode to generate a DC voltage across the output
resistor RL .
As the diode continues to bias itself, one of two possibilities will occur. One case is
shown in Figure 7a, where the circuit will reach steady state. At this point, the DC bias
increases until the input voltage is no longer greater than the voltage threshold. As long
as the incident signal remains at the same power level, the DC bias will remain constant.
In the other possibility, the incident RF signal will eventually begin to clip Vbr which will
short some current to ground as illustrated in Figure 7b. At this point, the DC bias can
no longer increase even though the incident signal is above the voltage threshold. This
limitation occurs because the signal is also below the reverse breakdown threshold.
A similar behavior is exhibited when the incident voltage level decreases. When the
inequality Vp + VD,DC > VT is no longer satisfied as illustrated in Figure 7d, the DC bias
31
(b) Energy-harvesting circuit at maximum volt-
(a) Energy-harvesting circuit at steady-state.
age.
(c) Energy-harvesting circuit with increasing bias. (d) Energy-harvesting circuit with decreasing bias.
Figure 7: A diode DC bias shifts according to the input voltage level. 7a shows a steady-
state DC-bias level where the peak voltage of the input is equal to the turn-on threshold.
7b shows the maximum bias level, 7c shows a scenario where the DC bias will increase, and
7c shows a circumstance where the DC bias will decrease.
32
will begin to decrease. This decrease will continue until the incident voltage is able to
satisfy the equality, surpass the turn-on voltage, and resume current transport to maintain
a suitable DC bias.
Vbr
VDC,max = . (9)
2
Once the DC bias reaches this level, the input will swing both positive toward VT and
negative towards Vbr . Since current will short at voltages below Vbr , no additional charge
can bias VDC . Consequently, the maximum DC power than can be produced is
Vbr 2
2 Vbr 2
PDC,max = = . (10)
RL 4RL
Previously, it was assumed that only the fundamental/carrier frequency and DC existed
in the circuit analysis. However, additional frequency components are created due to the
non-linear diode response. For a small-signal approximation where the diode voltage V is
equal to
V = Vo + v, (11)
where Vo is the self DC bias across the diode and v is the small AC voltage, then the taylor
v2 0 v3
I(V ) = I0 + i ≈ I0 + vGd + Gd + G00d + ... (12)
2! 3!
1
where I0 = I(V0 ) is the DC bias current, Gd = αIs eαVo = Rj , G0d = αGd , G00d = α2 Gd and
Thus, assuming steady-state operation so that a self biasing DC voltage Vo is across the
diode and for a typical CW excitation with v = A cos(ωo t), the diode current is
33
This variation illustrates the harmonic products generated by a semiconductor-based
rectifying circuit. As can be observed in (13), a sinusoidal excitation at ω0 not only produces
a DC signal as desired, but also frequency components at ω0 , 2ω0 , 3ω0 , and so on. It is for
Energy-harvesting circuits are typically characterized by two different metrics, efficiency (or
sensitivity1 ) and output ripple, to evaluate their performance. Efficiency can be expressed as
the desired power PoutDC is the DC voltage VoutDC across a load resistance RL defined
in (14)
2
VoutDC
PoutDC = . (14)
RL
This output power represents the useful power that can be provided to the device circuitry.
If the received power at the input to the energy-harvesting circuit is defined as PinEH , then
where Pref lected is the power reflected from the energy-harvesting circuit due to mismatches
harvesting circuit and rather focuses on the intrinsic ability of the energy harvester to
convert RF power to DC. Note that the largest ηEH can ever be is ηP C . This instance
would only occur if the antenna is perfectly matched to the energy-harvesting circuit at the
1
While related to energy-harvester efficiency, sensitivity is defined as the minimum power necessary to
power an integrated circuit. Inherently, this definition depends on the IC technology and application as
different sensors and wireless protocols can cause the sensitivity number to be larger. As many RFID
researchers operate in this semiconductor regime, they often cite sensitivity as their evaluation metric.
34
Voltage ripple is another important metric when discussing energy-harvesting circuit
where Vripple is the ripple voltage (expressed here as a percent of DC) and Ṽωi is the
possible for the ripple voltage to exceed 100% for low power levels when DC levels may be
While highly efficient energy harvesters are desirable, a variety of loss mechanisms make it
difficult to achieve high efficiencies, especially in non-linear devices such as the diodes and
FETs which are used as rectifying elements. Figure 8 helps to illustrate the relationship
between some of these losses and the energy-harvesting efficiency. The four main energy-
• Impedance matching
• Device parasitics
• Harmonic generation
Typically the most important parameter for diode efficiency is the threshold or turn-on
voltage VT . This parameter limits efficiency at low powers as noted by ‘VT Effect’ in Figure 8.
Unless sufficient power is incident on the energy harvester, there is not enough energy to
overcome this barrier and charge the output capacitor. Diode reverse breakdown voltage
Vbr also limits diode efficiency as this will allow energy to short the diode demonstrated by
the curve ‘Vbr Effect.’ This situation will occur at larger power levels when the diode DC
35
Figure 8: The general relationship between the efficiency and losses in microwave energy
conversion circuits as a function of input power (Modified from [6]). At low powers, the
efficiency is limited by the input signal ability to surpass the diode turn-on threshold voltage.
As the power continues to increase, the efficiency will increase along with the effect of higher
order harmonics. At higher powers, the diode reverse breakdown voltage limits the efficiency.
A maximum conversion efficiency occurs between these points.
bias is equal to half the breakdown voltage as previously discussed. Note that while the
efficiency curve decreases above this point, DC output power will remain constant.
When an energy-harvesting circuit is not properly matched to its antenna, part of the
incident power from the antenna will be reflected back to the environment and not absorbed.
This reflection decreases the available power for rectification. Matching is especially difficult
Diode junction resistance Rj and junction capacitance Cj are due to the metal-semiconductor
junction and change as a function of input power. Additional static reactive parasitics ex-
ist, but they are assumed to be tuned out with matching circuitry. Therefore, matching
not only becomes important as a function of frequency as with typical microwave circuits
due to reactive elements, but also with power due to the variable junction resistance and
36
Figure 9: A standard Schottky diode model. A resistance Rs is in series with a variable
junction resistance Rj in parallel with a variable junction capacitance Cj which change as
a function of input power.
capacitance.
Figure 10 shows an example of the wide range of impedance values that an energy-
harvesting circuit can experience. This single shunt rectenna was designed to operate at
input impedance. Only a simple open-circuited stub was used for matching. Additional
details for the design and simulation of this circuit can be found in Appendix E.1. In this
simulation, the input power was varied from -30 dBm to 30 dBm, typical input power for
an RFID tag; and the frequency was varied from 5.725 GHz to 5.875 GHz, an ISM band. It
would be difficult to match this entire range of impedances, thus an energy harvester must
be designed to operate with a given power and frequency typically specified by the intended
application.
Device parasitics can also cause a significant decrease in efficiency. Figure 9 and Figure 11
show a diode model and SOT-23 package parasitic model, respectively. A variety of packages
exist with varying numbers of diodes in that package. A full energy-harvesting circuit design
will have to decide on the number of diodes in the respective package as this will all affect
37
Figure 10: Impedance of single shunt rectenna using an Avago HSMS-2860 diode with a
316 Ω load matched at 5.8 GHz and 16 dBm input power. The large range of possible
impedances makes it difficult for an energy-harvesting circuit to work efficiently over a wide
range of input powers and frequencies because of the reflected power due to impedance
mismatch [7].
through the diode will dissipate power in the semiconductor junction [6]. Diode junc-
tion capacitance typically limits the maximum frequency at which the diode can operate
as high frequency current will short through this parasitic. Additional package parasitics
are defined as lead frame inductance LL , lead frame capacitance CL , package capacitance
Cp , coupling capacitance Cc , and bond wire inductance LB . Each of these values can cause
second or third order resonances which will decrease energy conversion efficiency. Further-
more, traditional losses in the substrate and transmission lines may also play a significant
role in the energy harvester efficiency depending on the substrate and line lengths chosen.
38
Figure 11: Parasitics in a SOT-23 package [8]. These parasitic values cause additional
impedance mismatch which can translate to additional power lost due to reflections. Note
that this package can be outfitted with one or two (shown in dotted lines) diodes.
While providing a means to convey RF energy to DC, the diode non-linearity is also a
source of loss. When operating, the diode will produce frequency harmonics from the inci-
dent power, which reduces the proportion of energy which gets converted to DC resulting in
a lower energy harvester efficiency. As the incident voltage continues to increase, the energy
lost to harmonics further increases. An optimal efficiency level corresponds to tradeoffs be-
tween harmonic generation, reverse breakdown effects, and threshold voltage. Additionally,
these diodes can create non-linear reflections which can corrupt the tag backscattered signal
As the diode is the primary component associated with energy rectification, it is important
to choose an appropriately designed diode for maximum efficiency given expected operating
conditions. However, as there are numerous types, vendors, and packages of these diodes,
it is often difficult to make a selection. The following section will help to describe some
diode semiconductor fundamentals and help to illustrate how these pieces can impact energy
harvester performance.
As mentioned before, Schottky diodes are typically chosen because of their low turn-
on threshold levels. This low level helps the diode to be more sensitive to low power
39
levels than a similar PN diode. While the basic diode equation was mentioned in (8), an
explanation of some of the more important parameters, along with their interdependencies,
will be discussed. Some of the most important diode parameters for microwave energy-
harvesting efficiency are the junction capacitance Cj , series resistance Rs , turn-on voltage
The junction capacitance Cj is typically associated with limiting the maximum operation
frequency of the diode. For microwave applications, this value is typically less than 3 pF
the charge separation. This equation follows from a parallel plate capacitor model where
the charge separation is defined in [79]. Note that this value will change as a function of
The semiconductor bulk series resistance Rs is usually the limiting factor in the diode energy
conversion efficiency at high powers [6]. This resistance is due to the inability of charges to
semiconductor [80].
The threshold, built-in, or turn-on voltage of a Schottky diode is a function of the metal
40
where ΦB is the surface potential-energy barrier height of the metal, and (Ec − EF )|F B is
the energy difference between the conductor band Ec and Fermi band EF under flat band
conditions. Thus, ΦB can be varied by changing the metal in the Schottky diode, and Ec
and EF can be changed with the choice of alternate semiconductors and doping. This value
limits the energy conversion efficiency at low powers as charge cannot be pumped over this
barrier voltage.
As mentioned in Section 3.2.1, the breakdown voltage limits the maximum DC output power
1
Vbr α , (21)
NB
−ΦB
Is = AAR T 2 e kT , (22)
where AR is the Richardson constant, a material and temperature dependent quantity. All
other parameters are as previously described. A high saturation current allows the diode
to more easily conduct at low-power levels [81]. This property is typically associated with
p-type diodes, which have a low barrier height. However, this also corresponds to a larger
series resistance.
The choice of the diode is one of the most important design decisions in energy-harvesting
circuits. While off-the-shelf components are often used, custom diode design may be con-
sidered for maximum performance. While a diode with an infinite breakdown voltage would
be helpful for high power applications, tradeoffs in other important diode parameters would
41
• Increase the doping ND - When the semiconductor is further doped, this will increase
the junction capacitance, reverse breakdown, and turn-on voltage, and decrease the
series resistance. Consequently, the maximum efficiency would increase, but would
occur at a higher power. Additionally, the maximum DC voltage would increase, and
• Increase the charge separation area A - By increasing the size of the charge separation
area, the capacitance and saturation current will increase and the series resistance will
decrease. Consequently, the efficiency would increase, but the maximum frequency
• Increase the metal barrier height ΦB - Since ΦB is a material parameter, the only way
to change this value is to physically change the choice of metal in the Schottky diode.
When this occurs, the threshold voltage will increase and the saturation current will
decrease. These changes mean a larger input signal would be required to turn on the
• Use a p-type Schottky diode - The use of a p-type diode typically means that the
saturation current Is will be larger than when compared to an n-type. This larger
saturation current means that a larger voltage can be generated for a small input
power level. However, the performance of a p-type diode is hampered at high input
As a consequence of these choices, it can be seen that the microwave performance of the
Schottky diode will be lower than for UHF bands. This result is due to the need for a lower
capacitance Cj . As this capacitance is lowered, the series resistance increases which will
increase the energy conversion efficiency at high powers. This analysis has neglected the
could affect the turn-on voltage and consequently could improve the capacitance while
keeping the series resistance low. A full analysis is outside the scope of this thesis, but
42
3.3.3 Diode Parameter Behavior
5.8 GHz. Each subplot shows the variation in performance as a single parameter changes
and all other parameters are held constant according to the values the table inset in the
figure and using a load resistance RL = 327 Ω. In [6], Yoo and Chang derive the maximum
energy conversion efficiency of a single diode under CW excitation with a resistive load.
Appendix B.1 shows this derivation in detail along with the addition of the effects of the
reverse breakdown voltage. This derivation assumes no reactive parasitic elements and no
harmonic generation. The two loss factors that are considered are the diode built-in voltage
VT and the diode series resistance Rs . Thus, actual conversion efficiencies will be lower
this analysis extends to find the input impedance of the diode. This method can assist when
Figure 12 shows several interesting trends. Figure 12a demonstrates that a low turn-on
threshold is required for efficient low-power operation. As the turn-on threshold voltage
is decreased, the energy conversion efficiency at a given power increases. This phenomena
occurs because the same power level can more easily overcome the turn-on voltage limita-
tion. While this may be a desirable feature, there are other tradeoffs in obtaining this low
threshold, such as reduced breakdown voltage, as mentioned before. Equations (18) - (22)
can help show the tradeoffs in changing some of the basic diode parameters.
Figure 12b shows that a large reverse breakdown voltage allows a larger allowable in-
put/output power as noted by (9). As the reverse breakdown voltage increases, the maxi-
mum output power level, increases while conversion efficiencies at lower power levels remains
the same.
Figure 12c shows that the junction capacitance limits the maximum frequency for which
a diode can operate. As shown, 2.4 GHz and 5.8 GHz have approximately the same energy
conversion efficiency because the 0.18 pF junction capacitance along with the 6 Ω series
resistance does not filter out much power at these frequencies. However, when the frequency
43
1 1
Vf=0.004 V Vbr=3 V
Vf=0.04 V Vbr=7 V
Energy Harvesting Efficiency
0.6 0.6
0.4 0.4
0.2 0.2
0 0
−50 −40 −30 −20 −10 0 10 20 30 −30 −20 −10 0 10 20 30
Input Power (dBm) Input Power (dBm)
0.6 0.6
0 Cj , and the diode series resistance0 Rs . Thus, actual conversion efficiencies will be lower because of
−30 −20 −10 0 10 20 30 −30 −20 −10 0 10 20 30
Input Power (dBm) Input Power (dBm)
generation, parasitic elements, and impedance mismatch.
(c) Variation of input frequency. (d) Variation of series resistance Rs .
1
RL=100 Ω TABLE I: Avago HSMS286x device parameters [10].
RL=327 Ω
Energy Harvesting Efficiency
0.8 RL=1000 Ω
RL=5000 Ω
Parameter Value
0.6
Vt 0.4 V
0.4
Vbr 7V
0.2 Rs 6⌦
0 Cj 0.18 pF
−30 −20 −10 0 10 20 30
Input Power (dBm)
efficiency decreases. This result is due to normal resistive losses in the series resistance.
Finally, Figure 12e demonstrates the energy conversion efficiency dependence on the load
resistance. As this resistance increases, the efficiency increases, but the input at which the
maximum efficiency occurs decreases. This decrease in maximum power is due to the larger
resistance producing the same voltage for a smaller current. The increase in efficiency is
circuit is limited by impedance matching, device parasitics, and harmonic generation. Using
the same method outlined in Appendix B.1, the efficiency of an ideal diode is plotted in
(to permit maximum efficiency as a function of power) and a 300 Ω load. Various values of r
Rs
are also plotted r = R L
. In this case, an ideal diode is defined as one that has an infinite
breakdown voltage and zero turn-on voltage. All other assumptions from this derivation
hold as well. Actual diode performance will typically be less as additional loss mechanisms
will be encountered; breakdown and turn-on voltages will cause additional losses.
Two main points can be seen in Figure 13. First, a frequency selective behavior is
exhibited by the efficiency. This behavior is due to the diode junction series resistance Rs
functioning in series with the diode junction capacitance Cj shown in the diode equivalent
the low-pass filter response of this circuit. Secondly, it can be seen that for small values
larger efficiency is a result of a lower loss due to this junction series resistance. However,
this lower resistance also means that for a fixed junction capacitance, the diode is only
efficient at low frequencies. Therefore, there will be a tradeoff in diode performance at high
frequencies assuming a constant junction capacitance. Highly efficient diodes are possible,
45
Maximum Ideal Diode Efficiency, R =300 Ω, V =4 V
L o
1
r=0.0001
0.9 r=0.001
r=0.01
0.8 r=0.1
0.7
0.6
Efficiency
0.5
0.4
0.3
0.2
0.1
0 −3 −2 −1 0
10 10 10 10
ω Rs Ceff
but this performance will occur at low frequencies. High frequency operation will require
a larger junction series resistance which will degrade the maximum possible efficiency. A
deeper relationship between these parameters can be investigated using the semiconductor
equations (18) and (19) where physical dimensions, semiconductor dielectric, doping, and
Tables 12, 13, 14, and 15 in Appendix A show this data in tabular form and also contain
additional information regarding the type of rectifying element and citation. The device
46
efficiencies shown use a variety of topologies and matching methods and use varying loads
and fabrication processes. This graph and tables present efficiencies for single frequency
excitation only; they do not include efficiencies for multi-frequency excitation waveforms.
It should be noted that when excitation signals are designed, energy harvester efficiency
100
80
Efficiency (%)
60
40
900 MHz
20
2.4 GHz
5.8 GHz
Above 5.8 GHz
0
−30 −20 −10 0 10 20 30 40 50 60 70
Input Power (dBm)
A peculiarity discerned from the graph along with Table 12 is that UHF charge pumps are
typically designed for low-power operation below 0 dBm. This design choice is a result of
UHF energy harvester research being driven by the RFID community that typically operates
in the 865 - 928 MHz regions. As passive RFID tags must scavenge energy in multi-path
environments far away from their readers, available energy levels are very low. Consequently,
the reported energy levels along with the reported energy harvester efficiencies in the figure
correspond to power level cutoffs around approximately -20 dBm. Also noteworthy, to
accommodate sufficiently high voltage levels to power digital electronics at these very low
47
power levels, UHF energy harvesters have utilized multi-stage rectifying circuits, such as
This low-power operation is possible due to the heavy use of CMOS technology as
demonstrated in Table 12. Not only do CMOS processes allow custom-built electronics,
which are more efficient and require lower operating voltage than microcontrollers or other
external digital devices, but custom rectifying elements are possible. While most standard
CMOS processes are incompatible with Schottky diodes due to concerns over cost, some
UHF energy harvester designs have leveraged CMOS-based Schottky diodes [82, 83]. More
commonly, diode connected transistors are used in these topologies. Diode connected tran-
sistors can be designed to have a near-zero threshold (‘zero threshold’ diodes or transistors
are typically those with a forward voltage level of less than 150 mV) which makes them
ideal for ultra-low power energy harvesting. While this threshold can be adjusted by ap-
propriately varying typical transistor properties such as the gate length and width and/or
doping level, different topologies have been suggested to further lower the threshold voltage
to increase the harvester efficiency at low power levels. Most of these techniques involve
some method of providing additional voltage bias which effectively reduces the threshold
plying a constant DC voltage to a transistor body, the source of this bias is difficult to
justify in many energy harvesting applications as supplying this energy source will degrade
the overall efficiency. Additionally, the use of non-RF power sources to bias the transistors
such as a battery defeats the purpose of traditional, passive RFID tags. Other on-chip
techniques better lend themselves to the same goal. In [84], Lin et al. used two charge
pumps, a primary charge pump and a secondary to bias the primary. However, this tech-
nique requires additional chip space and the limitations of the secondary charge pump are
still prevalent. In [85], Le et al. utilized a high-Q resonator along with floating-gate tran-
sistors to reduce the threshold voltage. Charge injected into the gate lowers the threshold
voltage, but a bias must be held constant to maintain this level. In [86], Papotto et al.
48
body biasing from connected transistor gates to the source of the previous stage.
Both SPS and RFID applications have numerous advantages in moving to microwave fre-
quencies. Microwave antennas are substantially smaller than some of their UHF and VHF
counterparts, and this smaller size helps to reduce the antenna aperture required and to
focus the antenna beam more easily. Since the antenna size is the dominating component in
the size of a rectenna or RFID-enabled sensor, space can be used more efficiently. This re-
duction in size (and therefore weight) can help SPS applications reduce the amount of land
required for ground-based, power collecting stations; and help RFID applications minimize
the profiles of their tags and sensors. While 2.45 GHz has been the traditional frequency
of choice for SPS, the world-wide availability of the unlicensed 5.8 GHz microwave band
has led research in microwave-power transfer to focus on this frequency. Moreover, work at
It is important to note that the majority of UHF charge-pump research has focused on
CMOS implementations. This fact is not surprising as an RFID tag cost drops tremendously
when the charge pump is integrated with the tag’s electronics. However, CMOS processes
have yet to mature to an acceptable price point, power handling ability, and efficiency
required for microwave energy-harvesting applications. Thus, as discerned from Tables 13,
14, and 15, most of these applications rely on discrete, Schottky diode rectifying elements.
These tables provide the data from Figure 14 at 2.45 GHz, 5.8 GHz, and greater than
5.8 GHz, respectively. It can also be observed from Figure 14 that the majority of microwave
work has been done at relatively higher power levels than most of the UHF work. This fact is
due to these devices being used for SPS or other predominantly WPT-focused applications,
where total power transfer is paramount. As a result, the multi-stage topologies used in
UHF energy harvester design is unnecessary. Single stage or single diode rectifiers such as
Since the majority of these devices rely on discrete Schottky diodes and not custom
CMOS transistors, much of the work in these energy harvesters focuses on choosing the
49
most efficient Schottky diode for the application and on antenna/circuit design and layout
optimization. Custom Schottky diodes have been investigated by McSpadden and Chang
in [29] and by Brown in [87]. These diodes have performed remarkably well for their
applications and have yielded some of the most efficient harvesters to date. More commonly,
commercial off the shelf components have been used in rectifying elements as their costs
are considerably lower. Often times, diodes are used in parallel in an attempt to lower the
resistive losses associated with the semiconductor junction. This setup is difficult at higher
frequencies because junction capacitances will increase as the diodes are put in parallel.
highly resonant microwave structures [88] and harmonic terminations [89, 90]. Highly res-
onant structures help to increase the voltage incident on the rectifying element. However,
practical considerations often limit the Q of these resonators. Harmonic terminations such
as class-E and class-F rectifiers help to ensure that harmonic frequencies of the excitation
frequency are properly terminated and reflected back to the rectifying element to eliminate
As previously noted and evident in Figure 14, as frequency increases, the efficiency of
these devices decreases. This observation can be attributed to the higher parasitic losses
figure that UHF charge pumps are typically designed for low-power operation as their
design is driven by the RFID community. In all cases, a general trend can be seen. At high
powers, large efficiencies are possible as the energy-harvesting devices are operating in a
linear state, far above their diode turn-on voltage. As the power level is reduced, the device
efficiency also decreases because less current is able to flow through the diodes.
DC. While there are multiple methods of doing this, diode-based rectifier circuits are most
50
(b) Single Shunt Rectenna.
(a) Half-wave rectifier.
Figure 15: Various energy-harvesting circuit topologies used in RFID and SSP applications.
51
3.4.1 Topology Descriptions
Long range WPT encompasses both energy transfer driven applications and ultra-low power
sensors. While these two communities have the same general goal in mind, their motivations
are different.
The SSP community typically utilizes rectenna-based designs as high efficiency is paramount
and large power densities are available. In contrast, battery-free sensor designers and the
RFID community use charge pump topologies for energy harvesting. While rectennas can
efficiently harvest RF energy, the DC voltages produced are not sufficient to drive the logic
Charge pumps not only rectify RF to DC, but also step up the voltage to usable levels. The
The half-wave rectifier in Figure 15a is mentioned here to build an understanding of recti-
fiers. Typically, the half-wave and full-wave rectifiers are studied in an introductory circuits
course as a means of learning how to convert AC to DC power. While these circuits tend
to be acceptable for relatively low frequency, high voltage applications, RF and microwave
In any case, half-wave and full-wave rectifiers demonstrate some basic principles of
energy harvesting. First, the input signal must be larger than the voltage turn-on threshold
of the diode to get a useful output voltage. Second, the maximum output voltage will be
limited by the reverse breakdown voltage of the diode as previously explained. Finally, an
output capacitor forms a low-pass filter which helps to smooth out the output voltage.
For energy transfer driven applications, the efficient, large-scale transfer of power (typically
microwave in frequency) is the ultimate goal. As these systems aim to transfer power
for commercial uses, high total system efficiency is necessary for the largest return on
investment. The single shunt rectenna in Figure 15b is the energy-harvesting circuit of
52
choice for these sorts of applications.
Single shunt rectennas are preferable because of their high efficiency, small size, and low
cost. These attributes are realized by their simple, single diode design. Due to the presence
of a single diode, there is only a single location for parasitic series losses. Furthermore, the
quarter wave transformer between the diode and the output low-pass filter allow harmonics
to reflect back to the diode and effectively trap energy there to increase the amount of
power converted to DC. The requirement of a single diode decreases the total system cost
To generate large amounts of power, rectennas are grouped in massive arrays to generate
large amounts of power. Because of power density limitations due to the diode power hand-
ing and safety reasons, the antenna aperture must be fairly large. Higher frequencies are
thus preferable as the array aperture will be smaller. For this reason, there is more available
carrier frequency of 2.45 GHz is typically favored within the community, but research has
also been done at 5.8 GHz [91, 29, 28, 30, 92], 8.51 GHz [93], 10 GHz [6], 20 GHz [94],
35 GHz [95, 96, 6], and 94 GHz [95]. Furthermore, size constraints may limit the useful-
ness of single shunt rectennas at UHF and lower frequencies because of the necessity of a
quarter-wave transformer.
Figure 15c shows the single stage voltage multiplier which forms the basic building block of
many more complex charge pumps, including the Dickson, Cockcroft-Walton, and Greinacher
charge pumps. Charge pumps differ from typical rectifier circuits in that they step up the
voltage of the received signal. This capability is necessary as the received signal is usually
incapable of producing voltage levels high enough to power the logic level of modern digital
devices.
The single stage voltage multiplier shown in Figure 15c is also called a voltage doubler
as it will double the peak input voltage under ideal conditions. When an input signal
greater than the threshold voltage is driven across diode D1 , it will rectify that current
53
and generate harmonics according to the discussion in Section 3.2.2. Capacitor C1 stores
the DC charge and biases diode D2 which acts as a peak detector to smooth the output.
Finally, the output capacitor and resistor form a low-pass filter which produces an output
One of the most well-known charge pump circuits is the Cockcroft Walton or Greinacher or
Villard charge pump shown in Figure 15d. Cockcroft and Walton had designed this circuit
in order to obtain high voltages necessary for their experiments in high velocity positive
ions [97]. Using cascading stages of the single stage voltage multiplier with thermionic
diodes, mechanical switches, and capacitors, they were able to generate up to 800 kV.
While this style of circuit is capable of producing large voltages, it has several drawbacks
which make it impractical for energy-harvesting applications. First, stage capacitances must
be much larger than any parasitic capacitances to ground in order to efficiently pump charge.
While this typically is not a problem for lumped element circuit designs, for integrated
circuit designs, practical capacitor sizes are not sufficiently large enough to overcome this
criteria [98]. Furthermore, as the number of stages in the Cockcroft-Walton charge pump
increase, the output impedance increases. This result can cause difficulties in successful
The Dickson charge pump was originally used as a voltage multiplier to create high voltages
for writing to non-volatile memory circuits [98, 99]. It was designed to overcome some of the
aforementioned challenges associated with the Cockcroft-Walton charge pump. The original
Dickson circuit used an arrangement of diodes and capacitors as shown in Figure 16 with a
digital clock connected to alternating stages to pump charge from one capacitor to another.
Assuming a completely discharged initial state, charging will begin when a DC signal is
present on the input and a clock signal, and its inverse are on Φ and Φ. During the low
phase of the clock signal Φ, diodes D1 and D3 will be turned on (assuming the input is
greater than the diode threshold), and DC current will flow charging capacitors C1 and
54
Figure 16: The original Dickson charge pump was a voltage multiplier which increased a
DC input signal by pumping charge across pairs of diodes and capacitors. This pumping
was controlled by an external clock signal which controlled the rate.
C3 . Also during this phase, all other diodes are effectively off, and no charge flows through
them. During the high part of the clock cycle Φ, the reverse occurs. Diodes D2 and DN −1
will be turned on, and current will flow charging capacitors C2 and CN −1 . As each stage
charges, the DC voltage level of the previous stage biases that stage and allows a larger
voltage to be generated. Diode DN and output capacitor Cout act as a peak detector and
provide smoothing.
In this topology, the output voltage of the Dickson charge pump is given as
(N + 1)(Vin − VT )
Vout = N
, (23)
1 + f CRL
where Vout is the DC output voltage across the load, N is the number of stages in the
charge pump, Vin is the input to the charge pump, VT is the diode threshold voltage, f
is the frequency of the input clock signal, C is the stage capacitance, and RL is the load
output. While this expression is valid for DC to DC charge pumping circuits, it is not valid
for the Dickson charge pump topology used in RF energy harvesting circuits in Figure 15e.
Appendix B.
Figure 15e shows a typical charge pump, based off the Dickson topology, typically used
in RFID tags. Compared to Figure 16, this figure grounds the DC input before D1 , since
a passive tag has no DC voltage source, and reconfigures the clock inputs to be connected
to an RF source, usually an antenna. Therefore, the RF signal acts as both the clock and
voltage source. However, now the charge pump must simultaneously rectify the input signal
55
The Dickson charge pump is fundamentally different from the Cockcroft-Walton in the
fact its diode nodes are coupled via capacitors in parallel with the voltage source. Con-
versely, the Cockcroft-Walton has its diode nodes coupled in series. Due to its ability to
more efficiently multiply input voltage when implemented in CMOS, the Dickson charge
pump is most often used by the RFID community [100]. While the majority of Dickson
charge pumps have been built at UHF frequencies for RFID applications, there has also
been some limited work at 2.4 GHz. However, there has been no documented applications
While there are numerous unique topologies for charge pump design, the modified Cockcroft-
Walton provides some unique benefits and is discussed in [54, 101]. Figure 15f shows its
benefit can be useful for antenna matching. Furthermore, the symmetry helps to reduce
If the intended application requires a voltage larger than a single diode can provide, ad-
ditional diode stages in the form of a Dickson or Cockcroft-Walton charge pump may be
required. To investigate the performance of energy-harvesting circuits with more than one
diode, the maximum efficiency derivation for a single diode from [6] was re-derived for an
N-stage Dickson charge pump. The details of this derivation are shown in Appendix B.2.
Using these derivations, the performance of energy-harvesting circuits with increasing num-
Figures 17, 18, and 19 show the theoretical results for the DC output voltage, DC output
power, and energy harvesting efficiency for a single diode; two-,four-, six-, and eight-stage
Dickson charge pumps with 300 Ω, 600 Ω, 1200 Ω, 1800 Ω, and 2400 Ω loads, respectively.
56
These loads were chosen according to criterion in [6], and the fact that these loads allow
each circuit to operate the most efficiently at the maximum input power level before reverse
breakdown. Each circuit utilizes the same Avago HSMS-285x diodes with parameters in
Table 3. Note the asymptotes in the voltage and power graphs. This line is due to the
diode model used in the theoretically derivation and is the minimum power level where the
N=1
10
1 N=2
N=4
N=6
N=8
0
10
Vout (V)
−1
10
−2
10
−3
10
−20 −10 0 10 20 30 40
Input Power (dBm)
Figure 17: The theoretical energy-harvesting DC output voltage for a single diode; two-,
four-, six-, and eight-stage Dickson charge pumps with 300 Ω, 600 Ω, 1200 Ω, 1800 Ω, and
2400 Ω loads, respectively. These circuits use Avago HSMS-285x diodes under 5.8 GHz CW
excitation.
As is expected, as the number of stages in the charge pump increases, the output voltage
also increases as shown in Figure 17. While this fact is most pronounced at higher powers,
57
N=1
25 N=2
N=4
N=6
N=8
10
Pout,DC (dBm)
−5
−20
−35
−50
−20 −10 0 10 20 30 40
Input Power (dBm)
Figure 18: The theoretical energy-harvesting DC output power for a single diode; two-,
four-, six-, and eight-stage Dickson charge pumps with 300 Ω, 600 Ω, 1200 Ω, 1800 Ω, and
2400 Ω loads, respectively. These circuits use Avago HSMS-285x diodes under 5.8 GHz CW
excitation.
at lower powers the opposite is actually true. At a sufficiently low power level, a single diode
energy harvester will outperform a multi-diode energy harvester. This fact is a result of
additional losses in multi-stage energy-harvesting circuits. Since charge must flow through
each diode, there will be more losses for a charge pump with more stages. Above this
sufficiently low power level, there will be an optimal number of stages to produce the
maximum output voltage. However, this difference is not typically very large (usually only
a few percent difference). Above a particular level, the more number of stages will produce
a larger voltage.
The maximum DC output voltage is higher for charge pumps with additional stages
because the charge pump has a collectively larger breakdown voltage relative to the output.
It is also apparent that the input power for which the charge pump output voltage reaches
its maximum increases with the number of stages. This phenomena occurs because there is
increased loss associated with each diode in the charge pump. Consequently, a larger input
power is necessary to overcome these losses and reach the same state.
58
1
N=1
N=2
0.9 N=4
N=6
0.8 N=8
0.7
Efficiency 0.6
0.5
0.4
0.3
0.2
0.1
0
−20 −10 0 10 20 30 40
Input Power (dBm)
Figure 19: The theoretical energy-harvesting efficiency for a single diode; two-, four-, six-,
and eight-stage Dickson charge pumps with 300 Ω, 600 Ω, 1200 Ω, 1800 Ω, and 2400 Ω loads,
respectively. These circuits use Avago HSMS-285x diodes under 5.8 GHz CW excitation.
Figure 18 illustrates many of the previously mentioned points but describes them in
terms of DC output power, not voltage. Principally, at high powers, large numbers of stages
are beneficially because they allow the rectification of larger amounts of power. However, at
low powers, a fewer number of diodes is beneficially because there are less loss mechanisms.
Figure 19 also confirms these points by plotting the efficiency. Note that the maximum
energy harvesting efficiency for all the different number of stages is approximately the
same, but occurs at different input powers. The levels are the same because of the choice in
output resistances, and the levels at which they occur are different because of the additional
Figures 20, 21, and 22 show the theoretical results for the DC output voltage, DC output
power, and energy harvesting efficiency for a single diode; two-,four-, six-, and eight-stage
Dickson charge pumps all with the same 300 Ω load. While this value provides a maximum
energy harvesting efficiency for the single diode case, its useful from a design perspective
59
with a fixed load to see the effects of increasing the number of stages in the energy harvester.
N=1
1 N=2
10
N=4
N=6
N=8
0
10
Vout (V)
−1
10
−2
10
−3
10
−20 −10 0 10 20 30 40
Input Power (dBm)
Figure 20: The theoretical energy-harvesting DC output voltage for a single diode; two-,
four-, six-, and eight-stage Dickson charge pumps all with a 300 Ω load. These circuits use
Avago HSMS-285x diodes under 5.8 GHz CW excitation.
These three figures illustrate the same general points as discussed in the previous section.
However, there are several key differences. First, Figures 20 and 21 show that for a fixed
load, the input power level for which the energy-harvesting circuit enters reverse breakdown
can be much larger (or potentially smaller) than when each circuit is individually matched.
For example, in this case with a load of 300 Ω, a charge pump with 6 stages has a maximum
voltage at 33 dBm input power. This value compares to a maximum voltage at 25 dBm when
a load of 1800 Ω is used. This difference can be attributed to the difference in the ratio of
the total series resistance (N Rs ) to the load resistance RL explained in Figure 13. Likewise,
since the circuits do not have their optimal output resistance, their maximum efficiency level
decreases as shown in Figure 22. Overall, load conditions and operating power levels should
be taken into account when choosing the number of stages in an energy-harvesting circuit
design.
60
N=1
25 N=2
N=4
N=6
N=8
10
Pout,DC (dBm)
−5
−20
−35
−50
−20 −10 0 10 20 30 40
Input Power (dBm)
Figure 21: The theoretical energy-harvesting DC output power for a single diode; two-,
four-, six-, and eight-stage Dickson charge pumps all with a 300 Ω load. These circuits use
Avago HSMS-285x diodes under 5.8 GHz CW excitation.
1
N=1
N=2
0.9 N=4
N=6
0.8 N=8
0.7
0.6
Efficiency
0.5
0.4
0.3
0.2
0.1
0
−20 −10 0 10 20 30 40
Input Power (dBm)
Figure 22: The theoretical energy-harvesting efficiency for a single diode; two-, four-, six-,
and eight-stage Dickson charge pumps all with a 300 Ω load. These circuits use Avago
HSMS-285x diodes under 5.8 GHz CW excitation.
61
3.5 Energy-harvesting Circuits Under CW Excitation
3.5.1 Design Methodology
The design of energy-harvesting circuits can be daunting because of the variety of param-
eters, complicated simulations, and various operating environments in which the devices
could operate. The following methodology sets a direction to use for this design and provide
a tool for theoretically comparing the maximum performance between systems. Figure 23
shows a flow graph of this method. It is assumed that the designer has constraints for
sensitivity, load impedance, energy storage, and output voltage, among others.
3. Using the theory in Appendix B, determine if the goals are met. Repeat previous
steps if not. Keep in mind that the theory will be the maximum possible and not
4. Based on a desired operating point (frequency and input power, which will determine
Perhaps the most confusing, but important choice in the energy-harvesting circuit design is
the choice of a diode. While some circumstances may warrant the fabrication of a custom
must still be evaluated. For example, if the three diodes in Table 3 were considered, how
could one decide which may be the best suited for a particular application? By leveraging
the theoretical diode performance discussed in Section 3.3.2, each diode performance can be
observed as shown in Figure 24. For the purpose of this calculation, an excitation frequency
As can be seen, each diode has varying performance characteristics. If low-power opera-
tion was desired, the HSMS-285x should be selected because it has the largest efficiencies at
62
Figure 23: The methodology for designing an energy-harvesting circuit.
the lowest powers. This characteristic is due to its low turn-on voltage. However, it is un-
able to function well at larger powers above 12 dBm because of its small reverse breakdown
voltage. On the other hand, if the desired operation of the diode was high power operation,
either the HSMS-286x or the HSMS-282x should be selected. While the HSMS-282x can
operate at a higher power, its efficiency is usually lower than the HSMS-286x. This fact
is due to the larger junction capacitance degrading the energy conversion efficiency across
63
1
HSMS−286x
HSMS−285x
HSMS−282x
0.8
0.6
Efficiency
0.4
0.2
0
−30 −20 −10 0 10 20 30 40
Input Power (dBm)
Figure 24: The Avago HSMS-286x, HSMS-285x, and the HSMS-282x theoretical efficiencies
are plotted as a function of input power. These curves were evaluated assuming an excitation
frequency of 5.8 GHz and a 300 Ω load.
all frequencies. It is able to operate at higher powers, however, because of its large re-
verse breakdown voltage. By using this theoretical calculation and a few of the basic diode
parameters from a few test cases, an intelligent diode choice can be made. Additionally,
experimental load-pull measurements can be made as in [89] to determine the optimal diode
The choice of the energy-harvesting circuit is typically dependent on the voltage required
at the output. Based on the reverse breakdown voltage of the selected diode, the desired
output voltage may not be possible to produce with a rectenna or single diode configuration.
Therefore, it may be necessary to choose a charge pump design to increase the DC output
voltage to the required level. Alternatively, if size constraints allow, multiple rectennas
could be connected in series to achieve a larger DC output voltage at the expense of antenna
aperture.
64
Regardless of the criterion, at power levels below 10 dBm, the rectenna performs the
best in terms of output voltage, output power, and efficiency. This result is primarily a
matter of the lower losses encountered in a single diode energy harvesting circuit. If the
application required a larger voltage, energy harvesting circuits could be combined in series
or antenna arrays such as the staggered-pattern charge collector (SPCC) could be used to
greatly vary over its operating range of powers and frequency. Since it would be consid-
erably difficult to design a passive matching network capable of encompassing to the wide
ranges of circuit input impedances, assumptions must be made about the device operation.
Typically, this should be done at the fundamental frequency to ensure the most efficient
tent from radiating into the environment. This design choice will improve efficiency and
reduce potential electromagnetic interference. While many first iteration designs may use
lumped elements, higher frequency designs may be better suited to using microstrip real-
ized matching networks. It is recommended that this design utilize a full-wave or 2.5-D
electromagnetic field solve to take into account all fringing fields. In many cases, microstrip
Its also important to point out that using lumped element passives at microwave fre-
quencies can lead to unanticipated problems. Most data sheets do not contain RF data for
these circuit elements. Many times, these components will fail to operate as expected at
microwave frequencies because of parasitic values in them. For instance, standard imperial
0603 capacitors do not function well at 5.8 GHz for values above 2 pF because of the ca-
pacitor parasitic inductance. In fact, for values above 2 pF, these capacitors will actually
lize microstrip-implemented passives whenever possible. Note that this requires the use of
65
3.5.1.4 Computer simulation
At this point, the entire energy-harvesting circuit design should be optimized using a com-
mercial full-wave or 2.5-D electromagnetic field solver and include the antenna, matching
network, and energy harvesting circuit layout in complete form. It should include any rel-
for any lumped elements which are used. This optimization will ensure that all losses
are accounted for and any implementation errors can be managed before a prototype is
fabricated.
To properly simulate a non-linear microwave circuit like an energy harvester, the har-
monic balancing technique should be used to computer voltages at each node. This tech-
nique computes the voltage/current at each node in the circuit for every harmonic and
ensures there is a balance of current entering and leaving each node for all frequencies
taken into account (in the simulation setup, the number of frequencies and mixing terms
is defined). For non-linear elements in the simulation, transient analysis is performed and
frequency components are extracted. Large-scale S-parameter simulations are also useful
The computer optimization method run in Agilent ADS is summarized in Figure 25.
First, component models for the diodes, resistors, and any other package are created in ADS
to include all parasitic losses. Second, these components are used along with models of the
the desired substrate. Relevant dimensions are parameterized and have a starting value set
according to rules previously discussed in this thesis and related papers. Third, using the
harmonic balance simulator, the energy harvester is optimized according to the application
goals. When goals are met, the schematic model is imported into ADS Momentum. Once
again, relevant dimensions are parameterized and substrate parameters are set. Next, the
layout and EM model are created which can be exported as a part in ADS. The model
(diodes, resistors, etc.) are connected. After ensuring the Momentum EM model is selected
on the imported part, the optimizer is run again with the harmonic balance technique.
66
Finally, after ensuring all goals are satisfied, the optimized part is fabricated and measured.
As an example, a single-shunt rectenna was designed to operate in the 5.8 GHz ISM band ac-
cording to the procedure previously discussed. However, this circuit was optimized entirely
using microstrip models and did not utilize ADS Momentum, a method-of-moments-based
solver.
This rectenna was designed on the top two layers of a 4-layer FR-4 substrate using Avago
67
HSMS-2865 diodes in a SOT-143 package (parallel arrangement). Substrate parameters are
listed in Table 4. Package parasitics as well as substrate loses were included to make
5.8 GHz and an input power of 16 dBm, and the circuit was optimized using the ADS
‘random search method’ algorithm with the goal of maximizing efficiency. Parameters
included in this optimization were the output resistance and all transmission line lengths
and widths except for the length and width of the transmission line from the T-junction to
the output resistor. These parameters were fixed at 90 degrees and 50 Ω to ensure single
shunt functionality. This optimization yielded a load value of 316 Ω and the widths and
lengths shown in the rectenna layout in Figure 26. The full, optimized schematic is listed
in Appendix E.1.
Parameter Value
Dielectric constant (r ) 3.95
tan δ 0.019 @ 10 GHz
Dielectric thickness 9.3 mil
Conductor thickness 35µm
Figure 27 shows the simulated efficiency and reflected power levels from the rectenna
obtained. This value is far removed from the theoretical maximum because of the choice of
FR-4 as a substrate. Due to the large tan δ, there are significant substrate losses which result
in an average 13% loss. The remaining power in the circuit is dissipated in the substrate and
copper traces, converted to harmonics, and lost in device parasitics. If lossless transmission
lines are used and device parasitics are removed, the maximum efficiency increases to 83%.
This value falls near the theoretical maximum predicted by the theory in Appendix B.
While the highest measured efficiency reported for a 5.8 GHz energy-harvesting circuit was
82% for an input power of 50 mW [29], typical efficiencies in this power range are around
60-70%.
68
Figure 26: The layout of a 5.8 GHz single-shunt rectenna on a 4-layer FR-4 substrate. The
input impedance was matched to 50 Ω at an input power of 16 dBm. Schematic elements
shown were simulated as SMDs.
At 5.8 GHz and 16 dBm input power, the rectenna is matched and experiences a max-
imum efficiency and minimum reflected power as confirmed in Figure 27. Below 16 dBm,
a majority of this power is reflected as the rectenna becomes more inductive mainly due
to the diode impedance change. At higher powers, not only does the diode become less
matched as is evident by the increase in reflected power, voltage levels also surpass the
reverse breakdown threshold as noted by the steep drop off in efficiency above 16 dBm.
69
0.8
Theoretical Efficiency
0.7 Simulated Efficiency
Simulated Reflected
% Total Power 0.6
0.5
0.4
0.3
0.2
0.1
0
−30 −20 −10 0 10 20 30
Pin (dBm)
Figure 27: The percentage of input power converted to DC (energy conversion efficiency)
and the percentage of power reflected into the source of the designed rectenna. Note that
the maximum efficiency of 70.5% at 16 dBm occurs when the reflected power is minimized.
The large discrepancy between the simulated results and theory is due to the FR-4 substrate
losses.
10e1
10e0
10e−1
10e−2
Vout (V)
10e−3
10e−4
10e−5
Theoretical
Simulated
10e−6
−30 −20 −10 0 10 20 30
Pin (dBm)
Figure 28: The theoretical and simulated output voltage of the single shunt rectenna follow
a similar trend. Much of the difference at high powers is due to substrate losses in the FR-4.
At lower powers, impedance mismatch causes a much greater deviation.
70
CHAPTER IV
The previous chapters have addressed improvements and tradeoffs associated with rec-
tifying element characteristics & topology and physical layout of energy harvesting circuits.
These methods are essential for efficient energy conversion, but all will eventually have a
minimum turn-on power level below which very little energy conversion will occur. Mod-
ifying the excitation waveform permits this minimum turn-on power level to be further
decreased. By modifying the waveform to transmit a higher power for a short amount of
time, the RF-to-DC conversion efficiency can be increased. One of these techniques known
power ratio (PAPR) of a transmitted signal to overcome the limitations of the diode
threshold voltage level in a charge pump [63]. Similar techniques called multi-sine exci-
tations [65, 66], chaotic waveforms [104], and intermittent transmission [105] have also been
71
discussed in literature. All these techniques aim to deliver additional power to the energy-
harvesting circuit in an attempt to increase its efficiency. POWs, multi-sines, and chaotic
waveforms shape their spectra so that individual frequency components can add in-phase
to produce large voltage spikes in the time domain. Similarly, the intermittent transmission
POWs are characterized by the number, amplitude, phase, and spacing of subcarriers, which
affect their PAPR and RMS bandwidth. These parameters are typically expressed in the
where ak is the voltage of the k th subcarrier, fc is the carrier frequency, Φk is the phase,
POW gain GP OW is an important metric which POWs are evaluated under and is
defined in (26) as
2
Vo,P
Po,P OW OW
GP OW = = 2 . (26)
Po,CW Vo,CW
In this definition, Po,P OW and Vo,P OW are the time-averaged output power and voltage,
respectively when under a POW excitation P OW (t) in (24). Likewise, Po,CW and Vo,CW
are the time-averaged output power and voltage, respectively when under CW excitation.
This gain gives the improvement of the designed POW over a CW waveform. POWs are
also characterized by their PAPR, RMS bandwidth, and RMS time width, all of which rely
72
4.1.2 Principles of POWs
Figure 29 shows an example of a POW with four equally weighted subcarriers. As evident
in the figure, compared to CW, POWs have a larger bandwidth due to their multi-carrier
nature. Furthermore, one can see the large spike in the time domain that allows the energy
harvester to turn on at a lower power than when using CW. The total power of the POW
can be made equal to the power of the CW signal by decreasing the relative power level of
each subcarrier. This fact allows POWs to continue to meet governmental regulations such
4 0.8
CW
POW
3 0.7
2 0.6
|POW(f)| (V/Hz1/2)
1 0.5
Voltage (V)
0 0.4
−1 0.3
0.2
−2
0.1
−3
0
−4
0 0.25 0.5 0.75 1 5.75 5.8 5.85
Time (s) x 10
−7 Frequency (Hz) x 10
9
As an RFID tag moves in an environment, the RF power available to the tag will vary
due to small- and large-scale fading. Therefore, it is possible there will come a point where
the induced voltage on the tag antenna is no longer greater than the diode threshold level
on the charge pump. Thus, the tag will fail to turn on. A POW creates a higher PAPR than
73
is exhibited by a CW waveform that allows the diodes to remain switched on for a greater
distance and thus, increase the range and reliability of RFID tags as shown in Figure 30.
As can be seen, while the CW signal is incapable of surpassing the diode voltage turn-on
threshold, the POW is able to at the same power. In the CW case, this would mean that
the DC bias level would decrease. Meanwhile, the POW could sustain the same high level
for a lower input power. Trotter et. al have verified this behavior in [64]. By using POWs,
a commercial 915 MHz RFID tag range was increased by 2.5 dB. Fernandes in [66] also
Figure 30: A POW can cause a larger DC bias across a diode for a lower average power
than a CW. This figure shows an equal power 4-POW and a CW for a set DC bias level.
Note that since the CW does not have sufficient voltage to turn-on the diode, its DC bias
level would decrease.
While these previous works have discussed the advantages of POWs and multisines,
these works have neglected to study the dynamics of the waveform and energy-harvesting
circuit in detail. Since the POW is designed to improve the energy-harvesting efficiency of
74
this circuit, it is advantageous to investigate the relationship between the circuit topology
POWs are characterized by the number, amplitude, and spacing of subcarriers they contain,
which affect their PAPR and RMS bandwidth. By changing the subcarrier spacing and
amplitudes, different POWs can be created. A variety of POWs such as M-POW, Gaussian
|POW(f)| (V/Hz1/2)
5 0.5
Voltage (V)
2.5 0.4
0.3
0 0.2
−2.5 0.1
−5 0
0 0.25 0.5 0.75 1 5.725 5.7625 5.8 5.8375 5.875
Time (s) x 10
−7 Frequency (Hz) x 10
9
5 0.5
Voltage (V)
2.5 0.4
0.3
0 0.2
−2.5 0.1
−5 0
0 0.25 0.5 0.75 1 5.725 5.7625 5.8 5.8375 5.875
Time (s) x 10
−7 Frequency (Hz) x 10
9
Gaussian−POW, ∆ f=10 MHz, BRMS=15 MHz Gaussian−POW, ∆ f=10 MHz, BRMS=15 MHz
|POW(f)| (V/Hz1/2)
5 0.5
Voltage (V)
2.5 0.4
0.3
0 0.2
−2.5 0.1
−5 0
0 0.25 0.5 0.75 1 5.725 5.7625 5.8 5.8375 5.875
Time (s) −7
x 10 Frequency (Hz) x 10
9
Figure 31: Three examples of power-optimized waveforms and their parameters. From top
to bottom, M-POW, Square POW, and Gaussian POW. Notice the different PAPR of each
waveform along with the frequency content and shape of their respective spectra.
where M is the number of cosine terms summed together, Pavg is the average power in V 2 ,
75
and TPOW is the POW time period. The corresponding baseband frequency spectrum is
described by
r M
Pavg X k k
POW(f ) = δ f− +δ f + . (28)
2M TPOW TPOW
k=1
2M
B3dB = . (30)
TPOW
For comparison purposes, the PAPR of CW is 2. Thus, simply adding one subcarrier
Pavg
increases the PAPR by a factor of 2. Note that each M-POW subcarrier has energy 2M V2 .
Furthermore, for spectral mask considerations, even M’s are preferred as odd M’s have non-
For a POW to be useful, it must meet several criteria. First, the POW must have the
necessary PAPR so that the DC power requirements of the RFID tag can be met when using
a charge pump [5]. Second, the time-domain transition must be smooth so that capacitors
can appropriately charge to the peak voltage. Finally, the 3-dB bandwidth of the POW
must be narrower than the RFID tag RF front end and fit in regulatory limits.
While it is known that POWs can increase the range of RFID tags, it is useful to have a
metric to quantify this increase. A modified forward-link budget from (3) is shown in (31)
which includes a POW gain term [76]. The inclusion of this term will help to illustrate
the tradeoffs in received power, increased distance, etc. when using a POW. Also note that
The multi-tone nature of POWs increases the complexity with backscatter communi-
cation. Instead of a single frequency as in the case of CW, POWs use multiple, lower
76
powered frequencies. Since RF tags communicate via modulating their RCS, each of these
subcarriers will contain copies of the modulated data. The simplest way of demodulating
the data would be to filter out all but one of the subcarriers1 . Assuming N equally powered
PT
PTN = = 10 log10 (PT ) − 10 log10 (N ). (32)
N
If the backscatter link budget in (2) is modified to include these effects, the result in (33)
is obtained.
4πrf 4πrb
PR = PT +GT +2Gt +GR −20 log10 −20 log10 +M −Fb −10 log10 (N ) (33)
λ λ
The backscatter link gives the reader received power PR which ultimately relates to the
number of bit errors that a reader will incur. Thus, for a reader to receive data from
an RFID tag with a certain bit error rate, the received power PR must be greater than
a minimum reader sensitivity PR,min . For POWs, the lower subcarrier power will result
in a lower SNR and consequently higher bit error rate. Recall that most passive tags are
forward-link limited. Thus, the increase in bit error rate will be minimal for low PAPR.
While (31) and (33) document all of the link budget terms and potential benefits and
tradeoffs, range is typically the most common RFID metric. According to the forward-link
budget, an RFID tag that uses a POW should have an increase in range according to (34).
rf
GP OW = 20 log10 , (34)
rf + ∆rf
where ∆rf is the range increase obtained when using POWs over CWs.
However, (33) shows that the backscatter link limit will be reduced by using POWs.
Analyzing this equation shows that an RFID tag should have a reduced range according to
(35).
√
rb ( 4 N − 1)
∆rb = √
4
, (35)
N
where ∆rb is the reduction in distance of the backscatter link.
It should be noted that this range reduction could be reduced if signal processing meth-
ods such as coherent integration are used to combine the signal energy from each subcarrier.
1
Other methods of demodulating POWs are discussed in Appendix D.
77
However, this would require additional filtering because of unmodulated backscatter and
signal leakage. POW communication is another research area which will greatly affect their
applicability and usefulness, but the majority of this discussion is outside the scope of this
thesis. Appendix D provides a short discussion of some of the challenges and potential
solutions.
It was shown in Chapter 3 that when a diode-based energy harvester is under CW excitation;
DC, the fundamental frequency, and its harmonics are generated. However, when POWs are
used as an excitation source, there are significantly more frequencies generated which can
produce substantial voltage ripple. Assuming the same small-signal approximately where
V = Vo + v, (36)
where Vo is the self DC bias across the diode and v is an arbitrary POW with two subcarriers,
By substituting V into the taylor series expansion of the diode I-V relationship in (8), the
78
result is
G0d (a1 + a2 )2
I(V ) = I0 +
4
+G0d (a1 a2 )2 cos(2∆ω)
G00d a1 a22
3 2
+ a1 + 4a2 a1 + + Gd a1 cos({ωc − ∆ω}t)
6 2
00
a2 a21
Gd 3 2
+ a2 + 4a1 a2 + + Gd a2 cos({ωc + ∆ω}t)
6 2
G00 a1 a22
+ d 2a22 a1 + cos({ωc + 3∆ω}t)
6 2
G00d a2 a21
2
+ 2a1 a2 + cos({ωc − 3∆ω}t)
6 2
G0 a 2 (38)
+ d 1 cos({2ωc − 2∆ω}t)
4
G0d a22
+ cos({2ωc + 2∆ω}t)
4
+G0d a1 a2 cos(2ωc t)
G00d a31
+ cos({3ωc − 3∆ω}t)
18
G00 a3
+ d 2 cos({3ωc + 3∆ω}t)
18
00
a1 a22
G
+ d 2a22 a1 + cos({3ωc − ∆ω}t)
6 2
G00 a2 a21
+ d 2a21 a2 + cos({3ωc + ∆ω}t) + ...
6 2
Similar to the result shown in (13) with a CW excitation, when a POW is used across
a diode, DC is produced along with the fundamental and harmonic frequency content.
However unlike a CW, POWs generate intermodulation products as a result of the subcarrier
spacing ∆ω. While the expansion in (38) is for a 1-POW, higher order POWs will have
where m and n are integers from zero to infinity. While all of these frequencies should
be considered when evaluating a non-linear device (especially with a POW), usually the
first several harmonics and first couple mixing orders are important. In other words, as the
number of subcarriers increases, the number of harmonics and intermodulation products will
79
also increase. Therefore, it is especially important to consider the number of harmonics and
While many of these frequencies will be removed via the low-pass output filter as shown in
Figure 32, intermodulation products less than the fundamental frequency may appear as
voltage ripple on the output unless the filter has a sufficient cut-off frequency. For practical
microwave circuits, the capacitance required for this filter may be difficult to construct
because of its resonance. While Figure 32 shows a single frequency which falls below the
cutoff frequency, if more subcarriers are added and/or higher orders of intermodulation
products are considered, any integer multiple of the subcarrier spacing n∆ω will have to be
considered.
A theoretical model expressing energy-harvester efficiency under high PAPR signals was
developed based on the theory in [6] to aid in the understanding of high PAPR signals.
Computationally, these signals can be difficult to simulate because of their spectral content.
Harmonic balance simulations typically used to simulate this type of non-linear, microwave
circuit encompass the effects of the diode, transmission lines, and matching considerations.
However, to ensure accuracy, several orders of harmonics and mixing terms are required.
Thus, for high PAPR signals that are composed of a large number of frequencies (for
instance, a 5-POW), the number of frequencies for the harmonic balance simulator to take
into account can easily surpass 100,000. Likewise, transient simulators require very small
time steps to take into account the high harmonic content and long run times to ensure
steady-state operation. At the time of this thesis, a 4-POW signal simulated in Agilent
ADS 2011.10 with a harmonic balance simulator (4 orders of harmonics and 3 orders of
mixing terms considered) takes approximately twenty four hours to simulate on an Intel
The derivation found in Appendix B.3 gives a theoretical model that gives the maximum
performance for an energy harvester which utilizes a high PAPR excitation signal. While
the method used has been shown to accurately predict the efficiencies of energy harvesters
80
(a) CW Excitation.
Figure 32: Harmonic content generated with a CW excitation (a) and POW excitation
(b) when used with a diode-based energy-harvesting circuit. Like a CW excitation, DC
and harmonics are generated along with the fundamental excitation frequency. However,
POWs also cause intermodulation products at integer multiples of the POW period TPOW .
A low-pass filter should help to remove some of this content, but intermodulation products
may many times be too low to effectively filter out.
81
which use high PAPR signals, it does have some limitations which are discussed in the
derivation.
Figures 33, 34, and 35 show the DC output voltage, DC output power, and power conver-
sion efficiency, respectively for an energy-harvesting circuit with a single Avago HSMS-2860
Schottky diode with a 300 Ω load under several different PAPR. Table 5 shows the relation-
ships between the duty cycle and the PAPR of the theoretical excitation signals. The duty
cycle of the POW is discussed in the appendix derivation, but corresponds to the percentage
of time that the excitation waveform is ‘on.’ Note that this means a 100% duty cycle is the
Figures 33, 34, and 35 demonstrate the benefits and limitations of using POWs. Two
major points can be discerned from these plots. First, POWs permit a diode to ‘turn on’ at
a lower power than a CW signal permits. This point is illustrated by the vertical asymptotes
of the signals with decreasing duty cycle (increasing PAPR) moving to the left on the input
power axis. As PAPR increases, the turn-on power decreases. Recall that these asymptotes
exist because of the simple diode model used in the derivation. Additionally, a POW is able
to generate more power and therefore higher energy-conversion efficiency at low powers than
a CW. Secondly, as duty cycle decreases, the output voltage, power, and efficiency at higher
powers decreases. This behavior occurs because the larger PAPR drives a larger current
through the diode (for a brief period of time) than a CW. This larger current translates to
additional energy lost across the diode series resistance Rs . Moreover, a large PAPR means
that the waveform will achieve the reverse breakdown voltage and subsequent clipping at
a lower input power than a CW. This is easily observed by the first drop in efficiency in
Figure 35. This early clipping means that the POW will more slowly increase its output
82
1
10
Duty Cycle=100 %
Duty Cycle=25 %
Duty Cycle=10 %
0 Duty Cycle=1 %
10
Output Voltage (V)
−1
10
−2
10
−3
10
−4
10
−40 −30 −20 −10 0 10 20 30 40
Input Power (dBm)
Figure 33: The theoretical DC output voltage for an energy-harvesting circuit with a single
Avago HSMS-2860 diode and 300 Ω load under 5.8 GHz POW excitation of various PA-
PR/duty cycles. As the PAPR increases, the diode is able to rectify DC voltage more easily
at lower powers while at higher powers, the DC voltage is lower because of an increased
voltage drop across the diode series resistance.
83
30
Duty Cycle=100 %
Duty Cycle=25 %
Duty Cycle=10 %
10 Duty Cycle=1 %
Output Power (dBm)
−10
−30
−50
−70
−40 −30 −20 −10 0 10 20 30 40
Input Power (dBm)
Figure 34: The theoretical DC output power for an energy-harvesting circuit with a single
Avago HSMS-2860 diode and 300 Ω load under 5.8 GHz POW excitation of various PAPR/-
duty cycles. As the PAPR increases, the diode is able to rectify DC power more easily at
lower powers while at higher powers, the DC power is lower because of an increased voltage
drop across the diode series resistance.
84
1
Duty Cycle=100 %
0.9 Duty Cycle=25 %
Duty Cycle=10 %
0.8 Duty Cycle=1 %
0.7
0.6
Efficiency
0.5
0.4
0.3
0.2
0.1
0
−40 −30 −20 −10 0 10 20 30 40
Input Power (dBm)
Figure 35: The theoretical energy-conversion efficiency for an energy-harvesting circuit with
a single Avago HSMS-2860 diode and 300 Ω load under 5.8 GHz POW excitation of various
PAPR/duty cycles. As the PAPR increases, the efficiency increases at lower powers while
at higher powers, the efficiency decreases due to power lost in the diode series resistance.
85
1
0.9
0.8
0.7
Maximum Efficiency
0.6
0.5
0.4
0.3
0.2
0.1
0
0 500 1000 1500 2000 2500 3000
PAPR
To further illustrate how different PAPR signals affect energy-harvesting circuits, the
Figure 36 using the same Avago diode and operating parameters in the previous theoret-
ical illustrations. As expected and as previously explained, as the PAPR increases, the
the maximum efficiency is at its largest value. As the PAPR increases, it asymptotically
approaches zero as more and more energy is lost across the diode series resistance. In
practicality, the diode would burn out far before it reaches a maximum efficiency of zero.
As mentioned before, a POW decreases the average turn-on power for a diode to reach
its voltage threshold. This relationship is visualized in Figure 37. Theoretically, this pro-
gression translates to a constant peak voltage VT with a decreasing average voltage. As this
86
peak voltage continues to increase, the turn-on voltage level will always be reached. How-
ever, holding the average power constant and increasing the PAPR will eventually damage
the diode. Although it may turn the diode on, the power level and efficiency which results
−15
−20
−25
Turn−on Power (dBm)
−30
−35
−40
−45
−50 1 2 3 4
10 10 10 10
PAPR
Figure 37: The input power for which the diode turns on for increasing PAPR in an energy-
harvesting circuit with a single Avago HSMS-2860 diode and 300 Ω load under 5.8 GHz
POW excitation.
The tradeoff between maximum efficiency at high powers and diode turn-on power is
illustrated in Figure 38. In this figure, increasing PAPR follows the curve moving right
to left. This relationship shows that there is an ideal PAPR for which the turn-on power
is decreased without significantly impairing the maximum efficiency of the diode. This
location is where the second derivative of the curve in Figure 38 is equal to zero. For this
diode configuration, this point corresponds to an optimal PAPR of 55 which yields a turn-on
power of -32 dBm and maximum efficiency of 30% - an decrease in turn-on power of 14 dB
87
1
0.9
0.8
0.7
Maximum Efficiency
0.6
0.5
0.4
0.3
0.2
0.1
0
−50 −45 −40 −35 −30 −25 −20 −15
Turn−on Power (dBm)
Figure 38: The maximum energy-conversion efficiency plotted against the turn-on power in
an energy-harvesting circuit with a single Avago HSMS-2860 diode and 300 Ω load under
5.8 GHz POW excitation. The PAPR increases as the efficiency decreases. An optimal
PAPR would be found at the inflection point. In this case, an optimal PAPR of 55 yields
a turn-on power of -32 dBm and a maximum efficiency of 30%, a 14 dB and 45% decrease,
respectively.
While a POW with a PAPR of 55 is the optimal point in terms of tradeoff between
decrease in turn-on power and reduction in maximum efficiency, the applicability of this
excitation signal would still need to be evaluated against the power/voltage requirements
of the application. Furthermore, the exact design of the POW and potential changes in the
The multi-tone, periodic nature of POWs increases the POW gain of energy harvesters
under low power excitation. However, the use of POWs has additional effects on the energy
harvester when compared to CW excitation. One additional complexity that occurs with
the use of POWs is the increased likelihood of ripple voltage. This ripple is a result of
88
1
the periodic pulsing of the POW which occurs every ∆f seconds. While this pulsing is
undesirable from a voltage ripple point of view, the harmonic content which causes the ripple
is a result of the same mechanism the increases POW gain. As the pulse becomes narrower
(PAPR is increased), the ripple will increase as the amount of time the output capacitor
can charge is reduced while the discharge time is increased as illustrated in Figure 39 and
Table 6.
Figure 39: As the POW pulse becomes narrower (PAPR is increased), the RMS time
width decreases which reduces the amount of time the output capacitor has to charge.
Furthermore, the time spent discharging is increased. This change results in increased
voltage ripple [7].
Peak voltages normalized to the 4-POW peak are listed to show the voltage increase as
a result of the subcarriers added in phase. Charge and discharge times are measured taking
only the largest peak into account as smaller peaks will typically be below the diode voltage
threshold in low-power circumstances. The charge time is thus defined as the time from
89
Table 6: POW subcarrier timing information.
zero volts to the maximum voltage, and the discharge time is defined as the time from the
maximum voltage until zero voltage immediately before the next maximum peak.
To help illustrate the behavior of an energy-harvesting circuit under POW excitation, the
single shunt rectenna from Section 3.6 was simulated under various POW excitations. Fig-
ure 40 illustrates the POW gain, DC output voltage, and ripple voltage for the single shunt
rectenna under 1-POW excitation (two frequencies) with a subcarrier spacing of 10 MHz.
Figure 40 shows at powers above -2 dBm, the POW gain is negative. Recall that a positive
POW gain corresponds to better performance than a CW. Negative POW gain means the
POW performs worse. While the POW gain is negative in this region it also has the highest
For powers between -2 dBm and -30 dBm, the POW gain is positive. These results
show that in this region, the energy harvester is performing more efficiently under POW
excitation as the PAPR helps to surpass the diode voltage threshold limitation and rectify
for a longer period of time. The DC output voltage also decreases as the power decreases,
and the rectifier becomes less efficient. Consequently, the voltage ripple increases. Below
-6 dBm, the voltage ripple surpasses 100 % ripple as the periodic nature of the POW
Below -30 dBm, the POW gain goes to zero as the power levels are too low, even with a
POW, to efficiently rectify the input voltage. DC voltage also goes to zero as there is very
little useful input power. Ripple voltage will continue to increase as power decreases as the
ripple is plotted as a percentage of DC voltage. The harmonics generated by the diode will
90
8
GPOW (dB)
6
4
2
0
−2
−30 −20 −10 0 10 20 30
Pin (dBm)
3
VDC (V)
0.1
0.01
0.001
0.0001
0
−30 −20 −10 0 10 20 30
Pin (dBm)
1000
VRipple (%)
600
400
200
100
−30 −20 −10 0 10 20 30
Pin (dBm)
Figure 40: Simulated POW gain, output DC voltage, and output ripple voltage for a 5.8 GHz
single shunt rectenna excited with a 1-POW with 10 MHz subcarrier spacing. At low
powers, voltage ripple is large as the rectifier cannot efficiently convert RF to DC and
intermodulation products dominate the output (From [7]).
POWs suffer from additional voltage ripple when compared to CW because output low-
pass filters on rectennas are typically designed to filter out the RF carrier. While these
output filters will remove the POW carrier waveform, intermodulation products from the
POW frequencies will cause frequency components at integer multiples of the subcarrier
frequency (n∆f ) to appear at the output. While increasing the time constant would lower
the cutoff frequency, these changes are difficult at microwave frequencies. Increasing the
load resistance may not be possible because of the device application and will affect the
energy conversion efficiency as the rectenna is very sensitive to changes in the resistive load.
in these filters must resonate above the RF carrier to filter it out. Since parasitic induc-
tances in the capacitor fabrication are set by its geometry, large capacitances are difficult
91
to obtain [78].
A POW could have any number of subcarriers each with its own unique amplitude, sub-
carrier spacing, and phase. Varying these parameters affects the PAPR, RMS bandwidth,
and RMS time width. These results, in turn, affect the POW gain and energy-harvesting
excitations, different M-POWs were simulated using the single shunt rectenna example.
Subcarriers produce the maximum PAPR when they are evenly spaced at integer multiples
of a single subcarrier frequency with all subcarriers in-phase. This choice ensures that in a
flat-phase fading environment, all subcarriers will combine in-phase at the same time and
produce a maximum PAPR. If subcarriers have different phases or the subcarrier spacing
is not done at integer multiples of a single frequency, the time domain waveform will have
a reduced PAPR in its main peak with that energy being shifted to secondary peaks. This
The optimal value of the subcarrier frequency spacing, however, is less clear. While the
signal bandwidth determined by the number of subcarriers and the subcarrier spacing should
fit within the regulatory band, this chosen value may not provide the best performance. To
investigate the effects on energy harvester efficiency, the single shunt rectenna was simulated
using a 1-POW (two subcarriers spaced by 2∆f ). Figure 41 shows the POW gain and
Figure 41 shows three distinct regions. The trade off region consists of large POW gain,
but also large voltage ripple. However, the small subcarrier spacing allows the POW to
fit in a more narrow bandwidth. If the application can tolerate more impulsive power or
add additional circuitry to smooth this output without affecting the rectenna performance,
this region may be acceptable. The transition region also has large POW gains and lower
voltage ripple as the subcarrier spacing increases. Subcarriers are further spaced, but are
close enough impedance matched to be absorbed. The larger subcarrier spacing also means
92
Figure 41: The POW gain of the designed rectenna was simulated with a 1-POW whose
subcarrier spacing was varied (top). It was determined that the maximum POW gain was
obtained when the subcarrier spacing was made equal to the rectenna load time constant.
The percent ripple voltage was also plotted (bottom).
their intermodulation products begin to be filtered out by the low-pass filter. However,
increased subcarrier spacing also means that bandwidth or matching constraints may limit
use in this region. Finally, the decay region has decreasing POW gains with minimal voltage
ripple. Intermodulation products are filtered out, but subcarriers are too far apart to be
impedance matched. This reduction in gain along with the large subcarrier spacing make
It was determined that the subcarrier spacing that gave the largest POW gain of 7.2 dB
at -11.5 dBm input power was approximately 31.6 MHz, the time constant of the low-pass
filter on the rectenna. At this value, there is a voltage ripple of 35 %. A maximum POW
gain occurs here because the subcarrier spacing is inversely proportional to the POW period.
Impedance matching and voltage ripple are of paramount importance when considering
subcarrier spacing. Subcarrier spacing must be balanced so all subcarriers are sufficiently
93
matched (minimal spacing), but also sufficiently spaced (maximum spacing) to minimize
voltage ripple. As shown in this simulation, this value will typically occur around the low-
pass filter cutoff frequency. Moreover, energy harvesters with a large number of stages may
be bandwidth limited due to the large number of diode junction capacitances in parallel
Although it has been shown in [63, 65] that increasing the number of subcarriers increases
the POW gain, these previous studies had not investigated the subcarrier spacing or energy-
harvesting circuit parameters. Recall the increasing the number of subcarriers increases
the PAPR. Figures 42 and 43 show the POW effect on the rectenna when additional equal
amplitude subcarriers (M-POW) are added at a subcarrier spacing of 31.6 MHz, the spacing
determined in the previous section. Although some of these POWs will take a bandwidth in
excess of the FCC allowed limit and may experience amplitude shifting because of the input
that an implementation of this method would require a POW designed simultaneously with
As expected, as the number of subcarriers increases, so does the maximum POW gain.
Additionally, as the number of subcarriers increases, the power at which the maximum gain
occurs decreases. This change corresponds to a lower power generating a higher PAPR
which can overcome the diode voltage threshold limitation better than a signal with a lower
PAPR. One can also see a decrease in improvement when adding additional subcarriers
due to the impedance mismatch of the larger bandwidth signal. To reduce these effects
and further increase POW gain, a smaller subcarrier spacing could be used. However, this
reduction will lead to an increase in ripple voltage. Thus, there should be an optimal
94
0
10 M=1
−1
10 M=2
(V)
−2
10 M=3
outDC
−3 M=4
10
−4 CW
V
10
−5
10
0 M=1
(dBm)
M=2
M=3
−50 M=4
outDC
CW
P
−100
−30 −20 −10 0 10 20 30
Pin (dBm)
Figure 42: DC output voltage and DC output power of the single-shunt rectenna with
increasing numbers of equal amplitude subcarriers with 31.6 MHz spacings. Larger number
of subcarriers allow the diode to operate more efficiently at low power as evident from the
larger voltages and powers between -30 dBm and -8 dBm compared to CW.
In the past, rectennas have been designed to be excited with a CW waveform. When
these design rules are followed, and a POW is used instead of a CW waveform, the rectenna
performance suffers. To increase the rectenna performance, the following design rules should
1. POW Design - For maximum POW gain, subcarriers should be spaced at integer
multiples of a single subcarrier frequency and arrive at the energy harvester all with
the energy-harvesting circuit time constant to produce maximum gain and minimum
ripple.
2. Output capacitance - The rectenna output capacitor (or resulting low-pass filter cutoff
95
M=1
20 M=2
GPOW (dB)
15 M=3
M=4
10
5
0
−5
−30 −20 −10 0 10 20 30
Pin (dBm)
0.8
M=1
M=2
0.6
M=3
Efficiency
M=4
0.4 CW
0.2
0
−30 −20 −10 0 10 20 30
Pin (dBm)
Figure 43: POW gain of M-POW as the number of subcarriers is increased. Subcarrier
spacing was set to be 31.6 MHz as determined from the previous section. Note that the
maximum gain occurs where the energy conversion efficiency is typically around 10%. Also
note that POWs do decrease the energy conversion efficiency at high powers.
sized appropriately to charge during the POW charging period. As the number of
subcarriers increases, the time constant must be adjusted via the output capacitor so
that these values are on the same order. Additionally, varying the output capacitance
will affect the input matching of the rectenna so these adjustments must be taken into
3. Input matching network - When the rectenna input matching network is designed,
an input power must be specified at which to match the circuit. This power should
be chosen based upon the intended device operating current, charging time, range,
and reliability. Figure 14 can help determine an approximate efficiency to make cal-
culations. It may be necessary to choose a low input power for matching to allow
the energy-harvesting circuit to absorb most of the power at this level. Although this
will result in more reflected power at high levels, the absolute input power will be
greater and energy-harvesting efficiency will also be greater since low input powers
96
When POWs are taken into account, the matching network design becomes signifi-
cantly more challenging. The increased bandwidth associated with POWs means that
the matching network must also increase its bandwidth. Designers must now match
the energy-harvesting circuit impedance over both multiple frequencies and powers.
tation, the input matching network should focus on matching at a low power. This
design choice will allow the most benefit from the POW excitation.
efficiency for high-power CW excitation, their design may have to be altered for POW
POWs, this structure will have reduced functionality since the full-wave rectifier is
Subcarriers that are farther away from this frequency will not reflect the same way
and the PAPR incident on the diode will be reduced. Likewise, the addition of mul-
tiple stages may make the use of multi-stage rectifiers difficult because of bandwidth
restrictions. Another issue with the rectenna topology and POWs is the increased
ripple voltage due to low POW duty cycle. Additional components may be required
Designing an energy harvester along with a POW adds significant complexity to an already
difficult problem. Not only must the energy harvester and POW be built in tandem, prac-
tical considerations such as realizing the designed POW must be taken into account. While
the design process follows that from Chapter 3.5.1, additional concerns are discussed which
3. Using the theory in Appendix B, determine if the design goals are met with a CW.
97
Keep in mind that the theory will be the maximum possible and not include many
losses.
4. If design goals are not met, use the POW theory in Appendix B.3 to determine if a
5. If a POW could benefit the circuit, determine how reasonable the PAPR would be. If
6. If the POW PAPR is reasonable, design the POW in its entirety and ensure goals are
satisfied.
7. Based on a desired operating point (frequency and input power, which will determine
8. Optimize the energy-harvesting circuit design using computer-assisted tools using the
Although POWs can provide large improvements in energy-harvesting efficiency, there are
several challenges and limitations that must be addressed before choosing to use a POW.
While some of these were mentioned as design rules and have been considered in this thesis,
4.6.1 Communication
Due to the additional number of subcarriers and the nature of backscatter communication,
POWs present unique challenges. Appendix D takes a quick look at some possible issues
for communicating with POWs, but this section is far from exhaustive. Some challenges
98
Figure 44: The methodology for designing an energy-harvesting circuit and POW.
The large PAPR of POWs is the reason why they can increase the efficiency of energy
harvesting circuits. However, generating these types of signals places great demands on the
dynamic ranges of ADCs and DACs and on the linearity and efficiency of power amplifiers
and low noise amplifiers. Appendix D also briefly investigates several transmitter and
receiver topologies.
While an ideal POW requires all subcarriers to be in-phase and at a relative amplitude level,
typical backscatter channels will fade frequency selectively in both phase and amplitude.
99
These channels could potentially cause a POW to arrive at an energy harvesting circuit
POW to fit within a channel frequency and time coherence bandwidth so as to maintain
a flat-fading channel. Additionally, channel sounding and other cognitive radio techniques
may be used to dynamically pre-distort a POW shape to ensure it arrives at the desire
Figure 45: The channel effects on a POW may vary the relative amplitude and phase of
each POW subcarrier. This variation will usually result in a sub-optimal shape will may be
less beneficial to the energy harvester than a CW.
Due to the non-linear nature of the rectifying elements in energy-harvesting circuits, when a
POW with multi-spectral content excites the rectifying device, many harmonics and mixing
terms will be present. Provided the circuit is correctly designed, none of these harmonics or
mixing terms should reflect back into the environment. However, if the matching network,
for instance, does not reject this content, the energy-harvesting circuit could generate inter-
ference across the RF spectrum despite it not requiring FCC certification as it is a passive
device.
4.6.5 Safety
The recommendations and limitations for electromagnetic fields has been an ever evolving
field due to the prevailing use of radio waves in consumer’s everyday lives. There has been
100
much discussion on the proper way to evaluate the safety of radio waves on human physiology
with many different methods having been suggested including: plane wave electric and
magnetic field limitations, specific absorption rate (SAR), maximum permissible exposure
(MPE), among others [106]. While POWs satisfy the requirement that the time-averaged
power over a single period remains at 1 W per FCC requirements [1], care must still be taken
to ensure that a SAR of 4 W/kg is not exceeded for POWs in the bandwidth of 1 MHz to
10 GHz per the ICNIRP requirement [107]. It should be noted that different organizations,
namely the IEEE, ICNIRP, FCC, and NCRP, all suggest different limitations in regards to
RF radiation exposure limits for biological organisms. This topic of safety is too large to
101
CHAPTER V
In an effort to validate the theoretical and simulated models presented in this thesis, pro-
totype energy-harvesting circuits were designed, and laboratory measurements were carried
out. The proceeding sections document the effectiveness of realizing these types of circuits
using the presented methodology. Two microwave energy-harvesting circuits were built at
5.8 GHz as examples: rectenna matched at 16 dBm, and rectenna matched at -10 dBm.
cost, and has acceptable parameters [2] for 5.8 GHz energy-harvesting circuits. To verify
the accuracy of the Avago HSMS-2860, a model complete with its SOT-23 package par-
asitics [8] was created in Agilent’s Advanced Design System (ADS) 2011.10 as shown in
Appendix E.2.1. The diode DC I-V curve was measured against several different batches
of diodes and compared to the simulated model and data sheet supplied model as shown in
Three main points can be discerned from the graphs. First, the ADS simulation matches
the diode data sheet. Second, while the diode curves agree relatively well around the turn
on voltage, notable differences exist for higher and lower powers. At lower powers, it is
102
0.08
Batch 1 − Diode A
0.07 Batch 1 − Diode B
Batch 2 − Diode A
0.06 Batch 2 − Diode B
Batch 3 − Diode A
Batch 3 − Diode B
0.05 Datasheet
Current (A)
Simulation
0.04
0.03
0.02
0.01
0
0 0.2 0.4 0.6 0.8 1 1.2
VoltageForward (V)
Figure 46: Plot of the measured, simulated, and data sheet supplied I-V curves from an
Avago HSMS-2682 series diode in a SOT-23 package.
difficult to accurately measure low levels of current and voltage and thus, a greater variation
is shown. At higher powers, parasitic resistances on the order of a couple ohms lead to a
reduction in the slope of the curves. Finally, the variation between batches of diodes is
apparent in these plots. While differences between each batch of diodes is apparent, diodes
from the same batch are relatively the same. Thus, while the ADS simulation will be used
for charge-pump design, one can expect a variation in performance due to the variation in
the fabrication process. Moreover, the observed difference is conducted at DC. Variations
This section outlines the software simulation and laboratory measurement setup used to
characterize all tested devices. Both circuits utilized the same dielectric substrate and
diodes from a single batch. Therefore, any variation between them will be solely due to
103
Batch 1 − Diode A
−2 Batch 1 − Diode B
10
Batch 2 − Diode A
Batch 2 − Diode B
−3
10 Batch 3 − Diode A
Batch 3 − Diode B
Datasheet
Current (A)
−4
10 Simulation
−5
10
−6
10
−7
10
−2 −1 0
10 10 10
VoltageForward (V)
Figure 47: Log plot of the measured, simulated, and data sheet supplied I-V curves from
an Avago HSMS-2682 series diode in a SOT-23 package.
As previously discussed, both circuits were simulated in Agilent ADS 2011.10 using the
harmonic balance and large-scale S-parameter techniques. Harmonic balance was utilized
to generate CWs and POWs and measure the circuit DC output voltage. Four harmonics
of each fundamental frequency were considered along with three orders of intermodulation
products. It was found that increasing each of these values did not significantly affect the
to simulate as the total number of frequency terms considered grows exponential with
the number of subcarriers. The large-scale S-parameter method was used to generate S-
parameter plots and simulate the impedance of the non-linear energy-harvesting circuits.
104
specified for which the S-parameters are generated.
Each circuit utilized microstrip lines implemented with the ADS microstrip model with a
Rogers Duroid 5880 dielectric substrate and parameters shown in Table 8 [108].
Parameter Value
Dielectric constant (r ) 2.20
tan δ 0.0009 @ 10 GHz
Dielectric thickness 0.031”
Conductor thickness 9µm
While theoretical predictions of microstrip length and width were estimated with the
appropriate equations, the large number of parasitic elements present in the diode package
was optimized using the ADS built-in ‘Random Search’ method according to the following
and the constraint function Hk (x). Optimization was completed at a single power Po to
optimize performance around this point. A fixed microstrip line width Wn was set to limit
x = [RLoad Ln ] (42)
105
under the conditions
100 Ω RLoad 10 kΩ
≤ ≤ (43)
0 mil Ln 400 mil
These optimization criteria can be summarized as follows. The circuit will be optimized
the square of the DC output voltage, this goal will also yield the maximum voltage. The
optimization will be constrained to a single frequency at 5.8 GHz and a power Po dBm
chosen per the specific circuit. Parameters to be swept are the output resistance microstrip
lengths. The resistance is allowed to vary across a wide range of values as any is permissible.
Finally, lengths can vary from zero to 400 mils. This maximum length was chosen to limit
While ADS is capable of optimizing specific circuits under a given excitation, its simulator
and optimizer have certain limitations. Some of these limitation include only simulating
a single excitation signal per design and inability to vary component types (i.e. different
diodes) in optimization. To address these limitations and to reduce the simulation time,
MATLAB was used to script and control the ADS simulations. The script follows the steps
listed below.
2. Supplement netlist with functions to control desired parameters (POW design, com-
6. Make decisions based on the data or run another instance of the simulation by changing
106
Using these steps, a streamlined process to simulate the performance of energy-harvesting
Input capacitors are used in energy-harvesting circuits to prevent DC currents from un-
necessarily flowing towards the antenna. As lumped element capacitors can be difficult
to completely control at 5.8 GHz, a microstrip interdigitated capacitor was designed and
optimized according to the procedure in Figure 25. The S-parameter simulation had goals
to maximize the return loss and minimize the insertion loss. The optimization yielded the
dimensions detailed in Appendix E.2.2 and the S-parameters shown in Figure 48. Although
not lossless, the resultant capacitor performs well at 5.8 GHz with a return loss of -33.1 dB
and an insertion loss of -0.51 dB. An image of the layout is also shown in Figure 49 for
completeness.
circuits under CW excitation. As show in Figure 50, an Agilent E8247 signal generator is
the 5.8 GHz source for all measurements. This signal passes into a Mini-Circuits ZVE-8G
power amplifier with a gain of 30 dB. To suppress any out-of-band noise or harmonics gener-
ated by the amplifier, a Mini-Circuits VBFZ-5500-S+ bandpass filter (4.9 GHz - 6.2 GHz)
follows the amplifier before the signal passes through two different variable attenuators.
These two attenuators allow the measurement system to operate at a fixed, linear point
on the amplifier gain curve. The first attenuator, an HP-8495B, is a step attenuator that
FSCM 12457, is a rotary attenuator that provides continuous attenuation from 0-40 dB.
To measure the true input power into the rectenna or charge pump, an M/A-COM
2025-4139-10 coupler was used immediately after the variable attenuators. Following this,
a Wentec F2658-0600-67 circulator (4.0 GHz - 8.0 GHz) was placed between the coupler
and the circuit under test. An Agilent E4404B spectrum analyzer was placed at the coupler
107
0
−5
−10
Return Loss (dB)
−15
−20
−25
−30
−35
0 5 10 15 20 25 30
Frequency (GHz)
−5
Insertion Loss (dB)
−10
−15
−20
−25
0 5 10 15 20 25 30
Frequency (GHz)
Figure 48: Simulated S-parameters from interdigitated capacitor used in the prototyped
rectennas.
108
Figure 49: Physical layout for the input interdigitated capacitor used for the prototyped
rectenna circuits. The length of the IDC is 740 mil and the maximum width is 689 mil.
-10 dB port to measure the reference signal. Reflected signals from the circuit (fundamental
and harmonics) are then passed back through the circulator where they were measured by
measures the signal across the load. An oscilloscope is used as its possible that the load
To calibrate the system and calculate the real input power into the test circuit and the
value of the reflected power and harmonics, all of the circuit components after the variable
attenuators had to be characterized with their losses. To accomplish this, the coupler and
circulator with all connecting SMA cables and barrel connectors were disconnected from the
system and measured with an Agilent E5071B vector network analyzer. A block diagram
109
Figure 50: Block diagram of the energy-harvesting circuit measurement system for CW
excitations. A circulator provides a means of measuring reflected power to determine the
matching of the circuit while a coupler allows the input power level to be monitored.
showing the measured components is shown in Fig. 51. The measured parameters are shown
in Figs. 52, 53, and 54. Their values at 5.8 GHz are summarized in Table 9.
Figure 51: Block diagram of calibrated section of the charge-pump measurement system.
By measuring the losses of the various signal paths, an accurate value of the input and
reflected power can be rectified.
Using the information in Table 9, the true input power into the energy-harvesting circuit
under test is
where Pin is the input power at the fundamental frequency to the energy-harvesting circuit
under test, and PB is the coupler output power at port B measured by the spectrum
110
0
−2.5
−5
SBA (dB)
−7.5
−10
−12.5
−15
4 4.5 5 5.5 6 6.5 7 7.5 8 8.5
Frequency (Hz) x 10
9
Figure 52: Measured insertion loss SBA of the coupler-circulator sub-circuit shown in Fig-
ure 51.
−0.5
SCA (dB)
−1
−1.5
−2
4 4.5 5 5.5 6 6.5 7 7.5 8 8.5
Frequency (Hz) x 10
9
Figure 53: Measured insertion loss SCA of the coupler-circulator sub-circuit shown in Fig-
ure 51.
where Pref lected is the power reflected by the energy-harvesting circuit, and PD is the power
measured by the spectrum analyzer at the output of the circulator at port D. Note that
these measurements were conducted at 5.8 GHz. While reflected power measurements are
discussed at harmonic frequencies, 11.6 GHz, 17.4 GHz, and 23.2 GHz, the system has not
been calibrated up to these frequencies. Furthermore, the Wentec circulator is not rated
past 8.0 GHz. Consequently, the reported levels are listed merely as a verification of their
existence.
111
0
−0.5
−1
−1.5
SDC (dB)
−2
−2.5
−3
−3.5
4 4.5 5 5.5 6 6.5 7 7.5 8 8.5
Frequency (Hz) x 10
9
Figure 54: Measured insertion loss SDC of the coupler-circulator sub-circuit shown in Fig-
ure 51.
Previous work has already developed a system for measuring charge-pump output power
and efficiency for various load impedances, frequencies, and input powers from a CW exci-
tation [109]. This system has been described in the previous section as well. The previous
work was extended to measure the effects of POWs on energy-harvesting circuits. Figure 55
shows a block diagram of the constructed system to characterize circuit performance when
On a PC, MATLAB was used to create a baseband POW and then program an Agilent
33250A function generator with the waveform via GPIB. The function generator then output
this signal to a Mini-Circuits ZMX-7GR mixer (3.7 GHz - 7.0 GHz) where the POW was
upconverted to 5.8 GHz before being amplified. After this point, the characterization system
follows from the CW system. For POWs, it is especially important at the coupler output
to verify the POW shape and signal levels as the amplifier non-linearity may cause clipping
112
Figure 55: Block diagram of the system used to generate POWs and characterize their
performance with energy harvesters. This system requires additional attention to ensure
the amplifiers operate linearly so as to not distort the POW.
output signal because of the presence of both DC and intermodulation products causing a
ripple.
Because the ladder part of the characterization system is identical to the CW system,
(44) and (45) will both be valid for measuring the input and reflected signal powers, respec-
tively.
Two single diode rectifying circuits were prototyped to demonstrate the feasibility of us-
ing the aforementioned design methodology and co-validate the theoretical predictions and
matched at -10 dBm were chosen to showcase the differences in energy harvester behavior
at these extremes. While a high-power rectenna might be chosen for SPS applications or
harvested power might be beneficial, its functionality at the expected low-power area of op-
eration would be compromised. Hence, a low-power rectenna will provide additional energy
at low powers while sacrificing performance a high powers. Using two energy harvesters in
this configuration has been suggested in [110], and an optimization routine is outlined to
113
Figure 56: Photograph of the experimental setup for measuring energy-harvester efficiency
under POW excitation.
The rectenna in Figure 57 was built and designed using a single Avago HSMS-2860 Schot-
tky diode on a 31 mil Rogers 5880 substrate and impedance matched for an input power
of 16 dBm at 5.8 GHz. Complete details for this rectenna are listed in the netlist in Ap-
pendix E.2.3. Measured results from an Agilent E5071B network analyzer show that this
rectifier has the best resonant frequency at 6.06 GHz as shown in Figure 58. While the
harvester could convert energy from anywhere in this band, the low return loss at 6.06 GHz
Using the measurement setup discussed in Sections 5.3.1 and 5.3.2, the output voltage
of the high-power rectenna was measured for CW, 1-POW, 2-POW, and 3-POW excitations
for a center frequency at 6.06 GHz and a subcarrier spacing of 10 MHz. From this voltage,
the DC output power and energy harvester efficiency were computed as shown in figure 59.
Higher order POWs were not utilized because as the Mini-Circuits power amplifier was
not sufficiently linear for 4-POWs and higher. The 1-POW, 2-POW, and 3-POW had
114
Figure 57: Annotated photograph of the high-power rectenna. Note the rectenna was placed
on standard engineering graph paper with a grid spacing of 0.25 inches or 6.35 mm.
intermodulation products at 27 dB, 25.5 dB, and 18 dB below the excitation frequencies
measured at the coupler output, respectively. The subcarrier spacing of 10 MHz was chosen
because of function generator limitations, and its falling within the transition region of the
The measurements in Figure 59 follow the expected trend and are summarized in Ta-
ble 10. At high powers, the CW excitations performs the most efficient, but its efficiency
drops off the most quickly and turns off at the highest power. As POWs with increasing
PAPR are utilized, the maximum efficiency drops, but the turn-on power decreases. Note
below several hundred micro-volts. Note that this is why the 2-POW and 3-POW never
appear to become more efficient than the 1-POW. However, it is apparent that if their
Figures 60 through 63 show the same measured results compared against the theoretical
model from Appendix B and simulated results using the method in Section 5.2.1. These
graphs show the same basic trends. First, the simulations and the measurements never
surpass the performance of the theoretical model. Since the theory is a best-case model,
this fact is not surprising. Next, the measured results line up relatively well with the
115
0
−2
−4
Return Loss (dB)
−6
−8
−10
−12
−14
5 5.2 5.4 5.6 5.8 6 6.2 6.4 6.6
Frequency (GHz)
Figure 58: Measured return loss of the high-power rectenna at a power of -10 dBm.
simulations. Although the efficiency results may appear significantly off, output powers are
predicted within 2 dB. These differences can be attributed to the use of an infinite ground
plane in the simulations and the diode variations observed in Figures 46 and 47.
Although not widely discussed, a tradeoff when using POWs is the ripple voltage on the
output caused by the periodic voltage peaks. Figure 64 shows some examples of what this
ripple may look like for 1-POW, 2-POW, and 3-POW at arbitrary power levels. Recall from
Section 4.4.1 that this ripple will changes as a function of power and subcarrier spacing.
116
Measured
20
10
−10
Pout (dBm)
−20
−30
−40 CW
1−POW
−50 2−POW
3−POW
−60
−20 −10 0 10 20
Pin (dBm)
0.5
0.4
0.3
0.2
0.1
0
−20 −10 0 10 20
Pin (dBm)
(b) Efficiency.
Figure 59: Measured output power and energy-harvester efficiency for the fabricated high-
power rectenna under CW, 1-POW, 2-POW, and 3-POW excitations with a center frequency
of 6.06 GHz and subcarrier spacing of 10 MHz.
117
CW
20
10
0
−10
Pout (dBm)
−20
−30
−40
−50
−60
Measured
−70 Theoretical
Simulated
−80
−20 −10 0 10 20 30
Pin (dBm)
0.5
0.4
0.3
0.2
0.1
0
−20 −10 0 10 20 30
Pin (dBm)
(b) Efficiency.
Figure 60: Theoretical, simulated, and measured output power and energy-harvester effi-
ciency for the fabricated high-power rectenna under CW excitation with a center frequency
of 6.06 GHz.
118
1−POW
20
10
0
−10
Pout (dBm)
−20
−30
−40
−50
−60
Measured
−70 Theoretical
Simulated
−80
−20 −10 0 10 20 30
Pin (dBm)
0.5
0.4
0.3
0.2
0.1
0
−20 −10 0 10 20 30
Pin (dBm)
(b) Efficiency.
Figure 61: Theoretical, simulated, and measured output power and energy-harvester ef-
ficiency for the fabricated high-power rectenna under a 1-POW excitation with a center
frequency of 6.06 GHz and subcarrier spacing of 10 MHz.
119
2−POW
20
10
0
−10
Pout (dBm)
−20
−30
−40
−50
−60
Measured
−70 Theoretical
Simulated
−80
−20 −10 0 10 20 30
Pin (dBm)
0.5
0.4
0.3
0.2
0.1
0
−20 −10 0 10 20 30
Pin (dBm)
(b) Efficiency.
Figure 62: Theoretical, simulated, and measured output power and energy-harvester ef-
ficiency for the fabricated high-power rectenna under a 2-POW excitation with a center
frequency of 6.06 GHz and subcarrier spacing of 10 MHz.
120
3−POW
20
10
0
−10
Pout (dBm)
−20
−30
−40
−50
−60
Measured
−70 Theoretical
Simulated
−80
−20 −10 0 10 20 30
Pin (dBm)
0.5
0.4
0.3
0.2
0.1
0
−20 −10 0 10 20 30
Pin (dBm)
(b) Efficiency.
Figure 63: Theoretical, simulated, and measured output power and energy-harvester ef-
ficiency for the fabricated high-power rectenna under a 3-POW excitation with a center
frequency of 6.06 GHz and subcarrier spacing of 10 MHz.
121
3
2.5
Vout (Volts)
1.5
0.5
0
−500 −400 −300 −200 −100 0 100 200 300 400 500
Time (nanoseconds)
(a) 1-POW.
3
2.5
2
Vout (Volts)
1.5
0.5
0
−500 −400 −300 −200 −100 0 100 200 300 400 500
Time (nanoseconds)
(b) 2-POW.
3
2.5
2
Vout (Volts)
1.5
0.5
0
−500 −400 −300 −200 −100 0 100 200 300 400 500
Time (nanoseconds)
(c) 3-POW.
Figure 64: Example measured outputs from the prototyped high-power rectenna for different
POWs. Note how the ripple increases with increasing PAPR.
122
5.4.2 Low-power Rectenna
While the high-power rectenna demonstrates some POW improvements, POWs are designed
a circuit which is matched at low powers will be much more substantial and representative.
Figure 65 shows an annotated version of the prototyped low-power charge pump matched
at -10 dBm of input power at 5.8 GHz using the same diode and substrate as the high
power rectenna. The only difference with this rectenna was that the ADS optimization
input power condition was changed. Complete details for this rectenna are listed in the
netlist in Appendix E.2.4. Figure 66 shows this rectenna resonates at 5.46 GHz as well.
Figure 65: Annotated photograph of the low-power rectenna. Note the rectenna was placed
on standard engineering graph paper with a grid spacing of 0.25 inches or 6.35 mm.
Output power and efficiency measurements for the low-power rectenna are shown in
Figure 67 using CW, 1-POW, 2-POW, and 3-POW. For this device, the improvement
using POWs is much more pronounced because of the better impedance matching. As the
power decreases, higher PAPR POWs generate more power and are also more efficient than
mismatch as expected. Also note that the breakdown voltage occurs for progressively lower
powers with increasing PAPRs. Table 11 lists some of the important measured values from
123
0
−5
−10
Return Loss (dB)
−15
−20
−25
−30
5 5.2 5.4 5.6 5.8 6 6.2 6.4 6.6
Frequency (GHz)
Figure 66: Measured return loss of the low-power rectenna at a power of -10 dBm.
the low-power rectenna under 3-POW excitation. This excitation was chosen because it
Figures 68 through 71 compare the low-power rectenna measurements with the ADS
simulations and theoretical predictions. While the basic trends appear correct in these com-
measurements seem to more closely approach the theoretical maximum. This discrepancy
can also be attributed to the variation in the diodes from Figure 47. Specifically, this plot
shows that for input voltages below 100 mV, the measured current is much greater than
the data sheet or simulation predicts. Consequently, since the data sheet values were used
for the theoretical prediction and the simulations (and are lower than the measurements),
124
the measured output power and efficiency are better than the simulations. Likewise, if the
correct diode parameters were extracted for this batch of diodes, the theoretical curve would
Solving this wide variation between batches of diodes could be solved in several ways.
Different batches of diodes could be purchased, characterized, and compared. Diodes from
the best batch could then be utilized in the energy-harvesting circuits. These characteri-
zations could be done by a simple IV curve measurement as was done in this thesis, or a
While the previous sections have compared the measured results to theoretical limits and
circuits stack up against the current state-of-the-art. To demonstrate this, the energy
harvesting efficiencies from the high-power rectenna under CW and the low-power rectenna
Figure 72 shows the prototyped high-power rectenna falls within the expected energy-
harvesting efficiencies for a 5.8 GHz as does the low-power rectenna under CW excitation.
However, when the low-power rectenna is excited with a 3-POW, its efficiency surpasses
the best energy harvesting efficiency at low powers for 5.8 GHz. It should be noted that it
is possible that the other circuits used to obtain these efficiencies may also improve under
POW. However, to our knowledge, this is the most efficiency 5.8 GHz rectifying circuit ever
reported. Recall that some of these important values are listed in Table 11.
125
Measured
20
10
−10
Pout (dBm)
−20
−30
−40 CW
1−POW
−50 2−POW
3−POW
−60
−20 −10 0 10 20
Pin (dBm)
0.5
0.4
0.3
0.2
0.1
0
−20 −10 0 10 20
Pin (dBm)
(b) Efficiency.
Figure 67: Measured output power and energy-harvester efficiency for the fabricated low-
power rectenna under CW, 1-POW, 2-POW, and 3-POW excitations with a center frequency
of 5.46 GHz and subcarrier spacing of 10 MHz.
126
CW
20
10
0
−10
Pout (dBm)
−20
−30
−40
−50
−60
Measured
−70 Theoretical
Simulated
−80
−20 −10 0 10 20 30
Pin (dBm)
0.5
0.4
0.3
0.2
0.1
0
−20 −10 0 10 20 30
Pin (dBm)
(b) Efficiency.
Figure 68: Theoretical, simulated, and measured output power and energy-harvester effi-
ciency for the fabricated low-power rectenna under CW excitation with a center frequency
of 5.46 GHz.
127
1−POW
20
10
0
−10
Pout (dBm)
−20
−30
−40
−50
−60
Measured
−70 Theoretical
Simulated
−80
−20 −10 0 10 20 30
Pin (dBm)
0.5
0.4
0.3
0.2
0.1
0
−20 −10 0 10 20 30
Pin (dBm)
(b) Efficiency.
Figure 69: Theoretical, simulated, and measured output power and energy-harvester ef-
ficiency for the fabricated low-power rectenna under a 1-POW excitation with a center
frequency of 5.46 GHz and subcarrier spacing of 10 MHz.
128
2−POW
20
10
0
−10
Pout (dBm)
−20
−30
−40
−50
−60
Measured
−70 Theoretical
Simulated
−80
−20 −10 0 10 20 30
Pin (dBm)
0.5
0.4
0.3
0.2
0.1
0
−20 −10 0 10 20 30
Pin (dBm)
(b) Efficiency.
Figure 70: Theoretical, simulated, and measured output power and energy-harvester ef-
ficiency for the fabricated low-power rectenna under a 2-POW excitation with a center
frequency of 5.46 GHz and subcarrier spacing of 10 MHz.
129
3−POW
20
10
0
−10
Pout (dBm)
−20
−30
−40
−50
−60
Measured
−70 Theoretical
Simulated
−80
−20 −10 0 10 20 30
Pin (dBm)
0.5
0.4
0.3
0.2
0.1
0
−20 −10 0 10 20 30
Pin (dBm)
(b) Efficiency.
Figure 71: Theoretical, simulated, and measured output power and energy-harvester ef-
ficiency for the fabricated low-power rectenna under a 3-POW excitation with a center
frequency of 5.46 GHz and subcarrier spacing of 10 MHz.
130
100
80
Efficiency (%)
60
900 MHz
40
2.4 GHz
5.8 GHz
Above 5.8 GHz
20
Low−power, CW
Low−power, 3−POW
High−power, CW
0
−30 −20 −10 0 10 20 30 40 50 60 70
Input Power (dBm)
Figure 72: Measurements of high-power and low-power rectenna efficiencies plotted with
the state-of-the-art.
131
CHAPTER VI
CONCLUSIONS
Energy-harvesting circuits permit the flow of power without wires between great distances
and are an enabling technology for SPS and passive sensors alike. Improvements in semicon-
ductor technology, circuit topologies, and POWs are pushing the possibilities of applications
such as smart dust, insect-like drones, and augmented reality-based contact lenses. It has
been shown through theoretical models, computer simulations, and laboratory measure-
ments that POWs help to enable the efficient transfer of power at very low power levels.
This thesis has demonstrated a systematic optimization approach for designing energy-
harvesting circuits that utilize CWs and POWs alike. Using this approach, a low-power
5.8 GHz rectenna showed an output power improvement of over 20 dB at -20 dBm input
power using a 3-POW compared to CW. The resultant efficiency is the highest reported
efficiency for low-power 5.8 GHz energy harvesters. While design challenges exist, this the-
sis has helped to outline some of the issues and discuss the sources and tradeoffs involved
in solving these challenges. Theoretical models developed in this thesis predict improve-
ments in diode turn-on power of over 20 dB. This improvement in turn-on power includes
132
an improvement in output power of hundreds of dB when compared to CW. However, this
improvement comes at the tradeoff of lower maximum efficiency which approaches zero per-
cent as PAPR increases. Furthermore the difficulty in generating and communicated with
the POWs which enable these performance increases will have to be addressed. Along with
technical improvements from other disciplines, Tesla’s dream of truly wireless world can
become reality.
veloping and evaluating RFID technology. This testbed is modular in both hardware
and software and lets users evaluate different RF front ends, microwave structures,
cal daughter card ports. These results help to determine the power requirements of a
passive RFID tag which leads to the design of the energy-harvesting circuit.
lator was implemented on an Ettus Research USRP N200 using GNUradio. This
system demodulates digital data from an RFID tag using a variety of modulations
cations protocol. Paired with a R.E.S.T. and a direct down-conversion receiver board,
ducted to investigate the various methods of RF energy harvesting and the maximum
conversion efficiencies that were obtained. Using this data, theoretical trends were
demonstrate the maximum realistic efficiency point for an energy harvester has been
added to the theory. Previously, the model by Yoo and Chang [6] made no mention
of the inclusion of reverse breakdown voltage, which limits the maximum DC voltage,
133
and therefore efficiency, of a rectifying element.
• Theoretical energy-harvesting efficiency for N-stage Dickson charge pump - The the-
oretical energy-harvesting efficiency for a rectenna was extended for the case of an
N-stage Dickson charge pump. This derivation allows designers to theoretically vary
the diode parameters and number of stages without performing large circuit-based
simulations.
• Theoretical energy harvesting efficiency for a POW - While there are an infinite num-
large PAPR. A duty-cycled sinusoid or square POW was used to approximate a large
derivation allows for a comparison of the first-order improvements that a POW could
provide. Moreover, the use of this theory allows for predicting energy harvester perfor-
mance at very large PAPR which would be nearly impossible with current computer
modeling tools.
harvesting circuit designs relied on extensive testing and simulation to determine the
best diode and circuit design for their energy-harvesting application. A clear method
utilizing the theoretical methods previously discussed helps designers to better under-
there has been little discussion on the circuit effects from a POW. This thesis has
discussed the implications of POW use on the circuit design and noted critical pieces
which must be addressed when a POW is to be used. These considerations will help
• Highest reported low-power efficiency for a 5.8 GHz energy harvester - Using the
134
5.8 GHz excited with a 3-POW using a subcarrier spacing of 10 MHz yielded a con-
• Diode evaluation - In this thesis, some of the basic diode parameters were broken
down into their respective semiconductor pieces and a brief mention of their relation-
ships were discussed. However, it would be very useful to determine the theoretical
ing efficiency.
dertaken to determine the optimal POW for a given energy-harvesting circuit. This
balance simulation of the circuit under POW excitation. However, it was discovered
after discussions with Agilent that the simulator is incapable of parallelizing the de-
sired simulation and will run solely on one CPU core. Since the harmonic balance
technique takes into account the number of subcarriers, their harmonics, and all pos-
greater than six, simulations take longer than 20 hours using an Intel Core i7. To
complete this work, a custom harmonic balance simulator could be created to use
• Integrated circuit design - As with most circuits, the largest improvements typically
result when the design is moved to an integrated circuit. Certainly moving away from
135
micro controller designs will greatly increase the efficiency and lower the power and
voltage requirements of the tag digital logic. Furthermore, the energy harvester will
benefit from lower implementation losses. All of these improvements should results in
a tag that should operate at a far greater range than any previous passive 5.8 GHz
RFID tag.
• C.R. Valenta, M.M. Morys, and G.D. Durgin, “Theoretical Energy-conversion Effi-
• C.R. Valenta and G.D. Durgin, “Harvesting Wireless Power: Survey of Energy-
Microwave Magazine, Vol. 15, No. 4, June 2014, pp. 108 - 120.
• L. Blanca, J. Block, M. Almada, J. Gonzalez, C.R. Valenta, and G.D. Durgin, “Char-
acterization of a 5.8 GHz 4-Stage Dickson Charge Pump with Resistive Loads,” Geor-
• C.R. Valenta, P.A. Graf, M.S. Trotter, G.A. Koo, G.D. Durgin, W.G. Daly, and B.J.
IEEE Antennas and Propagation Magazine, August 2011, pp. 231-240. (Invited).
• C.R. Valenta and G.D. Durgin, “Link Budgets for Backscatter Radio and RFID Sys-
• B.R. Marshall, C.R. Valenta, and G.D. Durgin, “RF Energy Harvesting Compari-
son between N-by-N Staggered Pattern Charge Collectors and N 2 Rectennas,” IEEE
Wireless Power Transfer Conference 2013, Perugia, Italy, May 15 - 16, 2013.
136
• G.D. Durgin, C.R. Valenta, M.B. Akbar, M.M. Morys, B.R Marshall, and Y. Lu,
“Modulation and Sensitivity Limits for Backscatter Receivers,” IEEE RFID 2013,
• C.R. Valenta and G.D. Durgin, “Rectenna Performance Under Power-optimized Wave-
form Excitation,” IEEE RFID 2013, Orlando, Florida, April 30 - May 2, 2013.
• C.R. Valenta and G.D. Durgin, “R.E.S.T. - a Flexible, Semi-Passive Platform for
Developing RFID Technologies,” IEEE Sensors 2012, Taipei, Taiwan, October 28-31,
2012.
• C.R. Valenta, R. Hasse, M.B. Akbar, W. Hunsicker, K. Naishadham, and G.D. Durgin,
“Omni-directional Loop Antenna for a 5.8 GHz Microwave Backscatter Tag,” IEEE
International Symposium on Antennas & Propagation, Chicago, IL, July 8-14, 2012.
gin, “Power Conversion Gain as a Design Metric for RFID Systems,” IEEE Interna-
tional Symposium on Antennas & Propagation, Chicago, IL, July 8-14, 2012.
• M.B. Akbar, M.M Morys, C.R. Valenta, and G.D. Durgin, “Range Improvement of
Backscatter Radio Systems at 5.8 GHz using Tags with Multiple Antennas,” IEEE
International Symposium on Antennas & Propagation, Chicago, IL, July 8-14, 2012.
• M.S. Trotter, C. R. Valenta, G.A. Koo, B.R. Marshall, and G.D. Durgin, “Multi-
antenna Techniques for Enabling Passive RFID Tags and Sensors at Microwave Fre-
• C.R. Valenta, P.A. Graf, M.S. Trotter, G.A. Koo, B. J. Schafer, and G.D. Durgin.
“Backscatter Channel Measurements at 5.8 GHz Across High Voltage Corona,” IEEE
• C.R. Valenta, P.A. Graf, M.S. Trotter, G.A. Koo, W.G. Daly, B.J. Schafer, and
G.D. Durgin. “Transient Backscatter Channel Measurements at 5.8 GHz Across High
137
Voltage Insulation Gaps,” 32nd Annual AMTA Symposium, Atlanta, GA, October
10-15, 2010.
6.4.3 Presentations
• Ultra-low Power Energy Harvester Design for Passive RFID-enabled Sensing Devices
138
– Tokyo Institute of Technology – Tokyo, Japan (December 19, 2012).
• Electromagnetics in a Nutshell
ary 2011).
139
APPENDIX A
• Data in tabular form showing the current state-of-the-art energy harvesting efficiencies
plotted in Figure 14.
140
Table 12: State-of-the-art UHF energy conversion efficiencies.
Efficiency (%) Input Power (dBm) Frequency (MHz) Rectifier Element Source
1.2 -14 950 0.3µm CMOS transistor [111]
5.1 -14.1 920 0.18µm CMOS transistor [112]
10 ∗ -22.6 906 0.25µm CMOS transistor [85]
11 -14 915 90nm CMOS transistor [86]
12.8 -19.5 900 0.18µm CMOS, CoSi2 − Si Schottky [82]
13 -14.7 900 0.35µm CMOS transistor [113]
16.4 -9 963 0.35µm CMOS transistor [114]
141
18 -19 869 0.5µm CMOS, SiT i Schottky [83]
26.5 -11.1 900 0.18µm CMOS transistor [115]
36.6 -6 963 0.35µm CMOS transistor [114]
47 -8 915 0.18µm CMOS transistor [55]
49 ∗ -1 900 Skyworks SMS7630 Si Schottky [116]
60 ∗ -8 906 0.25µm CMOS transistor [85]
60 -3 915 0.13µm CMOS transistor [117]
69 -3 915 0.18µm CMOS transistor [55]
∗Efficiencies include antenna effects in their efficiency calculations. Non-asterisked efficiencies include efficien-
cies only related to the rectifier.
Table 13: State-of-the-art RF energy conversion efficiencies in the 2.45 GHz band.
Efficiency (%) Input Power (dBm) Frequency (MHz) Rectifier Element Source
2.01 ∗ -13.3 2450 Skyworks SMS7630 Si Schottky [118]
9∗ -13 2450 Skyworks SMS7630 Si Schottky [116]
10.5 ∗ -20.4 2450 Skyworks SMS7630 Si Schottky [119]
15 -20 2450 Avago HSMS-2852 Si Schottky [120]
20 ∗ 2.4 3000 Skyworks SMS7630 Si Schottky (array) [121]
28 -20 2450 Avago HSMS-2852 Si Schottky [101]
37 ∗ -25.7 2450 Silicon on Saphire 0.5 µm CMOS transistors [122]
45 -10 2450 Avago HSMS-2852 Si Schottky [120]
53 10 2450 Avago HSMS-2852 Si Schottky [120]
142
55 ∗ 62.7 2450 Thermionic [24]
57 0 2450 Avago HSMS-282x Si Schottky [56]
60 -5 2450 Avago HSMS-2852 Si Schottky [120]
66.8 10 2450 Avago HSMS-2860 Si Schottky [123]
72 0 2100 Avago HSMS-282x Si Schottky [56]
72.8 8 2450 Skyworks SMS7630 Si Schottky [90]
83 ∗ 20 2450 M/A-COM 4E1317 GaAs Schottky [30]
85 ∗ 15 2450 GaAs Schottky [87]
90.6 ∗ 39 2450 GaAs − P t Schottky [29]
∗Efficiencies include antenna effects in their efficiency calculations. Non-asterisked efficiencies include efficien-
cies only related to the rectifier.
Table 14: State-of-the-art microwave-energy conversion efficiencies at 5.8 GHz.
Efficiency (%) Input Power (dBm) Frequency (MHz) Rectifier Element Source
18∗ 0 5800 M/A-COM 4E1317 GaAs Schottky [30]
23 -10 5800 Unspecified Schottky [28]
34 ∗ 0 5800 Avago HSMS-8202 Si Schottky [91]
50 ∗ 10 5800 M/A-COM 4E1317 GaAs Schottky [92]
51.5 10 5800 Avago HSMS-2860 Si Schottky [123]
54 0 5800 Unspecified Schottky [28]
59.3 18 5800 M/A-COM 4E2054 GaAs Schottky [124]
143
65.3 19 5800 M/A-COM 4E2054 GaAs Schottky [124]
68.5 ∗ 18 5800 Avago HSMS-8202 Si Schottky [91]
71.4 ∗ 24.2 5800 HP 5082-2835 GaAs Schottky [125]
76 ∗ 20 5800 M/A-COM 4E1317 GaAs Schottky [92]
80 ∗ 38 5870 Si Schottky [29, 126]
82 ∗ 17 5800 M/A-COM 40150-119 Si Schottky [29]
82.7 ∗ 16.9 5800 M/A-COM 4E1317 GaAs Schottky [30]
∗Efficiencies include antenna effects in their efficiency calculations. Non-asterisked efficiencies include efficien-
cies only related to the rectifier.
Table 15: State-of-the-art microwave-energy conversion efficiencies above 5.8 GHz.
Efficiency (%) Input Power (dBm) Frequency (MHz) Rectifier Element Source
62.5 20.2 8510 M/A-COM 40401 Schottky [93]
60 ∗ 21.5 10000 Alpha Industries DMK6606 GaAs Schottky [6]
47.9 23.2 24000 M/A-COM 4E1317 GaAs Schottky [127]
54.2 21.1 24000 M/A-COM MADS-001317-1320AG GaAs Schottky [128]
4.5 ∗ -2 25700 M/A-COM 4E2502L Si Schottky [129]
15.5 ∗ 8 25700 M/A-COM 4E2502L Si Schottky [129]
144
35 ∗ 10.6 35000 M/A-COM 4E1317 GaAs Schottky [130]
37 ∗ 17 35000 Alpha Industries DMK6606 GaAs Schottky [6]
39 ∗ 20.8 35000 Alpha Industries DMK6606 GaAs Schottky [6]
53 ∗ 39.4 35000 0.13 µm CMOS Schottky [95]
70 ∗ 7.7 35000 GaAs Schottky [96]
37 ∗ 29.9 94000 0.13 µm CMOS Schottky [95]
∗Efficiencies include antenna effects in their efficiency calculations. Non-asterisked efficiencies include efficien-
cies only related to the rectifier.
APPENDIX B
• Adaption of derived model for an energy harvesting circuit with N diodes under CW
excitation.
• Derivation of the maximum energy harvesting efficiency of a diode for POW excitation.
Much of this derivation follows from Yoo and Chang in [6], but will be repeated for com-
pleteness as the next two sections will largely draw from this work. Moreover, much of
the notation used conforms to this work. Figure 73 shows the circuit model for which this
1. The voltage across the diode and the load consists for DC and the fundamental only.
No harmonics exist.
145
Figure 73: A standard diode model with parallel output resistance. This model assumes
that all parasitic values have been tuned out with reactive elements when possible.
B.1.1 Definitions
Using the previous assumptions, the output voltage Vout can be shown to be
where V0 is the DC component on the output and V1 is the peak fundamental voltage of the
incident waveform. Recall from Section 3.2.1 that the maximum input voltage before a diode
Vbr Vbr
goes into reverse breakdown is 2 . This implies that for a lossless diode V0 = V1 = 2 .
where Vd0 is the DC component across the diode and Vd1 is the peak fundamental voltage
across the diode. The relationship between voltage across the diode and the output voltage
is visualized in Figure 74. Note that the voltage across the diode translates into a model
where there is infinite current through the diode when the voltage is above VT and no
current through the diode when the voltage is below this value.
B.1.2 Derivations
Using Kirchoff’s voltage law, the DC output voltage can be related to the DC voltage drop
Vd,DC
V0 = Rs
, (48)
1+ R L
146
Figure 74: The voltage across the diode and the output voltage. Note how Vd is delayed
by Φ and when the voltage is greater than VT , it remains constant at this value between
−Θoff and Θoff (Adapted from [6]).
where Vd,DC is the time averaged value of the diode voltage Vd given by
γ+2π
Θ
Θ
Vd1
Z
1 − off sin Θoff − VT off ,
Vd,DC = Vd dΘ = Vd0 + (49)
γ π π π
where Θoff is the phase angle where the diode is turned off and is defined by Θ = ωt − Φ.
Stated other ways, 2Θoff is the total amount of time in radians that the diode is turned on.
Its value can be calculated from the interception point for where the input signal level is
equal to the bias level plus the diode threshold voltage shown in
VT + Vd0
cos Θoff = . (50)
Vd1
Note that this relationship sets a lower limit for the operational range of this model. The
lowest value the input voltage is allowed to be is the threshold voltage VT . If the current
d(Cj Vd )
Rs = V − Vd . (51)
dt
Cj0
Cj = γ = C0 + C1 cos(ωt − Φ) + C2 cos(2ωt − 2Φ) + ... (52)
V0
1− VT
147
where Cj0 is the junction capacitance at zero bias, and γ is the grading coefficient. Substi-
tuting (52) into (51) and removing any term higher than the second harmonic gives
Assuming that (53) must hold when the diode is off, like terms must sum to zero as shown
Vd0 = V0 , (54)
Substituting (50) and (54) into (56), dividing both sides by (55) and solving for the
where Ceff is the effective capacitance shown in the first line. Note that setting Ceff = Cj
is also acceptable.
The losses encountered in the energy conversion efficiency for the modeled diode arise from
The diode input impedance and input power can be solved for by relating the input
current to the input voltage V1 . The current at the fundamental frequency can be expressed
as
148
where
"Z #
Θ 2π−Θ
1
Z
off off
IωR = (V − VT ) cos(Θ + Φ) dΘ + (V − Vd ) cos(Θ + Φ) dΘ , (60)
πRs −Θ Θ
off off
"Z #
Θ 2π−Θ
1
Z
off off
Iωi = (V − VT ) sin(Θ + Φ) dΘ + (V − Vd ) sin(Θ + Φ) dΘ . (61)
πRs −Θ Θ
off off
Assuming the output DC voltage is specified, the input fundamental voltage can be
Vd1 VT + Vd0
V1 = = . (62)
cos (Φ) cos (Φ) cos Θoff
V1
Zd = . (63)
IωR − jIωi
Rd = R {Zd } . (64)
The input power can be solved by using the RMS voltage of the fundamental input frequency
Vbr
From [9], the maximum value the output voltage can be is 2 , where Vbr is the reverse
breakdown voltage of the diode. Therefore, when the output voltage is greater than this
value, the efficiency expression will change to (67). This adjustment allows for the dip in
149
B.1.4 Method
Solving for the diode efficiency is not necessarily straightforward, so a brief explanation of
the process is outlined here. Using this method, the output or input voltage, V0 or V1 , is
specified along with diode parameters Rs , RL , VT , and the frequency of operation. Also
Vbr
note that the maximum input voltage is limited to 2 and RL is typically 1.3 to 1.4 times
the real part of the diode impedance (100Ω - 400Ω) as discussed in 3.2.1. Typically, this
method is used to plot the diode efficiency as a function of the input power.
1. To solve, the following parameters are required: threshold voltage VT , load resis-
output voltage is varied from near zero to a reasonable large voltage (10 V).
2. Using a numeric solver (such as f solve in MATLAB), solve for θoff in (58).
4. Assuming the output voltage V0 is specified, solve for the input voltage in (62).
5. Solve (60) and (61) to obtain the complex input current at the fundamental frequency.
6. With the complex impedance, solve for the complex diode impedance in (63) and
7. Using the input impedance, solve for the input power (65)
8. At this point, the efficiency can be plotted as a function of input power according
to (66).
9. Ensure that if the voltage level is high that the limitations in (67) are applied.
This section will discuss how to use the derivation in B.1 to obtain the maximum efficiency
for an ideal diode as plotted in Section 3.3.4, the maximum diode efficiency as a function
of ωRs Ceff . By using these parameters as a single, independent variable for which the
150
efficiency is evaluated, the maximum diode efficiency can be calculated for any frequency.
While (46) - (65) are still used, the method used to generate this result is slightly different.
1. Since this is an ‘ideal diode,’ the threshold voltage VT is set to zero and the reverse
Rs
2. A frequency ω, load resistor RL , DC output voltage V0 , and resistor ratio r = RL
3. The effective capacitance Ceff is varied over several orders of magnitude (i.e. fempto-
Farads to nano-Farads).
4. Follow steps 3.) to 9.) as in Section B.1.4 to calculate the efficiency and input power,
if desired.
The previous derivation is repeated for an N -stage Dickson charge pump. A 2-stage Dickson
charge pump is shown in Figure 75 as an example, but this derivation will hold for any
1. The voltage across the diode and the load consists for DC and the fundamental only.
No harmonics exist.
151
Figure 75: A 2-stage Dickson charge pump diode model with output resistance. This model
assumes that all diode parasitic values have been tuned out with reactive elements, but
includes the circuit capacitances for RF grounding.
B.2.1 Definitions
Using the previous assumptions, the input voltage Vin and output voltage Vout can be shown
to be
Vout = V0 , (68)
where V0 is the DC component on the output, and V1 is the peak fundamental voltage of
maximum output voltage is V0 = N V2br . The voltage Vd across each diode (A and B in this
152
where Vd0 is the DC component across the diode and Vd1 is the peak fundamental voltage
across the diode. The difference in this case from the single diode case is a 180 degree phase
shift. While diode A is on during the positive portion of the input, diode B is off. Likewise,
when the input swings negative, diode A turns off while diode B is on. Thus, the same
expression (save for a change in sign) can be used to describe their behavior. Note that the
voltage across the diode translates into a model where there is infinite current through the
diode when the voltage is above VT and no current through the diode when the voltage is
B.2.2 Derivations
Using Kirchoff’s voltage law, the DC output voltage can be related to the DC voltage drop
V0 Vd,DC
= Rs
, (72)
N 1+NR L
where Vd,DC is the time averaged value of the diode voltage Vd given by
γ+2π
Θ
Θ
Vd1
Z
1 − off sin Θoff − VT off ,
Vd,DC = Vd dΘ = Vd0 + (73)
γ π π π
where Θoff is the phase angle where the diode is turned off and is defined by Θ = ωt − Φ.
Its value can be calculated from the interception point for where the input signal level is
equal to the bias level plus the diode threshold voltage shown in
VT + Vd0
cos Θoff = . (74)
Vd1
Note that this relationship sets a lower limit for the operational range of this model. The
lowest value the input voltage is allowed to be is the threshold voltage VT . If the current
d(Cj Vd )
Rs = Vn − (Vd + Vn−1 ) . (75)
dt
Cj0
Cj = V0
γ = C0 + C1 cos(ωt − Φ) + C2 cos(2ωt − 2Φ) + ... (76)
1 − VT
N
153
where Cj0 is the junction capacitance at zero bias, and γ is the grading coefficient. Substi-
tuting (52) into (75) and removing any term higher than the second harmonic gives
V0
Vd0 = , (78)
N
Substituting (74) and (78) into (80), dividing both sides by (79) and solving for the
where Ceff is the effective capacitance shown in the first line. Note that setting Ceff = Cj
is also acceptable.
The losses encountered in the energy conversion efficiency for a Dickson charge pump arise
from the same sources, the series resistances Rs and the diodes themselves. A single diode
input impedance and input power can be solved for by relating the input current to the
154
where
Z Θ
1 off
IωR = (Vn − (VT + Vn−1 )) cos(Θ + Φ) dΘ
πRs −Θ
off (84)
Z 2π−Θ
1 off
+ (Vn − (Vd + Vn−1 )) cos(Θ + Φ) dΘ,
πRs Θ
off
Z Θ
1 off
Iωi = (Vn − (VT + Vn−1 )) sin(Θ + Φ) dΘ
πRs −Θ
off (85)
Z 2π−Θ
1 off
+ (Vn − (Vd + Vn−1 )) sin(Θ + Φ) dΘ.
πRs Θ
off
Assuming the output DC voltage is specified, the input fundamental voltage can be
Vd1 VT + Vd0
V1 = = . (86)
cos (Φ) cos (Φ) cos Θoff
The impedance of the charge pump can consequently be described by multiplying the current
of a single diode by the total number of diodes and solving for the impedance as in
V1
Zd = . (87)
N (IωR − jIωi )
Rd = R {Zd } . (88)
The Dickson charge pump input power can be solved by using the RMS voltage of the
fundamental input frequency along with the charge pump resistance as shown by
2
V1
√
2 V12
Pd,in = = . (89)
Rd 2Rd
Vbr
From [9], the maximum value the output voltage can be is 2 , where Vbr is the reverse
breakdown voltage of the diode. Therefore, when the output voltage is greater than this
155
value, the efficiency expression will change to (91). This adjustment allows for the dip in
The method to solve for the efficiency follows nearly identically to Section B.1.4.
A theoretical model expressing energy-harvester efficiency under high PAPR signals was
developed based on Yoo and Chang’s theory in [6] to aid in the understanding of high
PAPR signals. While similar to the previous used terminology, additional terms are added
to address the high PAPR. Some of these definitions do have their limitations and most
closely approximate an intermittent transmission [105] type of waveform or square POW [64]
with infinite frequency content, but their PAPR should still illustrate the usefulness of any
Figure 73 will still be used as the circuit model for this derivation. The following
assumptions are made, some follow from the previous derivation, some are new:
1. The voltage across the diode and the load consists of DC and the fundamental only.
No harmonics or mixing terms exist. Figure 76 shows the excitation waveform. Note
that while only a single period of the sinusoid is shown, the derivation would be
POW will have non-zero bandwidth, for the purposes of this derivation, the entire
5. There are no reflections due to impedance mismatch. All frequencies are absorbed.
156
B.3.1 Definitions
Using the previous assumptions, the output voltage Vout can be shown to be
−V0 + V1 cos(ωt)
: 0 < t < 2π
Vout = (92)
−V0
: 2π < t < 2π + ψ
where V0 is the DC component on the output, V1 is the peak fundamental voltage of the
incident waveform, and ψ is the amount of time the waveform is ‘off.’ Thus, the excitation
waveform is on between zero and two pi and off from two pi to two pi plus psi. Note that ψ
is bounded from 0 (when the waveform is continuously ‘on’ and reduces to the form shown
in [6], also in Section B.1) to approaching ∞ when the waveform is on for an infinitesimal
amount of time. The voltage Vd across the diode can also be described by
−Vd0 + Vd1 cos(ωt − Φ) : if diode is off & waveform is on
Vd = −Vd0 : if diode is off & waveform is off (93)
V
: if diode is on
T
where Vd0 is the DC component across the diode and Vd1 is the peak fundamental voltage
across the diode. The relationship between voltage across the diode and the output voltage
is visualized in Figure 76. Note that the voltage across the diode translates into a model
where there is infinite current through the diode, though still limited by Rs , when the
voltage is above VT and no DC current through the diode when the voltage is below this
value.
Figure 76: The voltage across the diode and the output voltage. For a POW or signal with
intermittent transmission, there is a period of Ψ when the waveform is ‘off’ and only the
rectified DC component is present.
157
B.3.2 Derivations
Using Kirchoff’s current law, the DC output voltage can be related to the DC voltage drop
Vd,DC
V0 = Rs
, (94)
1+ R L
where Vd,DC is the time averaged value of the diode voltage Vd given by
3π
+Ψ
1
Z
2
Vd,DC = Vd dΘ =
2π + Ψ − π2
(95)
1
2Θoff VT − 2Vd1 sin(Θoff ) − Vd0 2π + Ψ − 2Θoff ,
2π + Ψ
where Θoff is the phase angle where the diode is turned off and is defined by Θ = ωt − Φ.
Its value can be calculated from the interception point for where the input signal level is
equal to the bias level plus the diode threshold voltage shown in
VT + Vd0
cos Θoff = . (96)
Vd1
Note that this relationship sets a lower limit for the operational range of this model. The
lowest value the input voltage is allowed to take is the threshold voltage VT . Again using
Kirchoff’s current law to equate the currents across Rs and the diode junction when the
d(Cj Vd )
Rs = Vout − Vd . (97)
dt
yields
Cj0
Cj = γ = C0 + C1 cos(ωt − Φ) + C2 cos(2ωt − 2Φ) + ... (98)
V0
1− VT
where Cj0 is the junction capacitance at zero bias, and γ is the grading coefficient. Substi-
tuting (98) into (B.3.2) and removing any term higher than the second harmonic gives
158
Assuming that (99) must hold when the diode is off, like terms must sum to zero as shown
Vd0 = V0 , (100)
Substituting (96) and (100) into (102), dividing both sides by (101) and solving for the
where Ceff is the effective capacitance shown in the first line. Note that setting Ceff = Cj
is also acceptable.
The losses encountered in the energy conversion efficiency for the modeled diode arise from
the series resistance Rs and the diode itself. This calculation is further complicated in the
case of a POW because it is an intermittent transmission. Thus, RMS currents and voltages
The RMS voltage of the fundamental is taken using the classical definition shown in
s
1 T
Z
VRM S = V (Θ)2 dΘ, (105)
T 0
where T is the period of the waveform. Since we are concerned about only the RMS value
159
Note that the period is 2π + Ψ, but the waveform is only on for a period of 2π.
Likewise, the RMS current of the fundamental is taken using the classical definition
shown in
s
T
1
Z
IRM S = Iω (Θ)2 dΘ. (107)
T 0
The real and imaginary currents of the fundamental are calculated as in previous sections,
but now with more complicated integrals due to the intermittent transmission.
−Θ
1 h
Z
off
IωR = (V − Vd ) cos(Θ + Φ) dΘ
πRs −π
2
Z Θ
off
+ (V − VT ) cos(Θ + Φ) dΘ (110)
−Θ
off
Z 3π
2
i
+ (V − Vd ) cos(Θ + Φ) dΘ ,
Θ
off
−Θ
1 h
Z
off
Iωi = (V − Vd ) sin(Θ + Φ) dΘ
πRs −π
2
Z Θ
off
+ (V − VT ) sin(Θ + Φ) dΘ (111)
−Θ
off
Z 3π
2
i
+ (V − Vd ) sin(Θ + Φ) dΘ ,
Θ
off
Substituting (110) and (111) into (109) will give the total RMS current into the circuit.
To calculate the real power, however, only the real part of the current must be taken into
160
account as shown by
v
u 3π
1
Z
u 2
IRM S =t (IωR cos (Θ))2 dΘ. (112)
2π + Ψ −π
2
Assuming the output DC voltage is specified, the input fundamental voltage can be rewritten
Vd1 VT + Vd0
V1 = = . (113)
cos (Φ) cos (Φ) cos Θoff
The input power can be solved by using the RMS voltage of the fundamental input frequency
breakdown voltage of the diode. Therefore, when the output voltage is greater than this
value, the efficiency expression will change to (116). This adjustment allows for the dip in
B.3.4 Method
Solving for the diode efficiency is not necessarily straightforward, so a brief explanation of
the process is outlined here. To solve, the following parameters are required: threshold
1. Using a numeric solver (such as f solve in MATLAB), solve for θoff in (104).
3. Assuming the output voltage V0 is specified, solve for the input voltage in (113).
161
4. Solve (110) to obtain the real input current at the fundamental frequency.
5. With the real input current, solve for the input RMS current (112) and voltage (106).
6. Solve for the input power using the RMS voltage and current (114)
7. At this point, the efficiency can be plotted as a function of input power according
to (115).
8. Ensure that if the voltage level is high that the limitations in (116) are applied.
B.4 Validation
B.4.1 CW and N-diode Validation
To validate the theoretical models, Agilent ADS simulations were run and compared to the
theoretical results. Figures 77 and 78 show the DC output voltage and energy conversion
efficiency of a single diode, one-, two-, and three-stage Dickson charge pump with 300 Ω,
600 Ω, 1200 Ω, and 1800 Ω load, respectively. These loads tend to produce the highest
efficiency at the maximum power level before breakdown. The ADS simulation used ideal
transmission lines, the standard Avago HSMS-286x diode model, and no package parasitics.
Each circuit was optimized for its maximum efficiency with ADS ‘Random’ optimization
and using the harmonic balance simulation engine with four orders of harmonics and 3
orders of mixing terms. To minimize the number of frequencies in the simulation, M-POW
were used as opposed to the square-POWs used in the theoretical computation [64]. The
As can be seen, the theoretical model of the DC output voltage in Figure 77 describe the
behavior of the simulated circuit quite well. Since the theoretical model assumes impedance
matching across all power levels, it generally has a larger voltage level across all power levels.
This observation is especially true for low powers. To illustrate this case, two different
simulated cases (one circuit matched at 16 dBm and one at -20 dBm) were plotted. While
the 16 dBm matched case agrees well with the simulated circuit for power levels above
0 dBm, below this value, the difference begins to grow large due to reflections causes by
impedance mismatch. Likewise, the -20 dBm matched case agrees well for lower power levels,
162
Rectenna/Single Diode 1−Stage Dickson/Two Diodes
1 1
10 10
0 0
10 10
Vout (V)
Vout (V)
−1 −1
10 10
−2 −2
10 10
Simulated 16 dBm Match Simulated 16 dBm Match
−3 −3
10 Simulated −20 dBm Match 10 Simulated −20 dBm Match
−4 Theoretical −4
Theoretical
10 10
−30 −20 −10 0 10 20 30 −30 −20 −10 0 10 20 30
P (dBm) P (dBm)
in in
2−Stage Dickson/Four Diodes 3−Stage Dickson/Six Diodes
1 1
10 10
0 0
10 10
Vout (V)
Vout (V)
−1 −1
10 10
−2 −2
10 10
Simulated 16 dBm Match Simulated 16 dBm Match
−3 −3
10 Simulated −20 dBm Match 10 Simulated −20 dBm Match
−4
Theoretical −4
Theoretical
10 10
−30 −20 −10 0 10 20 30 −30 −20 −10 0 10 20 30
P (dBm) P (dBm)
in in
Figure 77: Theoretical and simulated DC output voltages of N-stage Dickson charge pumps.
Theoretical results were produced with the aforementioned theory. Simulated results were
computed with Agilent ADS using ideal components and Avago HSMS-286x Schottky
diodes.
but tends to have worse performance at higher power levels. Also note that the limitation
of the diode modeling technique is apparent in the asymptote in the theoretical prediction.
Recall that this behavior occurs because of the turn-on voltage threshold limitation on the
Figures 79 and 80 show validation results for the energy harvester theory for POW excita-
tion. Once again, the theoretical model lines up well with the simulated results where the
simulated results are best matched per input power. The differences between the theory and
the simulations can primarily be attributed to the model neglecting voltage ripple. Thus
as expected, the theoretical model produces a higher expected power and efficiency than
163
Rectenna/Single Diode 1−Stage Dickson/Two Diodes
1 1
Simulated 16 dBm Match Simulated 16 dBm Match
Theoretical Theoretical
0.75 0.75
Efficiency
Efficiency
0.5 0.5
0.25 0.25
0 0
−30 −20 −10 0 10 20 30 −30 −20 −10 0 10 20 30
P (dBm) P (dBm)
in in
2−Stage Dickson/Four Diodes 3−Stage Dickson/Six Diodes
1 1
Simulated 16 dBm Match Simulated 16 dBm Match
Theoretical Theoretical
0.75 0.75
Efficiency
Efficiency
0.5 0.5
0.25 0.25
0 0
−30 −20 −10 0 10 20 30 −30 −20 −10 0 10 20 30
P (dBm) P (dBm)
in in
Figure 78: Theoretical and simulated efficiencies of N-stage Dickson charge pumps. Theoret-
ical results were produced with the aforementioned theory. Simulated results were computed
with Agilent ADS using ideal components and Avago HSMS-286x Schottky diodes.
the simulated results. Actual results will be lower because of additional parasitic losses,
impedance mismatches, and voltage ripple. Using these closed-form equations, circuit de-
signers can determine if high PAPR signals are beneficial for their applications by quickly
164
CW 1−POW, 50% Duty Cycle
25 25
0 0
Pout (dBm)
Pout (dBm)
−25 −25
−50 −50
Simulated 16 dBm Match Simulated 16 dBm Match
−75 Simulated −20 dBm Match −75 Simulated −20 dBm Match
Theoretical Theoretical
−100 −100
−30 −20 −10 0 10 20 30 −30 −20 −10 0 10 20 30
Pin (dBm) Pin (dBm)
2−POW, 25% Duty Cycle 3−POW, 16.7% Duty Cycle
25 25
0 0
Pout (dBm)
Pout (dBm)
−25 −25
−50 −50
Simulated 16 dBm Match Simulated 16 dBm Match
−75 Simulated −20 dBm Match −75 Simulated −20 dBm Match
Theoretical Theoretical
−100 −100
−30 −20 −10 0 10 20 30 −30 −20 −10 0 10 20 30
Pin (dBm) Pin (dBm)
Figure 79: Theoretical and simulated output powers of rectennas under various POW excita-
tions. Theoretical results were produced with the aforementioned theory using duty-cycled
sinusoids or square POWs. Simulated results were computed with Agilent ADS using ideal
components, Avago HSMS-286x Schottky diodes, and excited with M-POWs.
165
CW 1−POW, 50% Duty Cycle
1 1
Simulated 16 dBm Match Simulated 16 dBm Match
Theoretical Theoretical
0.75 0.75
Efficiency
Efficiency
0.5 0.5
0.25 0.25
0 0
−30 −20 −10 0 10 20 30 −30 −20 −10 0 10 20 30
Pin (dBm) Pin (dBm)
2−POW, 25% Duty Cycle 3−POW, 16.7% Duty Cycle
1 1
Simulated 16 dBm Match Simulated 16 dBm Match
Theoretical Theoretical
0.75 0.75
Efficiency
Efficiency
0.5 0.5
0.25 0.25
0 0
−30 −20 −10 0 10 20 30 −30 −20 −10 0 10 20 30
Pin (dBm) Pin (dBm)
Figure 80: Theoretical and simulated efficiencies of rectennas under various POW excita-
tions. Theoretical results were produced with the aforementioned theory using duty-cycled
sinusoids or square POWs. Simulated results were computed with Agilent ADS using ideal
components and excited with M-POWs.
166
APPENDIX C
• Listings of the computer code used to compute the maximum theoretical energy-
conversion efficiencies.
This code is the top-level module which controls the functions to get the input power and
%Diode Example
Rs =2;
Rl =300;
Vf = 0 . 3 ;
C e f f =0.18 e −12;
f r e q =5.8 e 9 ;
167
%Avago HSMS286z
Rs =6;
Rl = 2 7 5 . 5 ;
Vf = 0 . 4 ;
Vbr =7;
C e f f =0.18 e −12;
f r e q =5.8 e 9 ;
%S e t v a r y i n g DC o u t p u t voltage
%Vo=l i n s p a c e ( 1 e − 3 , 5 , 1 0 0 ) ;
Vo=l o g s p a c e ( − 3 , 1 , 1 0 0 ) ;
%Loops f u n c t i o n to p r o v i d e data
for j =1: l e n g t h ( Vo )
[ Pin ( j ) , E f f i c i e n c y ( j ) , Vop ( j ) ] = S i n g l e D i o d e C a l c u l a t i o n ( Rl , Rs , Vo ( j ) , Vf , f r e q , C e f f , Vbr ) ;
end
%P l o t functions
figure ();
p l o t ( 1 0 ∗ l o g 1 0 ( Pin /1 e −3) , E f f i c i e n c y , ’ b ’ ) ; h o l d on ;
hold off ;
x l a b e l ( ’ I n p u t Power (dBm ) ’ ) ;
ylabel ( ’ Efficiency ’ ) ;
g r i d on ;
C.1.2 SingleDiodeComparePlot.m
This code follows form ‘SingleDiodePlot.m,’ but generates outputs from multiple diodes
under test.
%D i o d e s t o compare
%Avago HSMS286x , 285 x , 282 x
Vf = [ 0 . 4 0.2 0.35];
Vbr =[7 3 . 8 15];
Rs =[6 25 6];
Ceff =[0.18 0.18 0 . 7 ] ∗ 1 e −12;
f r e q =5.8 e 9 ;
Rl =300;
%S e t v a r y i n g DC o u t p u t voltage
168
%Vo=l i n s p a c e ( 1 e − 3 , 5 , 1 0 0 ) ;
Vo=l o g s p a c e ( − 3 , 1 . 7 , 2 0 0 )
%P l o t settings
f s i z e =24;
Lwidth =3;
p l o t d ={’−−r ’ , ’ − b ’ , ’ − . g ’ , ’ . k ’ } ;
%Loops f u n c t i o n to p r o v i d e d a t a and p l o t
figure ();
f o r k =1: l e n g t h ( Vf )
for j =1: l e n g t h ( Vo )
[ Pin ( j , k ) , E f f i c i e n c y ( j , k ) , Vop ( j , k ) ] = . . .
S i n g l e D i o d e C a l c u l a t i o n ( Rl , Rs ( k ) , Vo ( j ) , Vf ( k ) , f r e q , C e f f ( k ) , Vbr ( k ) ) ;
end
p l o t ( 1 0 ∗ l o g 1 0 ( Pin ( : , k ) / 1 e −3) , E f f i c i e n c y ( : , k ) , p l o t d {k } , ’ LineWidth ’ , Lwidth ) ;
h o l d on ;
end
hold off ;
x l a b e l ( ’ I n p u t Power (dBm) ’ , ’ F o n t S i z e ’ , f s i z e ) ;
y l a b e l ( ’ E f f i c i e n c y ’ , ’ FontSize ’ , f s i z e ) ;
a x i s ([ −30 40 0 1]);
l e g e n d ( ’ HSHS−286x ’ , ’ HSMS−285x ’ , ’ HSMS−282x ’ ) ;
g r i d on ;
box on ;
s e t ( gca , ’ F o n t S i z e ’ , f s i z e ) ;
C.1.3 SingleDiodeCalculation.m
This code implements the theory described in Appendix B.1 and is called by the previously
169
%Vo − ( V o l t s ) − Output v o l t a g e value
%Vf − ( V o l t s ) − Diode p a r a m e t e r − Turn−on / T h r e s h o l d voltage
%f r e q − ( Hz ) − F r e q u e n c y o f operation
%C e f f − ( F a r a d s ) − Diode p a r a m e t e r − E f f e c t i v e diode junction capacitance , u s u a l l y Cj0
%Vbr − ( V o l t s ) − Diode p a r a m e t e r − Diode r e v e r s e breakdown v o l t a g e
%
%Output Arguments
%Pin − ( Watts ) − I n p u t power a t t h e f u n d a m e n t a l f r e q u e n c y
%E f f i c i e n c y − (%/100) − RF t o DC c o n v e r s i o n efficiency
%Vop − ( V o l t s ) − DC o u t p u t voltage w i t h breakdown t a k e n into account
gamma = 0 . 5 ;
%I m a g i n a r y p a r t of current of fundamental
V=−Vo+V1∗ c o s ( I 1 r a t t+p h i ) ;
I 1 i a x =(V−Vf ) . ∗ s i n ( I 1 r a t t+p h i ) ;
V=−Vo+V1∗ c o s ( I 1 r b t t+p h i ) ;
Vd=−Vo+V1∗ c o s ( p h i ) ∗ c o s ( I 1 r b t t ) ;
I 1 i b x =(V−Vd ) . ∗ s i n ( I 1 r b t t+p h i ) ;
I 1 i =1/( p i ∗Rs ) ∗ ( t r a p z ( I 1 r a t t , I 1 i a x )+ t r a p z ( I 1 r b t t , I 1 i b x ) ) ;
170
Yd=( I 1 r −1 i ∗ I 1 i ) / V1 ;
Zd=1/Yd ;
Rd=1/ r e a l (Yd ) ;
%P a s s t h e o u t p u t v o l t a g e back a s an o u t p u t
Vop=Vo ;
%E v a l u a t e t h e efficiency
E f f i c i e n c y =(Voˆ2/ Rl ) / Pin ;
C.1.4 SingleDiodeSolver.m
This basic code contains a function which is solved for using one of MATLAB’s numeric
solvers.
function f=S i n g l e D i o d e S o l v e r ( x , r , v f f )
%T h i s function contains the f u n c t i o n which relates thetaOff ( c a l l e d x ) to
%t h e r e s i s t a n c e s and v o l t a g e s . It is meant t o be u s e d a s follows :
%
%Example o f use
%f u n=@( x ) SingleDiodeSolver (x , r , vff ) ;
%[ t h e t a O f f , f v a l , e x i t f l a g ]= f s o l v e ( fun , 0 . 1 ) ;
%
%I n p u t p a r a m e t e r s
%x − ( r a d i a n s ) − v a l u e of thetaOff
%r − ( u n i t l e s s ) − r a t i o o f Rs/ Rl , diode series resistance over load
%r e s i s t a n c e
%v f f − ( u n i t l e s s ) − r a t i o of d i o d e t u r n−on v o l t a g e t o DC o u t p u t voltage
%
%Output p a r a m e t e r s
%f − t h e evaluated value of the function
171
C.2 Maximum Energy-conversion Efficiency of an N-stage Dickson
Charge Pump
C.2.1 NdicksonDiodePlot.m
This code is the top-level module which controls the functions to get the input power and
%F r e q u e n c y u n d e r test
f r e q =5.8 e 9 ;
%Avago HSMS−286x
Rs =6;
Vf = 0 . 4 ;
Vbr =7;
C e f f =0.18 e −12;
% %Avago HSMS−285x
Rs =25;
Vf = 0 . 2 ;
Vbr = 3 . 8 ;
C e f f =0.18 e −12;
%S e t v a r y i n g DC o u t p u t voltage
%Vo=l i n s p a c e ( 1 e − 3 , 5 , 1 0 0 ) ;
Vo=l o g s p a c e ( − 6 , 2 , 2 0 0 )
%P l o t p a r a m e t e r s
f s i z e =20;
Lwidth =3;
p l o t d ={’−−r ’ , ’ − b ’ , ’ − . g ’ , ’ . k ’ , ’ − −m’ } ;
%Loops f u n c t i o n to p r o v i d e data
f o r k =1: l e n g t h (N)
for j =1: l e n g t h ( Vo )
[ Pin ( j , k ) , E f f i c i e n c y ( j , k ) , Vop ( j , k ) ] = . . .
172
N d i c k s o n D i o d e C a l c u l a t i o n v 2 (N( k ) , Rl ( k ) , Rs , Vo ( j ) , Vf , f r e q , C e f f , Vbr ) ;
PoutDC ( j , k)=Vop ( j , k ) ˆ 2 / Rl ( k ) ;
PoutDC dbm ( j , k )=10∗ l o g 1 0 ( PoutDC ( j , k ) / 1 e − 3 ) ;
end
end
%V a r i o u s plots
figure ();
h1=p l o t ( 1 0 ∗ l o g 1 0 ( Pin ( : , 1 ) / 1 e −3) , E f f i c i e n c y ( : , 1 ) , p l o t d { 1 } , . . .
10∗ l o g 1 0 ( Pin ( : , 2 ) / 1 e −3) , E f f i c i e n c y ( : , 2 ) , p l o t d { 2 } , . . .
10∗ l o g 1 0 ( Pin ( : , 3 ) / 1 e −3) , E f f i c i e n c y ( : , 3 ) , p l o t d { 3 } , . . .
10∗ l o g 1 0 ( Pin ( : , 4 ) / 1 e −3) , E f f i c i e n c y ( : , 4 ) , p l o t d { 4 } , . . .
10∗ l o g 1 0 ( Pin ( : , 5 ) / 1 e −3) , E f f i c i e n c y ( : , 5 ) , p l o t d { 5 } ) ;
x l a b e l ( ’ I n p u t Power (dBm) ’ , ’ F o n t S i z e ’ , f s i z e ) ;
y l a b e l ( ’ E f f i c i e n c y ’ , ’ FontSize ’ , f s i z e ) ;
a x i s ([ −20 40 0 1]);
l e g e n d ( [ ’ N=’ num2str (N ( 1 ) ) ] , [ ’ N=’ num2str (N ( 2 ) ) ] , [ ’ N=’ num2str (N ( 3 ) ) ] , . . .
[ ’ N=’ num2str (N ( 4 ) ) ] , [ ’ N=’ num2str (N ( 5 ) ) ] ) ;
g r i d on ;
box on ;
s e t ( gca , ’ F o n t S i z e ’ , f s i z e , ’ Y t i c k ’ , [ 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1]);
s e t ( h1 , ’ LineWidth ’ , Lwidth ) ;
figure ();
for j j =1: l e n g t h (N)
s e m i l o g y ( 1 0 ∗ l o g 1 0 ( Pin ( : , j j ) / 1 e −3) ,Vop ( : , j j ) , p l o t d { j j } , ’ LineWidth ’ , Lwidth ) ;
h o l d on ;
end
hold off ;
x l a b e l ( ’ I n p u t Power (dBm) ’ , ’ F o n t S i z e ’ , f s i z e ) ;
y l a b e l ( ’ V { o u t } (V) ’ , ’ F o n t S i z e ’ , f s i z e ) ;
a x i s ([ −20 40 1 e−3 2 5 ] ) ;
l e g e n d ( [ ’ N=’ num2str (N ( 1 ) ) ] , [ ’ N=’ num2str (N ( 2 ) ) ] , [ ’ N=’ num2str (N ( 3 ) ) ] , . . .
[ ’ N=’ num2str (N ( 4 ) ) ] , [ ’ N=’ num2str (N ( 5 ) ) ] ) ;
g r i d on ;
box on ;
s e t ( gca , ’ F o n t S i z e ’ , f s i z e , ’ Y t i c k ’ , [ 1 e−3 1 e−2 1 e−1 1 1 e 1 ] ) ;
figure ();
h3=p l o t ( 1 0 ∗ l o g 1 0 ( Pin ( : , 1 ) / 1 e −3) , PoutDC dbm ( : , 1 ) , p l o t d { 1 } , . . .
10∗ l o g 1 0 ( Pin ( : , 2 ) / 1 e −3) , PoutDC dbm ( : , 2 ) , p l o t d { 2 } , . . .
10∗ l o g 1 0 ( Pin ( : , 3 ) / 1 e −3) , PoutDC dbm ( : , 3 ) , p l o t d { 3 } , . . .
10∗ l o g 1 0 ( Pin ( : , 4 ) / 1 e −3) , PoutDC dbm ( : , 4 ) , p l o t d { 4 } , . . .
10∗ l o g 1 0 ( Pin ( : , 5 ) / 1 e −3) , PoutDC dbm ( : , 5 ) , p l o t d { 5 } ) ;
x l a b e l ( ’ I n p u t Power (dBm) ’ , ’ F o n t S i z e ’ , f s i z e ) ;
y l a b e l ( ’ P { out ,DC} (dBm) ’ , ’ F o n t S i z e ’ , f s i z e ) ;
a x i s ([ −20 40 −50 3 0 ] ) ;
l e g e n d ( [ ’ N=’ num2str (N ( 1 ) ) ] , [ ’ N=’ num2str (N ( 2 ) ) ] , [ ’ N=’ num2str (N ( 3 ) ) ] , . . .
[ ’ N=’ num2str (N ( 4 ) ) ] , [ ’ N=’ num2str (N ( 5 ) ) ] ) ;
g r i d on ;
box on ;
s e t ( gca , ’ F o n t S i z e ’ , f s i z e , ’ Y t i c k ’ , [ − 5 0 −35 −20 −5 10 25]);
s e t ( h3 , ’ LineWidth ’ , Lwidth ) ;
173
C.2.2 NdicksonDiodeCalculation.m
This code implements the theory described in Appendix B.2 and is called by the previously
gamma = 0 . 5 ;
174
%C a l c u l a t e peak v a l u e of fundamental input voltage
V1=(Vf+Vo ) / ( c o s ( p h i ) ∗ c o s ( t h e t a O f f ) ) ;
%I m a g i n a r y p a r t of current of fundamental
V=−Vo/N+V1∗ c o s ( I 1 r a t t+p h i ) ;
I 1 i a x =(V−Vf ) . ∗ s i n ( I 1 r a t t+p h i ) ;
V=−Vo/N+V1∗ c o s ( I 1 r b t t+p h i ) ;
Vd=−Vo/N+V1∗ c o s ( p h i ) ∗ c o s ( I 1 r b t t ) ;
I 1 i b x =(V−Vd ) . ∗ s i n ( I 1 r b t t+p h i ) ;
I 1 i =1/( p i ∗Rs ) ∗ ( t r a p z ( I 1 r a t t , I 1 i a x )+ t r a p z ( I 1 r b t t , I 1 i b x ) ) ;
%C a l c u l a t e e a c h d i o d e i n p u t impedance
Yd=( I 1 r −1 i ∗ I 1 i ) / V1 ;
Zd=1/Yd ;
Rd=1/ r e a l (Yd ) ;
%P a s s t h e o u t p u t v o l t a g e back a s an o u t p u t
Vop=Vo ;
%E v a l u a t e t h e efficiency
E f f i c i e n c y =(Vopˆ2/ Rl ) / Pin ;
C.2.3 NdicksonDiodeSolver.m
This basic code contains a function which is solved for using one of MATLAB’s numeric
solvers.
175
function f=N d i c k s o n D i o d e S o l v e r ( x , r , v f f , N)
%T h i s function contains the f u n c t i o n which relates thetaOff ( c a l l e d x ) to
%t h e resistances , v o l t a g e s , and numbers o f diodes in the Dickson charge
%pump . It is meant t o be u s e d a s follows :
%
%Example o f use
%f u n=@( x ) N d i c k s o n D i o d e S o l v e r ( x , r , v f f , N ) ;
%[ t h e t a O f f , f v a l , e x i t f l a g ]= f s o l v e ( fun , 0 . 1 ) ;
%
%I n p u t p a r a m e t e r s
%x − ( r a d i a n s ) − v a l u e of thetaOff
%r − ( u n i t l e s s ) − r a t i o o f Rs/ Rl , diode series resistance over load
%r e s i s t a n c e
%v f f − ( u n i t l e s s ) − r a t i o of d i o d e t u r n−on v o l t a g e t o DC o u t p u t voltage
%N − ( u n i t l e s s ) − number o f stages in t h e D i c k s o n c h a r g e pump
%
%Output p a r a m e t e r s
%f − t h e evaluated value of the function
This code is the top-level module which controls the functions to get the input power and
efficiency among other parameters from a single diode under POW excitation.
clear all ;
close all ;
clc ;
%Avago HSMS286x
Rs =6;
Rl =300;
Vf = 0 . 4 ;
Vbr =7;
C e f f =0.18 e −12;
f r e q =5.8 e 9 ;
%Skyworks SMS7630
% Rs =20;
176
% Rl =9000;
% Vf = 0 . 3 4 ;
% Vbr =2;
% C e f f =0.3 e −12;
% f r e q =2.4 e 9 ;
%S e t s v a r y i n g DC o u t p u t voltage
Vo=l i n s p a c e ( 1 e − 3 , 5 , 1 0 0 ) ;
Vo=l o g s p a c e ( − 6 , 1 , 1 0 0 0 ) ;
%Loops f u n c t i o n to p r o v i d e data
f o r k =1: l e n g t h ( P s i )
for j =1: l e n g t h ( Vo )
p l o t d ={’−−r ’ , ’ − b ’ , ’ − . g ’ , ’ . k ’ } ;
%P l o t functions
figure ();
for i =1: l e n g t h ( P s i )
p l o t ( 1 0 ∗ l o g 1 0 ( Pin ( i , : ) / 1 e −3) ,10∗ l o g 1 0 ( Vop ( 1 , : ) . ˆ 2 / Rl /1 e −3) , p l o t d { i } ) ; h o l d on ;
end
hold off ;
x l a b e l ( ’ I n p u t Power (dBm ) ’ ) ;
y l a b e l ( ’ Output Power (dBm ) ’ ) ;
l e g e n d ( [ ’ Duty C y c l e =’ num2str ( 1 0 0 ∗ P s i d u t y ( 1 ) ) ’ % ’] ,...
[ ’ Duty C y c l e =’ num2str ( 1 0 0 ∗ P s i d u t y ( 2 ) ) ’ % ’] ,...
[ ’ Duty C y c l e =’ num2str ( 1 0 0 ∗ P s i d u t y ( 3 ) ) ’ % ’] ,...
[ ’ Duty C y c l e =’ num2str ( 1 0 0 ∗ P s i d u t y ( 4 ) ) ’ % ’]);
g r i d on ;
a x i s ([ −40 40 −70 3 0 ] ) ;
figure ();
for i =1: l e n g t h ( P s i )
p l o t ( 1 0 ∗ l o g 1 0 ( Pin ( i , : ) / 1 e −3) , E f f i c i e n c y ( i , : ) , p l o t d { i } ) ; h o l d on ;
end
hold off ;
x l a b e l ( ’ I n p u t Power (dBm ) ’ ) ;
ylabel ( ’ Efficiency ’ ) ;
l e g e n d ( [ ’ Duty C y c l e =’ num2str ( 1 0 0 ∗ P s i d u t y ( 1 ) ) ’ % ’] ,...
[ ’ Duty C y c l e =’ num2str ( 1 0 0 ∗ P s i d u t y ( 2 ) ) ’ % ’] ,...
[ ’ Duty C y c l e =’ num2str ( 1 0 0 ∗ P s i d u t y ( 3 ) ) ’ % ’] ,...
[ ’ Duty C y c l e =’ num2str ( 1 0 0 ∗ P s i d u t y ( 4 ) ) ’ % ’]);
g r i d on ;
177
C.3.2 SingleDiodePOWCalculation.m
This code implements the theory described in Appendix B.3 and is called by the previously
178
I 1 r b t t=l i n s p a c e (− p i /2 , − t h e t a O f f , 1 e 4 ) ;
I 1 r c t t=l i n s p a c e (− t h e t a O f f , t h e t a O f f , 1 e 4 ) ;
I 1 r d t t=l i n s p a c e ( t h e t a O f f , 3 ∗ p i / 2 , 1 e 4 ) ;
I 1 o t h e r=l i n s p a c e ( 3 ∗ p i / 2 , 3 ∗ p i /2+ P s i , 1 e 4 ) ;
%R e a l p a r t of current of fundamental
V=−Vo+V1∗ c o s ( I 1 r b t t+p h i ) ;
Vd=−Vo+V1∗ c o s ( p h i ) ∗ c o s ( I 1 r b t t ) ;
I 1 r b x =(V−Vd ) . ∗ c o s ( I 1 r b t t+p h i ) ;
V=−Vo+V1∗ c o s ( I 1 r c t t+p h i ) ;
I 1 r c x =(V−Vf ) . ∗ c o s ( I 1 r c t t+p h i ) ;
V=−Vo+V1∗ c o s ( I 1 r d t t+p h i ) ;
Vd=−Vo+V1∗ c o s ( p h i ) ∗ c o s ( I 1 r d t t ) ;
Vd(Vd<−Vbr)=−Vbr ; %L i m i t t h e negative voltage s w i n g t o −Vbr
I 1 r d x =(V−Vd ) . ∗ c o s ( I 1 r d t t+p h i ) ;
I 1 r =1/(( p i ) ∗ Rs ) ∗ ( t r a p z ( I 1 r b t t , I 1 r b x )+ t r a p z ( I 1 r c t t , I 1 r c x )+ t r a p z ( I 1 r d t t , I 1 r d x ) ) ;
%I m a g i n a r y p a r t of current of fundamental
V=−Vo+V1∗ c o s ( I 1 r b t t+p h i ) ;
Vd=−Vo+V1∗ c o s ( p h i ) ∗ c o s ( I 1 r b t t ) ;
I 1 r b x =(V−Vd ) . ∗ s i n ( I 1 r b t t+p h i ) ;
V=−Vo+V1∗ c o s ( I 1 r c t t+p h i ) ;
I 1 r c x =(V−Vf ) . ∗ s i n ( I 1 r c t t+p h i ) ;
V=−Vo+V1∗ c o s ( I 1 r d t t+p h i ) ;
Vd=−Vo+V1∗ c o s ( p h i ) ∗ c o s ( I 1 r d t t ) ;
Vd(Vd<−Vbr)=−Vbr;% L i m i t t h e negative voltage s w i n g t o −Vbr
I 1 r d x =(V−Vd ) . ∗ s i n ( I 1 r d t t+p h i ) ;
I 1 i =1/(( p i ) ∗ Rs ) ∗ ( t r a p z ( I 1 r b t t , I 1 r b x )+ t r a p z ( I 1 r c t t , I 1 r c x )+ t r a p z ( I 1 r d t t , I 1 r d x ) ) ;
%C a l c u l a t e t h e i n p u t power
Pin=r e a l ( V rms ∗ I 1 r e a l r m s ) ;
%C a l c u l a t e t h e o u t p u t power
Pout=Voˆ2/ Rl ;
Pout ( Vo>Vbr / 2 ) = ( ( Vbr / 2 ) ˆ 2 / Rl );% A d j u s t t h e o u t p u t power i n case the voltage
%s u r p a s s e s t h e d i o d e breakdown
179
%A s s i g n t h e i n p u t p a r a m e t e r o u t p u t v o l t a g e Vo t o a n o t h e r v a r i a b l e Vop
Vop=Vo ;
%%A d j u s t t h e o u t p u t voltage in case it t h e d i o d e breakdown limitation
Vop ( Vo>(Vbr /2))= Vbr / 2 ;
end
C.3.3 SingleDiodePOWSolver.m
This basic code contains a function which is solved for using one of MATLAB’s numeric
solvers.
function f=SingleDiodePOWSolver ( x , r , v f f , P s i )
%T h i s function contains the f u n c t i o n which relates thetaOff ( c a l l e d x ) to
%t h e r e s i s t a n c e s and v o l t a g e s . It is meant t o be u s e d a s follows :
%
%Example o f use
%f u n=@( x ) SingleDiodePOWSolver ( x , r , v f f ) ;
%[ t h e t a O f f , f v a l , e x i t f l a g ]= f s o l v e ( fun , 0 . 1 ) ;
%
%I n p u t p a r a m e t e r s
%x − ( r a d i a n s ) − v a l u e of thetaOff
%r − ( u n i t l e s s ) − r a t i o o f Rs/ Rl , diode series resistance over load
%r e s i s t a n c e
%v f f − ( u n i t l e s s ) − r a t i o of d i o d e t u r n−on v o l t a g e t o DC o u t p u t voltage
%P s i − ( R a d i a n s ) − Waveform p a r a m e t e r − Time when waveform is ’ off ’
%
%Output p a r a m e t e r s
%f − t h e evaluated value of the function
f =( r ∗ ( 2 ∗ p i+P s i ) ) / ( 2 ∗ ( 1 + v f f )) − t a n ( x)+x ;
end
180
APPENDIX D
vesting to power their on-board electronics. This task is usually implemented by means of
a charge pump. This circuit converts the incoming carrier wave to DC and then increases
the voltage by means of cascaded diode-capacitor stages. Traditionally, RFID systems rely
on continuous wave (CW) excitation for communication and for energy harvesting. How-
ever, Trotter et. al have shown that by exciting a charge pump with a Power Optimized
Waveform (POW), the charge pump efficiency, and thus range and reliability of the charge
While it has been shown that a POW excitation is preferable for charge pump efficiency
and multi-path mitigation on the forward link, the demodulation of a POW and its impact
on communication performance has not yet been discussed. This appendix investigates
the different possibilities of demodulating a POW and sets POW design criterion from
a physical layer communication point of view. It is demonstrated that using POWs will
provide additional resistance from frequency selective fading and multi-path interference
due to their multi-carrier nature. Furthermore, it is feasible that a POW would allow
181
multiple tags to simultaneously harvest RF energy from the entire POW waveform while
As previously mentioned, backscatter radio links have a forward and a backscatter link.
Each of these links can independently be treated as a traditional one-way link characterized
by either a Rayleigh or Rician fading channel to model the multi-path interference effects.
As a consequence, the backscattered signal received at the reader is a product of both the
forward and backscatter link [10]. Additionally, the free space pass loss expected for a
backscatter radio system is proportional to 1/d4 where d is the distance between the tag
and the reader. Kim et. al have also shown that the path loss exponent of a backscatter
link is approximately twice that of a traditional one-way link in the same environment and
that a product Rician best represents a backscatter radio channel in terms of the mean
squared error of the cumulative distribution function [131]. In this case, both the forward
and backscatter links are modeled by Rician fading with the same K factor (Recall that as
the Rice factor K approaches zero, Rician fading becomes Rayleigh). This product nature
further implies that the small-scale fading exhibited by a backscatter link is especially severe.
channel model expresses M transmitting antenna(s), L RFID tag antenna(s), and N receiv-
ing antenna(s).
+∞ Z
+∞
˜(t, ~r ) = 1
Z
~y ˜(t − τb − τf )dτb dτf + ~n
H̃b (τb ; t, ~r )S̃(t − τb )H̃f (τf ; t − τb , ~r )~x ˜(t) (117)
2
−∞ −∞
narrow band L × L tag signaling matrix; H̃f (τf ; t, ~r ) is the L × M , complex, baseband-
˜(t − τb − τf ) is an M × 1 vector of
channel impulse-response matrix of the forward link; ~x
˜(t) is an N × 1 vector of noise signals.
signals transmitted to the tags from the reader; and ~n
If this equation is simplified to the case of a single transmitting and receiving antenna in
182
mono-static or bi-static configuration (1 × L × 1) and the forward and backscatter links are
where A is the random channel envelope, α is the index of the PDF, Γ(·) is the gamma
function, K ν (·) is a modified Bessel function of the second kind with order ν = 1 − L, and
σb and σf are the standard deviations of the backscatter and forward channels, respectively.
Figure 81 shows a plot of Equation (118) for an increasing number of RF tag antenna(s) L.
Note that the fades for product Rayleigh that are exhibited in backscatter communication
channels are significantly more severe than simple Rayleigh fading in traditional communi-
radio communication.
Channel coherence is the opposite of channel selectivity. Coherence describes a part of the
channel that does not change as a function of time, space, or frequency [132].
1. Temporal fading is most often caused by the movement of the transmitter or receiver,
but also by scatterers in the environment. In order to have temporal coherence, data
symbols should be transmitted at a rate much slower than the channel coherence time.
When this occurs, a channel can be said to be constant over the temporal coherence
time.
2. Frequency fading occurs due to radio waves arriving with delays after reflecting off
objects in the environment. For channels where the signal bandwidth is less than the
coherence bandwidth, the channel can be said to be frequency flat and this fading
can be ignored. Trotter [133] has reported in Table 16 some common coherence
3. Spatial fading occurs because of the constructive and destructive interference of the
radio waves arriving at different positions in space with different amplitudes. A chan-
nel is said to be spatially coherent if a device can move in a channel and not have
183
1.5
1x1x1
1x2x1
1x4x1
1x6x1
Rayleigh
1.25
1
PDF,fα (α)
0.75
0.5
0.25
0
0 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.25 2.5 2.75 3
Envelope, α
Figure 81: The normalized probability density function of a product Rayleigh distribution
for increasing numbers of RFID tag antennas in an 1 × L × 1 Dyadic Backscatter Channel
with fully correlated channels. Note that as the number of tag antennas increase, the
envelope PDF becomes a Rayleigh distribution. Note that these fades are more severe than
for a simple Rayleigh. Modified from [10].
fluctuations in power.
POWs have already been shown to both improve the range [63, 64] and reliability [133] of
RFID tags on the forward-link though their multi-carrier nature. Furthermore, they are
well-suited for use in communication because they provide resistance to multipath interfer-
POWs use multiple carriers in order to create non-linear waveforms which increase the effi-
ciency of charge pumps to enhance the range and reliability of RFID tags. Since RFID tags
modulate their data by changing their radar cross section, when a multi-carrier waveform
184
Table 16: Coherence bandwidths encountered in different environments in a conventional
one-way wireless link
impinges on an RFID tag, it will modulate its data onto each spectral component within
the bandwidth of its RF front end. Thus, it is important that the POW bandwidth fall
within the bandwidth of the RF front end of the RFID. Using a POW allows an RFID tag
to simultaneously reflect identical data on multiple carriers back to the reader at the same
time. This technique can be compared very similarly to orthogonal frequency division mul-
tiplexing (OFDM). In OFDM, a high data rate signal is broken up into N slower signals and
transmitted simultaneously on N subcarriers [134]. This method allows for each subcarrier
to function on a narrow band, frequency flat channel which has better characteristics for
communication. Figure 82 shows a sketch of a data signal modulated onto a POW. In order
to minimize inter-carrier interference, a POW must be designed so that the data on each of
the subcarriers are orthogonal to each other. Similar to OFDM, this requires the symbol
185
1
0.8
0.6
PSD
0.4
0.2
−0.2
Figure 82: A 3-POW with modulated data. Notice how orthogonality is preserved similar
to OFDM. At the peak of each subcarrier, the side lobes from the other subcarriers are
zero. Provided that the signals are properly sampled, no self inter-carrier interference will
occur.
n
∆f = , (119)
T
where ∆f is the subcarrier spacing, 1/T is the tag symbol rate, and n is the subcarrier
are typically unable to do any pulse shaping such as root-raised cosine to limit their inter-
carrier and inter-symbol interference. This fact is due to the nature of the simple state of
the RFID tag’s RF front end switching between (usually) two distinct states. Thus, each
symbol and subcarrier will cause interference with one another. In OFDM systems, clock
synchronization can potentially cause issues when trying to correctly sample the received
waveform. However in backscatter radio systems, coherent receivers are typically employed
which can mitigate these problems. Still, the low-cost oscillators employed for RFID tags
186
Figure 83: An RF tag configured to communicate on a single POW subcarrier. The dotted
line corresponds to the frequency response of the bandpass filter. Note how the charge
pump receives all subcarriers in order to increase charge pump efficiency. However, the RF
tag will only backscatter on subcarrier f3 .
can have frequency deviations which could potentially cause timing problems.
By making a small change to the RF front end of an RFID tag, it is possible to perform
a multiple access technique similar to OFDM frequency division multiple access (OFDM-
FDMA) passively in a backscatter radio system. As previously stated, for an RFID tag to
receive the maximum benefit from a POW, its RF front end must have a bandwidth greater
than or equal to the POW’s bandwidth. However, while it is important that the RFID
charge pump receives this entire bandwidth to maximize charge pump efficiency, from a
communications point of view, only a single subcarrier is needed for data modulation as
they would all carry the same data. Thus, a filter can be placed in line with the tag RF
switch so that the tag will only modulate data onto the subcarrier(s) which fall(s) within
the passband of the filter. Figure 83 illustrates this concept. The filter design is outside the
scope of this paper, but its Q may be too stringent to easily implement on a tag.
187
pow(t)
reader data
reader data signal
signal
N-way
splitter
LO
LO
(a) (b)
A1 f1 f2 f3 f4 f5 f6
f1
A2
f2
Passband
POW(t)
AN
fN
reader data
signal
(c)
Figure 84: Different methods for transmitting POWs. Method (a) is the most simplistic
and uses a DAC with an inefficient and expensive Class A amplifier. Method (b) also uses a
DAC but uses N parallel Class AB or C amplifiers for the POW signal. Finally, method (c)
creates, amplifies, and combines each POW at RF reducing the need for POW upconversion.
From [5].
Several methods of transmitting POWs have been briefly discussed by Trotter in [5] and
are summarized in Figure 84. The major differences between these designs lay in where the
POW is generated, at baseband or RF, and how it is amplified. The large PAPR of POWs
creates difficulties in amplifying these signals, thus the choice of amplifier and transmitter
D.3.3.2 Receiver
Several different receiver designs can be used for POW demodulation and vary in flexibility
and complexity. As backscatter radio systems have a single RF oscillator, all receivers will
be coherent receivers.
188
1. Homodyne (Direct Down-Conversion Receiver) - The simplest method of demodulat-
ing a POW (for a single tag backscattering on all subcarriers) would involve using
ceiver would require a single modification of a low-pass filter around a single POW
subcarrier centered at baseband. However, this received signal will have a lower signal-
to-noise ratio (SNR) than a CW signal would. This disadvantage is due to the fact
that a POW spreads it signal power out into additional subcarriers. Thus, when a
single subcarrier is used, other signal power is wasted. However, passive RFID sys-
tems are forward-link limited and have extra signal margin to spare on the backscatter
link. While this method is the most simple to implement, most of the subcarriers are
thrown away. Additional SNR can be gained by using this information, and the ability
would mix the received bandpass signal with a copy of the original POW at RF
instead of a single frequency carrier wave. The resultant signal would then have a
single carrier low-pass filtered and sampled. The advantage of this technique is that it
averages the subcarriers’ received powers so that if one carrier is in a fade, the signal
will still be maintained. However, there are additional intermodulation products which
spread out the received power across even more frequencies. Thus, this approach is
not recommended.
3. Heterodyne - The most complicated receiver option also provides the most flexibility
taken to ensure that the unmodulated backscatter component of the signals is filtered
out using a comb filter or other type of interference cancellation technique when
189
utilizing a heterodyne receiver. If each RF tag is allowed to have multiple carriers
use a diversity combining technique such as selective ratio combining, maximal ratio
combining, or equal gain combining. In [135], Stüber derives the bit error probabilities
independent of the chosen modulation technique for a Rayleigh fading channel. While
backscatter systems exhibit product Rayleigh or product Rician fading channels and
not a Rayleigh distribution, it is expected that the results will follow Stüber’s work.
That is, the largest gain in improvement results from going from a single branch to
In OFDM, several techniques such as predistortion, power back-off, and nonlinear equaliza-
tion/detection techniques are undertaken in order to reduce the PAPR of a waveform [134].
A high PAPR causes non-linearities with amplifiers due to clipping and dynamic range
issues with analog to digital converters (ADC). However, it is the high PAPR and conse-
quent non-linearity which is beneficial for charge pumps to operate more efficiently. Thus,
any attempt at reducing the PAPR would, in turn, reduce the range of the RFID tag.
Only those techniques which could be implemented on the receiver side, such as nonlinear
D.4 Simulations
D.4.1 Multipath Mitigation of the Backscatter Link
Trotter has shown in [133] that POWs lead to reduced multipath in the forward propagation
channel for turning on RFID tags. The following simulations will demonstrate that POWs
also reduce the effects of multipath in the backscatter channel for communication purposes.
If we operate under the assumption that the subcarrier spacing of the POW is greater than
the coherence bandwidth of the channel, each subcarrier can be treated as an independent
product Rayleigh fading channel. In a 1 × 1 × 1 fully correlated channel, the envelope PDF
190
−0.1
10
Pα2(α<δ)
−0.2
10
−0.3
10
α=0.25δ
α=0.5δ
α=1δ
−0.4
10
0 1
10 10
Number of Subcarriers
Figure 85: Probability that at least one subcarrier is not in a fade for the fully correlated,
product Rayleigh fading channel for various fade margins. Note that as the number of
subcarriers increase, the probability of at least one of the subcarriers not being in a fade
goes to unity.
1 −α
fA (α) = exp( ) (120)
δ δ
This equation can be integrated to get the CDF as shown in Equation (121).
−α
FA (α2 ) = 1 − exp( ) (121)
δ
If an RF tag has M subcarriers, the probability that at least one is not in a fade is
given by Equation (122). Figure 85 shows the probability that at least one subcarrier is
not in a fade for different fade margins. As can be observed, as the number of subcarriers
increases, the probability that at least one of the subcarriers won’t be in a fade goes to
unity. Moreover, the probability increases as the level of the fade increases.
−α M
PF ade (α < δ) = 1 − FA (α2 )M = 1 − (1 − exp( )) (122)
δ
191
D.4.2 POW Design and Modulation/Demodulation Example
An example POW will be designed for use in the 5.8 GHz ISM band (5.725 - 5.875 GHz)
temporal coherence, the maximum doppler shift must be calculated. The doppler
shift is given by
v
∆f = , (123)
λo
where ∆f is the doppler shift, v is the maximum velocity of the object, and λo is the
of 20 m/s, this corresponds to a doppler shift of 386 Hz. The inverse will give the
coherence time of about 2.5ms. Thus the maximum symbol period should be much
less than 2.5ms. Usually, a good rule of thumb for packetized data is that the entire
less than the coherence bandwidth. For simplicity, assume a 1 M-symbol/second sym-
bol rate where each packet is 100 symbols long. Thus, one packet is 100µseconds long.
If binary phase shift keying (BPSK) is assumed, then 1 symbol = 1 bit. Therefore,
the 100µsecond packet meets both the coherence bandwidth criteria as well as the
frequency coherence criteria from Table 16. This lower symbol rate will also help to
also be at least a coherence bandwidth away from the other. Thus, a subcarrier
4. POW Type: For simplicity and to allow the possibility of future multiple access, a
3-POW is chosen. From Trotter’s measurements in [63], a 5.8 GHz, 4-stage Dickson
192
Table 17: Parameters used for the example 5.8 GHz link budget example
charge pump can provide -10 dBm of DC power for an RF input power of 0 dBm
when using a 3-POW. For a CW, the same charge pump requires 5 dBm of input
power to provide the same DC power. Thus using Equation (1) with the parameters
specified in Table 17, the maximum range of the RF tag is approximately 1 meter.
Note that this distance reduces to 50 cm for CW illumination. All of these parameters
Now that the maximum range of the tag has been determined, it is advisable to check
the current link margin for each of the subcarriers. As there are 3 subcarriers, the transmit
power PT is split equally between them. Thus, each subcarrier has approximately 25 dBm
of transmit power. Table 19 outlines the results of the link budget including a link margin
of 44.5 dB. This number means that given the stated parameters, each subcarrier can fade
an additional 44.5 dB before an outage occurs. Thus, this system should be largely robust
against any fades or multipath effects. Alternatively, the tag symbol rate could increase
by using some of the additional link margin. The link margin is so large primarily due to
193
Table 19: Link budget results
the large power required by the RFID tag to perform processing. This large link margin
shows that the range can significantly increase provided that the charge pump becomes
Power-optimized Waveforms (POWs) have been shown to increase the range and reliability
of passive RFID tags by using waveforms with high peak-to-average power ratios (PAPRs)
to allow the RF energy harvesters to operate more efficiently at low power levels. This
appendix has suggested a method for using POWs for communication. It has been shown
that POWs also provide some of the same benefits as OFDM to RFID systems such as
price of complexity and increased baseband processing. However, the increase in demand
for wireless sensor networks (WSNs) and for 100% read rates for RFID systems warrants
Additional work will include channel sounding measurements to determine the relative
phase differences between POW subcarriers. These subcarriers, if out of phase, will degrade
the POW gains exhibited by the charge pumps. The degree to which the subcarriers are
POWs could still be designed so that subcarriers would be in phase, but this would add
an additional constraint to the design criteria. Furthermore, the bit error probability in a
product Rician channel using diversity combining techniques should be investigated with
more detail and compared to the bit error probability for a Rician channel.
194
Pending these results, a POW transmitter should be prototyped and receiver platform
tested using laboratory mixers, filters, and amplifiers along with a software defined radio.
195
APPENDIX E
• Agilent ADS netlists for the simulated example single-shunt rectenna from Chapter 3
and for the prototyped circuits in Chapter 5.
L l =0.5 e−9
C c =0.06 e −12
C p =0.08 e −12
L b=1e−9
end ”HSMS286x SOT−143”
C : C1 N 15 N 2 5 C=Cin F
C : C2 0 N 2 C=Cout F
R : R1 Vout 0 R=Rout Ohm N o i s e=y e s
Options : Options1 Temp=25 Tnom=25 TopologyCheck=y e s F o r c e S P a r a m s=y e s G i v e A l l W a r n i n g s=y e s \
196
MaxWarnings=10 ForceM Params=y e s I n i t i a l G u e s s A n n o t a t i o n =0 T o p o l o g y C h e c k M e s s a g e s=no \
NumThreads=4 DatasetMode=1 doDeltaAC=y e s R e d u c e S P o r t R a t i o =0.5 WarnSOA=y e s MaxWarnSOA=5 \
Census=no
HB: HB2 MaxOrder=4 Freq [ 1 ] = 5 8 0 0 0 0 0 0 0 0 Order [ 1 ] = 3 S t a t u s L e v e l =2 SweepVar=”Pin ” \
SweepPlan=”HB2 stim ” OutputPlan=”HB2 Output ”
MLIN2 : TL9 N 18 N 2 S u b s t=”MSub1” W=W1 m i l L=L1 m i l Wall1 =1.0E+30 m i l Wall2 =1.0E+30 m i l Mod=1
MLIN2 : TL11 N 11 N 3 0 S u b s t=”MSub1” W=W3 m i l L=L3 m i l Wall1 =1.0E+30 m i l Wall2 =1.0E+30 m i l \
Mod=1
MTAPER: Taper2 N 26 N 1 0 S u b s t=”MSub1” W1=W2 m i l W2=Wt m i l L=50.0 m i l
Pin=10
d f =10 e 6
E1=90 n o o p t { 75 t o 105 }
f 0 =5.8 e 9
Z0=50 n o o p t { 25 t o 225 }
MTEE ADS : Tee2 N 25 N 2 9 Vc S u b s t=”MSub1” W1=Wt2 m i l W2=W2 m i l W3=Wt2 m i l
MLIN2 : TL10 N 29 N 2 6 S u b s t=”MSub1” W=W2 m i l L=L2 m i l Wall1 =1.0E+30 m i l Wall2 =1.0E+30 m i l \
Mod=1
P o r t : PORT2 Vin 0 Num=1 Z=Z0 Ohm Freq [ 1 ] = 5 8 0 0 0 0 0 0 0 0 P[ 1 ] = p o l a r ( dbmtow ( Pin ) ∗ 1 , 0 ) N o i s e=y e s \
Pac=p o l a r ( dbmtow ( 0 ) , 0 )
Rout = 3 1 5 . 8 4 2 o p t { 50 t o 500 }
Cout=1e −010 n o o p t { 1 e −013 t o 0.0001 }
Cin=1e −012 n o o p t { 1 e −013 t o 2 e −012 }
model HSMS286x Diode I s =5e−8 Rs=6 N=1.08 Cjo =.18 e −12 Vj =.65 M=.5 Bv=7 I b v=1e−5 A l l o w S c a l i n g =0 \
X t i=2 Eg=.69
197
Short : I o u t N 2 Vout Mode=0 S a v e C u r r e n t=y e s
#u s e l i b ” c k t ” , ” Circulator ”
C i r c u l a t o r : CIR1 N 23 N 1 5 V r e f L o s s 1 =0. dB L o s s 2 =0. dB L o s s 3 =0. dB L o s s 4 =0. dB VSWR1=1. \
VSWR2=1. VSWR3=1. VSWR4=1. I s o l a t =100. dB Z1=Z0 Z2=Z0 Z3=Z0 C h e c k P a s s i v i t y =1
Short : I r e f Vref N 6 Mode=0 S a v e C u r r e n t=y e s
R : R2 N 6 0 R=Z0 Ohm N o i s e=y e s
MTEE ADS : Tee1 N 10 N 1 7 Vd1 S u b s t=”MSub1” W1=Wt m i l W2=Wt m i l W3=Wt m i l
MTAPER: Taper1 N 17 N 1 8 S u b s t=”MSub1” W1=Wt m i l W2=W1 m i l L=50.0 m i l
L1=300
L2 =72.5 o p t { d i s c r e t e 0.5 t o 600 by 0 . 5 }
L3=453 o p t { d i s c r e t e 0.5 t o 600 by 0 . 5 }
L4 =162.5 o p t { d i s c r e t e 0.5 t o 600 by 0 . 5 }
L5 =546.5 o p t { d i s c r e t e 0.5 t o 600 by 0 . 5 }
W1=18.3
W2=10.5 o p t { d i s c r e t e 4 t o 35 by 0 . 5 }
W3=26 o p t { d i s c r e t e 4 t o 35 by 0 . 5 }
W4=25 o p t { d i s c r e t e 4 t o 35 by 0 . 5 }
W5=12.5 o p t { d i s c r e t e 4 t o 35 by 0 . 5 }
Wt=15 o p t { d i s c r e t e 4 t o 35 by 0 . 5 }
Wt2=23 o p t { d i s c r e t e 4 t o 35 by 0 . 5 }
”HSMS286x SOT −143”: I 37 N 30 N 30 N 20 N 20
Short : I i n Vin N 2 3 Mode=0 S a v e C u r r e n t=y e s
model MSub1 MSUB H=9.3 m i l Er =3.95 Mur=1 Cond =5.96E+7 Hu=3.9 e +034 m i l T=1.378 m i l TanD=.019 \
Rough=0 m i l D i e l e c t r i c L o s s M o d e l =1 FreqForEpsrTanD =1.0 GHz LowFreqForTanD =1.0 kHz \
HighFreqForTanD =1.0 THz RoughnessModel=2
MTAPER: Taper3 Vd1 N 1 1 S u b s t=”MSub1” W1=Wt m i l W2=W3 m i l L=50.0 m i l
MTAPER: Taper4 Vc N 2 2 S u b s t=”MSub1” W1=Wt2 m i l W2=W5 m i l L=100.0 m i l
MLSC2 : TL12 N 2 0 S u b s t=”MSub1” W=W4 m i l L=L4 m i l Wall1 =1.0E+30 m i l Wall2 =1.0E+30 m i l Mod=1
MLOC2: TL13 N 2 2 S u b s t=”MSub1” W=W5 m i l L=L5 m i l Wall1 =1.0E+30 m i l Wall2 =1.0E+30 m i l Mod=1
198
L l =0.5 e−9
C c =0.06 e −12
C L=0
C p =0.08 e −12
L b=1e−9
C : C6 0 N 1 2 C=C L F
C : C7 0 N 4 C=C L F
Pin=10
d f =10 e 6
f 0 =5.8 e 9
Z0=50 n o o p t { 25 t o 225 }
MLIN2 : TL2 N 25 N 2 S u b s t=”MSub1” W=W m i l L=L3 m i l Wall1 =1.0E+30 m i l Wall2 =1.0E+30 m i l Mod=1
MLIN2 : TL1 N 5 N 3 1 S u b s t=”MSub1” W=W m i l L=L1 m i l Wall1 =1.0E+30 m i l Wall2 =1.0E+30 m i l Mod=1
W=94
199
Ww=30 o p t { d i s c r e t e 10 t o 100 by 1 }
Wt=30 o p t { d i s c r e t e 10 t o 100 by 1 }
G=5 o p t { d i s c r e t e 5 t o 20 by 1 }
Np=5 o p t { d i s c r e t e 3 t o 20 by 1 }
Ge=5 o p t { d i s c r e t e 5 t o 20 by 1 }
L1=400 n o o p t { d i s c r e t e 20 t o 400 by 1 }
L i =536 o p t { d i s c r e t e 20 t o 500 by 1 }
L3=20 o p t { d i s c r e t e 20 t o 400 by 1 }
P o r t : Term1 N 5 0 Num=1 Z=50 Ohm N o i s e=y e s
MICAP1 : C1 N 31 N 2 5 S u b s t=”MSub1” W=Ww m i l G=G m i l Ge=Ge m i l L=L i m i l Np=Np Wt=Wt m i l \
Wf=W m i l
L l =0.5 e−9
200
C c =0.06 e −12
C L=0
C p =0.08 e −12
L b=1e−9
C : C6 0 N 1 2 C=C L F
C : C7 0 N 4 C=C L F
end ” HSMS2860single SOT −23 v 1 ”
model MSub1 MSUB H=31 m i l Er =2.2 Mur=1 Cond =5.96E+7 Hu=3.9 e +034 m i l T=.035 mm TanD=.0004 \
Rough=0 m i l D i e l e c t r i c L o s s M o d e l =1 FreqForEpsrTanD =1.0 GHz LowFreqForTanD =1.0 kHz \
HighFreqForTanD =1.0 THz RoughnessModel=2
OptimGoal : OptimGoal3 Expr=\
”mag ( I P r o b e 1 . i [ : : , 0 ] ) ∗ mag ( Vout [ : : , 0 ] ) / ( 0 . 5 ∗ mag ( Vin [ : : , 1 ] ) ∗ mag ( I P r o b e 6 . i [ : : , 1 ] ) ) ” \
SimInstanceName=”HB4” Weight=1 \
SpecLimitLine [1]=” OptimGoal3 limit1 ”
S p e c L i m i t L i n e : ” O p t i m G o a l 3 l i m i t 1 ” Type=”GreaterThan ” Min =0.9 Weight=1
Optim : Optim1 OptimType=”random ” ErrorForm=”L2” M a x I t e r s =2000 P=2 D e s i r e d E r r o r =0.0 \
S t a t u s L e v e l =4 F i n a l A n a l y s i s =”None ” N o r m a l i z e G o a l s=y e s S e t B e s t V a l u e s=y e s S a v e S o l n s=y e s \
S a v e G o a l s=y e s SaveOptimVars=no U p d a t e D a t a s e t=y e s SaveNominal=no S a v e A l l I t e r a t i o n s=no \
U s e A l l O p t V a r s=y e s U s e A l l G o a l s=y e s SaveCurrentEF=no I n i t i a l T e m p =0.1 N u m S h o o t s P e r I t e r =20 \
E n a b l e C o c k p i t=y e s S a v e A l l T r i a l s=no
R : R2 N 3 0 R=Z0 Ohm N o i s e=y e s
P o r t : PORT1 Vin 0 Num=1 Z=Z0 Ohm P[ 1 ] = p o l a r ( dbmtow ( Pin ) , 0 ) Freq [ 1 ] = f 0 Hz \
N o i s e=y e s Pac=p o l a r ( dbmtow ( 0 ) , 0 )
C : C5 N 11 N 7 C=0.0403 pF
R : R1 N 7 N 4 R=Rout Ohm N o i s e=y e s
Pin=16
d f =10 e 6
f 0 =5.8 e 9
Z0=50 n o o p t { 25 t o 225 }
L : L1 N 4 N 1 1 L=0.0267 nH N o i s e=y e s
Short : I Probe6 Vin N 6 Mode=0 S a v e C u r r e n t=y e s
Short : I Probe4 Vref N 3 Mode=0 S a v e C u r r e n t=y e s
#u s e l i b ” c k t ” , ” Circulator ”
C i r c u l a t o r : CIR1 N 6 N 2 8 V r e f L o s s 1 =0. dB L o s s 2 =0. dB L o s s 3 =0. dB L o s s 4 =0. dB VSWR1=1. \
VSWR2=1. VSWR3=1. VSWR4=1. I s o l a t =100. dB Z1=Z0 Z2=Z0 Z3=Z0 C h e c k P a s s i v i t y =1
uStrip rectenna v3 Momentum Optim 16dBm : X1 N 28 N 3 1 Vout N 9 N 1 1 L3 m=171 L4 m=93
L5 m=186 L6 m=556 L7 m=105 L8 m=105
” HSMS2860single SOT −23 v 1 ” : I 19 N 9 N 9 N 31
W=94
Rout=370 o p t { d i s c r e t e 150 t o 400 by 10 }
Short : I Probe1 Vout N 7 Mode=0 S a v e C u r r e n t=y e s
HB: HB4 MaxOrder=4 Freq [ 1 ] = f 0 Order [ 1 ] = 5 S t a t u s L e v e l =2 SweepVar=”Pin ” SweepPlan=”HB4 stim ” \
OutputPlan=”HB4 Output ”
201
U s e S a v e d E q u a t i o n N e s t L e v e l=y e s \
S a v e d E q u a t i o n N e s t L e v e l=2 \
U s e D e v i c e C u r r e n t N e s t L e v e l=no \
D e v i c e C u r r e n t N e s t L e v e l =0 \
D e v i c e C u r r e n t D e v i c e T y p e=” A l l ” \
D e v ic e C u r r en t S y m Sy n t a x=y e s \
U s e C u r r e n t N e s t L e v e l=y e s \
C u r r e n t N e s t L e v e l =999 \
U s e D e v i c e V o l t a g e N e s t L e v e l=no \
D e v i c e V o l t a g e N e s t L e v e l =0 \
D e v i c e V o l t a g e D e v i c e T y p e=” A l l ”
202
X t i=2 Eg=.69
L : L6 N 7 N 1 2 L=L l H N o i s e=y e s
C : C1 N 12 N 1 0 C=C c F
C : C2 N 10 N 4 C=C p F
C : C4 N 12 N 4 C=C p F
L l =0.5 e−9
C c =0.06 e −12
C L=0
C p =0.08 e −12
L b=1e−9
C : C6 0 N 1 2 C=C L F
C : C7 0 N 4 C=C L F
end ” HSMS2860single SOT −23 v 1 ”
model MSub1 MSUB H=31 m i l Er =2.2 Mur=1 Cond =5.96E+7 Hu=3.9 e +034 m i l T=.035 mm TanD=.0004 \
Rough=0 m i l D i e l e c t r i c L o s s M o d e l =1 FreqForEpsrTanD =1.0 GHz LowFreqForTanD =1.0 kHz \
HighFreqForTanD =1.0 THz RoughnessModel=2
OptimGoal : OptimGoal3 Expr=\
”mag ( I P r o b e 1 . i [ : : , 0 ] ) ∗ mag ( Vout [ : : , 0 ] ) / ( 0 . 5 ∗ mag ( Vin [ : : , 1 ] ) ∗ mag ( I P r o b e 6 . i [ : : , 1 ] ) ) ” \
SimInstanceName=”HB4” Weight=1 \
SpecLimitLine [1]=” OptimGoal3 limit1 ”
S p e c L i m i t L i n e : ” O p t i m G o a l 3 l i m i t 1 ” Type=”GreaterThan ” Min =0.9 Weight=1
Optim : Optim1 OptimType=”random ” ErrorForm=”L2” M a x I t e r s =2000 P=2 D e s i r e d E r r o r =0.0 \
S t a t u s L e v e l =4 F i n a l A n a l y s i s =”None ” N o r m a l i z e G o a l s=y e s S e t B e s t V a l u e s=y e s S a v e S o l n s=y e s \
S a v e G o a l s=y e s SaveOptimVars=no U p d a t e D a t a s e t=y e s SaveNominal=no S a v e A l l I t e r a t i o n s=no \
U s e A l l O p t V a r s=y e s U s e A l l G o a l s=y e s SaveCurrentEF=no I n i t i a l T e m p =0.1 N u m S h o o t s P e r I t e r =20 \
E n a b l e C o c k p i t=y e s S a v e A l l T r i a l s=no
R : R2 N 3 0 R=Z0 Ohm N o i s e=y e s
P o r t : PORT1 Vin 0 Num=1 Z=Z0 Ohm P[ 1 ] = p o l a r ( dbmtow ( Pin ) , 0 ) Freq [ 1 ] = f 0 Hz \
N o i s e=y e s Pac=p o l a r ( dbmtow ( 0 ) , 0 )
C : C5 N 11 N 7 C=0.0403 pF
R : R1 N 7 N 4 R=Rout Ohm N o i s e=y e s
Pin=−10
d f =10 e 6
f 0 =5.8 e 9
Z0=50 n o o p t { 25 t o 225 }
L : L1 N 4 N 1 1 L=0.0267 nH N o i s e=y e s
Short : I Probe6 Vin N 6 Mode=0 S a v e C u r r e n t=y e s
Short : I Probe4 Vref N 3 Mode=0 S a v e C u r r e n t=y e s
#u s e l i b ” c k t ” , ” Circulator ”
C i r c u l a t o r : CIR1 N 6 N 2 8 V r e f L o s s 1 =0. dB L o s s 2 =0. dB L o s s 3 =0. dB L o s s 4 =0. dB VSWR1=1. \
VSWR2=1. VSWR3=1. VSWR4=1. I s o l a t =100. dB Z1=Z0 Z2=Z0 Z3=Z0 C h e c k P a s s i v i t y =1
” u S t r i p r e c t e n n a v 3 M o m e n t u m O p t i m −10dBm” : X1 N 28 N 3 1 Vout N 22 N 1 1 L3 m=171
L4 m=93 L5 m=186 L6 m=556 L7 m=105 L8 m=105
” HSMS2860single SOT −23 v 1 ” : I 19 N 22 N 22 N 31
W=94
Rout =2200 o p t { d i s c r e t e 2000 t o 4000 by 50 }
Short : I Probe1 Vout N 7 Mode=0 S a v e C u r r e n t=y e s
HB: HB4 MaxOrder=4 Freq [ 1 ] = f 0 Order [ 1 ] = 5 S t a t u s L e v e l =2 SweepVar=”Pin ” SweepPlan=”HB4 stim ” \
OutputPlan=”HB4 Output ”
203
OutputPlan : HB4 Output \
Type=”Output ” \
U s e N o d e N e s t L e v e l=y e s \
N o d e N e s t L e v e l=2 \
U s e E q u a t i o n N e s t L e v e l=y e s \
E q u a t i o n N e s t L e v e l =2 \
U s e S a v e d E q u a t i o n N e s t L e v e l=y e s \
S a v e d E q u a t i o n N e s t L e v e l=2 \
U s e D e v i c e C u r r e n t N e s t L e v e l=no \
D e v i c e C u r r e n t N e s t L e v e l =0 \
D e v i c e C u r r e n t D e v i c e T y p e=” A l l ” \
D e v ic e C u r re n t S y m Sy n t a x=y e s \
U s e C u r r e n t N e s t L e v e l=y e s \
C u r r e n t N e s t L e v e l =999 \
U s e D e v i c e V o l t a g e N e s t L e v e l=no \
D e v i c e V o l t a g e N e s t L e v e l =0 \
D e v i c e V o l t a g e D e v i c e T y p e=” A l l ”
204
REFERENCES
[1] FCC Rules and Regulations Part 15 Section 247 (15.247), Operation within the bands
902 - 928 MHz, 2400 - 2483.5 MHz, and 5725 - 5850 MHz, FCC.
[2] HSMS-286x, Surface Mount Microwave Schottky Barrier Diodes, Avago Technolo-
gies, http://www.avagotech.com/, August 2009.
[3] HSMS-285x, Surface Mount Zero Bias Schottky Detector Diodes, Avago Technolo-
gies, http://www.avagotech.com/, May 2009.
[5] M. S. Trotter, “Range finding in passive wireless sensor networks using power-
optimized waveforms,” Ph.D. dissertation, Georgia Institute of Technology, November
2011.
[6] T. Yoo and K. Chang, “Theoretical and experimental development of 10 and 35 GHz
rectennas,” Microwave Theory and Techniques, IEEE Transactions on, vol. 40, no. 6,
p. 8, June 1992.
[8] Linear Models for Diode Surface Mount Packages, Avago Technologies,
http://www.avagotech.com/, July 2010.
[10] J. D. Griffin and G. D. Durgin, “Gains for rf tags using multiple antennas,” IEEE
Transactions on Antennas and Propagation, vol. 56, no. 2, pp. 563 – 570, February
2008.
[11] N. Tesla, “The transmission of electrical energy without wires as a means for furthering
peace,” Electrical World and Engineer, pp. 21 – 24, January 1905.
[12] G. Crawford, Flexible Flat Panel Displays. John Wiley & Sons Company, 2005.
[13] B. Parviz, “Augmented Reality in a Contact Lens,” IEEE Spectrum, September 2009.
[15] P. E. Glaser, “Power from the Sun: Its Future,” Science, vol. 162, no. 3856, pp. 857
– 861, November 1968.
205
[16] Shinohara, N., “Wireless Power Transmission for Solar Power Satellite (SPS),” Space
Solar Power Institute, Tech. Rep., 2000.
[17] R. M. Dickson, “Power in the sky: Requirements for microwave wireless power beam-
ers for powering high-altitude platforms,” IEEE Microwave Magazine, vol. 14, pp. 36
– 47, April 2013.
[18] D. M. Dobkin and S. M. Weigand, “UHF RFID and Tag Antenna Scattering, Part I:
Experimental Results,” Microwave Journal, Euro-Global Edition, vol. 49, no. 5, pp.
170 – 190, 2007.
[19] J. P. Barrett, Electricity at the Columbian Exposition: Including an Account of the
exhibits in the Electricity Building, the power plant in Machinery Hall, the arc and
incandescent lighting of the grounds and buildings, etc. R. R. Donnelley & Sons
Company, 1894.
[20] W. Brown, “A survey of the elements of power transmission by microwave beam,”
IRE International Convention Record, vol. 9, no. 3, pp. 93 – 106, 1961.
[21] N. Tesla, Colorado Springs Notes, 1899-1900.
[22] ——, Experiments with alternate currents of high potential and high frequency. Mc-
Graw Publishing Company, 1904.
[23] W. Brown, “The history of power transmission of radio by waves,” IEEE Transactions
on Microwave Theory and Techniques, vol. MTT-32, no. 9, pp. 1230 – 1242, September
1984.
[24] ——, “Experiments in the transportation of energy by microwave beam,” IRE
International Convention Record, vol. 12, no. 2, pp. 8 – 17, 1964.
[25] W. Brown, J. Mims, and N. Heenan, “An experimental microwave-powered heli-
copter,” IRE International Convention Record, vol. 13, no. 5, pp. 225 – 235, 1965.
[26] J. Schlesak, A. Alden, and T. Ohno, “A Microwave Powered High Altitude Platform,”
IEEE MTT-S Digest, pp. 283 – 286, January 1988.
[27] K. Takahashi, J. Ao, Y. Ikawa, C. Hu, H. Kawai, N. Shinohara, N. Niwa, and Y. Ohno,
“GaN Schottky Diodes for Microwave Power Rectification,” Japanese Journal of
Applied Physics, vol. 48, no. 4, pp. 1 – 4, April 2009.
[28] K. Nishida, Y. Taniguchi, K. Kawakami, Y. Homma, H. Mizutani, M. Miyazaki,
H. Ikematsu, and N. Shinohara, “5.8 GHz high sensitivity rectenna array,” in
Proceedings of the IMWS-IWPT, 2011, pp. 19–22.
[29] J. McSpadden, L. Fan, and K. Chang, “Design and ex-
periments of a high-conversion-efficiency 5.8-GHz rectenna,”
Microwave Theory and Techniques, IEEE Transactions on, vol. 46, no. 12, p. 8,
December 1998.
[30] Y. Suh and K. Chang, “A high-efficiency dual-frequency
rectenna for 2.45- and 5.8-GHz wireless power transmission,”
IEEE Transactions on Microwave Theory and Techniques, vol. 50, no. 7, pp.
1784–1789, July 2002.
206
[31] D. M. Dobkin, The RF in RFID. Passive UHF RFID in Practice. Elsevier, 2008.
[34] D. Harris, “Radio transmission systems with modulatable passive responder,” U.S.
Patent 2 927 321, August 16 1952.
[35] J. Landt, “The History of RFID,” IEEE Potentials, vol. 24, no. 4, pp. 8 – 11, Octo-
ber/November 2005.
[37] A. Koelle, S. Depp, and R. Freyman, “Short-range radio-telemetry for electronic iden-
tification using modulated RF backscatter,” Proceedings of the IEEE, vol. 63, no. 8,
pp. 1260 – 1261, August 1975.
[42] R. Chaudhri, J. Lester, and G. Borriello, “An RFID Based System for Monitoring
Free Weight Exercises,” The 6th ACM Conference on Embedded Networked Sensor
Systems, November 2008.
[44] H. Chae, D. Yeager, J. Smith, and K. Fu, “Maximalist Cryptography and Computa-
tion on the WISP UHF RFID Tag,” Proceedings of the Conference on RFID Security,
July 2007.
[45] B. Ransford, S. Clark, M. Salajegheh, and K. Fu, “Getting things done on com-
putational rfids with energy-aware checkpointing and voltage-aware scheduling,” in
USENIX Workshop on Power Aware Computing and Systems, November 2008.
207
[46] Y. Ko, S. Roy, J. Smith, H. Lee, and C. Cho, “An Enhanced RFID Multiple Access
Protocol for Fast Inventory,” IEEE Globecom 2007, pp. 4575 – 4579, November 2007.
[47] D. Holcomb, W. Burleson, and K. Fu, “Initial SRAM State as a Fingerprint and
Source of True Random Numbers for RFID Tags,” Proceedings of the Conference on
RFID Security, July 2007.
[48] A. Sample, J. Braun, A. Parks, and J. Smith, “Photovoltaic enhanced UHF RFID
tag Antennas for dual purpose energy harvesting,” in IEEE International Conference
on RFID, April 2011, pp. 146 – 153.
[49] Roundy, S. and Wright, P.K. and Rabaey, J.M, “A study of low level vibrations as a
power source for wireless sensor nodes,” Computer Communications, vol. 26, no. 11,
pp. 1131 – 1144, July 2003.
[50] A. Sample and J. Smith, “Experimental results with two wireless power transfer
systems,” in IEEE Radio and Wireless Symposium, January 2009, pp. 16 – 18.
[51] G. Durgin, “The Hidden Benefits of Backscatter Radio and RFID at 5.8 GHz,” URSI
2008, January 2008.
[53] L. Yang, A. Rida, R. Vyas, and M. Tentzeris, “RFID Tag and RF Structures on a
Paper Substrate Using Inkjet-Printing Technology,” IEEE Transactions on Microwave
Theory and Techniques, vol. 55, no. 12, pp. 2894 – 2901, December 2007.
[55] D. Liu, F. Li, X. Zou, Y. Liu, X. Hui, and X. Tao, “New analysis and design of a RF
rectifier for RFID and implantable devices,” Sensors, pp. 6494–6508, June 2011.
[56] S. Mbombolo and C. Park, “An improved detector topology for a rectenna,” in
Proceedings of the IMWS-IWPT, 2011, pp. 23–26.
[60] ——, “Reflected electro-material signatures for self-sensing passive RFID sensors ,”
in 2011 IEEE International Conference on RFID, 2011, pp. 62 – 69.
208
[61] C. Valenta, P. Graf, M. Trotter, G. Koo, G. Durgin, and B. Schafer, “Backscatter
Channel Measurements at 5.8 GHz Across High-Voltage Corona,” in IEEE Sensors
Conference, November 2010, p. 2400.
[62] C. Valenta, P. Graf, M. Trotter, G. Koo, G. Durgin, W. Daly, and B. Schafer, “Tran-
sient Backscatter Channel Measurements at 5.8 GHz Across High-Voltage Insulation
Gaps,” in AMTA 2010 Symposium, October 2010.
[67] EPCglobal, EPC RF Identity Protocols Class-1 Generation-2 UHF RFID Protocol
for Communications Version 1.2.1, October 2008.
[68] J. D. Griffin and G. D. Durgin, “Complete link budgets for backscatter radio and
RFID systems,” IEEE Antennas and Propagation Magazine, April 2009.
[72] S. Thomas and M. Reynolds, “QAM backscatter for passive UHF RFID tags,” in
IEEE International Conference on RFID, April 2010, pp. 210 – 214.
[73] P. Nikitin and K. V. S. Rao, “Antennas and Propagation in UHF RFID Systems,” in
IEEE International Conference on RFID, April 2008, pp. 277 – 288.
209
[76] C. Valenta and G. Durgin, “Link Budgets for Backscatter Ra-
dio and RFID Systems Using Power-optimized Waveforms,” in
IEEE International Symposium on Antennas & Propagation, July 2013, pp. 237
– 244.
[78] D. M. Pozar, Microwave Engineering, 3rd Edition. John Wiley & Sons, Inc., 2005.
[80] S. Sze, Semiconductor Devices: Physics and Technology, 2nd ed. John Wiley & Sons,
Inc., 2002.
[81] Designing the Virtual Battery: Application Note 1088, Avago Technologies,
http://www.avagotech.com/, July 2010.
[82] J. Lee, B. Lee, and H. Kang, “A high sensitivity CoSi2 -Si Schot-
tky diode voltage multiplier for UHF-band passive RFID tag chips,”
IEEE Microwave and Wireless Components Letters, vol. 18, no. 12, pp. 830–832,
December 2008.
[83] U. Karthause and M. Fischer, “Fully integrated passive UHF RFID transponder
IC with 16.7-µW minimum RF input power,” IEEE Journal of Solid-State Circuits,
vol. 38, no. 10, pp. 1602 – 1608, October 2003.
[84] H. Lin, K. Chang, and S. Wong, “Novel high positive and negative pumping circuits
for low supply voltage,” in IEEE International Symposium on Circuits and Systems,
1999, pp. 238 – 241.
[85] D. Kim, M. A. Ingram, and W. W. Smith Jr., “Efficient far-field radio frequency
energy harvesting for passively powered sensor networks,” IEEE Journal of Solid-State
Circuits, vol. 43, no. 5, pp. 1287 – 1302, May 2008.
[88] T. Ungan, X. Le Polozec, W. Walker, and L. Reindl, “Rf energy harvesting design
using high q resonators,” in IEEE IMWS 2009, 2009, pp. 1 – 4.
[89] J. Hagerty, “Nonlinear circuits and antennas for microwave energy conversion,” Ph.D.
dissertation, University of Colorado, January 2003.
210
[91] C. Chin, Q. Xue, and C. Chan, “Design of a 5.8-GHz rectenna incorporating a new
patch antenna,” IEEE Antennas and Wireless Propagation Letters, vol. 4, pp. 175–
178, 2005.
[92] W. Tu, S. Hsu, and K. Chang, “Compact 5.8-GHz rectenna using stepped-impedance
dipole antenna,” IEEE Antennas and Wireless Propagation Letters, vol. 6, pp. 282 –
284, 2007.
[94] Brown, W.C, “Rectenna technology program: ultra light 2.45 GHz rectenna and
20 GHz rectenna,” Raytheon Company, Tech. Rep., 1987.
[96] P. Koert and J. Cha, “Millimeter Wave Technology for Space Power Beaming,” IEEE
Transactions on Microwave Theory and Techniques, vol. 40, no. 6, pp. 1251 – 1258,
June 1992.
[97] J. Cockcroft and E. Walton, “Experiements with High Velocity Positive Ions (I)
Further Developments in the Method of Obtaining High Velocity Positive Ions,” in
Proceedings of the Royal Society of London, June 1932.
[99] T. Tanzawa and T. Tanaka, “A dynamic analysis of the dickson charge pump circuit,”
IEEE Journal of Solid-State Circuits, vol. 32, no. 8, pp. 1231–1240, August 1997.
[100] F. Pan and T. Samaddar, Charge Pump Circuit Design. McGraw-Hill, 2006.
[101] U. Olgun, C. Chen, and J. Volakis, “Investigation of rectenna array configurations for
enhanced RF power harvesting,” IEEE Antennas and Wireless Propagation Letters,
vol. 10, pp. 262–265, 2011.
[105] H. Matsumoto and K. Takei, “An Experimental Study of Passive UHF RFID Sys-
tem with Longer Communication Range,” in Proceedings of Asia-Pacific Microwave
Conference, 2007, pp. 1 – 4.
211
[106] J. Osepchuk and R. Petersen, “Historical Review of RF Exposure Standards and
the International Committee on Electromagnetic Safety (ICES),” Bioelectromagnetics
Supplement, vol. 6, pp. 7 – 16, May 2003.
[107] ICNIRP Guidelines for Limiting Exposure to Time-Varying Electric, Magnetic and
Electromagnetic Fields (Up to 300 GHz), ICNIRP.
[113] Y. Yao, J. Wu, Y. Shi, and F. Dai, “A fully integrated 900-MHz passive RFID
transponder front end with novel zero-threshold RF-DC rectifier,” IEEE Transactions
on Industrial Electronics, vol. 56, no. 7, pp. 2317–2325, July 2009.
[115] J. Yi, W. Ki, and C. Tsui, “Analysis and design strategy of UHF
micro-power CMOS rectifiers for micro-sensor and RFID applications,”
IEEE Transactions on Circuits and Systems, vol. 54, no. 1, pp. 153–166, Jan-
uary 2007.
[118] T. Paing, E. Falkenstein, R. Zane, and Z. Popovic, “Custom ic for ultralow power rf
energy scavenging,” IEEE Transactions on Power Electronics, vol. 26, no. 6, pp. 1620
– 1626, June 2011.
212
[119] G. Vera, A. Georgiadis, A. Collado, and S. Via, “Design of a 2.45 GHz Rectenna for
Electromagnetic (EM) Energy Scavenging,” in IEEE Radio and Wireless Symposium,
2010, pp. 61 – 64.
[123] D. Wang and R. Negra, “Design of a dual-band rectifier for wireless power transmis-
sion,” in 2013 IEEE Wireless Power Transfer, May 2013, pp. 127 – 130.
[124] S. Imai, S. Tamaru, K. Fujimori, M. Sanagi, and S. Nogi, “Efficiency and harmonics
generation in microwave to DC conversion circuits of half-wave and full-wave rectifier
types,” in Proceedings of the IMWS-IWPT, 2011, pp. 15–18.
[130] Y. Ren, M. Li, and K. Chang, “35 GHz rectifying antenna for wireless power trans-
mission,” Electronics Letters, vol. 43, no. 11, pp. 602 – 603, November 2007.
213
[133] M. S. Trotter, “Mitigation of multipath fading in passive backscatter networks with
power-optimized waveforms,” Georgia Institute of Technology, Tech. Rep., 2010.
[134] H. Rohling, Ed., OFDM: Concepts for Future Communication Systems. Springer,
2011.
214
VITA
Christopher R. Valenta received the BS in Electrical and Computer Engineering (ECE) and
the MSECE from the Georgia Institute of Technology in 2010, and the PhD in ECE from
the Georgia Institute of Technology in 2014. In 2008, Christopher joined The Propagation
Group under the direction of Prof. Gregory Durgin where he helped to commercialize
the first 5.8 GHz backscatter sensor system for use in high-voltage environments and has
studied backscatter radio, software defined radio, RFID tag design, and microwave-energy
harvesting. During his time at Georgia Tech, he also worked with the Electro-Optical
Systems Laboratory (EOSL) at the Georgia Tech Research Institute (GTRI) under the
RFID applications, and systems level engineering. In 2009, he was awarded an ATLANTIS
Fellowship and participated in an 18-month research program with the Politecnico di Torino
and the Technical University of Munich. From 2011 to 2013, Christopher was awarded a
Shackelford Fellowship from GTRI to help support his work in RFID-enabled sensors. He
also spent three months with Sampei Laboratory at Osaka University in Osaka, Japan as
a visiting scholar in the fall of 2012. After graduation, Christopher accepted a research
faculty position with the Electro-Optical Systems Laboratory (EOSL) at the Georgia Tech
215