《MIMO多进多出多天线》课程教学资源（文献资料）Polarization Diversity in Mobile Communications.pdf

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 39, NO. 3, AUGUST 1990 177 Polarization Diversity in Mobile Communications RODNEY G. VAUGHAN, MEMBER, IEEE Abstmct- Signals in the vertical and horizontal polarizations at the base station have been measured by transmitting from a principally vertically polarized mobile. There was no direct line-of-sight path between the mobile and base. The envelopes were uncorrelated and the means differed by 7 dB and 12 dB when the mobile was in urban and suburban areas, respectively. The discussion of the results includes theoretical curves showing the relationship between the envelope correlation coefficient and the mean levels difference of Rayleigh distributed signals in orthogonal linear polarizations at the base station. The variable parameters are the rotation angle of the base station antenna and the cross polar discrimination of the incident fields. I. INTRODUCTION LARIZATION diversity seems to have received dispro- p" portionately little attention in the literature. Lee and Yeh's [4] polarization path diversity proposal was the first step in this direction, but the idea of using just one transmit antenna at the mobile and receiving orthogonal polarizations at the base station was not considered. Kozono et al. [3] measured linear polarizations at a f 45" orientation, received from a vertical dipole antenna on a mobile in Tokyo. Their base station antenna was variable in the sense that the arms could rotate in opposite directions, from both arms lying in the horizontal plane to being orthogonal to each other at f45"; they also accounted for the azimuthal dependence. Herein, base station elements that are always orthogonal and rotatable together are considered. The measurements described are confined to a vertical-and-horizontal configuration. The azimuthal dependence is not considered: incident signals are assumed to be broadside to the plane of the antenna. Preliminary results were published by Vaughan and Bach Andersen [8]. Bergmann and Arnold [l] have also discussed measurements for (handheld) polarization diversity. The aforementioned publications seem to be the only ones that regard polarization diversity. This is surprising, considering the advantages and simplicity of the scheme. At the base station, space diversity is considerably less practical than at the mobile because the narrow angle of incident fields calls for large spacings of the antennas. For example, Lee [5, p. 2011 notes that for an azimuthal angle width of sources of 0.4" and an envelope correlation coefficient of 0.7, base station antennas must be spaced by 25 wavelengths for the broadside propagation case and well over 100 wavelengths for the in-line propagation case. The high cost of space diversity at the base station prompts the consideration of using orthogonal polarizations. This may permit only two diversity Manuscript received August 16, 1986; revised February 1990. The author is with the Department of Scientific and Industrial Research, Physics and Engineering Laboratory, Gracefield Road, Gracefield, Private Bag, Lower Hutt, New Zealand. IEEE Log Number 9035261. branches (without resorting to antenna spacing), but it does allow the antenna elements to be colocated. Currently, there is considerable interest in many-branched diversity at the base station motivated by the potential for reduced interference. Glance and Greenstein [2] discuss six branches, for example. If the polarization diversity scheme can be made to work, then the physical extent of base station diversity antennas is halved relative to conventional space diversity. At the mobile, use of orthogonal polarizations only to produce diversity branches has not been very successful. Measured horizontal and vertical polarization paths between a mobile and a base station are reported to be short-term uncorrelated by Lee and Yeh [4]. Their mobile polarization diversity antenna consisted of colocated vertical electrical and magnetic (implemented as a horizontal wire loop) elements sited 1.5 h above the vehicle's conducting roof. The presence of the vehicle roof gives rise to an array pattern, which is discussed in the appendix. For an infinite, perfectly conducting groundplane, it turns out that the array patterns alone are sufficient to decorrelate the signals received by the mobile antenna elements. The mechanism of decorrelation for the signals in each polarization is the multiple reflections undertaken between the mobile and base antennas. The reflection coefficient for each polarization is in general different (cf. Fresnel's formulas), which results in the phases of orthogonal polarizations undergoing different changes for each (or at least some) of the reflections. The path of the signal occupies three dimensions and includes reflection and refraction, which causes coupling between orthogonal polarizations. After sufficient random reflections, the polarization state of the signal will be independent of the transmitted polarization. This is ideally what happens to signals propagating through an urban environment. In practice, and as noted from the measurements of Lee and Yeh, there is apparently some dependence of the received polarization on the transmitted polarization, even in urban environments. The multiple scattering cannot be sufficient for a given polarization to decouple half its power into the orthogonal polarization. Still, an antenna at the mobile need not be of pure polarization (in fact, achieving true polarization purity from an antenna mounted on an average car would be virtually impossible). A sloping monopole and the roof-mounted circular patch antenna [7] offer some possibility of designing for a given (approximate) ratio of radiated polarizations. However, this may not be necessary to get a considerable return from a polarization diversity system. There are too many unknowns regarding the propagation from the mobile to the base station to establish from theoretical considerations how well a polarization diversity sys- 0018-9545/90/08oO-0177$01 .oO 0 1990 IEEE

178 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 39, NO. 3, AUGUST 1990 I E+I2 * Y J x Fig. 1. Pseudo-three-dimensional far-field linear power patterns for a sloping monopole of length 0.6 X and elevation angle 60'. Infinite ground plane and sinusoidal current distribution is assumed. Vertical polarization is clearly dominant at lower elevation angles. tem will work, so empirical techniques are called for. The measurement scheme is relatively simple to implement and is described in the following section. The ensuing discussion includes an analysis using Rayleigh distributed signals incident on the rotatable base station antenna. This shows the tradeoff between the envelope correlation coefficient and the mean signal level different between the signals of the orthogonal polarizations, with the cross-polar discrimination as a parameter. Maximum ratio combining is assumed. 11. MEASUREMENTS AND DISCUSSION A. Measurement Setup Any two orthogonal polarizations for receiving at the base station will suffice for the measurements. There seems little justification for not using vertical and horizontal components. The transmitting mobile antenna was a sloping monopole of length 0.6 X and elevation angle 60". The antenna pattern, assuming an infinite groundplane, is depicted in Fig. 1. The polarization is dominantly vertical for the lower elevation angles. The monopole was mounted on the center of an aluminium groundplane, which was larger than 2 x 1 wavelengths at the measurement frequency of 463 MHz. The base station antenna comprised two identical fourelement dipole arrays, one for vertical and the other for horizontal polarization reception. These antennas were mounted adjacently, at a height of 100 m (-160 A), on a mast at Frejlev, Denmark. The land near the mast is basically rural. The mobile was driven in urban and suburban areas of the city of Aalborg, which is about 20 km from Frejlev. The terrain between Frejlev and Aalborg is that of slightly rolling plains with occasional foliage and dwellings. Fig. 2 is a map giving the layout around the measurement zone. The urban measurement run was taken at Vestergade, in Ngrresundby

VAUGHAN: POLARIZATION DIVERSITY IN MOBILE COMMUNICATIONS 179 Fig. 2. Map showing layout. (a) Between base station measurement site at Frejlev and urban transmitting route. (b) Between base station measurement site at Frejlev and suburban transmitting route. (Open terrain is gently rolling plains.) which is parallel to the water edge on the other side of a wide strip of water, the Limfjord, to Frejlev. The street is lined on both sides with 4-5 story buildings and runs at right angles to the direction of the base station. There is no direct line-ofsight path to the base station. The suburban measurement was taken in Hobrovej, beside the Southern Hospital. Hobrovej lies in a shadow area, caused by a hill, and this street runs at near right angles to the direction of the base station. Looking toward the base station, there is suburban housing and some large buildings on the hill. In the opposite direction, there is a row of large trees in front of suburban houses. B. Measurement Results The measured envelopes, their lossless maximum ratio combination, and the Rayleigh diagrams are given in Figs. 3 and 4 for the suburban and urban measurement runs. The noise level was chosen to be the lowest value of signal power that could be measured. The standard deviation noted in the figures is defined by U dB = 10 log (: ~ : $9, where p and U are the conventional mean and standard eviations of the envelope power. A Rayleigh signal has U dB = 4.77, and an uncorrelated two-branch maximum ratio has U dB = 3.2 1. For the suburban measurement, the horizontal polarization is limited at the noise level, restricting its dynamic range to about only 25 dB. The dynamic range of the vertically polarized signal is about 31 dB, and departs from the Rayleigh distribution at about 15 dB below its mean level. The envelope correlation coefficient is 0.02. For the urban measurements, the horizontal polarization curve follows a Rayleigh distribution over its full dynamic range of about 36 dB. The vertical polarization departs from the Rayleigh distribution at about 18 dB below its mean value. The envelope correlation coefficient is - 0.003. The results are summarized in Table I. The exact mechanism causing domination by the vertical polarization is not clear. There are several possible contributing factors: the urban and suburban environments are both rather transparent at 450 MHz so that an effective line-of-sight exists between mobile and base station; there is insufficient polarization coupling in the multipath reflections for the vertically polarized transmission to decouple into equal levels of vertical and horizontal polarizations; and the open terrain between the base station and the mobile environment plus the height of the antenna act to favor propagation of the vertical polarization to the base station. This suggests that future measurements could involve repeating the experiment with mobile antennas of varying polarization. The mobile antenna need not be in a vehicle- a simple handheld apparatus could be sufficient to find some useful information. It would also be useful to compare these measurements with those taken with the base station sited in the same urban environment as the transmitting mobile, rather than being well separated from the urban area. In the Rayleigh diagrams of Figs. 3 and 4, the reference (SNR) is that of the vertically polarized signal. The diversity gain, which is referred to as a “vertical-vertical’’ system, can be read directly off the diagrams. The curves have not been smoothed; they represent a direct mapping from the finite number (4600) measurement points. For the suburban measurements, the diversity gain is less than l dB at the 80% level (i.e., for 80% of the time), and increases to about 3 dB at the 95% level, and nearly 5 dB at the 95% level (a smoothed curve for the distribution is imagined). In the urban measurements, there is over 3 dB diversity gain at the 80% level, increasing to nearly 7 dB at the 99.5% level. In practice, there will be combination losses of perhaps 1 dB, so that the return from polarization diversity in suburban environments is negligible for this definition of diversity gain

VAUGHAN: POLARIZATION DIVERSITY IN MOBILE COMMUNICATIONS 181 ISNRI/<SNR> IN OB. (b) Fig. 3. (Continued) In both the urban and suburban measurements, the received horizontal polarization appears genuinely Rayleighdistributed. The vertical polarization looks more Rician, caused by a dominant component, as discussed in the preceding. The lack of a dominant component in the horizontal polarization is quite reasonable, since essentially all the received power results from polarization coupling by reflection and refraction, of which no individual contribution continually dominates. It is noteworthy that the figure found here for the polarization cross-coupling agrees with that of Lee and Yeh [4]. This indicates that the effects of the open terrain may not be particularly relevant. The figure measured here also agrees with the measurements of Kozono et al. [3], albeit in an indirect sense (see Table I), further supporting this possibility. C. Discussion In Kozono et al.'s [3] results, the base station was in urban Tokyo. Their base station antenna was arranged so that the received polarizations were f 45", in order to equalize the received mean signal levels. This equalization works, of course, but only at the expense of raising the correlation coefficient. Note that the total energy in the incident signal is being collected, assuming maximum ratio combining, irrespective of the angles (or senses) of the orthogonal polarizations. A trade-off between the mean level difference and the branch signal correlations is evident. An investigation similar to that of Kozono et al. [3] can be arranged by assuming an incident signal of horizontal polarization and vertical polarization in which rl and r2 are Rayleigh-distributed and uncorrelated. 41 and 42 are assumed to be random, uniformly distributed, and uncorrelated. An antenna receiving orthogonal linear polarizations, such as crossed dipoles broadside to the incident propagation vector (the situation is depicted in Fig. 5(a), receives voltages proportional to VI =Ey cos a +Ex sin a (3) = (r2 cos a cos 42 + rl sin a cos 41) cos ut - (r2 cos a sin 42 + rl sin a sin 41) sin ut (4) and V2 =Ey sin a -E, cos a (5) = (r2 sin a cos 42 + rl sin a cos 41) cos at - (r2 cos a sin 42 - rl sin a sin 41) sin ut (6) whose respective envelopes are RI = [r,' cos2 a + r: sin2 a +2r1r2 cos a sin a cos(41 - 42)1'/~ (7) and R2 = [r: cos2 a +ri sin2 a +2rlr2 cos a sin a cos(41 - 42)]1/2. (8) The problem of the square root can be circumvented by considering the power correlation coefficient (known to be similar to the envelope correlation coefficient [6]) The required moments and moment products are (R:) = (r,' cos2 a + r: sin2 a + 2r1r2 cos a sin a cos (41 - 42)) (R:)(R,~) = (r:)(t,2)(cos4 a + sin4 a) + cos2 a sin2 a( (r:)2 + (r:)2); (13)