Paper Identification Number: 09-2189-TIE

1

Abstract In this paper a direct AC-DC power electronic
converter topology is proposed for efficient and optimum energy
harvesting from low voltage microgenerators. The converter
utilizes the bidirectional current conduction capability of
MOSFETs to avoid the use of a front-end bridge rectifier. It is
operated in discontinuous conduction mode (DCM) and offers a
resistive load to the micro-generator. Detailed analysis and
modeling of the converter is presented. In such low power
applications, the power consumption of gate drive and control
circuits should be minimal. In this paper, they are specifically
designed to consume very low power. A suitable start-up circuit
and auxiliary DC supply circuit is proposed for implementation
of the converter. A low voltage microgenerator is used to verify
the performance and operation of the converter and the gate
drive circuits.
Index Terms AC-DC power conversion, boost converter,
energy harvesting, low power, and low voltage.
I. INTRODUCTION
HE development of energy-efficient semiconductor
devices has reduced the power requirements of electronic
circuits. This has led to the development of wireless electronic
devices like sensor nodes, medical implants etc. which require
only a few milliwatts for their operation. They can be powered
by harvesting ambient energy from the environment in the
form of light, vibration, heat etc [1-9]. Vibration, being a
ubiquitous source of low power has been a research focus for
many years [10-20]. Electromagnetic microgenerators are
particularly popular due to high energy density and are hence,
considered for this work.
Such microgenerators are typically spring-mass systems, in
which mechanical energy is converted to electrical energy by
electromagnetic damping (Fig. 1). The output of an inertial

Manuscript received December 29, 2009, modified April 13 2010.
Copyright (c) 2009 IEEE. Personal use of this material is permitted. However,
permission to use this material for any other purposes must be obtained from
the IEEE by sending a request to pubs-permissions@ieee.org.
Rohan Dayal is with Department of Electrical, Computer and Systems
Engineering, Rensselaer Polytechnic Institute, Troy, NY-12180 USA
(email:dayalr2@rpi.edu).
Suman Dwari is with United Technology Research Center, Hartford, CT,
USA (email:dwaris@gmail.com).
Leila Parsa is with Department of Electrical, Computer and Systems
Engineering, Rensselaer Polytechnic Institute, Troy, NY-12180
(email:parsa@ecse.rpi.edu).

(a)

(b)
Fig. 2. Block diagram: (a) two-stage power conversion, (b)reported dual
polarity boost converter [29]
microgenerator is typically around a few hundred millivolts
AC. The conventional power converters (Fig. 2(a)) reported
for vibration energy harvesting mostly consist of a front end
diode bridge rectifier followed by a standard buck or boost
converter [21-28]. This arrangement of two stage power
conversion has several disadvantages for the electromagnetic
microgenerator (1) diode voltages in a bridge rectifier are
difficult to overcome for low input voltage, (2) input current is
much higher than output current leading to more losses in
diodes, (3) rectifier offers a non-linear load which makes the
converter unsuitable for energy harvesting. For energy
Design and Implementation of a Direct AC-DC
Boost Converter for Low Voltage Energy
Harvesting
Rohan Dayal, Student member, IEEE, Suman Dwari, and Leila Parsa, Member, IEEE
T

Fig. 1. Resonance based inertial electromagnetic microgenerator.
Paper Identification Number: 09-2189-TIE

2
harvesting applications, a power converter is required which
can condition the AC output of the microgenerator to a
suitable DC level and offer an adjustable resistive load to the
microgenerator [18]. Direct AC-DC boost converters can be
used to address these issues. A dual polarity boost converter
(Fig. 2(b)) topology for such AC-DC conversion was recently
reported in [29]. This converter uses two inductors and the
output DC bus is split into two series connected capacitors.
The fundamental principle behind the proposed topology is
the use of two boost converters each operating in one half
cycle of the AC voltage. Each capacitor is charged only in the
respective half cycle. However, they discharge to the load
continuously causing large voltage drops. Extremely large
capacitors are needed to make the voltage ripple acceptable.
This makes the converter response very slow. The control of
the converter also requires input line polarity sensing adding
further complexity. The efficiency of the reported converter is
only 50% using simulation results.
In [30], the authors presented a direct AC-DC converter
which utilizes the bidirectional current conduction capability
of MOSFETs for direct AC-DC conversion. The converter
utilizes only one inductor and charges the output capacitor
continuously without any line voltage sensing. The paper
presented only open loop simulation and experimental results
using an ideal sinusoidal voltage source. The reported
converter also utilized two n-channel MOSFETs to form a
bidirectional switch. However, floating gate drivers and
indirect feedback control were needed for closed loop
operation. For a low power application, it is difficult to realize
such drivers. In the present paper, the proposed split capacitor
topology utilizes an n and p-MOSFET pair to form the
bidirectional switch which does not need floating gate drivers.
This work also includes detailed analysis and modeling for the
converter. Dedicated control and driver circuits that consume
very low power suitable for such an application have been
designed. A simple control strategy for energy harvesting and
output voltage regulation is presented. Finally, experimental
results for the converter with a low voltage microgenerator are
presented.
The rest of the work is organized as follows: Section II
presents a brief review of the converter. In section III, the
analysis and modeling of the converter is described. Section
IV presents the design of the control and driver circuits. A
prototype of the converter is developed to verify its operation
and performance with the designed controller. A suitable start-
up circuit along with a dual DC supply for stand-alone
operation of the energy harvesting system is also presented.
The simulation and experimental results are presented in
Section V. Finally the paper is concluded.
II. REVIEW OF THE CONVERTER
The converter topology utilizes a bidirectional switch for its
operation. However, a single semiconductor device capable of
both bidirectional conduction and blocking capability does not
exist. A MOSFET channel is typically capable of conduction
in both directions when it is sufficiently turned ON.
However, due to the presence of the inherent body diode it
cannot block current in the reverse direction. A bidirectional
switch in the present case is realized by connecting the drain
of a n-MOSFET to the source of a p- MOSFET so that their
body diodes block the current in opposite direction. The
MOSFETs are turned on and off at the same instants and thus
can conduct and block currents in both directions. This
bidirectional switch is referred as S1 throughout the paper.
It should also be noted that the converter is operated in
discontinuous conduction mode (DCM). This operation has
many advantages (1) a constant duty cycle extracts constant
power from the source enabling a simple control, (2) a
converter operating with a constant duty cycle has only
fundamental and switching harmonic frequency components
(much higher than fundamental) thus, offers a resistive load to
the microgenerator; (3) DCM operation reduces switching
losses which are significant in such low power applications.
The circuit diagram for the split capacitor topology is
shown in Fig. 3. A single inductor L is used for the boost
operation in both half cycles. The converter utilizes three
capacitors to boost the low AC voltage. The split capacitors
C2 and C3 are charged in alternate half-cycles as explained
below.
A. Positive half-cycle: The inductor current increases
linearly from zero when the switch S1 turns ON. When
S1 is turned OFF, the body diodes block the circulating
current. The diode D1 is forward biased and the current
flows into the capacitor C2 to complete the charging
process.
B. Negative half-cycle: In the negative half-cycle, the current
rises in opposite direction when S1 is turned ON.
However, this time when S1 is turned OFF, diode D1
remains OFF and diode D2 is forward biased. The
inductor energy is transferred to the capacitor C3.
The three capacitors share energy through charge recycling.
It should be noted that even though voltages across capacitors
C2 & C3 show large variations, the duty cycle can be
effectively controlled to maintain a steady voltage across the
output capacitor C1.
III. CONVERTER ANALYSIS
A. Input-side analysis
A bidirectional switch in the input side is used to build the
inductor current [31]. The input current waveform of the
converter can be considered as shown in Fig. 4(a). Consider
any k
th
switching cycle of the boost converter as shown in Fig.
Fig. 3. Proposed direct AC-DC converter: split capacitor topology.
Paper Identification Number: 09-2189-TIE

3
4(b); where T
s
is the time period of the switching cycle, D is
the duty cycle of the converter, d
f
T
s
is the boost-inductor
current fall time (or the output diode conduction time), V
i
is
the voltage of the microgenerator with amplitude V
p
, and V
0
is
the converter output voltage. The converter switching
frequency is much higher than the generator output AC
voltage frequency. Therefore, T
s
is much smaller than the time
period of the input AC cycle (Ti). In the analysis presented
henceforth, circuit parasitics are ignored and it is assumed that
output voltage remains constant over a switching cycle. The
peak value of the inductor current (i
pk
) for a general switching
cycle can be obtained as in (1).


(a)


(b)
Fig. 4. Converter input current: (a) Input current over one cycle of
microgenerator output voltage , (b) input current over one switching cycle
1

L DT V DT m i
s ik s Pk
= = .
1


(1)
The inductor current fall time can be found as:
) (
2 ik o pk pk s f
V V L i m i T d = =
(2)
During this switching cycle, the energy (
k
E
) transferred
from the input to the output can be obtained as:
2 ) ( . .
f s pk ik k
d D T i V E + =

(3)
where, ( )
i s p ik
T T k V V . . 2 sin t = . Defining N=T
i
/T
s
, the
average input power,
i
P , of the converter can be obtained as in
(4).

= =
+ = =
N
k
f s pk ik i
N
k
k i i
d D T i V T E T P
1
.
1
2 ) ( . . ). 1 ( ). 1 (

(4)
Furthermore, for large N, the summation in (4) can be
approximated as integration. Therefore, the average input
power, P
i
, can be derived as in (5).
|
1
2 2
4L
T D V
P
s p
i
=
( ) ( ) | | dt V V
p
. sin . 1 . sin . 1
1
0
2
0
2
=
}
u u t |
t

(5)
where
i
T t . 2t u =
. It can be noted that in the above equation,
(5),

is constant for fixed V
p
and V
o
. Under steady state
condition, the above equation can be used to estimate the
operating duty cycle of the converter for a particular load.
This information is further used to design the MOSFETs,
inductor and diodes of the converter.
B. Output-Side Analysis
A switch-diode low frequency model can be formed by
averaging over the switching frequency as shown in Fig. 5.
This allows us to study the input current and output voltage
related characteristics.
The port currents I
1
(t) and I
2
(t) can be averaged over the
switching cycle according to the DCM operation. Their
waveforms are depicted in Fig. 6. The three time periods D.T
s
,
d
f
.T
s
and d
s
.T
s
correspond to the three states of the converter in
each switching cycle. During period DT
s
, the switch S1 is ON.
The period d
f
T
s
corresponds to the time when S1 is turned off
but the inductor current is still non-zero and d
s
T
s
is the
discontinuous state of the converter. The waveforms for port 1
voltage and current are shown in Fig. 6(a). Their averaged
values can be derived as:
| |
s
s f peak
avg
i i o f avg
T
T d D i
t I
t V D t V t V d t V
) ( 5 . 0
) (
) ( ) 1 ( ) ( ) ( ) (
1
1
+
=
+ =



(6)

The current I
2
(t) corresponds to the diode current. The
voltages V
2
(t) and I
2
(t) are similarly averaged as:
s
s f peak
avg
o avg
T
T d i
t I
t V t V
5 . 0
) (
) ( ) (
2
2
=
~



(7)



Fig. 5. DCM Switch model
Paper Identification Number: 09-2189-TIE

4

(a) (b)

Fig. 6. (a)Port 1:Voltage V
1
& Current I
1
(b)Port 2: Voltage V
2
& Current I
2

The values of peak current and fall time have already been
calculated in (1) & (2) and can be used to simplify the
equations in (6) & (7). These equations are further used to
model the converter as in Fig. 7.

Fig.7. Converter equivalent circuit
The obtained values for the emulated resistance (R
e
) and the
dependent current source (I
2avg
(t)) are as:

|
|
.
|

\
|
= =
) (
) (
1
2
) (
) (
) (
2
1
t V
t V
T D
L
t I
t V
t R
o
i
s avg
i
e


(8)
|
.
|

\
|
(

|
.
|

\
|
=
L
T
t V
t V
t V
t V
D
t I
s
o
i
o
i
avg
2
) (
) (
1
) (
) (
) (
2
2
2



(9)
As long as the converter is operated for a high step-up ratio
(V
o
>>V
i
), the only controllable variable for the emulated
resistance in (8) during operation of the converter is the duty
cycle. This also shows that the input current contains only
fundamental component and high switching frequency
components. The output current model can be easily used to
predict the low frequency behavior of the converter output
voltage. The basic relation governing the output voltage for
load resistance, R
L
is:
L
o
avg
o
R
t V
t I
t
t V
C
) (
) (
) (
2 1
=
c
c


(10)
The square of input voltage in (9) is the reason for the
second order frequency ripple in the output voltage waveform.
This large signal model can be used to predict the output
voltage behavior. The non-linear equations ((9) & (10)) can be
solved in real-time using any equation solver like simulink. A
numerical simulation of such a model for the predesigned
parameters (TABLE I, [30]) is shown in Fig. 8. The output
voltage waveform is very similar to the open loop results
obtained in [30] and the experimental results presented later.
Further insight into the circuit configuration is needed to
predict the output voltage of the split capacitors, C2 and C3. It
should be noted that the model described below is valid for
only half-line cycle. This is because of the fact that the split
capacitors C2 and C3 are charged only in positive and
negative half cycle respectively. Since, the aim of this model
is to predict this transient behavior further averaging over full
line-cycle is not attempted. An equivalent circuit for the
positive half-cycle can be shown as in Fig 9.
In the positive half-cycle, the capacitor C2 is charged and
the capacitor C3 discharged such that the output capacitor C1
holds steady voltage with a second order harmonic ripple.

Fig. 8. Low order frequency model for output voltage


Fig. 9. Split capacitor model

The capacitors C2 and C3 are chosen to be of equal value.
Each capacitor is charged and discharged in alternate half-
cycles. Therefore, due to symmetry; in each half-line cycle,
the initial voltage of one capacitor will be equal to the final
voltage of the other capacitor and vice-versa. Therefore, for a
general half-cycle, the following relationship holds true:

Paper Identification Number: 09-2189-TIE

5
|
.
|

\
|
=
|
.
|

\
|
=
2
) 0 ( ;
2
) 0 (
2 3 3 2
i
c c
i
c c
T
V V
T
V V


(11)
A state-space model for voltages V
C2
(t) and V
C3
(t) can be
obtained for the circuit in Fig. 9. The details of the model are
presented in Appendix. The input is the controlled current
source which is a function of the output voltage V
C1
(t) as in
(9). A numerical simulation of the state space model for half-
cycle is done using simulink (Fig. 10). The difficulty with
such a simulation is that the initial conditions of the state
variables need to be accurately known. The initial conditions
are taken from the actual SPICE simulation results of the
converter. The waveform in Fig. 10 closely matches the
experimental results as well. In the next half cycle, capacitor
C3 is charged instead of C2. Therefore, the split capacitors
show large variations in voltage.
Fig. 10. State-space simulation of split capacitor voltages

Fig. 11. The proposed energy harvesting system
IV. CONVERTER IMPLEMENTATION
In low power applications, the implementation of auxiliary
circuits (gate driver and control circuits) is very important.
They should be chosen such that they consume very low
power and are able to drive the circuit in steady state. In this
section, a gate driver circuit and an auxiliary DC supply with
start-up circuit are presented. A control strategy with its
analog implementation is also described. The schematic of the
power converter along with the auxiliary circuits is shown in
Fig. 11.
A. Feedback and control circuit
A simple PI controller is implemented to regulate the output
voltage. As can be seen in Fig 11, the negative rail of the
output voltage is not the same as the ground of the control
circuit which is the common node of the split capacitors.
Therefore, a specific feedback circuit for the converter has to
be designed.
For the split capacitor topology, the capacitors C2 and C3
are chosen to have equal value. The voltages across these two
capacitors are shown in Fig. 10. It can be found from previous
analysis that the average voltage across any of these split
capacitors is half of the output voltage. Therefore, this voltage
can be used as feedback to the controller. However, due to
charge recycling, the voltages across the split capacitors have
designed to eliminate the voltage ripple and extract the DC


Fig. 12. A single op-amp based PI controller
Paper Identification Number: 09-2189-TIE

6
large AC ripple. Two cascaded single pole low pass filters are
voltage information (V
f
).
The PI controller is used to regulate the output voltage to a
nominal value of 3.3V. A single op-amp in differential mode
(Fig. 12) is used to implement the controller and the error
amplifier. According to the normal gain in differential mode:
( )
f ref duty
V V
Z
Z
V
|
|
.
|

\
|
=
1
2


(12)
where Z
2
=R
p
+1/(jC
I
) and Z
1
=R
1
The various components of Z
2
and Z
1
are defined in Fig
12. The impedances Z
2
and Z
1
are chosen such that the
expression in (13) implements a PI controller with a
proportional gain K
p
and integral gain K
I
where,
) /( 1
/
1
1
I I
p p
C R K
R R K
=
=


(13)
B. Gate Driver Circuit
The bidirectional switch is realized using an n-MOSFET
(M
n
) and a p-MOSFET (M
p
) connected in series (Fig. 13). The
source of the n-MOSFET is connected to the ground. The
source of the p-MOSFET is connected to the drain of the n-
MOSFET. Such an arrangement gives the switch S1
bidirectional current capability and ability to block the reverse
conduction through body diodes. The schematic of the gate
driver circuits is shown in Fig. 11 along with the overall
converter system. In this converter, the MOSFETs are driven
with respect to common node of the split capacitors, C2 and
C3. Comparator C
P1
is used to drive the n-MOSFET. Since,
the source of the n-MOSFET M
n
is connected to the ground; it
can be driven with a conventional low-side driver.
Comparator C
P2
is used to drive the p-MOSFET M
p
using a
negative gate pulse
.
It should be noted that the MOSFET M
p
is
driven with respect to ground instead of its source S
p
.
However, since the voltage drop across MOSFET M
n
is very

Fig.13. Bidirectional Switch S1
small during conduction; the gate drive voltage can turn on the
MOSFET M
p
properly. The inputs to the comparator C
P2
are
connected in such a way that the output is negative when the
value of the duty cycle is higher than the sawtooth waveform.
Therefore, both the MOSFETs are turned on at the same time.
C. Start-up Circuit
It can be observed that the controller and driver circuits
require a dual DC supply for their operation. The low AC
input voltage cannot be used to start the converter. A suitable
circuit, as indicated in Fig. 11, is used as start-up circuit. The
voltage nodes V
A
+
and V
A
-
denote the positive and negative
DC voltages which power the controller and gate driver in the
converter system. The batteries E1 and E2 provide the start-up
power to charge the capacitors C
a
and C
b
through diodes D
c

and D
d
. With the controller and driver circuits operating, the
capacitors C
a
and C
b
start getting charged by the
microgenerator through diodes D
a
and D
b
. This boost
mechanism is similar to the charging of capacitors C2 and C3
in the split capacitor topology. Further, these capacitors are
designed to maintain steady DC voltage while powering the
auxiliary circuits. In steady state, the voltages across the
capacitors C
a
and C
b
can be approximately related as: V
A
+
=V
0

and V
A
-
= -V
0
. The nominal value of battery voltages are
chosen to be less than steady state voltages of capacitors C
a

and C
b
. Therefore, when the converter is operating, capacitors
C
a
and C
b
are charged by the energy harvesting converter.
Under this condition, the diodes D
c
and D
d
become reverse
biased and batteries E1 and E2 are cutoff from the circuit.
It should be noted that the batteries utilized for start-up are
of really small capacity and footprint. Once the converter starts
operating, capacitors C
a
and C
b
can be used to keep the
batteries, E1 and E2 charged. If they get discharged, the DC
bus voltages V
A
+ and V
A
- on capacitors C
a
and C
b
charge these
batteries through diodes D
R1
and D
R2
. These diodes are so
chosen that they only get forward biased when the battery
voltage drops below a nominal voltage of 3V. The connecting
resistors R
R1
and R
R2
are chosen to limit the charging current in
the batteries.
V. RESULTS
The prototype for the converter was developed to verify its
operation. The values of the key components of the converter
are presented in the Table-I.
The auxiliary circuits were developed using micropower
components. The gate driver circuit is realized using a low
voltage comparator, TLV2760 from Texas Instruments which
has a nominal current requirement of 20A per channel.
TABLE I
Parameter Value Parasitic

Switching Frequency 10kHz -
Output Voltage (targeted)
3.3
V@R=1k
(nominal
load)
-
Capacitor (C1) 22F R
esr
= 33m
Capacitor (C2 & C3) 4.7F R
esr
= 5m
Inductor (L) 10H R
esr
= 80m
N & P channel MOSFET (M
n

& M
p
)
20V, 6A
R
ds_on
(N-FET
)

= 30m
R
ds_on(
P-FET
)
=
40m
Schottky Diode (D1,D2, ,Da,
Db, Dc , Dd)
23V, 1A
V
f
=0.25V
@15mA
Paper Identification Number: 09-2189-TIE

7
The micropower op-amp used as error amplifier and
controller is LT1541 from Linear Technology with an internal
reference of 1.2V. A resonance-based linear microgenerator
capable of producing 400 mV AC at 108 Hz is used as input
to the converter. An electrodynamic shaker is used to produce
vibrations for the electromagnetic microgenerator. The
converter is operated in DCM to reduce switching losses. The
gate pulses for the MOSFETs M
n
and M
p
, generated by the
comparators, C
p1
and C
p2
are shown in Fig. 14. It can be seen
that the gate voltage of n-MOSFET (V
gn
) is positive when p-
MOSFET gate voltage (V
gp
) is negative. Therefore, both the
MOSFETs are turned ON at the same instants. The gate pulses
for the MOSFETs are not perfectly square in nature. This is
caused by the low slew rate (0.2 V/s) of the low-power
comparator (TLV2760) being used as the pulse generator. This
nature of gate pulses does not affect the normal operation of
the converter. The threshold voltage (V
tn
and V
tp
) of the
MOSFETs is . 2 . 1 V Therefore, as shown in Fig. 14; the on-
time (t
on
) of the MOSFETs is the period for which the gate
pulse value is more than the threshold voltage. The use of a
faster comparator would have led to high losses in the gate
driver circuits.
The converter is operated at a switching frequency of 10
kHz with a nominal load of R
L
=1k. The input voltage (V
i
)
and the input current (I
i
) for the converter are indicated in Fig.
15. Due to the non-linear microgenerator operation, the
Fig. 14. PWM waveforms for MOSFETs Mn and Mp (1V/div)
Fig. 15. Input voltage (0.5V/div) & Current(0.5A/div)


Fig. 16. (a) Input current (0.1A/div) (b) Gate pulses (Vgn) (2V/div)
input voltage is not an ideal sinusoid. It can be seen that the
converter operates in DCM and the input current follows the
input voltage profile. The input current over a few switching
cycles is shown in Fig. 16 along with positive gate signals.
The input current has an exponential rise instead of an ideal
ramp because of the L-R circuit formed by the on-state
resistance (R
ds-on
) of the MOSFETs and boost inductor L. The
output voltage is shown in Fig. 17 along with the steady state
duty cycle of the controller, D=0.2. The average output
voltage is estimated to be around 3.3 V. The output voltage
ripple is estimated to be within
%. 5
The output voltage is
shown in Fig. 18 along with the voltages of the split
capacitors, C2 and C3. It can be seen that the capacitor C1
holds a steady output voltage while the other two capacitors
(C2 & C3) charge and discharge in alternate half cycles. The
capacitors C2 and C3 have positive and negative voltages with
respect to the system ground, which is the basis for the dual
DC supply design for the controller and driver circuits. The
DC bus voltages, VA+ and VA- are shown in Fig. 19. The
bus is initially powered by the batteries and has a steady
voltage without any ripple. As the output voltage builds up,
the batteries are cut-off from the circuit and the capacitors C
a


Fig. 17. Output Voltage (1V/div) and Duty cycle (0.2V/div)
Paper Identification Number: 09-2189-TIE

8

Fig. 18. Voltages for Output Capacitors (V
C1
,V
C2
,V
C3
) (1V/div)
and C
b
are charged by the energy harvesting converter. It can
be seen that the capacitors are now charged in alternate
halfcycles while continuously powering the gate driver and
controller circuits. The slightly increased ripple on the
negative bus can be attributed to imbalance in the power
requirements of the gate drivers. The on-state resistance of the
chosen p-MOSFET is same as that of the n-MOSFET.
However, its gate charge is more than the gate charge of the n-
MOSFET. This increases the gate driver losses for comparator
C
P2
and hence, a bigger ripple is observed on the negative DC
bus. In Fig. 20, the gate pulses along with sawtooth waveform
and steady state duty cycle are depicted. The slight delay in
comparison is attributed to the finite propagation delay of the
comparator. To estimate the power consumed by the auxiliary
circuits, the converter is first operated using external DC
supplies. Under this condition, the output voltage is measured
for a known duty cycle value. Further, the converter is operated
at the same duty cycle without these supplies. It is found that
the difference in output power in both cases is around 400 W.
As discussed earlier in section III.A, for a fixed duty cycle the
input power remains same. Therefore, this difference in output
power can be attributed to the power consumed by the
auxiliary circuits.
For the load of 1k, the output power is 10mW. The
efficiency of the overall converter system for this nominal

Fig. 19. DC bus voltages V
A
+ and V
A
- (2V/div)

Fig. 20. (a) Sawtooth waveform (0.5/div), (b) Duty cycle (0.5V/div) and (c)
Gate pulses(0.5V/div)
load is around 55%. The performance of the converter should
also be estimated for different input and output conditions.
Real device models are utilized to obtain these simulation
results. The output voltage is always regulated at 3.3V. The
efficiency of the converter for different input voltages is shown
in Fig. 21. It can be seen that as the input voltage increases, the
efficiency of the converter increases as well. This trend is
expected for constant power loads, because increase in input
voltage decreases the current flowing in the circuit, and hence,
conduction losses are reduced. The efficiency of the converter
for different load resistances is shown in Fig. 22. The

Fig. 21. Efficiency Curves for different input voltages

Fig. 22. Efficiency of the converter with variation of load
Paper Identification Number: 09-2189-TIE

9
efficiency of the converter is higher at lighter loads. However,
at really light loads, the duty cycle of operation becomes very
low (2-5%) making the conversion process inefficient A
comparison of the proposed converter with other low power
converters is depicted in TABLE II. The boost converter
system proposed by Cao et. al [21] consists of a rectifier
followed by a boost converter. It is designed for an output
power of 35mW. However, this set-up does not work well for
low input voltages. The dual polarity boost converter proposed
by Mitcheson et. al [29] works for low input voltages.
However, the reported efficiency of the converter is only 50%
through simulation results. No experimental results were
reported and details about the control circuit implementation
were not presented.
TABLE II

Electromagnetic Energy Harvesting
Converters
Input Voltage (V
i
) Output Voltage
(V
0
)
Power
(P
0
)
Efficiency
( )

Boost Converter with Feeedforward Control
[21]
0.4V-1.4V DC(after
rectifier)
3.3V 35 mW 35-65%

Dual Polarity Boost Converter [29] 0.3V AC 3.3V 50 mW 50% (simulation)

Proposed Split Capacitor Topology 0.2V-0.8V AC 3.3V 10 mW 55% (including auxiliary
circuits)
VI. CONCLUSION
The paper presents a split-capacitor based AC-DC boost
converter for low power, low voltage energy harvesting. A
bidirectional switch, based on, series connected n-MOSFET
and p-MOSFET, is proposed in this work. The converter
utilizes this bidirectional switch to boost the low AC
microgenerator voltage to a steady DC voltage in both
theinput half cycles. The modeling and analysis for the
converter has been presented in detail. The auxiliary circuits
in the energy harvesting converter - gate driver circuits and
the control circuit, have been designed for low power
operation. A suitable start-up circuit, an auxiliary DC supply
and a feedback circuit are proposed for the implementation of
the converter. Experimental results for a low voltage
microgenerator have been presented to verify the operation of
the converter and the proposed auxiliary circuits. The
designed auxiliary circuits draw minimal power and are able
to operate the converter at a high efficiency.
APPENDIX: DERIVATION OF STATE SPACE MODEL FOR SPLIT
CAPACITORS
The following equations can be written from Fig. 9 using
circuital analysis,
) ( ) ( ) ( ) (
3 2 1 0
t V t V t V t V
C C C
+ = =
L C C C avg
R V V C V C t I / ) (
1 1 1 2 2 2
+ + =

3 3 2 2 2
) (
C avg C
V C t I V C

+ =


(14)
Simplifying, by substituting for V
C1
(t)
L C C avg C C
R V V t I V C V C C / ) ( ) ( ) (
2 3 2 3 1 2 2 1
+ = + +

) (
2 3 3 2 2
t I V C V C
avg C C
=



(15)

Since I
2avg
(t) is a function of V
C1
(t) as indicated in (9),
equation (16) represents a system that can be solved for V
C2
(t)
and V
C3
(t). As the present system is not in the form of
conventional linear state space model, a change in state
variables was employed to solve for V
C2
(t) and V
C3
(t). The
new state variables, W
1
(t) and W
2
(t) can be defined as:


3 1 2 2 1 1
) (
C C
V C V C C W + + =
3 3 2 2 2 C C
V C V C W =



(16)

The resulting linear state space system model can be

represented as:
| | ) ( ) (
) (
1
2 2 2 1 2 1
3 1 3 2 2 1
1
t I W C W C C
C C C C C C R
W
avg
L
+ +
+ +
=

) (
2 2
t I W
avg
=

(17)
This model can be solved for a half-line cycle to get the
waveforms for V
C2
(t) and V
C3
(t).
REFERENCES
[1] J. A. Paradiso and T. Starner, "Energy scavenging for mobile and
wireless electronics", IEEE Pervasive Computing, vol. 4, pp. 18- 27,
March 2005.
[2] J. Krikke, "Sunrise for energy harvesting products", IEEE Pervasive
Computing, vol. 4, pp. 4 5, March 2005.
[3] S. Meninger, J. O. Mur-Miranda, R. Amirtharajah, A. P. Chandrakasan,
and J. H. Lang, "Vibration-to-electric energy conversion," IEEE Trans
Very Large Scale Integration (VLSI)Systems, vol. 9, pp. 64-76, 2001.
[4] M. El-Hami, P. Glynne-Jones, N.M. White, M. Hill, S. Beeby, E. James,
A. D. Brown, and J. N. Ross, "Design and fabrication of a new
vibration-based electromechanical power generator," Sensors and
Actuators A: Physical, vol. 92, Issues 1-3, pp. 335-342, August 2001.
[5] Travis M. Thul, Suman Dwari, Robert D. Lorenz, and Leila Parsa,
"Energy Harvesting and Efficient Power Generation from Human
Paper Identification Number: 09-2189-TIE

10
Activities," presented at Proc. Center for Power Electronics Systems
(CPES) Seminar, April 2007.
[6] J. R. Amirtharajah and A. P. Chandrakasan, Self-powered signal
processingusing vibration-based power generation, IEEE Journ. Solid-
State Circuits, vol. 33 No. 5, pp. 687-695, May 1998.
[7] D. Dondi, A. Bertacchini, D. Brunelli, L. Larcher, and L. Benini,
"Modeling and Optimization of a Solar Energy Harvester System for
Self-Powered Wireless Sensor Networks," IEEE Trans. Industrial
Electronics, vol. 55, no. 7, pp. 2759-2766, July 2008.
[8] S. P. Beeby, R. N. Torah, M. J. Tudor, P. Glynne-Jones, T. ODonnell,
C. R. Saha, and S. Roy, "Micro electromagnetic generator for vibration
energy harvesting," Journ. of Micromechanics and Microengineering,
vol. 17, pp. 12571265, July 2007.
[9] A. Nasiri, S.A. Zabalawi, and G. Mandic, Indoor Power Harvesting
Using Photovoltaic Cells for Low-Power Applications, IEEE Trans.
Industrial Electronics, vol. 56, Issue 11, pp. 4502-4509, July 2008.
[10] A. Khaligh, P. Zeng and C. Zheng, "Kinetic Energy Harvesting Using
Piezoelectric and Electromagnetic TechnologiesState of the Art,"
IEEE Trans.Industrial Electronics, vol. 57, no. 3, pp. 850-860, March
2010.
[11] G.K Ottman, H.F. Hofmann, A.C. Bhatt, and G.A. Lesieutre, "Adaptive
piezoelectric energy harvesting circuit for wireless remote power
supply," IEEE Trans. Power Electronics, vol. 17, pp. 669- 676, Sep
2002.
[12] B. H. Stark, P. D. Mitcheson, M. Peng, T. C. Green, E. Yeatman, and A.
S. Holmes,"Converter circuit design, semiconductor device selection and
analysis of parasitics for micropower electrostatic Generators," IEEE
Trans. Power Electronics, vol. 21, pp. 27 37, Jan 2006.
[13] C. B. Williams and R. B. Yates, "Analysis Of A Micro-electric
Generator For Microsystems," Sensors and Actuators A: Physical, Vol.
52, Issues 1-3, pp. 8-11, March-April 1996.
[14] P. D. Mitcheson, T. C. Green, E. E. Yeatman, and A. S. Holmes,
Architectures for Vibration-driven Micropower Generators, Journal of
Microelectomechanical Systems, vol. 13, pp. 429-440, June 2004.
[15] H. Minakata, H. Seki, and S. Tadakuma, "A Study of Energy-Saving
Shoes for Robot Considering Lateral Plane Motion," IEEE Trans.
Industrial Electronics, vol. 55, no. 3, pp. 1271-1276, March 2008.
[16] Jiang Bing, J.R.Smith,M. Philipose., S. Roy,; K Sundara-Rajan, A.V.
Mamishev,; Energy Scavenging for Inductively Coupled Passive RFID
Systems , IEEE Trans. Instrumentation and Measurement, Vol.
56, Issue 1, pp. 118 125, Feb. 2007.
[17] H. Kulah,H. Najafix; Energy Scavenging From Low-Frequency
Vibrations by Using Frequency Up-Conversion for Wireless Sensor
Applications. , IEEE Sensors Journal, Vol. 8, Issue 3, pp. 261 268,
March 2008.
[18] N.G. Stephen,"On energy harvesting from ambient vibration", Journal of
Sound and Vibrations, vol. 293, Issues 1-2, pp. 409-425, May 2006.
[19] M.E Kiziroglou, Cairan He, and Eric M. Yeatman, Rolling Rod
Electrostatic Microgenerator, IEEE Trans. Industrial Electronics, Vol.
56, Issue 4, pp. 1101-1108, April 2009.
[20] C.R. Saha, T. O'Donnell, H. Loder., S. Beeby, and J. Tudor;
Optimization of an Electromagnetic Energy Harvesting Device, IEEE
Trans. Magnetics, Vol. 42, Issue 10, pp. 3509 3511, Oct. 2006.
[21] Xinping Cao, Wen-Jen Chiang, Ya-Chin King, and Yi-Kuen Lee
Electromagnetic Energy Harvesting Circuit With Feedforward and
Feedback DCDC PWM Boost Converter for Vibration Power
Generator System, IEEE Trans. Power Electronics, Vol. 22, pp. 679
685, March 2007.
[22] J. Elmes, V. Gaydarzhiev, A. Mensah, K. Rustom, J. Shen, I. Batarseh,
Maximum Energy Harvesting Control for Oscillating Energy
Harvesting Systems, in Proc. 2007 IEEE Power Electronics Specialists
Conference, pp. 2792 2798.
[23] Ngok-Man Sze, Feng Su, Yat-Hei Lam, Wing-Hung Ki, and Chi-Ying
Tsui, Integrated single-inductor dual-input dual-output boost converter
for energy harvesting applications, in Proc. 2007 IEEE Power
Electronics Specialists Conference, pp. 2792 2798.
[24] T. Paing, J. Shin, R. Zane, Z. Popovic, "Resistor Emulation Approach to
Low-Power RF Energy Harvesting", IEEE Tranasactions on Power
Electronics, vol. 23, pp. 1494 1501, May 2008.
[25] L. Garbuio, M. Lallart, D. Guyomar, C. Richard, and D. Audigier,
Mechanical Energy Harvester With Ultralow Threshold Rectification
Based on SSHI Nonlinear Technique, IEEE Trans. Industrial
Electronics, Vol. 56 Issue 4, pp. 1048-1056, April 2009.
[26] J. Colomer-Farrarons, P. Miribel-Catala, A. Saiz-Vela, M. Puig-Vidal, J.
Samitier, Power-Conditioning Circuitry for a Self-Powered System
Based on Micro PZT Generators in a 0.13-m Low-Voltage Low-Power
Technology, IEEE Trans. on Industrial Electronics, Vol. 55 Issue 9, pp.
3249-3257, Sept. 2008.
[27] Guanghui Wang, Luo Cheng, H. Hofmann, and L. Rome, Power
electronic circuitry for energy harvesting backpack, in Proc. 2009 IEEE
Energy Conversion Congress and Exposition (ECCE), pp: 3544 - 3549.
[28] A. Tabesh and L. G. Frechette, "A Low-Power Stand-Alone Adaptive
Circuit for Harvesting Energy From a Piezoelectric Micropower
Generator," IEEE Trans. on Industrial Electronics, vol. 57, no. 3, pp.
840-849, March 2010.
[29] P. D. Mitcheson, T. C. Green, E. M. Yeatman, and A. S. Holmes,
Power Processing Circuits for Electromagnetic, Electrostatic and
Piezoelectric Inertial Energy Scavengers, Microsystem Technologies,
Vol. 13, pp. 16291635, May 2007.
[30] S. Dwari,R. Dayal, L. Parsa, and K. N. Salama, Efficient direct ac-to-dc
converters for vibration-based low voltage energy harvesting, in 2008
IEEE Conf. of Industrial Electronics (IECON), pp. 2320 2325.
[31] S. Dwari, L. Parsa, Efficient low voltage direct AC/DC converters for
self-powered wireless sensor nodes and mobile electronics, in Proc. Of
2008 IEEE Telecommunications Energy Conf., pp. 1-7.


Rohan Dayal (S09) received the B.Tech. degree in
Electrical Engineering from Indian Institute of
Technology, Roorkee, India in 2007. He completed
Masters of Science in Electrical Engineering from
Rensselaer Polytechnic Institute, USA in 2009. Since
then, he has been working towards his Phd. degree at
the same institute.
He did his summer internship at Electronics and
Radar Development Establishment (LRDE), India in
2006. He is currently working on vibration-based low
power energy harvesting microgenerators and converters. His research
interests include power quality, power converters, low power electronics,
electromagnetic microgenerators and energy harvesting.

Suman Dwari (S05) received the M.Tech. degree
in Electrical Engineering from the Indian Institute
of Technology, Bombay, India in 2003. He
received the Ph.D. degree in Electrical Engineering
from Rensselaer Polytechnic Institute, NY, USA in
2010. He worked in Power Electronics industry
from 2003-2005. At present, he is a Senior
Research Scientist in United Technology Research
Center, Hartford, CT, USA.
His current research interests are power electronic
systems and control, efficient power-converters, renewable energy systems,
energy harvesting, special machines, electromagnetics, and embedded control.

Leila Parsa (S00M05) received the Ph.D. degree
in Electrical Engineering from Texas A&M
University, College Station. In 2005 she joined the
department of Electrical, Computer, and Systems
Engineering at Rensselaer Polytechnic Institute,
Troy, NY as an assistant professor.
Her research interests are in design, analysis and
control of electromechanical energy converters and
power electronics converters for various
applications.
Dr. Parsa is the recipient of the 2010 RPI School of
Engineering Research Excellence Award, the 2009 Office of Naval Research
Young Investigator Award, the 2007 IEEE Industry Applications Society
Outstanding Young Member Award, and 2006 IEEE Industry Applications
Society Transactions Paper Award.