Adaptive Power Extraction from Micro Generators
with Implicit Voltage Up-Conversion
D. Maurath, M. Ortmanns and Y. Manoli
Chair of Microelectronics, Department of Microsystems Engineering (IMTEK), University of Freiburg, Germany
Abstract: This paper presents a novel approach of a generator interface and voltage up-converter for
energy scavenging micro generators. The converter will implicitly adapt to the generator impedance and
convert the generator ac-voltage to a much higher buffered dc-voltage. The applied technique uses
integrated and adaptively configured switch-capacitor arrays in order to achieve maximum efficiency
regarding the power transfer from the generator to an application load. As simulations express, this fully
integrated converter is able to up-convert ac-voltages greater than 0.3 V. Thereby, high power transfer
efficiency of 82 - 94.5 % is achieved within the entire operational range.
Key Words: generator interface, impedance matching, maximum power extraction, harvesting efficiency,
voltage converter, conversion efficiency
1. INTRODUCTION
Energy harvesting micro-generator based
applications emerge more and more. One reason
certainly is the opportunity of long-term, batteryfree and wireless operation. However, the power
delivered by those generators is commonly very
rare, fluctuant and low in voltage amplitude. In
addition to this, applications often have very
specific and repetitive operation sequences with
repeating
power
supply
requirements.
Consequently, this asynchronous generatorapplication interaction requires an energy buffer,
which stores the energy difference between the
present generator output and the desired
application energy consumption. In order to
harvest continuously the maximum possible
energy, permanent optimal generator loading for
maximum generator output power as well as
efficient rectification, voltage conversion and
energy storage is required [1].
In order to obtain this, the power electronic is
separated in two stages (see Fig. 1). The first stage
is a generator interface and has to be optimized
for adaptive impedance matching so as to load the
generator as ideal as possible. Thus, maximum
energy could be extracted permanently out of the
micro-generator. In addition to that, the first stage
also has to transfer the harvested energy most
efficiently into a buffer. Thereby, the output
voltage of the first stage is neither fixed nor
regulated by the converter. For efficient energy
transfer, the output voltage is rather adapted to the
buffer voltage Vb(t). Whereas the buffer voltage is
rms
a result of the mean generator output power Pgen
versus the mean application electronic power
rms
consumption Papp
as well as of the external buffer
capacitance Cb. With appropriate dimensioning of
the components, the buffer voltage will vary only
little and slowly compared to the generator acvoltage. Thus, the second stage is realized as a dcdc converter. This converter boosts the dc-voltage
of the buffer into a regulated output voltage for
supplying an application.
Vgen
Vout
Cb
RL
Vb
generator rectifier
interface
buffer
dc/dc
application
Figure 1: energy-harvesting system with twostage power electronic.
In this paper a novel generator interface is
presented. This interface works basically as a
voltage converter with adaptive input impedance
adaptation, efficient energy transfer and implicit
voltage up-conversion. By applying power saving
circuit techniques, the power consumption of the
control circuitry and switch driving is less then
Ptotal d 50 PW . The generator interface was
especially designed to incorporate with inductive
micro-generators [1, 2]. The internal source
351
t
C2,1
C2,j
C2,max
array 2
c
Cb
start-up bypass
c
C1,1
C1,i
C1,max
array 1
t
Figure 2: Overview of the generator interface
with impedance adapting switched capacitor
arrays and passive start-up bypass.
resistance of these inductive generators is between
Ri 1 5k : . The generated open-circuit output
voltage has 2.8 Vpp at most and frequencies are
not higher then f gen d 500 Hz [3, 4]. As
simulations express, this fully integrated converter
is able to boost ac-voltages greater than |0.3 V| to
an output dc-voltage of 2.0 V - 2.5 V. Thereby,
high harvesting efficiencies above 90 % are
possible within the entire operational range.
2. ADAPTIVE CONVERSION PRINCIPLE
The proposed adaptive switching converter, as
shown in Fig. 2, is based on two capacitor arrays,
which toggle complementary-phased with
frequency fconv between a charging and a transfer
state, Sc and St respectively. Whereas, a
conversion-cycle is the single succession of Sc and
St. In the considered converter both arrays consist
of kmax = 6 capacitors, whereas kmax directly
corresponds to the maximum required voltage
conversion ratio. However, the number of arraycapacitors ni, which are applied within a single
conversion-cycle, is dynamically adapted
according to the current required conversion ratio
ki Vb / Vgen (t ) . Hence, while one array is in
charging state Sc, ni array capacitors out of the
kmax available array capacitors are connected to
the generator in parallel. At this state, the
capacitors are charged until the capacitor voltages
Varray reach a certain voltage limit Vstop, as shown
in Fig. 3. Afterwards, the array transitions to
transfer state St. Therefore, the array capacitors
have to be re-configured so as to obtain a stacked
voltage Vstack, which is higher than the buffer
voltage Vb of an external buffer. In order to obtain
the dynamic adaptation of the optimal conversion
352
Vstop
2.5
Vb
Varray
2.0
voltage [V]
rectifier
1.5
1.0
0.5
46.5
46.8
47.1
47.4
time
47.7 [µs]
Figure 3: This graph show one conversion cycle,
whereas Varray is the current highest voltage
within an array.
ratio ki, the converter works in an oversampling
manner. Thus, if f gen f conv is given, the
conversion ratio can be adapted quite
continuously within a generator voltage period
1/ f gen . Thereby, oversampling means that the acvoltage of the generator is divided into short time
slots (conversion cycles) whereas ki is adapted for
each time slot.
2.1 Adaptive Impedance Matching
Ideal generator loading and impedance
matching is especially important, because of
commonly high internal generator resistance Ri.
This high resistance requires a particular converter
design since an exceeding current load would
cause a dramatic voltage breakdown, which would
cause reduced generator output power.
Due to the low output currents of energy
harvesting generators (50 – 500 P A ) capacitive
voltage conversion is preferred rather than
inductive conversion. However, charging a
capacitor allows generally only at topt | W ln 2
maximum power transfer from the generator into
the capacitor. The generator-interface matching
situation at topt is equivalent to the impedance
matching situation where Pgen(Ri = Rload) = Pmax is
achieved. Therefore, the converter’s capacitors
must not be fully charged or discharged. Instead,
the capacitors charging level should stay around a
optimal voltage of Vch ,opt | 0.5 Vgen (t ) , which
corresponds to topt. This means, within Sc the
generator charges the capacitors up to
Vstop Vch,opt 'Vch . Then the capacitors are
toggled to St and discharged onto an output buffer
Cb until the capacitor voltage reaches
Vch ,opt 'Vdisch . Whereas, the smaller the overdrive
connected in parallel (Fig. 4). Thus, the maximum
capacitance for a certain conversion ratio is used,
which reduces the conversion frequency fconv.
voltages 'Vch and 'Vdisch are, the higher is the
conversion frequency fconv. However, impedance
matching is improved and generator output power
is increased. The consequence of keeping the caps
only in a small charging range of 0.5 Vgen (t ) r 'V
requires an increased number of capacitors. Thus,
twice as many capacitors are implemented to
boost the generator voltage up to a certain voltage.
All simulations were obtained with Spectre
Cadence circuit simulator with a standard
0.35 P m n-well double-poly CMOS process. The
generator was included as a sine wave source with
an ohmic serial resistance. Because energy
harvesting applications are considered, there are
two important efficiencies:
2.2 Charge Transfer and Array Configuration
As is generally known, charge transfer from
one capacitor to another one causes intrinsic
losses due to series resistances [2]. But by
minimizing the initial voltage gap between Vstack
and Vb the charge transfer losses are minimized.
Hence, the optimal conversion ratio ki is obtained
by
Vb
Vb
ki o
Vch ,opt 'Vch Vstop
(1)
The strategy for stacking is that a maximum
quantity ni o kmax of capacitors is used, whereas
within each stack-stage ni / ki capacitors are
3 SIMULATION RESULTS
Kin
Pgen
Pmax
Kconv
Pout
Pin
³
t2
Vgen I gen dt
1
4 Ri
Vˆgen
t1
³ VI
³ V I
Papp
t2
t1
Pgen
t2
b app
gen
dt
gen dt
(2)
(3)
Where (2) defines the input efficiency Kin ,
which compares the theoretical maximum
generator power to the actual converter input
power. The higher Kin is, the better achieved is
impedance matching. In addition to this, the
efficiency Kconv in (3) gives a measure of the
conversion efficiency as usually applied for
voltage converters. Therefore, this efficiency is
calculated by comparison of Pin versus Pout of the
interface. Here, the conversion efficiency is
mainly affected by internal converter losses and
capacitor charge transfer losses.
As Fig. 5 illustrates, high efficiencies are
achieved over the entire input voltage range.
Moreover, Fig. 6 exhibits that the interface input
power almost equals the maximum possible
output power. Thereby, the power and current Imax
changes quadratically with the currently supplied
generator voltage Vgen(t):
I max (t )
Figure 4: Schematic and state diagram of a
conversion cycle. By the transition to state St all
capacitors are stacked in ki = 2stages with
ni/ ki = 3 parallel capacitors at each stage.
1
t2 t1
t1
Vgen (t )
2
4 RV
i b (t )
(4)
Both graphs prove that this novel interface
approach and conversion method operates as
required. Generally, in order to characterize the
efficiency by interfacing a generator, not only
353
Kconv can be considered, but also Kin is of
importance. Therefore, Fig. 7 shows the total
efficiency over the generator voltage range.
1.0
PN
Pmax
Pin
0.8
0.6
Pout
0.4
0.2
0.0
0.3
0.5
0.7
0.9
1.1
Vgen
[V]
harvesting efficiency [%]
Figure 5: This graph illustrates high input and
conversion efficiency due to adaptive impedance
matching within the first 90° of a sinusoid.
93
2µA
Iopt
Imax
92
ACKNOWLEDGMENT
This work is supported by the German Research
Foundation (Deutsche Forschungsgemeinschaft DFG) under Grant Number GR1322.
REFERENCES
90
[1] Xu Shengwen, K.D.T. Ngo, T. Nishida, Ch.
Gyo-Bum, A. Sharma, Converter and
controller for micro-power energy harvesting,
20th Annual Applied Power Electronics
Conference and Exposition, vol. 1, pp. 226230, March 2005.
[2] L. Gobbi, A. Cabrini, and G. Torelli, Impact of
Parasitic Elements on CMOS Charge Pumps:
a Numerical Analysis, IEEE International
Symposium on Circuits and Systems (ISCAS),
May 2006
[3] D. Spreemann, B. Folkmer, D. Maurath and
Y. Manoli, Tuneable transducer for low
frequency vibrational energy harvesting,
EuroSensors 2006, vol. , September 2006
[4] S. Kulkarni, E. Koukharenko, J. Tudor, S.
Beeby, T. O’Donnell, S. Roy, Fabrication and
Test of Integrated Micro-Scale Vibration
Based
Electromagnetic
Generator,
International Solid-State Sensors Actuators
and Microsystems Conference, 2007
89
88
87
Vgen
0.6
0.8
1.0
1.2
1.4
1.6 [V]
Figure 6: The interface input power almost equals
the maximum possible output power.
total efficiency [%]
It was shown, that high efficient power
extraction and voltage up-conversion is feasible in
light-load energy harvesting environments.
Especially the realization of adaptive impedance
matching and almost continuous maximum
generator power enhances the usage of microgenerators further. These advantages give the
opportunity of using smaller generators to supply
autonomous applications with un-changed
performance. One drawback might be the
relatively large quantity of capacitances, which
need to be integrated. Thus, a lot of chip area is
necessary and parasitic effects could decrease the
overall efficiency.
91
0.4
2 µA
Imax
95
85
Vgen
75
0.5
0.8
1.1
1.4
1.7
2.0 [V]
Figure 7: This graphs show the total efficiency
Kin uKconv for standby-load and full load.
354
4 CONCLUSIONS