Dual-polarization, wideband microstrip antenna array for

the bandwidth requirement are inversely proportional: where B is the bandwidth of the transmitted pulse and c is the speeb of Ii ht The resent EMISAR ...

0 downloads 77 Views 570KB Size
Downloaded from orbit.dtu.dk on: Jan 11, 2020

Dual-polarization, wideband microstrip antenna array for airborne C-band SAR

Granholm, Johan; Skou, Niels Published in: Proceedings of Phased Array Systems and Technology Link to article, DOI: 10.1109/PAST.2000.858949 Publication date: 2000 Document Version Publisher's PDF, also known as Version of record Link back to DTU Orbit

Citation (APA): Granholm, J., & Skou, N. (2000). Dual-polarization, wideband microstrip antenna array for airborne C-band SAR. In Proceedings of Phased Array Systems and Technology (pp. 243-246). Dana Point, CA, USA: IEEE. https://doi.org/10.1109/PAST.2000.858949

General rights Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.  Users may download and print one copy of any publication from the public portal for the purpose of private study or research.  You may not further distribute the material or use it for any profit-making activity or commercial gain  You may freely distribute the URL identifying the publication in the public portal If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

DUAL-POLARIZATION,WIDEBAND MICROSTRIP ANTENNA ARRAY FOR AIRBORNE C-BAND SAR Johan Granholm & Niels Skou Danish Center for Remote Sensing, Department of ElectromagneticSystems,Technical University of Denmark, Building 348,DK-2800 Lyngby, Denmark ([email protected] .dtu .dk, nsDemi .dtu.dk)

Abstract The paper describes the development of a C-band, duallinear polarization wideband antenna array, for use in the next-generation of the Danish airbome polarimetric synthetic aperture radar (SAR) system. The array is made of probe-fed, stacked microstri atches. The design and performance of the basic stactel p t c h element, operating rom 4.9 GHz to 5.7 GHz, and a x2 element test-array of these, are described.

Introduction High-resolution airbome and spacebome imaging of the Earth is often carried out using remote sensing techniques, such as SAR. Early SAR systems were single-polarization instruments, but time has seen a rowing interest in dualpolarization (i.e. “polarimetric”) 8AR systems. The reason or this trend is the additional amount of geophysical information, which it is possible to extract from polarimetric SAR data, com ared to single- olarization data. It is well known, that ra&r signatures 0Pe.g. crops are polarization dependent. An intuitive physical explanation of this de endence is, that the vertical polarization primarily are relected by the vertical structures (e.g. straws, trunks), whereas the honzontal polanzation are in stead reflected by the predominantly horizontal structures (e.g. branches). Mapping an area with polarimetnc SAR thus provides more information, hence allows more d e t a h to be distinguished and increases the ability to classify targets. Future SAR systems for e.g. crop study and monitoring are therefore required to be polarimetric instruments. Several space agencies and other institutions are currently develo ing such next-generation polanmetnc SAR systems, [I] - [4f

Resolution requirements for future SAR systems Several polarimetric SAR systems have already been built and flown, incl. the Danish L- and C-band system, EMISAR, develo ed at Department of Electroma netic S stems, Technic3 University of Denmark (EMI), [5f The EhISAR system has a resolution of approx. 2 m in both range and elevation. At the time of design (1993) this was state-of-the-art, but not least due to the rapidly advancing digital technology (es ecially hi h speed data acquisition and storage systems), {igher resofuions are possible today. At the same time user demands, e.g. for surve in applications, continue to call for increased resolution. J o t 1 for scientific and commercial map in urposes, there are a ~ m. desire to achieve resolution of 0 . 2 9 8.f5 The maximum SAR azimuth resolution is not de endent on the bandwidth of the transmitted radar chirp, gut on the pulse-repetition frequenc and to a first ap roximation to the physical antenna yength, 1,. The {AR azimuthis limited to: resolution, BA

= 4/2 This (frequency inde endent) result shows, that to obtain an azimuth resolution ofO.25 m the antenna should be max. 0.5 m Ion . Lower frequency results in less ain with this limitef antenna length. High-resolution SAR systems therefore are difficult at lower microwave frequencies.

a,,and

The maximum range-resolution in SAR systems, the bandwidth requirement are inversely proportional:

where B is the bandwidth of the transmitted pulse and c is the speeb of Ii ht The resent EMISAR s stem has 100 MHz bandwidta in both e-band and C-banJ translating to

0-7803-6345-0/00/$10.00 0 2000 IEEE

a+prox. 2 m resolution in range, including proper weighting. o obtain 0.25 m resolution in range, the pulsed chirp must therefore have an eightfold increase in bandwidth. Although the range resolution of both the L-band SAR (with centre fre uency 1.25 GHz) and the C-band SAR (centered at 5.3 GI-? z) is desired to be increased in the next-generation EMISAR s stem, technology will not allow for this in Lband. In C-{and, however, it may be possible to achieve up to 800 MHz bandwidth. The goal therefore is to up rade the, today medium resolution, C-band EMISAR to a fighresolution system, capable of 800 MHz operation. To be compatible with existin and planned civilian and scientific air- and spacebome SAS’s (to, allow for data comparison), the centre frequency shall remain 5.3 GHz.

Requirements for polarimetric SAR antenna arrays Future polarimetric SAR-systems require compact and lightweight dual-linear polarization antenna arrays, which should preferably be flat to facilitate easy installation; e.g. conformal on the fuselage of aircrafts. The arrays’ elevation plane radiation pattems shall be shaped to resemble a modified cosecant-squared shape (to compensate for the range-dependence), while the azimuth Pane radiation pattems shall be narrow, with a moderate y low sidelobe evel, and symmetrical w.r.t. boresight. In order to obtain a hi h degree of polarization discrimination of the overall SWR system, it is a further requirement to the array, that it should have a low cross-polarization level. An antenna element, which com lies with these requirements, is the microstrip patch. Algough it is very narrowbanded as a single-layer structure, a stacked patch can be added to significantly increase the bandwidth. A erture coupled microstrip patches, although capable of oifering wide bandwidth, has the drawback that, in a ractical antenna, a closed cavity is required to be placed gehind the aperture. Such cavities may lead to a reduction of the bandwidth of the “naked” element but, worse, the cavity will take up valuable board s ace undemeath the aperture. Due to this “real-estate’’ probyem, this cavit may revent, or at least severely complicate, the (alread; chalknging) design and la out of the necessaril multilayer beamforming network (BFh), which must resi& under the groundplane, below the aperture. Furthermore, a ertures are known to excite surface waves far stronger, &an probe-fed patches. For these reasons aperture-coupled patches is not the optimum choice in this application. Probe-fed patches are attractive from the a feeding network point of view, since the probes does not take up any board space, and easily connect the patch to its beamforming network, which is e.g. located several layers down from the patch groundplane. For this reasons probe-fed patches are chosen. The existin C band EMISAR system [6],operating over an ap rox. 2 98 bandwidth, uses a sin le, probe-fed patch. The L-&andEMISAR s stem [7], whica is also operating over a 100 MHz bandwidlth (translating to 8 %), uses a stacked probe-fed patch. The idea thus was to investigate, if the stacked patch concept could be adopted and optimized to operate over an 800 MHz bandwidth in C-band.

Selection of dielectric substrate The bandwidth of microstrip patches depends strongly on the substrate thickness and -permittivity. For practical reasons, C-band patch probes shall be fabricated as “integrated via’s’’. Although electrical attractive, this mechanical requirement preclude the use of low loss, low permittivity foam matenals, since via’s can neither be rown, nor supported, by such soft materials. Hence, for fandwidth reasons, it is desirably to use a thick, low

243

Authorized licensed use limited to: Danmarks Tekniske Informationscenter. Downloaded on February 16,2010 at 07:16:00 EST from IEEE Xplore. Restrictions apply.

permittivity hard substrate. On the other hand, the maximum thickness is limited for several reasons: A thick substrate may support the (undesirable) propa ation of excessive surface waves, and will also be heavy.%urthermore, it will necessarily imply a lon feeding probe, which electrically act as a large series-inkctance, thus reducing the element bandwidth. Due to fabrication and reliability issues, the multi-layer stri line BFN and the radiating patch layer should referably be fabricated using the same t pe of substrate. h e wide bandwidth requirement will akmand that the elements must be fed through a binary-type network in the azimuth direction. In the elevation direction, it is desired to sha e the elevation pattem, so that it will resemble a mosified cosecant-squared pattem. This elevation BFN will be implemented using couplers and lines. It is desirable to be able to vary the characteristic impedance of the striplines over a wide range (w.r.t. 50 Q), within acceptable linewidths (e.g. 5 mm to 0.5 mm) to facilitate the design of these cou lers, lines and necessa impedance matching networks.his requirement also c 8 s for the use of a low permittivity substrate, although its thickness can be chosen independently of the patch substrate thickness. The desi n of the BFN, however, is a separate task, and will not %e covered in this paper. In this work, 32 mil (0.8128 mm) Rogers R04003 substrate was initially chosen for the driven patch, having a dielectric constant &=3.38 and loss tangent tan6=0.0027 @ 10 GHz. The R04603 material is constructed as woven lass cloth, impre nated with a ceramic loaded thermoset pfastic resin to yiefd a thermal1 stable rigid laminate with electrical microwave, frequencies Although the P p e r t i e s suitable, 04003 permittivity and dielectnc loss factor is somewhat hi her than PTFE-based materials (e. . Rogers RT/duroid 5850 having &=2.22 and tan6=0.0009 10 GHz), R04003 is attractive in' an antenna feasibility study like the present, since R04003 resembles FR4 in mechanical integrity and can be fabricated like basic FR4 material (and also because of its much lower price). Hence prototype development of stripline PCB's with integrated, via's is .much faster and cheaper if using R04003 matenals, than if using a PTFEmaterial.

6r

8

There is on1 one resonance in the band shown (approx. at 5.3 GHz). T i e impedance for the t w o ports is identical. An interesting experimental observation is, that the resonance frequency occurs at the same fre UInc where the two ports are best de-coupled (i.e. at zequency where the scattering parameter S, exhibits its minimum). At this frequency it can theretore be expected, that the crosspolarization level of the patch will be best, because S is minimum. When designing microstrip patches for ddalpolarization purposes, it is important to obtain a low level of the element S,,, since the "cross-coupled" power will directly affect the cross-polarization level. The impedance match of the single-layer patch to 50 Q ( i s . the scattering parameter S ) is quite oor, but this is a direct consequence of the patci'being f e z a t the edge. Here the input impedance is on the order of 300 R. If a single-layer patch is desired to have a good intrinsic match to a lower impedance level (e, . 50 Q), this can easily be accomplished simply by moving t8e feed points )towardsthe patch center. The measured radiation pattem of the single-layer robe-fed microstrip patch is shown i.n Fig. 3. A I f r t t e m measurements presented in this paper have been pe ormed in the spherical near-field test facility of EMI, [8]. II.pmlMYmUUIEYt141.1

\'-a" ,d.m"Lbw,+o->

Design of the C-band wideband element

The startin point in the design of a stacked atch is the design of tfe driven (i.e. single-layer) patch, slown'in Fig. 1. It is designed so that its resonance frequency is in the centre of the band, i.e. 5.3 GHz.

*m I

'

I

'

..-c.-

L L = 14.5 m m D = 0.3 mm a)

b)

Fig.1. Dual polarization probe-fed microstrip single patch; a) Top view of the patch, b) cross section.

The patch is fed close to its edges usinf 0:6 mm diameter robes (in the laboratory patch, the ro e is direct1 made from the center conductor of a 8MA-connectorr Two probes (designated "H" and 'fv"),located on the patch' center lines, are used to excite the patch in orthogonal modes. This orthogonality will have the effect, that the coupling between the H and V-ports (i.e. the scatterin parameter S ) remains fairly low in the vicinity of the patcl resonance fiAquencies ("resonance frequency" is defined as the frequency at which the real part of the patchs' input impedance achieves its maximum value, when the reference plane is at the upper side of the ground plane). The measured input impedance of the dnven patch alone is shown in Fig. 2.

Fig. 3. Radiation pattern of single (unstacked) microstrip patch (f = 5.3 GHz).

The co-polarized fields in the E-planes (i.e. H-port fed, azimuth cut and V-port fed, elevation cut) show the ty ical rippled behaviour due to diffraction at the edges of the {nite sized (0.5 m x 0.5 m) ground lane The co-polarized fields in the patch's H-plane, on tKe other hand, have a much smoother appearance. The crosss-polarization level i n the main beam is seen to be approximate1 of the (same magnitude as the level of the transmission tlrough the patch (approx. -30dB at 5.3,,GHz which is almost the same as the "intrinsic element-S,, , i.e. S?, measured, but corrected for mismatch loss). The peak diredvity is approx. 6 dBi. As shown in Fig. 2 a sinfle-layer patch can far from achieve the 800 MHz andwidt required, and a stacked patch is therefore added. This configuration is shown in Fig. 4.

244

Authorized licensed use limited to: Danmarks Tekniske Informationscenter. Downloaded on February 16,2010 at 07:16:00 EST from IEEE Xplore. Restrictions apply.

loss in a 40 R system will be lower than the loss in a 50 R system). If reducing the system impedance to 40 R and implementing a simple microstrip matching network at the patch feeding oint, this has been found to brin S of the patch down bekw -20 dB, over the 800 MHz banfiwidth.

L = 14.5 mm D = 0.3 mm b)

a)

Fig. 4. Dual polarization probe-fed microstrip stacked patch; a) Top view of the lower patch, b) cross section.

The driven patch is identical to the one shown in Fig. 1. The upper patch acts as a passive parasitic element. Although the im edance of the single (i.e. unstacked) driven patch, when f e l o n its edge, is very high (see Fig. 2), it is significantly reduced when bein loaded by the stacked patch. Hence, to obtain an a prox. f0 Ohm input impedance of the stacked patch, the &ven patch is fed on the edges. The dual linear polarization requirement implies, that the stacked microstrip patch must be constructed symmetrically. A pair of quadratic and co-axially aligned patches is therefore used. Since no direct synthesis method for the design of wideband stacked microstrip patches is known, an iterative numerical design process, using an electromagnetic simulator, was ado ted. The sizes of and the distance between the two patcPhes were varied, until the desired wideband performance was obtained. During this process it was found, that a good starting point was to design the driven patch first, having its resonance frequenc lyin in the centre of the band of interest. The stackec? patck. adds a second resonance, and the task is now to find the parameters so that the two resonances balance in value (i.e. are excited e ually strong), and are spaced in fre uency so, that the impeaance remans constant over the ban% Using this "numerical iterative design", combined with impedance measurements to validate the approach, a stacked C-band microstri patch yielding an near-optimum result was found, when t i e size of the uadratic lower and U per patches were 14.5 mm and 19.9 mm, respectively. &e upper patch is etched on a 0.1 mm FR4 substrate, and mounted inverted to let the FR4 matenal act as environmental protection. The s acing between the patches is 4.5 mm, using a Rohacell 31 h F foam material having E of 1.05. All layers are lued to ether usin a 0 1 mm thick typk 668. The measured adhesive film (supplie8 by scattering parameters for this element are shown in Fig. 5.

3d;

R r l m IOU and mmrmlrion Ihrmgh patch

11

26

I8

70

5:

<>

I6

Fmqucncy. IGHzJ

I"

LO

6:

R u t - and amaginar). pm uf pitch lmpdanco

11

11

i t

to '? I4 (6 Requcncy ICHzl

58

69

6:

Fig. 5 Measured S-parameters and input impedance of wideband, stacked C-band microstrip patch.

A wideband behaviour for the stacked patch is indeed observed. Note, that the two "peaks" of the real-part input impedance balance quite well. Also, note how the parasitic loading of the lower patch leads to a significant decrease of the input impedance, compared to the single-la er patch. It is seen, that the intrinsic impedance of the staczed patch is not very different from 50 R (S of the unmatched atch remains below -10 dB in a 50 R 'Aystem, over a 860 R H z bandwidth). This ty ical characteristic of the stacked patches eases the matciing of the element to 50 R.

Rather than matching the element to 50 R, it should be considered to design the entire BFN usin a slightly lower reference impedance (e. 40 Q). This wily only require one transformation from 40 to 50 R (at the in ut of the array), thus saving board space and reducing the Posses (since the

8

Note that the transmission through the patch (i.e. the scattering parameter S ) is seen to exhibit a similar "stagger-tuned" charactehic as seen in the real- art of the input impedance. The average level of S,, for t i e stacked y t c h is somewhat poorer than for the singre-layer atch, but rom a network point-of-view this stagger-tuned {ehaviour is not surprisin Due to this charactenstic of S,, it must be expected, that &e cross- olarization of the stacked patch is slightly worse, than for t i e single-layer patch. The "sta ger tuned behaviour of S of a stacked patch may probabfy be the reason why stackgd patches are sometimes claimed in the literature to have a poorer cross-polarization pformance. than single-layer atches. The above level of (between -20 dB and -25 $3) is typical for wideband, pi-obe-fed stacked microstrip patches. The measured radiation pattem of the stacked microstrip patch is shown in Fig. 6. H-pni d U,"..*

,IO

>U

m

II

0

0

1.1

M

t4Q.l

8

"-p"13.."muUirul~w~-oDI

.o

111

la

111

*

n

0

I 11.

"

(0

111

110

Fig. 6. Measured radiation pattern of stacked microstrip patch (f = 5.3 GHz).

The stacked patch radiation pattern has a sli htly hi her directivity, compared to the unstacked patch. #he H-pfane pattem of the stacked atch is somewhat asymmetrical (elevation-plane for the d p o r t fed; azimuth plane for the Vport fed). This behaviour is neither expected, nor understood,. and is believed to have been caused by either improper alignment of the two patches or by passive loadin of the driven p t c h by adjacent (non-driven, but stib mutually couple ) patches. A new measurement is planned to investigate this further, and will be reported at the conference ( reviously stacked patches deslgned .at EM1 have all hack) well-behaved pattems). Despite this slight gattem anomal , the CO-and cross-polarization pattems are 0th quite usaile for arra s The influence of the ground plane is much less signdcant, than for the sin le la er atch. The attem remain stable over the full 880-MLz ' ut the main beam narrows with increasing andwidth frequency, and the associated peak directivity increases from 6.5 dBi at 5.1 GHz to 10 dBi at 5.9 GHz.

,g

The cross-polarization level for the stacked patch is worse than for the single-la er patch, but this was expected, due to its hifher S,,-Ievel. ryf this level of cross-polarization is not comp iant with array requirements, simple, yet very efficient, techniques can be employed in the large array to substantially improve the overall array cross-polarization suppression over that of the basic stacked patch element. A cross-polarization improvement of ty ical 15 dB in microstri patch arrays has been achievegusing the method describetf In [9] and [lo]. A practical, example of the im lementation of this cross-polarization improvement tecKnique is found in [7].

245

Authorized licensed use limited to: Danmarks Tekniske Informationscenter. Downloaded on February 16,2010 at 07:16:00 EST from IEEE Xplore. Restrictions apply.

To investigate wideband array-issues, four stacked patch elements were combined into a 2 x 2 element group, as shown in FI 7. The S-parameters of the array was measured in t t e case where the elements were connected to two four-way Wilkinson power splitterskombiners using coaxial cables (i.e. all elements were fed in e ual amplitude and phase). The element spacing was approx. 3.7&,.

The radiation pattern is symmetrical, does not show any si n of anomaly, and has deep nulls. The peak directivit is 1 f 9 dBi at 5.3 GHz, increasin to 14 8 dBi at 5 9 GHz. cross- olarization level is on &e order of -25 dB, which is slight& better than for the individual element. Measurements confirm that the pattern is stable over a very wide frequency range (in excess of 1 GHz), and that the cross-polanzation level remains low.

he:

As demonstrated, low-cost stacked microstrip probe-fed Ytch arrays can be designed, capable of operating over a 15 o bandwidth in C-band. What still remains in order to arrive at a complete antenna, is the design of the azimuth and elevation beam forming networks.

Future work

The C-band element described here is resent1 being and $-band). integrated into a full dual-frequency array This array will be based on the use. of perforated, stacked Lband patches [I 11, inside the perforations of which the (3band elements are placed.

(E-

“.+,

“.*I

Conclusion

Fig. 7. Layout of 2 x 2 element stacked microstrip array.

The measured S-parameters are shown in Fi 8, and are com ared with the calculated S-parameters in tke two cases: a) &tu$ couplin between the elements is neglected in the calculation, b) Alf mutual couplings between the elements are taken into account. In the a) case only the four eigenimpedances of the patches were connected together (in the computer) with the measured S-parameters of the four-wa Wilkinson power splitters and associated cables. In the b3 case, all S-parameters of the 2 x 2 element dual-polarized array (i.e. the full ei ht port), all cables, and the power dividers were measure$ a i d connected.

The design and performance of a wideband, dualpolarization probe-fed microstrip C-band antenna element, and a small test array, have been described. The element will be used in an 800 MHz dual-polarization array, bein designed for use with the next- eneration of the Danisg airborne SAR system, EMISAR. The dual- olanzation probe-fed microstrip array descnbed here is beieved to be the most widebanded antenna of its kind yet reported, for use In polarimetric SAR systems.

Acknowledgements This work was supported by the Danish National Research Foundation.

References [ 11JPL’s “LightSAR’, see e.g. http:r’/lightsar.jpl.nasa.gov/

k2]

European

Space

ttp://envisat.estec.esa.nU

Agency’s

“Envisat”,

Canadian Space Agency’s “RADARSAT see I31ttp://~~~.space.gc.ca/eng/about/radarsat/rad8.html 11”.

Fig. 8. Measured S-parameters of the 2 x 2 element stacked Cband microstrip array

From Fig. 8 it is observed, that the bandwidth of the 2 x 2 element roup is slightly lower than for the individual stacked ekment. This is due to mutual coupling between the elements, and to a lesser extent due to the finite bandwidth of the Wilkinson divider. Note that S of the array is practically identical to S for the e1eme;t. Also note, that the mutual coupling c e r t h must be taken roperly into account, otherwise measuredl vs. calculated &a does not a ree very well. The measured radiation pattem of the 2 x 2 efement array is shown in Fig. 9. H - p a ,d..z,&W,w-,

_..,

z .$


YpDn kd. YlmhA#..

-

e.g.

De artment of Electromagnetic System, Technical University 02 8enmark’s next-generation SAR, “SAR++”; see http://sarppsrv.emi.dtu.dlJ-sarppl 14

51 E.L. Christensen, N. Skou, J. Dall, K. Woelders, J.H. 5cal%rated’ 0r ensen J. Granholm S.N. Madsen ”EMISAR: An absolutely oolarimetric ’L- and C-baid SAY, IEEE Trans. on

Geoscience‘and Remote Sensing Applications, Vol. 36, no. 6, Nov.

1998, pp. 1852-1865.

[6] J. Granholm, K. Woelders, M..Dich, E. Lintz Christensen, Microstn antenna for polanmetnc C-band SAR”, IEEE‘ Int. S mp. on Intennas and Propagation, Seattle, WA, June 1994, pp. 1g44-1847.

\8echnical 1 J.E. . Hansen, . F. . Jensen, “Spherical near-field scanning at the Universit of Denmark, IEEE Trans. on Antennas and

f.. E

>l

e.g.

[7] J. Granholm, K. Woelders, “Dual polarization microstrip antenna array with very low cross-polarization”, submitted for publication in IEEE Trans. on Antmnas ond Propagation.

0

5,

IO

see

I5

Propagarion., Vol. $6, June 1988, pp. 734-739.

.,lo

>a

II

I5

0

I,

x)

8 11‘ I I W ,d.N “ * h

I,,

,IO

1 0.

111

a

I,

(I

@

1,

a

11%

1.0

IrI

.,“,#eo*,

[9] J. Granholm, K. Woelders. “Dual-polarization antenna array with very low cross-polarization and low side lobes’, International PCT Patent Application no. PCT/DK97/00141, filed March 26, 1997. [ 101 K. Woelders, J. Granholm, “Cross-polarization and sidelobe suppression in dual linear polarization antenna arra s”, IEEE Trans. on Antennas and Propagation, Vol. 45, No. 12, &ec. 1997, pp. 1727-1740.

[l 1) J. Granholm “Dual-frequenc and

olarization antenna

arrays for future SAR systems”, 2-1“ &TEC &tenna Worksho on Array Antenna Technology (1-STEC ublication WPP-A2), Noordwijk, The Netherlands, 6-8 May 199%.pp. 29-35. Fig. 9. Radiation pattern of 2 x 2 element array of stack.ed microstrip patches (f = 5.3 GHz,uniform excitation).

246

Authorized licensed use limited to: Danmarks Tekniske Informationscenter. Downloaded on February 16,2010 at 07:16:00 EST from IEEE Xplore. Restrictions apply.