Broadband Rectenna Arrays for Randomly

Polarized Incident Waves

Joseph A. Hagerty, Néstor D. López, Branko Popović, and Zoya Popović
[email protected]
Department of Electrical and Computer Engineering, University of Colorado
Boulder, CO 80309-0425 USA

Abstract— This paper presents a new approach to efficient

rectenna arrays for arbitrarily polarized incident waves with
broad spectral content. The approach is validated experi-
mentally on a dense grid array that rectifies two orthogo-
nal linear polarizations, and on a self-similar spiral array
with alternating right-hand and left-hand circular polariza-
tions. The two arrays operate from 4.5 to 8 GHz and 8.5 to
15 GHz and have maximum open circuit voltages of 3.5 and
4.0 V,respectively. Their efficiencies increase above 35 % and
45 % ,respectively, for higher incident powers.

I. Introduction
Microwave rectennas – active antennas containing recti-
fication devices – have been investigated for power trans-
mission and detection over the past half century [1]. Appli- (a)
cations have included long distance power beaming [2,3,4],
signal detection [5] and wireless control systems [6]. In all of
these cases, the polarization, CW frequency, and power of
the incoming RF field were not time varying and were well
defined and known a priori. In this paper, we explore a new
application of RF rectennas: recycling of unused RF energy
in areas where RF radiated power densities are relatively
high. For example, the rooftop of the building at 1801
California Street in Denver, Colorado, houses a large vari-
ety of transmitting antennas for applications ranging from
police communications (at several hundred MHz), cellular
(900 MHz) and PCS (around 2 GHz) telephony, two-way
microwave radio communications, and up to millimeter-
wave satellite communications. A variety of radiated power
levels and polarizations are present in this environment,
and interference between the antennas, as well as health
safety of operating personnel due to RF power densities (b)
exceeding FCC regulations are existing problems [7]. In
addition, in such multi-path environments, the wave po- Fig. 1. The grid (a) and spiral (b) arrays. The latter is arrayed as
larization changes as the waves propagate. For example, alternating RHCP and LHCP elements, while the former rectifies one
half cycle of both polarizations equally.
for a vertically polarized transmitted wave, the waves after
propagation contain both polarizations at roughly the same
power level and with uncorrelated phases [8]. To mittigate
this problem, we investigate the prospect of efficiently cap- incident waves over an extended frequency band and with
turing power contained in fields with unknown and arbi- arbitrary elliptical polarization.
trary time varying spectral distribution and polarization. The following issues are relevant to the work presented
To this end, two rectenna arrays, a grid and a spiral array in this paper:
shown in Fig.1, have been designed and characterized with
1. In rectennas presented to date, a CW wave at a sin-
This work was funded in part by an NSF SPNC grant NCR9725778, gle frequency is incident on a resonant antenna, fol-
and by ITN Energy Systems, Inc. through ARO and DARPA con-
tract DAAG55-98-C-0037. The content of this information does not lowed by a matching circuit that helps deliver the re-
necessarily reflect the position of the United States Government. ceived power to a rectifying element, typically a Schot-
 tky diode. However, resonant antennas followed by field vector with respect to the page are labeled as U (up),
matching circuits are narrowband (at most 15% frac- D (down), R (right) and L (left). The rectified DC cur-
tional bandwidth), and arrays of these circuits take rent is directed through the horizontal and vertical lines
up considerable real estate which reduces overall aper- and collected at the diagonal ends of the array (Fig.1a).
ture efficiency, and therefore conversion efficiency. In Assume first a vertically polarized incident wave and all
the rectenna arrays presented here, resonant antennas the diodes in the horizontal lines (D11, D22 and D33) to
and matching circuits are absent, and a large number be unbiased or forward biased. During the half period in
of diodes load a nonresonant radiating array. which the electric field vector is in the U direction, diode
2. The rectification process is more efficient when the D12 conducts, and there is a current in leads 11’. In the
rectifier diode is biased to an operating point of high next half-period, the electric field is in the D direction, so
IV curve nonlinearity. When a large number of diodes that charges stored in strip 1 are discharged through diode
are connected in a combined series-parallel DC circuit, D11 into the DC line, and strip 1’ draws current from the
they provide self biasing, thereby increasing the con- horizontal line 3. The same happens to radiators 2 and 2’,
version efficiency. so that the two DC generators corresponding to antennas
3. In previously demonstrated rectennas, the polariza- composed of strips 1-1’ and 2-2’ are connected in series.
tion of the incident wave is well defined and lin- Consider now the horizontal conductor 3, when the diodes
ear in most cases (except in [5] where it is right in the vertical strips are unbiased or forward biased. If
hand circular). This enables polarization-matching the horizontally polarized component of the electric field
of the rectenna for maximized efficiency. Our pro- is in the R direction, diodes D11, D22 and D33 prevent
posed applications, however, involve generalized el- the induced current from flowing. In the next half-period,
liptical waves with arbitrary and time-varying polar- when the electric field vector is L-directed, a current will
ization. We accomplish efficient rectification of such flow through these diodes into the horozontal “DC collec-
waves by independently rectifying two orthogonal po- tion” line, simultaneously charging to some extent strips
larizations (either linear vertical and horizontal, or 1 and 1’. The entire structure receives both horizontally
RHCP and LHCP), and adding the rectified DC volt- and vertically polarized waves and adds the electromotive
ages and/or currents. forces after rectification. Note that the capacitance of the
antenna array has the effect of smoothing the rectified in-
II. Generalized Rectenna Design duced emf.
A. Physical Principles of Rectenna Operation
B. Design of Grid and Spiral Rectenna Arrays
The grid rectenna array (Fig.1a) was designed based on
the unit cell described above. The diodes in the grid are
all aligned in the horizontal and vertical leads, and there-
fore the grid rectenna acts as a half-wave rectifier. The
broad bandwidth results from the electrically-small period
of the grid (roughly a tenth of a wavelength at the cen-
ter of the band), as well as the non-resonant nature of the
metal pattern. For the measurements presented in this pa-
per, a 6 by 6 grid was fabricated on a Duroid substrate
(εr = 2.2, 0.508-mm thick) with a period of 4.5 mm and a
strip width of 0.5 mm. Eighty-four MA4E2054A M/ACom
low-cost Schottky diodes were used to populate the mesh.
The rectenna delivers DC power to a load between 4.5 and
Fig. 2. Schematic of a portion of the grid rectenna array used in the 8.0 GHz. The two diagonal corners are the DC collection
explanation of the physical principles of operation. leads, and the diode orientation results in a net current
from bottom right to top left corner of the grid. The recti-
An elliptically-polarized wave can be represented as two fied current and voltage is therefore a combination of series
orthogonal linearly polarized waves of appropriate ampli- and parallel connections of diodes. The grid rectenna array
tudes and phases. Therefore, there is an average 3-dB loss is roughly 0.5λ20 wide at the center of the band, and uses
associated with a single-feed rectenna. Separate rectifiers about 160 diodes per squared wavelength.
for each of the two orthogonal polarizations must be used to The spiral array shown in Fig.1b includes broadband
avoid effects of distructive phasor addition. A diode-loaded loaded elements with both orthogonal circular polariza-
antenna has nonlinear performance and does not conform tions (RHCP and LHCP). The antennas are self-similar
to standard linear antenna theory; therefore, a qualitative spirals with a single diode in the feed. The DC collection
analysis becomes useful. Fig.2 illustrates the physical prin- points are again at the diagonal corners, with the diodes
ciples of operation in the context of a unit cell of the grid connected in a hybrid series-parallel circuit. In this array,
rectenna array. The reference directions for the electric the antennas are approximately half of a free-space wave-
length at the largest diameter at the center of the operating In addition, the two arrays are compared to a linear 8-
band (from 8.5 to 15.0 GHz). In contrast to the very dense element dipole rectenna array with diodes connected in se-
grid rectenna arrays, the spiral rectenna array is 5λ20 in ries DC-wise. For these measurements, the incident field
area at the center of the band, with 3 diodes per square was circularly polarized by using two orthogonally polar-
wavelength. ized linear horn antennas in quadrature. For all three ar-
rays, the same incident power was used. The measure-
C. Figures of Merit for Rectenna Performance ments show that the grid and spiral areas indeed deliver
The most relevant figure of merit for rectenna arrays DC power to a load over a broad fractional bandwidth,
is the efficiency of the RF to DC conversion, given the while the dipole array is resonant, as expected.
physical area of the array and the incident power density.
B. Performance for Complex Incident Waves
Other significant figures of merit for the rectenna arrays
are: (1) frequency response; (2) DC power (efficiency) as Since we have designed the arrays to operate over a wide
a function of input RF power; (3) DC power (efficiency) range of incident frequencies and polarizations, it is of
as a function of input wave polarization; (4) DC power interest to measure the nonlinear response of the recten-
(efficiency) as a function of input wave spectral content; nas. As the rectification process is non-linear, superposi-
and (5) optimal DC load impedance for maximal power tion does not hold.
transfer. To quantify the produced DC power when waves at dif-
ferent frequencies are simultaneously incident on the ar-
III. Experimental Results rays, three measurements were performed using the spiral
array. The results with low incident power at 12.2 GHz and
varying power at 8.5 GHz are shown in Fig. 4. One would
expect that the power contained in one frequency will upon
rectification provide bias and therefore improve rectifica-
tion for a wave at another frequency. As can be seen in
Fig.4, this process is very nonlinear. Namely, when only
a vertically-polarized wave at 8.5 GHz is incident with a
power P , the received normalized DC power was measured
as shown with circular symbols in the figure. When a 12.2-
GHz horizontally-polarized wave of power (PdB − 17 dB)
was turned on, the increase in DC power measured in
the load was 5 dB (or, a factor of 6). As the power of
the 8.4 GHz signal is increased, the DC power follows this
trend, but not linearly.

Fig. 3. Measured frequency response of normalized DC power deliv-

ered to an optimal load for the grid rectenna array (square symbols),
spiral array (circular symbols) and linear 8-element resonant dipole
array (triangular symbols). The response is measured with an inci-
dent circularly polarized wave at boresight.

The measurements were performed with the rectenna ar-

rays in the far field of one or several transmitting antennas.
The power incident on the rectenna arrays was calculated
from the Friis transmission formula, using the geometrical
area of the arrays. Since the grid and the spiral arrays are
different in electrical size and operating frequency range,
the data were normalized w.r.t. the quantity of interest
to compare their performance. Below we present experi-
mental data for the frequency, polarization, and power re- Fig. 4. Measured normalized DC power for a vertically polarized in-
sponses of the grid and spiral arrays. cident wave at 8.5 GHz (circular symbols), at 12.2 GHz ( square sym-
bols), and for waves at both frequencies simultaneously (diamond
A. Frequency Response for CP Incident Wave symbols). In the latter case the DC measured power is increased by
3 to 5 dB over the simple sum of the first two cases.
The normalized frequency response is given in Fig.3.
Both the DC power delivered to an optimal load and the Another measurement was performed to investigate the
frequency of operation are normalized for fair comparison. advantageous nonlinear properties of the rectennas, and is
 Table 1: Nonlinear Power Combining
efficiency was calculated from measurements to be 35 %
at 5.7 GHz and for 7.78 mW/cm2 . The maximal open-
Horizontal Vertical Independent Dual circuit voltage of the spiral rectenna array was 4.0 V at
Pol. Pol. Sum Pol. 10.7 GHz, and the maximal conversion efficiency was cal-
culated from measurements to be 45% at 10.7 GHz and for
-12.1 dBm -8.32 dBm -3.97 dBm 0.0 dBm
1.56 mW/cm2 .

IV. Discussion
summarized in Table 1, demonstrated on the spiral array. This paper presents an experimental study of wideband
First a vertically-polarized wave of power P was incident rectenna arrays for arbitrarilly polarized input waves. The
on the rectenna and the rectified power measured in an study is based on physical principles. A theoretical (numer-
optimal load. Then the rectified DC power resulting from ical) investigation needs to combine sophisticated electro-
a horizontally-polarized wave of equal power P was mea- magnetic modelling of the passive antenna structure with
sured, and subsequently waves of both polarization with to- nonlinear circuit modelling of the diodes, and is in progress.
tal power 2P were incident on the two arrays. The relative Some remaining issues related to the rectenna designs that
phase between the two waves was varied in order to achieve we are also working on are: coverage of a very large fre-
different elliptical polarizations for the total incident wave. quency range, proper DC connections for very large arrays,
The DC power produced in this case is on average (aver- conformal low-cost (flexible) rectenna arrays, and reliabil-
aged over all relative phases) larger than the summ of the ity and degradation.
DC powers measured when only one linear polarization is The grid rectifiers have excellent reliability and graceful
present. This again is because the arrays are designed to degradation. The limiting factor on the size of the grid
rectify both polarizations, and power in one provides some rectifier is the current rating on the diodes. Note that
biasing of the diodes, resulting in a nonlinear increase, as in the grid in Fig.1a the four corner diodes are the most
expected from the diode nonlinear IV curve. critical ones, because one half of the current through the
DC terminals passes through each of the diodes. If these 4
C. Rectenna Array Optimal Load diodes are replaced by shorts, the current in the DC leads
As the incident RF power is increased, the efficiency in- can be twice as large. The current intensity in the next
creases, as expected. The optimal load for maximum deliv- diodes closest to the terminals is half of that through the
ered power decreases as a result of the fact that the deriva- four most critical diodes. Any overloaded diodes in the grid
tive of the IV curve of the diode increases as the diode are automatically eliminated if the fail as shorts, and the
is biased by rectified RF power. Since the optimal load is rest of the grid will function. If a diode fails as an open,
given as the ratio V /I, Fig.5 shows the decrease in mea- the current will find a path through the other diodes/shorts
sured load impedance as a function of normalized incident across the grid.
power.
V. References
[1] W.C. Brown, “The History of Power Transmission by Radio
Waves,” IEEE Trans. Microwave Theory and Techn., Vol.32, No.9,
pp.1230–1242, September 1984.
[2] N. Shinohara, H. Matsumoto, “Experimental Study of Large
Rectenna Array for Microwave Energy Transmission,” IEEE Trans.
Microwave Theory and Techn., Vol. 46, No.3, pp.261–267, March
1998.
[3] S.S.Bharj, R. Camisa, S. Grober, F. Wosniak, E. Pendleton,
“High-Efficiency C-band 1000-Element Rectenna Arrayfor Microwave
Powered Applications,” IEEE International Microwave Symposium
Digest, pp.301–303, June 1992.
[4] J.O. McSpadden, l. Fan, K. Chang, “A High Conversion Efficiency
5.8-GHz Rectenna,” IEEE International Microwave Symposium Di-
gest, pp.547–550, June 1982.
[5] R.H. Rasshofer, M.o. Thieme, E.M. Biebl, “Circularly Polarized
Millimeter-Wave Rectenna on Silicon Substrate,” IEEE Trans. Mi-
crowave Theory and Techn., Vol. 46, No.5, pp.715–718, May 1998.
[6] L.W. Epp, A.R. Khan, H.K. Smith, R.P Smith, “A Compact
Dual-polarized 8.51-GHz Rectenna for High-Voltage (50 V) Actuator
Applications,” IEEE Trans. Microwave Theory and Techn., Vol. 48,
No.1, pp.111–120, January 2000.
[7] Private communication, Amy Barnes Frey, Video Accessory Cor-
poration, Boulder, Colorado.
Fig. 5. Measured optimal load for maximum DC power transfer as a [8] Microwave Mobile Communications,] W.C. Jakes, Ed., IEEE
function of incident power for both the grid and the spiral rectenna Press, 1994, pp.125–158.
arrays.

The maximal open-circuit voltage of the grid rectenna

array was 3.5 V at 5.7 GHz, and the maximal conversion