Design and Performance Evaluation of Handset ... - Semantic Scholar
Design and Performance Evaluation of Handset MIMO Antenna Prototypes Qiong Wang1, Hui Zhang1, and Dirk Plettemeier1, Eckhard Ohlmer2 and Gerhard Fettweis2 1
2
Chair for RF and Photonics, Technische Universitaet Dresden, Germany Vodafone Chair Mobile Communications Systems, Technische Universitaet Dresden, Germany Email:{qiong.wang and eckhard.ohlmer}@ifn.et.tu-dresden.de
Abstract— This work presents performance evaluation results for four 2x2 multiple-input-multiple-output (MIMO) prototype antenna configurations with different antenna element spacings and polarization properties. The prototypes are based on a size optimized planar inverted-F antenna (PIFA) element which applies a meander structure etched on the metallic patch. Anechoic and reverberation chamber measurements have verified that the simulated performance translates into the manufactured prototypes. The diversity performance of different prototypes is calculated and compared based on different statistical propagation models. Realistic outdoor MIMO transmission experiments at 2.68 GHz show their suitability for spatial multiplexing transmission. Keywords- multiple input multiple output (MIMO); planar inverted-F antenna (PIFA); antenna diversity; spatial multiplexing
I.
INTRODUCTION
Multiple-input-multiple-output (MIMO) antenna systems have been broadly investigated since they can either increase the transmission systems robustness by adding diversity or increase the achievable data rate by transmitting several spatially separated data streams in parallel. However, achieving those gains requires the different transmission path to be sufficiently de-correlated which can be difficult to realize if multiple antenna elements have to be integrated close to each other in the limited volume of small handset devices. In that context, the planar inverted-F antenna (PIFA) is widely applied in mobile handsets in virtue of its compact size and good performance. In this work a meander PIFA structure which allows to significantly reduce the PIFA patch size, has been proposed for handset applications. Based on the optimized antenna element, four two element antenna array prototypes, characterized by different antenna locations and orientations on the printed circuit board (PCB), have been designed and manufactured in order to investigate the influence on the antennas diversity performance. Measurements in an anechoic chamber and in a reverberation chamber have been carried out and the results are used to verify the match of the simulated antenna parameters and fabricated prototypes. The diversity performance in terms of correlation and diversity gain has been compared for four prototypes using typical statistical propagation models. Finally, transmission experiments have been carried out in an urban outdoor environment for four prototype handset antennas to reveal how the differences seen
from the antenna design perspective translate into MIMO system performance under realistic propagation conditions. II.
ANTENNA DESIGN AND PERFORMANCE
A. Optimized Antenna Element Design The optimized PIFA applies the meander structure, etched in the metallic patch of the traditional PIFA, as shown in Figure 1. The meander PIFA is shorted to the FR-4 PCB (100x40 mm) by a metallic strip and fed by the standard 50Ω RG405 coaxial cable. The optimized patch size is 14.1mm (L) x 7mm (W) which produces a patch length reduction by around 30% compared with the traditional patch PIFA. Antenna element dimension optimization is advantageous since the accommodated handset PCB volume is limited. The meanders change the charge distribution on the antenna plane and increase the inductance resulting in a reduced size [1]. The resonance characteristic of the meander PIFA antenna depends on the meander geometry as well as the antenna element position on the PCB. Figure 2 shows the simulated (solid curve) and measured (dotted curve) return losses which show good agreement in the 2.2-3GHz frequency band. The centre resonant frequency is at 2.68 GHz, required for outdoor transmission experiments in UMTS band VII. The simulated 10 dB bandwidth (200 MHz) is slightly broader than the measured one (180 MHz). B. Dual-element Meander PIFA Array Prototypes Design MIMO performance strongly depends on antenna elements spacing and coupling. Four different handset configurations PT2, PT4, PT6 and PT8 in Figure 3, characterized by different antenna locations and orientations on the PCB, are designed to
Figure 1. Optimized Meander PIFA configurations
0
Return Loss [dB]
-5 -10 -15 -20
simulated measured
-25 2.2
2.3
2.4
2.5 2.6 2.7 Frequency [GHz]
2.8
2.9
3
Figure 2. Simulated and measured return loss of the optimized single PIFA
investigate the influence on the MIMO performance. PT2 is characterized by two meander PIFA elements parallel to each other and PT4 by perpendicular elements based on PT2. The PT6 arrangement consists of two anti parallel elements located diagonally on the PCB and PT8 of two diagonal elements perpendicular to each other based on PT6. Due to the small spacing of the two antennas on the PCB, PT2 and PT4 are supposed to introduce a high coupling between the antenna elements compared with PT6 and PT8. Prototypes characterized by perpendicular antenna arrangements, like PT4 and PT8, are expected to show a decoupling effect due to cross polarization. All of the antenna elements in these four sets of antenna arrays have the feeding point at the margin side of the PCB. Figure 4 shows the fabricated four prototypes and the white foam covered on the PCB is designed to enhance the robustness of the prototypes in outdoor transmission measurement which will be described in the following section.
C. Antenna Prototypes Characteristics First the S-parameters and radiation efficiencies of four prototypes were simulated in HFSS and measurements were performed in a reverberation chamber to validate the simulation results. Figure 5 shows the simulated and measured S-parameters and radiation efficiencies of four prototypes. PT2, PT4 and PT6 show good agreement for S-parameters between simulated and measured results. PT8 has a relatively large difference between simulated and measured S-parameters which is mainly caused by the imperfection during the fabrication process of small antennas. The simulated radiation efficiencies for two antenna elements in PT2 and PT6 are almost the same due to the parallel antenna elements arrangement. PT4 and PT8 have small difference for two elements in both the simulated centre frequencies and the simulated efficiency values due to perpendicular elements arrangement. Both simulated and measured radiation efficiencies in four prototypes are higher than 80% while PT4 has relatively low efficiency values due to two closed-spaced as well as perpendicular arranged elements. Radiation patterns of four prototypes were measured in an anechoic chamber with the double edged horn transmitter antenna. Figure 6 and Figure 7 show the comparisons of measured and simulated radiation patterns for antenna 1 of PT2 at the resonance frequency 2.68 GHz. The radiation patterns were obtained by applying an ideal voltage source to antenna 1 while antenna 2 is terminated with the characteristic impedance and vice versa. Figure 6 shows the patterns in the XY-plane. There are stronger cross-polarized radiations on the front of the PCB (i.e., between 0 to 180 degrees) than the back (i.e., between 0 to -180 degrees), which is similar to the radiation patterns of the traditional PIFA. Besides, there are two notches for the cross-polarized component along the negative Y-direction below the PCB due to the influence of the PCB ground. Figure 7 plots the measured and simulated patterns in the XZ-plane. The ripples in the measured patterns are mainly due to the coaxial feeding cable effect. The patterns of antenna 1 and antenna 2 are symmetrical to each other along the XY-plane when both antennas are mounted on the PCB. Such kind of complementary patterns produce a good pattern diversity characteristic. D. Antenna Prototypes Diversity Performance Evaluation
Figure 3. Configurations of four sets of dual-element meander PIFA arrays 40mm
P4
P6
P8
100mm
P2
Figure 4. Four prototypes of dual-element meander PIFA Array
1) Diversity Parameters: Diversity performance is the critical merit of multiple element antenna systems for handset applications. A number of typical parameters have been widely used to describe the diversity performance of a multiple antenna system in a mobile environment. These parameters are the mean effective gain (MEG) of the antenna elements, the envelope correlation coefficient (ECC) of the received signals, and the effective diversity gain (EDG) achieved by applying maximum ratio combining (MRC). The MEG is originally proposed and defined in [2] as the ratio between the mean received power of the antenna and the total mean incident power. It can be calculated from
0 -5
S11/S22
-15 -20
Efficiency
S [dB]
-10
S12/S21
-25 -30 -35 simulated measured
-40 -45 2.2
2.4 2.6 2.8 Frequency [GHz]
1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0
3
X
0
-60
2.4 2.6 2.8 Frequency [GHz]
Efficiency
S [dB]
-20 -25 -30
120
2.4 2.6 2.8 Frequency [GHz]
0.1 2.2
3
2.4 2.6 2.8 Frequency [GHz]
Efficiency
S [dB]
-30
2.2
2.4 2.6 2.8 Frequency [GHz]
Efficiency
S [dB]
-30 simulated measured
-40
0.6 0.5 0.4 meas. ant1 meas. ant2 sim. ant1 sim. ant2
0.2 0.1 0
-45 2.2
2.4 2.6 2.8 Frequency [GHz]
3
(b)
2.2
2.4 2.6 2.8 Frequency [GHz]
, ,
1
,
(1)
,
where the cross polarization ratio XPR is defined as the ratio of averaged vertical power to averaged horizontal power of the , and , are the and polarized incident field, components of the spherical power gain of the antenna, and , and , are the and polarized components of the angular probability density functions of the incoming plane waves, respectively. The antenna branch power ratio / is used to measure how different the signal power levels delivered by the antennas are. In good diversity system, the power levels of the signals received by the two antennas should be similar.
0.3
-35
150 -180
1 3
0.8
S12/S21
120 -150
1
0.7
-25
-120
150
meas. ant1 meas. ant2 sim. ant1 sim. ant2
0.9
-20
120
0.4
1
-15
90
0.5
3
S11/S22
90 -90
Figure 7. Measured (circle dashed) and simulated (triangle solid) radiation patterns of antenna 1 on the XZ-plane for (a) E and (b) E
0.6
(c) PT6 -5
60
(a)
0.1
0
-60
-180
0 2.4 2.6 2.8 Frequency [GHz]
60
-90
-150
0.2
-45
Z
-20.00
-120
0.3 simulated measured
30 -10.00
y
3
0.7
-25
-30
-20.00
0.9
S12/S21
0 30
-60
meas. ant1 meas. ant2 sim. ant1 sim. ant2
0.8
-10
(b)
-10.00
0.2
-10
2.2
150 -180
-30
1
-40
-150
X
0.4
S11/S22
-35
120
0
0.5
0
-20
-120
(a)
(b) PT4
-15
90
Figure 6. Measured (circle dashed) and simulated (triangle solid) radiation patterns of antenna 1 on the XY-plane for (a) E and (b) E
0
-5
-90
150
0.6
-45 2.2
60
-180
0.3 simulated measured
-40
-60
0.7
S12/S21
-35
60
90
-150
0.8
-15
Z
-20.00
-120
3
0.9
S11/S22
30 -10.00
y
-90
meas. ant1 meas. ant2 sim. ant1/2
1
-5
-30
-20.00
(a) PT2
-10
30 -10.00
2.2
0
0
-30
3
The ECC between the two antenna branches can be | | calculated from the complex correlation coefficient [3] ,
(d) PT8 Figure 5. Simulated and measured S-parameters and radiation efficiencies of four prototypes: (a) PT2, (b) PT4, (c) PT6 and (d) PT8
,
,
·
,
,
, ,
,
,
, ,
,
(2)
,
and denote the and polarized electric fields of the antennas. The correlation coefficient of the signals is related
where
with the propagation environment and the radiated far field characteristics. If there is no correlation beetween the signals received, then there is 0. However, forr a mobile handset application, it is almost impossible to have a zero correlation because of limited antenna spacing on the PC CB.
Both the MEG and ECC are based on thhe spherical power gain patterns and the propagation characterisstics of the mobile communication environment. In the muulti-path wireless propagation environment, a uniform distribuution is reasonably assumed in azimuth direction for the arrivval angles of the incident radio waves towards the mobile term minal antennas. In the elevation direction, Gaussian and Laplaciian distribution are typically used [4]. with different statistical parameters p for the incident waves in indoor and outddoor propagation environments [5]. This is reasonable sinnce in an indoor environment there are more incident waves reflected from the roof while an outdoor environment is charaacterized by more incident waves reflected from surrounding buuildings. The EDG is a widely used measure of the diversity i defined as [6] performance of different antenna systems. It is (3)
%
in which is the power after combining thhe signals received at different antenna branches and is thhe power received by one ideal antenna operating in the same environment. e Both parameters are calculated at the same probabbility level (p%) in a cumulative probability density (CDF) verssus relative power level. The CDF of follows the Rayyleigh distribution while the CDF of follows: 1
1
/
1
/
(4)
where , are the eigenvalues of the signals covariance matrix which is formulated using the MEGs and ECC Cs, . Λ
(5)
m ratio combining The upper bound of EDG for maximum (MRC) with 1% probability level can be derived as 11.7 dB with 1 and 0 0 [66]. 2) Diversity performance: Table I shoows the diversity parameters results for PT2 as an example in different d statistical propagation environments with a handset spatial orientation vertical to the ground plane as shown in Figuure 8 (XY-plane as the ground plane). The envelope correlation coefficient c of less
Figure 8. PT2 spatial orientation for f diversity performance evaluation
than 0.2 has been achieved in both Gaussian indoor and outdoor environments. For Lapplacian outdoor the ECC is 0.53 which is only for the current handset h vertical orientation. The MEG values of each antenna inn different environments vary by less than 1 dB and the differennce of the MEG values between two antenna elements is less than 0.1 dB. Good correlation characteristics with low ECC value and similar MEG values are favorable for a high diversiity gain. The EDG values range from 8.69 to 9.57 dB in differennt propagation environments. Different handset spatial orientations in the space will produce different diversity resuults since spherical patterns will interact with the propagation models. In order to simulate different handset spatial posittions in a real application, the radiation patterns of antenna 1 and antenna 2 are rotated in the whole space for four prototyypes. By averaging all of the spatial orientations, the mean values of diversity parameters are tabulated and compared inn Table II for four prototypes. Gaussian outdoor propagation environment is here used as the typical diversity evaluation ennvironment. As expected, PT8 gives the best correlation charaacteristics with a low ECC value 0.08 while the difference with PT6 is negligible. PT2 and PT4 show higher ECC values, but thhe deterioration is very small. TABLE I.
DIVERSITY PARAMEETERS RESULTS FOR PT2 IN DIFFERENT STATISTICAL PROPAGATION ENVIRONMENTS G [dB] MEG
Value
ECC
Isotropic Gaussian indoor Gaussian outdoor Laplacian indoor Laplacian outdoor
MEG ratio[dB]
EDG [dB]
MEG1
MEG2
0.0008
-3.01
-3.01
0
8.69
0.01
-2.69
-2.69
0
9.01
0.18
-2.37
-2.36
0.01
9.33
0.02
-2.59
-2.58
0.01
9.11
0.53
-2.13
-2.12
0.01
9.57
TABLE II.
DIVERSITY PERFFORMANCE IN GAUSSIAN OUTDOOR PROPAGATION ENVIRONMENT BY Y AVERAGE OF HANDSET SPATIAL ORIENTATIONS FOR FOUR PROTOTYPES AT 2.68 GHZ
Mean values ECC Antenna1 MEG [dB] Antenna2 EDG [dB]
PT2 0.13 -2.87 -2.87 8.76
PT4 0.12 -2.85 -3.40 8.48
PT6 0.09 -3.07 -3.11 8.56
PT8 0.08 -3.25 -3.12 8.47
3) Performance discussion: Prototypes characterized by perpendicular antenna arrangements, like PT4 and PT8, show more decoupling effect than PT2 and PT6 mainly due to cross polarization. This can be explained from the XPR of single antenna element comparing PT2 and PT4 as an example. XPR represents the cross-polarixation power gain which is defined as: ,
sin
,
sin
12
Antenna 1 Antenna 2
10 Total Gain [dB]
The parallel polarized prototypes PT2 and PT6 show coequal MEG characteristics for the two antenna elements, better than the perpendicular polarized prototypes PT4 and PT8. The difference in the EDG for all four prototypes is very small. On the whole, PT2 can provide a desirable diversity performance combined with the most compact arrangement of the MIMO antenna system which introduces possibilities for multiple antenna element applications in small handsets. Two or more antenna elements can be accommodated in very compact 2D or 3D arrangements on the PCB or in the chassis to increase data rate and SNR.
8 6 4 2 0 PT2 Gphi
III.
OUTDOOR TRANSMISSION EXPERIMENTS
Transmission experiments have been carried out in an
PT4 Gtheta
2
XPR [dB]
0
Further, we compare the theta component pattern and phi component pattern for PT2 in the whole sphere in Figure 11. Regarding PT2, we see from the 3D patterns that their shapes for both Etheta and Ephi are differently aligned in the coordinate system, comparing both antenna elements. Thus, e.g. the Ephi component of the incident waves will be differently attenuated at both antenna elements, depending on the angle of arrival. Several waves which arrive at the same time at both antenna elements will be attenuated differently and thus yield different superposition results at both elements which will lead to a de-correlation of the received signals. Therefore, for PT-4 we observe pattern diversity and polarization diversity at the same time. In summary, PT-2 exploits pattern diversity while PT-4 exploits pattern diversity and polarization diversity.
PT4 Gphi
Figure 9. Total polarization component gain comparision for each antenna element of PT2 and PT4
(6)
where , , , are the theta component gain and the phi component gain, respectively. These two polarization gains include the port mismatch. Figure 9 first shows total polarization component gain comparision for each antenna element of PT2 and PT4. It can be concluded that less power is received for the theta component of antenna 2 (PT4), while all other antenna elements seem to be more or less equally sensitive. This is well matched with the observation situation of experiment. Figure 10 shows the XPR comparision for each antenna element of PT2 and PT4. The XPR for antenna 2 of PT4 shows a low value while the other three antenna elements show XPR values close to 0 dB. This means that a certain degree of polarization diversity is employed for PT4 but not for PT2.
PT2 Gtheta
Antenna 1 Antenna 2
-2
-4
-6
-8 PT2
PT4
Figure 10. XPR comparision for each antenna element of PT2 and PT4
urban outdoor environment for four prototype handset antennas to reveal how the differences seen from the antenna design perspective translate into MIMO system performance under realistic propagation conditions. First, we briefly introduce a stochastic signal model, required for comparison with the measured data, followed by a brief description of the experimentation setup and finally the discussion of the measured results. A. Stochastical Channel Model We consider the transmission over a frequency flat fading channel using transmit and receive antennas. The is obtained from received signal vector (7) , and denoting the with MIMO channel matrix, the transmitted symbols with variance and complex valued additive with Gaussian noise with variance , respectively.
The elements of the MIMO channel matrix are assumed to be non amplifying, jointly complex Gaussian distributed with zero mean and correlation matrix . In favor of a computes from reduced parameter set we assume that the Kronecker product of transmitter and receiver correlation matrix [7]: . For the case of two transmit and two receive antennas (denoted 2 × 2) which were also observed in the experiments, and are matrices of dimension 2 × 2 which contain a single correlation coefficient and respectively. In the result section III-C we choose and use this model to assess the impact of the different receive antennas on the MIMO performance. An important figure of merit for a MIMO system is the achievable rate ; | . Without channel state information at the transmitter, the total transmit power is uniformly distributed among the transmit antennas and the achievable rate calculates as [8]
; | with singular value of
log
(8)
1
,
and
denoting the
-th
.
An important criterion for MIMO systems, which allows to draw conclusions if several data streams (two in the experimental setup) can be transmitted in parallel, is the condition number κ which is defined by the ratio of the maximum and the minimum singular value of Η. Ideally, κ should be close to 1 in order to maximize the achievable data rate. Large condition numbers might either stem from highly correlated propagation channels or from different path gains which might originate from the antenna design. The distribution of the random variable κ was computed from the measurements and compared to the distribution obtained from a theoretical model in order to assess the MIMO performance.
(a) PT2, antenna 1, E
(b) PT2, antenna 1, E
(c) PT2, antenna 2, E
(d) PT2, antenna 2, E
Figure 11. Polarization patterns for two antenna elements of PT2
Dipoles PT-2 PT-4 PT-6 PT-8 ρ=0 ρ=0.5 ρ=0.7 ρ=0.9
Figure 13. Dipole antenna pattern relative towards incident waves for different polar antenna alignments
Figure 12. Cumulative distribution function of the condition number. Black curves: theoretical model; colored curves: measurement.
B. Experimentation Setup The experimentation setup implemented a 2 × 2 MIMOOFDM system for downlink transmission in UMTS band VII at 2.68 GHz. The base station was located on a tower, 20 m above street level, surrounded by buildings of about 15 m height in a residential urban area in Dresden, Germany. It employed a typical commercial 45 cross polarized transmit antenna with a 3 dB half beam width of 60 and a down tilt of 6 . Measurements were carried out for the four prototype handset antennas shown in Figure 3 as well as for a two element antenna composed of commercial vertically polarized dipoles [9], mounted parallel to each other in a distance of 50 cm. The receiver hardware was fit into a bicycle rickshaw which pulled a trolley with the receive antennas mounted on a wooden tripod. Repetitive, OFDM based training sequences with a bandwidth of 20 MHz were constantly transmitted from both base station antennas in a frequency orthogonal fashion in order to measure the channel transfer functions. The receiver was slowly moved along the same short route four times for each antenna, varying the antennas azimuth angle orientation in order to emulate random user orientations. Those measurements have been carried out with each antenna in an upright position (polar angle = 0 ) and rotated by 45 (polar angle =45 ). For each antenna and each polar alignment, 800 channel snapshots were captured on different short measurement routes. Those routes, in a distance of 30 m - 130 m from the base station covered line-of-sight propagation conditions with additional reflections from the ground and close walls as well as pure non-line-of-sight propagation conditions. For a detailed description of the experimental setup refer to [10]. C. Results The cumulative distribution function of the condition number has been computed from the measured MIMO channel
transfer functions and is compared for the four prototype handset antennas and the dipoles in Figure 12. Additionally, results obtained from the theoretical model with equal transmitter and receiver correlation coefficients are shown. From a comparison of the stochastic model with the experiments it can be concluded if the prototypes would be suitable for spatial multiplexing transmission. For a polar angle of 0 it can be observed that the performance of the prototypes is comparable with the performance of the theoretical model with a spatial correlation coefficient of 0.3 0.5, whereas the prototypes with the larger antenna element spacing (PT-6 and PT-8) lead to the smallest condition number, i.e. the highest de-correlation. Thus, from a correlation perspective, even the smallest antenna spacing of about a quarter the wavelength would be sufficient to support two spatial streams with only minor performance degradations compared to the uncorrelated case [10]. The reason for the relatively low correlation becomes obvious from the discussion of the prototypes far field pattern in Section II D. A significant difference in shape can be observed, comparing the pattern of two antenna elements of the same prototype for a single polarization component, which can be referred to as pattern diversity. Thus, the same incident waves will lead to different superposition results at both antenna elements. Moreover, PT-4 for instance was found to exploit a certain degree of polarization diversity in addition to pattern diversity, which was not the case for PT-2. This can explain the smaller condition number of PT-4 seen in Figure 12 (left), although both prototypes exhibit almost the same antenna element spacing. Interestingly, the channels measured with the dipoles which are only sensitive to one polarization component are worse conditioned and the performance compares to the theoretical model with 0.7 which can be attributed to the fact that the dipoles cannot exploit pattern diversity. However, mounting the dipoles perpendicular to each other would allow to exploit polarization diversity and thus decrease the condition number. Additionally, it can be seen that changing the polar antenna alignment to 45 , decreases the condition number, most noticeable for the dipoles. In an upright antenna position and due to the positioning of the antenna w.r.t. the base station, the incident waves will most likely arrive at the dipole antenna
under polar angles where the antenna pattern is almost constant over a wide angular range (3 dB half beam width of about 40 ) as sketched in the upper part of Figure 13. However, by changing the polar antenna alignment, situations appear where incident waves in particular reflections from the ground and close walls arrive more frequently from directions where the antenna pattern changes rapidly within a few degrees as shown in the lower part of Figure 13. Since the inter antenna element distance (0.5 m) cannot be neglected compared to the distance towards ground and reflecting walls (2 m - 5 m) along the measurement routes, those reflections will arrive under different angles at the two antenna elements and will thus be differently attenuated which in turn will result in different superposition results of several incident waves at both antenna elements. IV.
SUMMARY AND CONCLUSIONS
Four handset MIMO antenna prototypes with different antenna element spacing distances and polarization properties have been designed and fabricated based on a meander PIFA element, size optimized for operation at 2.68 GHz. Anechoic and reverberation chamber measurements have verified that the simulated performance translates into the manufactured prototypes. It was found that even an antenna element spacing of smaller than quarter the wavelength can offer low correlation and high diversity gains. A comparison of measured channel data, obtained from outdoor MIMO transmission experiments using realistic base station antennas and the four prototypes, with the condition number distribution, obtained from the well known stochastic Kronecker model, reveals their suitability for spatial multiplexing transmission.
ACKNOWLEDGMENT The authors would like to acknowledge the support of Tyco Electronics, ‘s-Hertogenbosch, Netherlands. REFERENCES [1]
L. Zheng, X. Zhang and S. Lai, “Resonant Characteristics of the Meander-line Dipole Antenna,” Journal of Microwaves, vol. 3, 2006. [2] T. Taga, “Analysis for mean effective gain of mobile antennas in land mobile radio environments,” IEEE Transactions on Vehicular Technology, vol. 39, no. 2, pp. 117–131, 1990. [3] R. G. Vaughan and J. B. Andersen, “Antenna diversity in mobile communications,” IEEE Trans. Veh. Technol., vol. VT-36, pp. 149–172, Nov.1987 [4] K. Pedersen, P. Mogensen, and B. Fleury, “Power azimuth spectrum in outdoor environments,” Electron. Lett., vol. 33, no. 18, pp. 1583–1584, Aug. 1997. [5] Z. Ying, T. Bolin, V. Plicanic, A. Derneryd, and G. Kristensson, “Diversity antenna terminal evaluation,” in IEEE Antennas Propag. Soc. Int. Symp., Jul. 2005, vol. 2A, pp. 375–378. [6] P.-S. Kildal, K. Rosengren, J. Byun, and J. Lee, “Definition of effective diversity gain and how to measure it in a reverberation chamber,” Microwave and Optical Technology Letters, vol. 34, no. 1, pp. 56–59, 2002. [7] J. Kermoal, L. Schumacher, K. Pedersen, P. Mogensen, and F. Frederiksen, “A stochastic MIMO radio channel model with experimental validation,” IEEE Journal on Selected Areas in Communications, vol. 20, no. 6, pp. 1211–1226, Aug. 2002. [8] N. Chiurtu, B. Rimoldy, and E.Telatar, “On the Capacity of MultiAntenna Gaussian Channels,” in IEEE International Symposium on Information Theory, 2001. [9] Kathrein, “VPol Omni 1710 - 2700 360± 2dBi,” 800 10431 datasheet, 2007. [Online]. Available: http://www.kathrein.de [10] E. Ohlmer, J. Hofrichter, S. Bittner, G. Fettweis, Q. Wang, H. Zhang, K. Wolf, and D. Plettemeier, “Urban Outdoor MIMO Experiments with Realistic Handset and Base Station Antennas,” in IEEE Vehicular Technology Conference, VTC Spring , 2010.
2
Chair for RF and Photonics, Technische Universitaet Dresden, Germany Vodafone Chair Mobile Communications Systems, Technische Universitaet Dresden, Germany Email:{qiong.wang and eckhard.ohlmer}@ifn.et.tu-dresden.de
Abstract— This work presents performance evaluation results for four 2x2 multiple-input-multiple-output (MIMO) prototype antenna configurations with different antenna element spacings and polarization properties. The prototypes are based on a size optimized planar inverted-F antenna (PIFA) element which applies a meander structure etched on the metallic patch. Anechoic and reverberation chamber measurements have verified that the simulated performance translates into the manufactured prototypes. The diversity performance of different prototypes is calculated and compared based on different statistical propagation models. Realistic outdoor MIMO transmission experiments at 2.68 GHz show their suitability for spatial multiplexing transmission. Keywords- multiple input multiple output (MIMO); planar inverted-F antenna (PIFA); antenna diversity; spatial multiplexing
I.
INTRODUCTION
Multiple-input-multiple-output (MIMO) antenna systems have been broadly investigated since they can either increase the transmission systems robustness by adding diversity or increase the achievable data rate by transmitting several spatially separated data streams in parallel. However, achieving those gains requires the different transmission path to be sufficiently de-correlated which can be difficult to realize if multiple antenna elements have to be integrated close to each other in the limited volume of small handset devices. In that context, the planar inverted-F antenna (PIFA) is widely applied in mobile handsets in virtue of its compact size and good performance. In this work a meander PIFA structure which allows to significantly reduce the PIFA patch size, has been proposed for handset applications. Based on the optimized antenna element, four two element antenna array prototypes, characterized by different antenna locations and orientations on the printed circuit board (PCB), have been designed and manufactured in order to investigate the influence on the antennas diversity performance. Measurements in an anechoic chamber and in a reverberation chamber have been carried out and the results are used to verify the match of the simulated antenna parameters and fabricated prototypes. The diversity performance in terms of correlation and diversity gain has been compared for four prototypes using typical statistical propagation models. Finally, transmission experiments have been carried out in an urban outdoor environment for four prototype handset antennas to reveal how the differences seen
from the antenna design perspective translate into MIMO system performance under realistic propagation conditions. II.
ANTENNA DESIGN AND PERFORMANCE
A. Optimized Antenna Element Design The optimized PIFA applies the meander structure, etched in the metallic patch of the traditional PIFA, as shown in Figure 1. The meander PIFA is shorted to the FR-4 PCB (100x40 mm) by a metallic strip and fed by the standard 50Ω RG405 coaxial cable. The optimized patch size is 14.1mm (L) x 7mm (W) which produces a patch length reduction by around 30% compared with the traditional patch PIFA. Antenna element dimension optimization is advantageous since the accommodated handset PCB volume is limited. The meanders change the charge distribution on the antenna plane and increase the inductance resulting in a reduced size [1]. The resonance characteristic of the meander PIFA antenna depends on the meander geometry as well as the antenna element position on the PCB. Figure 2 shows the simulated (solid curve) and measured (dotted curve) return losses which show good agreement in the 2.2-3GHz frequency band. The centre resonant frequency is at 2.68 GHz, required for outdoor transmission experiments in UMTS band VII. The simulated 10 dB bandwidth (200 MHz) is slightly broader than the measured one (180 MHz). B. Dual-element Meander PIFA Array Prototypes Design MIMO performance strongly depends on antenna elements spacing and coupling. Four different handset configurations PT2, PT4, PT6 and PT8 in Figure 3, characterized by different antenna locations and orientations on the PCB, are designed to
Figure 1. Optimized Meander PIFA configurations
0
Return Loss [dB]
-5 -10 -15 -20
simulated measured
-25 2.2
2.3
2.4
2.5 2.6 2.7 Frequency [GHz]
2.8
2.9
3
Figure 2. Simulated and measured return loss of the optimized single PIFA
investigate the influence on the MIMO performance. PT2 is characterized by two meander PIFA elements parallel to each other and PT4 by perpendicular elements based on PT2. The PT6 arrangement consists of two anti parallel elements located diagonally on the PCB and PT8 of two diagonal elements perpendicular to each other based on PT6. Due to the small spacing of the two antennas on the PCB, PT2 and PT4 are supposed to introduce a high coupling between the antenna elements compared with PT6 and PT8. Prototypes characterized by perpendicular antenna arrangements, like PT4 and PT8, are expected to show a decoupling effect due to cross polarization. All of the antenna elements in these four sets of antenna arrays have the feeding point at the margin side of the PCB. Figure 4 shows the fabricated four prototypes and the white foam covered on the PCB is designed to enhance the robustness of the prototypes in outdoor transmission measurement which will be described in the following section.
C. Antenna Prototypes Characteristics First the S-parameters and radiation efficiencies of four prototypes were simulated in HFSS and measurements were performed in a reverberation chamber to validate the simulation results. Figure 5 shows the simulated and measured S-parameters and radiation efficiencies of four prototypes. PT2, PT4 and PT6 show good agreement for S-parameters between simulated and measured results. PT8 has a relatively large difference between simulated and measured S-parameters which is mainly caused by the imperfection during the fabrication process of small antennas. The simulated radiation efficiencies for two antenna elements in PT2 and PT6 are almost the same due to the parallel antenna elements arrangement. PT4 and PT8 have small difference for two elements in both the simulated centre frequencies and the simulated efficiency values due to perpendicular elements arrangement. Both simulated and measured radiation efficiencies in four prototypes are higher than 80% while PT4 has relatively low efficiency values due to two closed-spaced as well as perpendicular arranged elements. Radiation patterns of four prototypes were measured in an anechoic chamber with the double edged horn transmitter antenna. Figure 6 and Figure 7 show the comparisons of measured and simulated radiation patterns for antenna 1 of PT2 at the resonance frequency 2.68 GHz. The radiation patterns were obtained by applying an ideal voltage source to antenna 1 while antenna 2 is terminated with the characteristic impedance and vice versa. Figure 6 shows the patterns in the XY-plane. There are stronger cross-polarized radiations on the front of the PCB (i.e., between 0 to 180 degrees) than the back (i.e., between 0 to -180 degrees), which is similar to the radiation patterns of the traditional PIFA. Besides, there are two notches for the cross-polarized component along the negative Y-direction below the PCB due to the influence of the PCB ground. Figure 7 plots the measured and simulated patterns in the XZ-plane. The ripples in the measured patterns are mainly due to the coaxial feeding cable effect. The patterns of antenna 1 and antenna 2 are symmetrical to each other along the XY-plane when both antennas are mounted on the PCB. Such kind of complementary patterns produce a good pattern diversity characteristic. D. Antenna Prototypes Diversity Performance Evaluation
Figure 3. Configurations of four sets of dual-element meander PIFA arrays 40mm
P4
P6
P8
100mm
P2
Figure 4. Four prototypes of dual-element meander PIFA Array
1) Diversity Parameters: Diversity performance is the critical merit of multiple element antenna systems for handset applications. A number of typical parameters have been widely used to describe the diversity performance of a multiple antenna system in a mobile environment. These parameters are the mean effective gain (MEG) of the antenna elements, the envelope correlation coefficient (ECC) of the received signals, and the effective diversity gain (EDG) achieved by applying maximum ratio combining (MRC). The MEG is originally proposed and defined in [2] as the ratio between the mean received power of the antenna and the total mean incident power. It can be calculated from
0 -5
S11/S22
-15 -20
Efficiency
S [dB]
-10
S12/S21
-25 -30 -35 simulated measured
-40 -45 2.2
2.4 2.6 2.8 Frequency [GHz]
1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0
3
X
0
-60
2.4 2.6 2.8 Frequency [GHz]
Efficiency
S [dB]
-20 -25 -30
120
2.4 2.6 2.8 Frequency [GHz]
0.1 2.2
3
2.4 2.6 2.8 Frequency [GHz]
Efficiency
S [dB]
-30
2.2
2.4 2.6 2.8 Frequency [GHz]
Efficiency
S [dB]
-30 simulated measured
-40
0.6 0.5 0.4 meas. ant1 meas. ant2 sim. ant1 sim. ant2
0.2 0.1 0
-45 2.2
2.4 2.6 2.8 Frequency [GHz]
3
(b)
2.2
2.4 2.6 2.8 Frequency [GHz]
, ,
1
,
(1)
,
where the cross polarization ratio XPR is defined as the ratio of averaged vertical power to averaged horizontal power of the , and , are the and polarized incident field, components of the spherical power gain of the antenna, and , and , are the and polarized components of the angular probability density functions of the incoming plane waves, respectively. The antenna branch power ratio / is used to measure how different the signal power levels delivered by the antennas are. In good diversity system, the power levels of the signals received by the two antennas should be similar.
0.3
-35
150 -180
1 3
0.8
S12/S21
120 -150
1
0.7
-25
-120
150
meas. ant1 meas. ant2 sim. ant1 sim. ant2
0.9
-20
120
0.4
1
-15
90
0.5
3
S11/S22
90 -90
Figure 7. Measured (circle dashed) and simulated (triangle solid) radiation patterns of antenna 1 on the XZ-plane for (a) E and (b) E
0.6
(c) PT6 -5
60
(a)
0.1
0
-60
-180
0 2.4 2.6 2.8 Frequency [GHz]
60
-90
-150
0.2
-45
Z
-20.00
-120
0.3 simulated measured
30 -10.00
y
3
0.7
-25
-30
-20.00
0.9
S12/S21
0 30
-60
meas. ant1 meas. ant2 sim. ant1 sim. ant2
0.8
-10
(b)
-10.00
0.2
-10
2.2
150 -180
-30
1
-40
-150
X
0.4
S11/S22
-35
120
0
0.5
0
-20
-120
(a)
(b) PT4
-15
90
Figure 6. Measured (circle dashed) and simulated (triangle solid) radiation patterns of antenna 1 on the XY-plane for (a) E and (b) E
0
-5
-90
150
0.6
-45 2.2
60
-180
0.3 simulated measured
-40
-60
0.7
S12/S21
-35
60
90
-150
0.8
-15
Z
-20.00
-120
3
0.9
S11/S22
30 -10.00
y
-90
meas. ant1 meas. ant2 sim. ant1/2
1
-5
-30
-20.00
(a) PT2
-10
30 -10.00
2.2
0
0
-30
3
The ECC between the two antenna branches can be | | calculated from the complex correlation coefficient [3] ,
(d) PT8 Figure 5. Simulated and measured S-parameters and radiation efficiencies of four prototypes: (a) PT2, (b) PT4, (c) PT6 and (d) PT8
,
,
·
,
,
, ,
,
,
, ,
,
(2)
,
and denote the and polarized electric fields of the antennas. The correlation coefficient of the signals is related
where
with the propagation environment and the radiated far field characteristics. If there is no correlation beetween the signals received, then there is 0. However, forr a mobile handset application, it is almost impossible to have a zero correlation because of limited antenna spacing on the PC CB.
Both the MEG and ECC are based on thhe spherical power gain patterns and the propagation characterisstics of the mobile communication environment. In the muulti-path wireless propagation environment, a uniform distribuution is reasonably assumed in azimuth direction for the arrivval angles of the incident radio waves towards the mobile term minal antennas. In the elevation direction, Gaussian and Laplaciian distribution are typically used [4]. with different statistical parameters p for the incident waves in indoor and outddoor propagation environments [5]. This is reasonable sinnce in an indoor environment there are more incident waves reflected from the roof while an outdoor environment is charaacterized by more incident waves reflected from surrounding buuildings. The EDG is a widely used measure of the diversity i defined as [6] performance of different antenna systems. It is (3)
%
in which is the power after combining thhe signals received at different antenna branches and is thhe power received by one ideal antenna operating in the same environment. e Both parameters are calculated at the same probabbility level (p%) in a cumulative probability density (CDF) verssus relative power level. The CDF of follows the Rayyleigh distribution while the CDF of follows: 1
1
/
1
/
(4)
where , are the eigenvalues of the signals covariance matrix which is formulated using the MEGs and ECC Cs, . Λ
(5)
m ratio combining The upper bound of EDG for maximum (MRC) with 1% probability level can be derived as 11.7 dB with 1 and 0 0 [66]. 2) Diversity performance: Table I shoows the diversity parameters results for PT2 as an example in different d statistical propagation environments with a handset spatial orientation vertical to the ground plane as shown in Figuure 8 (XY-plane as the ground plane). The envelope correlation coefficient c of less
Figure 8. PT2 spatial orientation for f diversity performance evaluation
than 0.2 has been achieved in both Gaussian indoor and outdoor environments. For Lapplacian outdoor the ECC is 0.53 which is only for the current handset h vertical orientation. The MEG values of each antenna inn different environments vary by less than 1 dB and the differennce of the MEG values between two antenna elements is less than 0.1 dB. Good correlation characteristics with low ECC value and similar MEG values are favorable for a high diversiity gain. The EDG values range from 8.69 to 9.57 dB in differennt propagation environments. Different handset spatial orientations in the space will produce different diversity resuults since spherical patterns will interact with the propagation models. In order to simulate different handset spatial posittions in a real application, the radiation patterns of antenna 1 and antenna 2 are rotated in the whole space for four prototyypes. By averaging all of the spatial orientations, the mean values of diversity parameters are tabulated and compared inn Table II for four prototypes. Gaussian outdoor propagation environment is here used as the typical diversity evaluation ennvironment. As expected, PT8 gives the best correlation charaacteristics with a low ECC value 0.08 while the difference with PT6 is negligible. PT2 and PT4 show higher ECC values, but thhe deterioration is very small. TABLE I.
DIVERSITY PARAMEETERS RESULTS FOR PT2 IN DIFFERENT STATISTICAL PROPAGATION ENVIRONMENTS G [dB] MEG
Value
ECC
Isotropic Gaussian indoor Gaussian outdoor Laplacian indoor Laplacian outdoor
MEG ratio[dB]
EDG [dB]
MEG1
MEG2
0.0008
-3.01
-3.01
0
8.69
0.01
-2.69
-2.69
0
9.01
0.18
-2.37
-2.36
0.01
9.33
0.02
-2.59
-2.58
0.01
9.11
0.53
-2.13
-2.12
0.01
9.57
TABLE II.
DIVERSITY PERFFORMANCE IN GAUSSIAN OUTDOOR PROPAGATION ENVIRONMENT BY Y AVERAGE OF HANDSET SPATIAL ORIENTATIONS FOR FOUR PROTOTYPES AT 2.68 GHZ
Mean values ECC Antenna1 MEG [dB] Antenna2 EDG [dB]
PT2 0.13 -2.87 -2.87 8.76
PT4 0.12 -2.85 -3.40 8.48
PT6 0.09 -3.07 -3.11 8.56
PT8 0.08 -3.25 -3.12 8.47
3) Performance discussion: Prototypes characterized by perpendicular antenna arrangements, like PT4 and PT8, show more decoupling effect than PT2 and PT6 mainly due to cross polarization. This can be explained from the XPR of single antenna element comparing PT2 and PT4 as an example. XPR represents the cross-polarixation power gain which is defined as: ,
sin
,
sin
12
Antenna 1 Antenna 2
10 Total Gain [dB]
The parallel polarized prototypes PT2 and PT6 show coequal MEG characteristics for the two antenna elements, better than the perpendicular polarized prototypes PT4 and PT8. The difference in the EDG for all four prototypes is very small. On the whole, PT2 can provide a desirable diversity performance combined with the most compact arrangement of the MIMO antenna system which introduces possibilities for multiple antenna element applications in small handsets. Two or more antenna elements can be accommodated in very compact 2D or 3D arrangements on the PCB or in the chassis to increase data rate and SNR.
8 6 4 2 0 PT2 Gphi
III.
OUTDOOR TRANSMISSION EXPERIMENTS
Transmission experiments have been carried out in an
PT4 Gtheta
2
XPR [dB]
0
Further, we compare the theta component pattern and phi component pattern for PT2 in the whole sphere in Figure 11. Regarding PT2, we see from the 3D patterns that their shapes for both Etheta and Ephi are differently aligned in the coordinate system, comparing both antenna elements. Thus, e.g. the Ephi component of the incident waves will be differently attenuated at both antenna elements, depending on the angle of arrival. Several waves which arrive at the same time at both antenna elements will be attenuated differently and thus yield different superposition results at both elements which will lead to a de-correlation of the received signals. Therefore, for PT-4 we observe pattern diversity and polarization diversity at the same time. In summary, PT-2 exploits pattern diversity while PT-4 exploits pattern diversity and polarization diversity.
PT4 Gphi
Figure 9. Total polarization component gain comparision for each antenna element of PT2 and PT4
(6)
where , , , are the theta component gain and the phi component gain, respectively. These two polarization gains include the port mismatch. Figure 9 first shows total polarization component gain comparision for each antenna element of PT2 and PT4. It can be concluded that less power is received for the theta component of antenna 2 (PT4), while all other antenna elements seem to be more or less equally sensitive. This is well matched with the observation situation of experiment. Figure 10 shows the XPR comparision for each antenna element of PT2 and PT4. The XPR for antenna 2 of PT4 shows a low value while the other three antenna elements show XPR values close to 0 dB. This means that a certain degree of polarization diversity is employed for PT4 but not for PT2.
PT2 Gtheta
Antenna 1 Antenna 2
-2
-4
-6
-8 PT2
PT4
Figure 10. XPR comparision for each antenna element of PT2 and PT4
urban outdoor environment for four prototype handset antennas to reveal how the differences seen from the antenna design perspective translate into MIMO system performance under realistic propagation conditions. First, we briefly introduce a stochastic signal model, required for comparison with the measured data, followed by a brief description of the experimentation setup and finally the discussion of the measured results. A. Stochastical Channel Model We consider the transmission over a frequency flat fading channel using transmit and receive antennas. The is obtained from received signal vector (7) , and denoting the with MIMO channel matrix, the transmitted symbols with variance and complex valued additive with Gaussian noise with variance , respectively.
The elements of the MIMO channel matrix are assumed to be non amplifying, jointly complex Gaussian distributed with zero mean and correlation matrix . In favor of a computes from reduced parameter set we assume that the Kronecker product of transmitter and receiver correlation matrix [7]: . For the case of two transmit and two receive antennas (denoted 2 × 2) which were also observed in the experiments, and are matrices of dimension 2 × 2 which contain a single correlation coefficient and respectively. In the result section III-C we choose and use this model to assess the impact of the different receive antennas on the MIMO performance. An important figure of merit for a MIMO system is the achievable rate ; | . Without channel state information at the transmitter, the total transmit power is uniformly distributed among the transmit antennas and the achievable rate calculates as [8]
; | with singular value of
log
(8)
1
,
and
denoting the
-th
.
An important criterion for MIMO systems, which allows to draw conclusions if several data streams (two in the experimental setup) can be transmitted in parallel, is the condition number κ which is defined by the ratio of the maximum and the minimum singular value of Η. Ideally, κ should be close to 1 in order to maximize the achievable data rate. Large condition numbers might either stem from highly correlated propagation channels or from different path gains which might originate from the antenna design. The distribution of the random variable κ was computed from the measurements and compared to the distribution obtained from a theoretical model in order to assess the MIMO performance.
(a) PT2, antenna 1, E
(b) PT2, antenna 1, E
(c) PT2, antenna 2, E
(d) PT2, antenna 2, E
Figure 11. Polarization patterns for two antenna elements of PT2
Dipoles PT-2 PT-4 PT-6 PT-8 ρ=0 ρ=0.5 ρ=0.7 ρ=0.9
Figure 13. Dipole antenna pattern relative towards incident waves for different polar antenna alignments
Figure 12. Cumulative distribution function of the condition number. Black curves: theoretical model; colored curves: measurement.
B. Experimentation Setup The experimentation setup implemented a 2 × 2 MIMOOFDM system for downlink transmission in UMTS band VII at 2.68 GHz. The base station was located on a tower, 20 m above street level, surrounded by buildings of about 15 m height in a residential urban area in Dresden, Germany. It employed a typical commercial 45 cross polarized transmit antenna with a 3 dB half beam width of 60 and a down tilt of 6 . Measurements were carried out for the four prototype handset antennas shown in Figure 3 as well as for a two element antenna composed of commercial vertically polarized dipoles [9], mounted parallel to each other in a distance of 50 cm. The receiver hardware was fit into a bicycle rickshaw which pulled a trolley with the receive antennas mounted on a wooden tripod. Repetitive, OFDM based training sequences with a bandwidth of 20 MHz were constantly transmitted from both base station antennas in a frequency orthogonal fashion in order to measure the channel transfer functions. The receiver was slowly moved along the same short route four times for each antenna, varying the antennas azimuth angle orientation in order to emulate random user orientations. Those measurements have been carried out with each antenna in an upright position (polar angle = 0 ) and rotated by 45 (polar angle =45 ). For each antenna and each polar alignment, 800 channel snapshots were captured on different short measurement routes. Those routes, in a distance of 30 m - 130 m from the base station covered line-of-sight propagation conditions with additional reflections from the ground and close walls as well as pure non-line-of-sight propagation conditions. For a detailed description of the experimental setup refer to [10]. C. Results The cumulative distribution function of the condition number has been computed from the measured MIMO channel
transfer functions and is compared for the four prototype handset antennas and the dipoles in Figure 12. Additionally, results obtained from the theoretical model with equal transmitter and receiver correlation coefficients are shown. From a comparison of the stochastic model with the experiments it can be concluded if the prototypes would be suitable for spatial multiplexing transmission. For a polar angle of 0 it can be observed that the performance of the prototypes is comparable with the performance of the theoretical model with a spatial correlation coefficient of 0.3 0.5, whereas the prototypes with the larger antenna element spacing (PT-6 and PT-8) lead to the smallest condition number, i.e. the highest de-correlation. Thus, from a correlation perspective, even the smallest antenna spacing of about a quarter the wavelength would be sufficient to support two spatial streams with only minor performance degradations compared to the uncorrelated case [10]. The reason for the relatively low correlation becomes obvious from the discussion of the prototypes far field pattern in Section II D. A significant difference in shape can be observed, comparing the pattern of two antenna elements of the same prototype for a single polarization component, which can be referred to as pattern diversity. Thus, the same incident waves will lead to different superposition results at both antenna elements. Moreover, PT-4 for instance was found to exploit a certain degree of polarization diversity in addition to pattern diversity, which was not the case for PT-2. This can explain the smaller condition number of PT-4 seen in Figure 12 (left), although both prototypes exhibit almost the same antenna element spacing. Interestingly, the channels measured with the dipoles which are only sensitive to one polarization component are worse conditioned and the performance compares to the theoretical model with 0.7 which can be attributed to the fact that the dipoles cannot exploit pattern diversity. However, mounting the dipoles perpendicular to each other would allow to exploit polarization diversity and thus decrease the condition number. Additionally, it can be seen that changing the polar antenna alignment to 45 , decreases the condition number, most noticeable for the dipoles. In an upright antenna position and due to the positioning of the antenna w.r.t. the base station, the incident waves will most likely arrive at the dipole antenna
under polar angles where the antenna pattern is almost constant over a wide angular range (3 dB half beam width of about 40 ) as sketched in the upper part of Figure 13. However, by changing the polar antenna alignment, situations appear where incident waves in particular reflections from the ground and close walls arrive more frequently from directions where the antenna pattern changes rapidly within a few degrees as shown in the lower part of Figure 13. Since the inter antenna element distance (0.5 m) cannot be neglected compared to the distance towards ground and reflecting walls (2 m - 5 m) along the measurement routes, those reflections will arrive under different angles at the two antenna elements and will thus be differently attenuated which in turn will result in different superposition results of several incident waves at both antenna elements. IV.
SUMMARY AND CONCLUSIONS
Four handset MIMO antenna prototypes with different antenna element spacing distances and polarization properties have been designed and fabricated based on a meander PIFA element, size optimized for operation at 2.68 GHz. Anechoic and reverberation chamber measurements have verified that the simulated performance translates into the manufactured prototypes. It was found that even an antenna element spacing of smaller than quarter the wavelength can offer low correlation and high diversity gains. A comparison of measured channel data, obtained from outdoor MIMO transmission experiments using realistic base station antennas and the four prototypes, with the condition number distribution, obtained from the well known stochastic Kronecker model, reveals their suitability for spatial multiplexing transmission.
ACKNOWLEDGMENT The authors would like to acknowledge the support of Tyco Electronics, ‘s-Hertogenbosch, Netherlands. REFERENCES [1]
L. Zheng, X. Zhang and S. Lai, “Resonant Characteristics of the Meander-line Dipole Antenna,” Journal of Microwaves, vol. 3, 2006. [2] T. Taga, “Analysis for mean effective gain of mobile antennas in land mobile radio environments,” IEEE Transactions on Vehicular Technology, vol. 39, no. 2, pp. 117–131, 1990. [3] R. G. Vaughan and J. B. Andersen, “Antenna diversity in mobile communications,” IEEE Trans. Veh. Technol., vol. VT-36, pp. 149–172, Nov.1987 [4] K. Pedersen, P. Mogensen, and B. Fleury, “Power azimuth spectrum in outdoor environments,” Electron. Lett., vol. 33, no. 18, pp. 1583–1584, Aug. 1997. [5] Z. Ying, T. Bolin, V. Plicanic, A. Derneryd, and G. Kristensson, “Diversity antenna terminal evaluation,” in IEEE Antennas Propag. Soc. Int. Symp., Jul. 2005, vol. 2A, pp. 375–378. [6] P.-S. Kildal, K. Rosengren, J. Byun, and J. Lee, “Definition of effective diversity gain and how to measure it in a reverberation chamber,” Microwave and Optical Technology Letters, vol. 34, no. 1, pp. 56–59, 2002. [7] J. Kermoal, L. Schumacher, K. Pedersen, P. Mogensen, and F. Frederiksen, “A stochastic MIMO radio channel model with experimental validation,” IEEE Journal on Selected Areas in Communications, vol. 20, no. 6, pp. 1211–1226, Aug. 2002. [8] N. Chiurtu, B. Rimoldy, and E.Telatar, “On the Capacity of MultiAntenna Gaussian Channels,” in IEEE International Symposium on Information Theory, 2001. [9] Kathrein, “VPol Omni 1710 - 2700 360± 2dBi,” 800 10431 datasheet, 2007. [Online]. Available: http://www.kathrein.de [10] E. Ohlmer, J. Hofrichter, S. Bittner, G. Fettweis, Q. Wang, H. Zhang, K. Wolf, and D. Plettemeier, “Urban Outdoor MIMO Experiments with Realistic Handset and Base Station Antennas,” in IEEE Vehicular Technology Conference, VTC Spring , 2010.
