Propagation Characteristics of Polarized Radio ... - Semantic Scholar
Propagation characteristics of polarized radio waves in cellular communications Henrik Asplund, Jan-Erik Berg, Fredrik Harrysson, Jonas Medbo, Mathias Riback Ericsson Research Ericsson AB Stockholm, Sweden {firstname.lastname}@ericsson.com Abstract— Narrowband and wideband measurements of the radio channel using different combinations of transmit and receive polarization have been performed. The measurements cover a range of scenarios including urban, suburban and open terrain, as well as both outdoor and indoor terminals. The vertical-to-vertical (V-V) and horizontal-to-horizontal (H-H) polarization combinations are found to provide equal received power on average, while the cross-polarized combinations (V-H) and (H-V) typically provide 5-15 dB weaker received power due to the limited amount of cross-polarization scattering in the radio channel. Fast fading variations are further found to be uncorrelated between different combinations of transmit and receive polarization.
I. INTRODUCTION A plane radio wave has two degrees of freedom, commonly expressed as the polarization of the wave. These degrees of freedom can be exploited to provide diversity and therefore resilience towards fading dips in the radio channel. Such polarization diversity is common at the base station in cellular communication systems, but have until recently not been considered practical or cost-efficient in the mobile station. However, this has changed since the discovery of the possibility of large capacity gains by using multiple antennas at both transmitter and receiver, so called MIMO (Multiple Input Multiple Output) techniques. Utilizing polarization is now seen as a promising way to fit multiple antennas into a small terminal, thereby enabling higher bit rates. The propagation characteristics of differently polarized radio waves have been characterized in several measurement series [1]-[6], however in some of the measurements the orthogonally polarized antennas have had non-equal radiation patterns. An example is the commonly used two crossed electrical dipoles, which have the nulls of the respective antenna patterns oriented in different directions and where the polarizations are not orthogonal over the whole sphere. These differences in the antenna patterns and polarizations might be an explanation to the sometimes contradictory conclusions from the earlier measurements. The main source of confusion is whether vertically and horizontally polarized waves suffer the same path loss [1][2] or not [3],[5] in macrocellular environments.
1-4244-0264-6/07/$25.00 ©2007 IEEE
Given these uncertainties, it was decided to conduct a series of measurements to characterize the propagation of polarized radio waves in a cellular communication system. The purpose of the investigation was three-fold, namely: •
Determine if there is a difference in path loss for vertically and horizontally polarized waves in a normal cellular deployment
•
Gather statistics of the cross-polarization scattering, Vertical-to-Horizontal and Horizontal-to-Vertical, to be able to parameterize polarization models
•
Determine the fast fading correlations among different combinations of transmit and receive polarization The remainder of this paper deals with the measurements, the analysis of results and discusses the conclusions that can be drawn from the observations. II.
POLARIZATION MEASUREMENTS
A. Equipment As the purpose of the measurements was to isolate and characterize the polarization transfer in the wireless medium, special care was needed to ensure that the transmit and receive antennas did not invalidate the data by introducing polarization cross-coupling or power imbalances that could erroneously be interpreted as being part of the wireless channel. A dualpolarized (0°/90°) base station antenna with a 65° half-power beam-width in azimuth and a 6° electrical down-tilt was used at the transmitter. This antenna was carefully characterized in an anechoic chamber in order to ensure good cross-polar discrimination (XPD) and very small differences in radiation patterns between the two polarizations over the 120° sector of coverage. It was established that the radiated power on the two polarizations differed by less than 0.5 dB and that the XPD was in excess of 15 dB throughout the sector. Several dual-polarized antenna designs were considered for the mobile station; however these were all discarded due to the difficulty in getting orthogonal polarizations with omnidirectional radiation characteristics. Instead, two spatially separated single-polarized antennas were used, one being a λ/2 electric dipole and the other a magnetic dipole. For the majority of the measurements, a –3 dBi cloverleaf antenna [7] was used to approximate a magnetic dipole. However, this particular antenna was found to have a poor XPD and therefore some of the outdoor measurements were subsequently repeated using a
839
The measurements were conducted by transmitting 1 W continuous wave (CW) signals at 1877.5 MHz from the two transmit antennas. A frequency offset of 500 Hz between the two antennas was used to allow the receivers to distinguish between the two transmitted polarizations. The two receive antennas were connected to the two ports of a network analyzer in time sweep mode. Subsequent offline processing allowed the four possible polarization combinations to be extracted: sVV (t ) sVH (t ) H(t ) = sHV (t ) sHH (t )
(1)
Here the first sub-index denotes the base station (transmit) polarization and the second denotes the mobile station (receive) polarization. In the measurements with the cloverleaf antenna, sVH was corrupted due to poor antenna XPD and therefore unusable and the data has therefore been removed from the analysis. This problem did not exist when using the reference magnetic dipole. Continuous measurements of H(t) for hours of time were possible. B. Set-up and environment Both outdoor and indoor measurements were performed in and around Kista in Stockholm, Sweden. The base station was mounted on a 1.5m high tripod placed on the roof of an 8storey office building. Three different pointing directions approximating the sectors of a three-sector site were used to ensure that each measurement route was mainly within the main lobe coverage of the BS antenna. For the outdoor measurements, the receiver was installed in a van and the two receive antennas were mounted about 1m over the car roof to minimize the influence of the metallic car on the radiation patterns. Measurements were recorded along several routes totaling tens of kilometers of traveled distance through residential, commercial and rural areas at distances from the base station ranging from 100 m to 4500 m. The van was driven according to the traffic conditions in the area, with speeds ranging from standstill to 70 km/h (19 m/s) although the most common speed was approximately 30 km/h (8 m/s). Less than 10% of the results were obtained in line of sight conditions. The indoor measurements were conducted inside three office buildings within 300 m of the base station. Two of the
buildings consist mainly of long corridors with smaller office rooms with plasterboard walls between the rooms, glass walls and a wooden door towards the corridor, and reinforced concrete on the exterior side. The third and most distant building differs by having an open office-landscape with few interior walls, and heat-protective window coating on the external windows resulting in more than 20 dB attenuation of the radio waves. Measurements were recorded along numerous routes on different floors and sections of the buildings, at walking speeds ranging from standstill to roughly 1 m/s. III.
ANALYSIS
A. Average power for various polarization combinations The local average of the polarization transfer matrix H(t) was calculated as pVV P= pHV
{ {
2 pVH E sVV (t ) = 2 pHH E sHV (t )
} E{s } E{s
VH HH
(t ) 2 }
(2)
(t ) } 2
with the expectation taken over a time long enough to suppress fast fading variations but short enough to not remove the slower shadow fading. An averaging time of one second was used for the outdoor data, while 2.5 seconds was found suitable for the indoor data due to the lower speed. This gives different averaging lengths for different parts of the measurement routes due to the varying speed of the mobile, however the convenience of a time-based averaging was found to outweigh this disadvantage. The averaging length for the typical outdoor speed (30 km/h) was 50λ, while for the typical indoor speed (1 m/s) it was 15λ. Fig. 1 shows an example of the local averaged power along one of the outdoor measurement routes for the four possible combinations of transmit and receive polarization. -70 p HH pVH p HV p
-80 Path gain [dB]
high quality reference magnetic dipole. The spatial separation was introduced to minimize coupling between antennas, and the resulting disturbance of the omni-directional characteristics was therefore considered to be minimal. A separation of about 1 m (roughly 6λ) was used, except for the measurements with the reference magnetic dipole where the separation due to a different mounting was 30 cm (~2λ).The electric dipole and the reference magnetic dipole had very similar radiation patterns, with less than 0.5 dB difference over the range of azimuth and elevation angles of interest and an XPD in excess of 20 dB. In contrast, the cloverleaf antenna had an XPD of about 6 dB and also some ripple in the radiation pattern. Still, with the help of careful calibration it was possible to extract useful results also from this antenna.
VV
-90
-100
-110
-120 0
50
100
150 Time [s]
200
250
300
Figure 1. Average path gain (relative received power) per combination of transmit and receive polarization, from a representative measurement route where the reference magnetic dipole was used.
840
The behavior is very similar on all outdoor and indoor routes, and can be summarized as follows:
Locally there can be significant differences, as at about 180s in Figure 1 where pHH is almost 6 dB stronger than pVV
•
The two cross-polarized components, pVH and pHV, have the same power on average.
•
The cross-polarized components, pHV and pVH, are heavily suppressed, usually by 5-15 dB, compared to the two co-polarized components Results regarding the component pVH were only available in outdoor measurements where the reference magnetic dipole was used. While it is well-known that parallel and perpendicular polarizations have different coefficients for transmission, reflection, and diffraction, the average received power in the cellular scenario seems to be independent of whether a vertically or a horizontally polarized wave was transmitted. It is not obvious why, but it can be speculated that the differences tend to average out on multipath channels. This lack of difference between vertically and horizontally polarized waves is in contrast to some earlier results [3][5] but confirms others [1][2]. On the other hand, the strong attenuation of the crosspolarization scattering is easier to understand. Most man-made surfaces tend to be either horizontal or vertical, and thus will preserve the polarization of an impinging horizontally or vertically polarized wave that is propagating in either the vertical or the horizontal plane. Slanted or irregular surfaces cause cross-polar scattering, but such surfaces are relatively rare in the considered deployment. Table I lists the measured power differences, averaged over all measurement routes. For comparison, the values reported in earlier work has also been listed. The grey values are from indirect measurements. Further analysis of both the indoor and the outdoor data showed that the suppression of the cross-polarization component depends on the total received power, such that a larger path loss leads to a higher cross-polarization scattering. This is exemplified in Figure 2, where it can be seen how the ratio between the cross-polarized component pHV and pVV varies from -12 dB at higher total received powers down to about -4 dB at the lower power range. In contrast, the ratio between pHH and pVV remains close to zero for all values of pVV. TABLE I.
Outdoor routes Indoor routes Lee [1] Turkmani [2] Lotse [3] Kaunilainen [5] Knudsen [6]
AVERAGE PATH GAIN DIFFERENCES BETWEEN CO- AND CROSS -POLARIZED COMPONENTS Average of pHH/ pVV [dB] 0.6 0.1 0 ~0 [-6,-11] [-1.6,2.2]
Average of pHV/ pVV [dB] -8.9 -6.9 -6 [-10.8,-13.3] [-7,-13] [-2.3,-9.5] -5.5
[dB]
•
5
VV
The two co-polar channel components, pVV and pHH, have the same power on average
Path gain relative to p
•
0 -5
-10 -15
p /p HV VV p /p HH
-20 -150
VV
-140 -130 -120 -110 -100 -90 Path gain for co-polar vertical, p [dB] VV
Figure 2. Co- and cross-polar powers as a function of total received power (here given by pVV). Error bars denote +- one standard deviation per 5 dB pVV interval. Results are from all indoor measurements.
As can also be seen in Fig. 2, the ratio between the local average powers of two polarization components has a standard deviation of about 3 dB. This quantifies the observation of local differences in power from Fig. 1. These local power differences remain for several seconds even at driving speeds, corresponding to tens of meters of travel. Since the duration of an imbalance between e.g. pHH and pVV can be quite long, even co-polarized vertically or horizontally polarized antennas at transmitter and receiver is not a guarantee for optimal power transfer. The ability to adapt to whichever of pHH and pVV is currently strongest would be necessary, requiring polarization adaptability at both transmitter and receiver. B. Fast fading characteritics The fast fading variations of the elements of H(t) were found to follow the same statistics – in most cases Rayleigh fading was observed. Fig. 3 shows an example of the fast fading processes over a limited segment, where the four coefficients appear to undergo independent Rayleigh fading. The independence is confirmed by a correlation analysis, where the estimated correlation coefficients are summarized in Table II. Mean envelope correlations are invariably low, however both the spatial separation of the two terminal antennas as well as the orthogonal polarizations used at transmitter and receiver are possible explanations. The spatial separation of the terminal antennas is in itself sufficient to result in uncorrelated fast fading. Only in the case where coefficients correspond to the same terminal antenna can the lack of correlation uniquely be attributed to the independence of the polarization channels. Such correlation coefficients are indicated by bold-faced values in the table. In these cases it is evident that changing the base station polarization results in an uncorrelated fast fading process.
841
-80
0
Relative power [dB]
-90 Path gain [dB]
pVV(τ) pHH(τ) pHV(τ) pVH(τ)
-5
-100 -110
s VV sHH s VH s
-120
HV
-130 0
0.05
0.1 Time [s]
0.15
-10 -15 -20 -25 -30 -35
0.2
-40 -45
Figure 3. Example of fast fading variations on a short segment of an outdoor route.
TABLE II.
ρ(|s1|,|s2|) shh svh shv svv
MEAN FAST FADING ENVELOPE CORRELATIONS (OUTDOOR ROUTES)
shh 1 0.02 0.05 0.07
svh 0.02 1 0 0.07
shv 0.05 0 1 0.02
svv 0.07 0.07 0.02 1
C. Wideband polarization characteristics Wideband polarimetric channel measurements such as those reported in [8] provide further insight into the propagation mechanisms that give rise to the observed signal strength differences among various polarization combinations at transmitter and receiver. An example of how the power of the co- and cross-polarized components vary as a function of the propagation delay is given by the power delay profiles shown in Fig. 4. For shorter delays, where the waves carrying most power are received, there are quite rapid fluctuations, both in the total power as well as in the relation of co-polarized to cross-polarized power. For some peaks it is even so that the cross-polarized components are stronger than one of the copolarized, as for e.g. the first arriving peak seen near 200 ns in Fig. 4. Many of these peaks can be attributed to a few or only a single interaction along the propagation path. Some of these interactions have been identified as scattering from irregular objects and diffraction at large angles. In contrast, the behavior at larger delays shows an almost constant offset between coand cross-polarized components and minimal difference between the two possible co- and cross-polarized components respectively. These components are more likely a result of multiple specular reflections. The total average power per polarization component, integrated over all delays, is similar to that reported in the narrowband measurements described above. For a further analysis, including the identification of individual propagation paths and corresponding scattering objects, the reader is referred to [8].
200
400
600 800 1000 1200 1400 Time delay, τ [ns]
Figure 4. Power delay profiles from a 200 MHz bandwidth macrocell outdoor measurement.
IV. DISCUSSION The results presented above show that vertical and horizontal polarizations are mainly preserved during propagation from the transmitter to the receiver. This poses the question if this property is unique to vertical and horizontal polarization, or if any polarization will be preserved. To answer this question, first consider how the received polarization r = [rv rh]T can be expressed as
r = H ⋅t
(3)
where H is the polarization transfer matrix defined in (1), and t = [tv th]T is the transmit polarization vector. It has been shown that the off-diagonal elements of H are quite weak, and that each element experiences uncorrelated fading and therefore will have a random phase. It is therefore possible to approximate H as
1 0 H≈ 0 e jϕ
(4)
where ϕ is a random angle. Only the eigenvectors of (4) will be preserved, and these are simply vertical, tv = [1 0]T, and horizontal, th = [0 1]T, polarization. Other polarizations, such as slant 45°, t45 = [1 1]T, or circular, tRHCP = [1 i]T, will, depending on the random phase shift ϕ, be transformed into other polarizations including an orthogonal polarization when ϕ is equal to π. This is true also if amplitude fading of the components of H is included in the analysis, as the eigenvectors are unchanged. Therefore, it can be concluded from the analysis of these measurements that horizontal and vertical polarizations are the only polarizations that are preserved during propagation in a cellular environment.
842
V. CONCLUSIONS A series of narrowband and wideband measurements have been performed to characterize how polarized radio waves propagate through an environment typical of a cellular system deployment. The subsequent analysis supports the following observations: •
There is no fast fading correlation between signals received on (or transmitted from) vertically and horizontally polarized BS antennas, irrespective of the MS polarization.
•
Vertically and horizontally polarized radio waves experience the same path loss on average.
•
The amount of power that is scattered from vertical to horizontal polarization or vice versa is suppressed by on average 5-15 dB compared to co-polar propagation. The cross-polar scattering increases with increasing path loss.
•
Local variations of the co- and cross-polar power are much slower than the fast fading variations, and have a standard deviation of around 3 dB.
•
Vertical and horizontal polarizations are the only polarizations that are mainly preserved during propagation in the cellular environment.
REFERENCES [1]
[2]
[3]
[4]
[5]
[6]
[7] [8]
W. C. Y. Lee and Y. S. Yeh, “Polarization diversity system for mobile radio”, IEEE Transactions on Communications, Vol. COM-20, No. 5, pp. 912-923, Oct 1972. A. M. D. Turkmani, A. A. Arowojolu, P. A. Jefford and C. J. Kellet, “An experimental evaluation of the performance of two-branch space and polarization diversity schemes at 1800 MHz”, IEEE Transactions on Vehicular Technology, Vol. 44, No. 2, pp. 318-326, May 1995. F. Lotse, J-E. Berg, U. Forssén and P. Idahl, “Base station polarization diversity reception in macrocellular systems at 1800 MHz”, Proc. IEEE Vehicular Technology Conference (VTC’96), Atlanta, pp. 1643-1646, May 1996. T. B. Sørensen, A. Ø. Nielsen, P. E. Mogensen, M. Tolstrup and K. Steffensen, “Performance of two-branch polarisation antenna diversity in an operational GSM network”, Proc. IEEE Vehicular Technology Conference (VTC’98), Ottawa, Canada, pp. 741-746, May 1998. A. Kaunilainen, L. Vuokko and P. Vainikainen, “Polarization behaviour in different urban radio environments at 5.3 GHz”, COST 273 TD(05)018, Bologna, Italy, Jan 2005. M. B. Knudsen and G. F. Pedersen, “Spherical outdoor to indoor power spectrum model at the mobile terminal”, IEEE Journal on Selected Areas in Communications, Vol. 20, No. 6, pp. 1156-1169, Aug. 2002. P. H. Smith, ““Cloverleaf” antenna for F.M. Broadcasting”, Proc. of the IRE, Vol. 35, Issue 12, pp. 1556-1563, Dec. 1947. J. Medbo, M. Riback, H. Asplund and J-E. Berg, “MIMO channel characteristics in a small macrocell”, Proc. IEEE Vehicular Technology Conference (VTC’2005Fall), Dallas, Sept.2005.
843
I. INTRODUCTION A plane radio wave has two degrees of freedom, commonly expressed as the polarization of the wave. These degrees of freedom can be exploited to provide diversity and therefore resilience towards fading dips in the radio channel. Such polarization diversity is common at the base station in cellular communication systems, but have until recently not been considered practical or cost-efficient in the mobile station. However, this has changed since the discovery of the possibility of large capacity gains by using multiple antennas at both transmitter and receiver, so called MIMO (Multiple Input Multiple Output) techniques. Utilizing polarization is now seen as a promising way to fit multiple antennas into a small terminal, thereby enabling higher bit rates. The propagation characteristics of differently polarized radio waves have been characterized in several measurement series [1]-[6], however in some of the measurements the orthogonally polarized antennas have had non-equal radiation patterns. An example is the commonly used two crossed electrical dipoles, which have the nulls of the respective antenna patterns oriented in different directions and where the polarizations are not orthogonal over the whole sphere. These differences in the antenna patterns and polarizations might be an explanation to the sometimes contradictory conclusions from the earlier measurements. The main source of confusion is whether vertically and horizontally polarized waves suffer the same path loss [1][2] or not [3],[5] in macrocellular environments.
1-4244-0264-6/07/$25.00 ©2007 IEEE
Given these uncertainties, it was decided to conduct a series of measurements to characterize the propagation of polarized radio waves in a cellular communication system. The purpose of the investigation was three-fold, namely: •
Determine if there is a difference in path loss for vertically and horizontally polarized waves in a normal cellular deployment
•
Gather statistics of the cross-polarization scattering, Vertical-to-Horizontal and Horizontal-to-Vertical, to be able to parameterize polarization models
•
Determine the fast fading correlations among different combinations of transmit and receive polarization The remainder of this paper deals with the measurements, the analysis of results and discusses the conclusions that can be drawn from the observations. II.
POLARIZATION MEASUREMENTS
A. Equipment As the purpose of the measurements was to isolate and characterize the polarization transfer in the wireless medium, special care was needed to ensure that the transmit and receive antennas did not invalidate the data by introducing polarization cross-coupling or power imbalances that could erroneously be interpreted as being part of the wireless channel. A dualpolarized (0°/90°) base station antenna with a 65° half-power beam-width in azimuth and a 6° electrical down-tilt was used at the transmitter. This antenna was carefully characterized in an anechoic chamber in order to ensure good cross-polar discrimination (XPD) and very small differences in radiation patterns between the two polarizations over the 120° sector of coverage. It was established that the radiated power on the two polarizations differed by less than 0.5 dB and that the XPD was in excess of 15 dB throughout the sector. Several dual-polarized antenna designs were considered for the mobile station; however these were all discarded due to the difficulty in getting orthogonal polarizations with omnidirectional radiation characteristics. Instead, two spatially separated single-polarized antennas were used, one being a λ/2 electric dipole and the other a magnetic dipole. For the majority of the measurements, a –3 dBi cloverleaf antenna [7] was used to approximate a magnetic dipole. However, this particular antenna was found to have a poor XPD and therefore some of the outdoor measurements were subsequently repeated using a
839
The measurements were conducted by transmitting 1 W continuous wave (CW) signals at 1877.5 MHz from the two transmit antennas. A frequency offset of 500 Hz between the two antennas was used to allow the receivers to distinguish between the two transmitted polarizations. The two receive antennas were connected to the two ports of a network analyzer in time sweep mode. Subsequent offline processing allowed the four possible polarization combinations to be extracted: sVV (t ) sVH (t ) H(t ) = sHV (t ) sHH (t )
(1)
Here the first sub-index denotes the base station (transmit) polarization and the second denotes the mobile station (receive) polarization. In the measurements with the cloverleaf antenna, sVH was corrupted due to poor antenna XPD and therefore unusable and the data has therefore been removed from the analysis. This problem did not exist when using the reference magnetic dipole. Continuous measurements of H(t) for hours of time were possible. B. Set-up and environment Both outdoor and indoor measurements were performed in and around Kista in Stockholm, Sweden. The base station was mounted on a 1.5m high tripod placed on the roof of an 8storey office building. Three different pointing directions approximating the sectors of a three-sector site were used to ensure that each measurement route was mainly within the main lobe coverage of the BS antenna. For the outdoor measurements, the receiver was installed in a van and the two receive antennas were mounted about 1m over the car roof to minimize the influence of the metallic car on the radiation patterns. Measurements were recorded along several routes totaling tens of kilometers of traveled distance through residential, commercial and rural areas at distances from the base station ranging from 100 m to 4500 m. The van was driven according to the traffic conditions in the area, with speeds ranging from standstill to 70 km/h (19 m/s) although the most common speed was approximately 30 km/h (8 m/s). Less than 10% of the results were obtained in line of sight conditions. The indoor measurements were conducted inside three office buildings within 300 m of the base station. Two of the
buildings consist mainly of long corridors with smaller office rooms with plasterboard walls between the rooms, glass walls and a wooden door towards the corridor, and reinforced concrete on the exterior side. The third and most distant building differs by having an open office-landscape with few interior walls, and heat-protective window coating on the external windows resulting in more than 20 dB attenuation of the radio waves. Measurements were recorded along numerous routes on different floors and sections of the buildings, at walking speeds ranging from standstill to roughly 1 m/s. III.
ANALYSIS
A. Average power for various polarization combinations The local average of the polarization transfer matrix H(t) was calculated as pVV P= pHV
{ {
2 pVH E sVV (t ) = 2 pHH E sHV (t )
} E{s } E{s
VH HH
(t ) 2 }
(2)
(t ) } 2
with the expectation taken over a time long enough to suppress fast fading variations but short enough to not remove the slower shadow fading. An averaging time of one second was used for the outdoor data, while 2.5 seconds was found suitable for the indoor data due to the lower speed. This gives different averaging lengths for different parts of the measurement routes due to the varying speed of the mobile, however the convenience of a time-based averaging was found to outweigh this disadvantage. The averaging length for the typical outdoor speed (30 km/h) was 50λ, while for the typical indoor speed (1 m/s) it was 15λ. Fig. 1 shows an example of the local averaged power along one of the outdoor measurement routes for the four possible combinations of transmit and receive polarization. -70 p HH pVH p HV p
-80 Path gain [dB]
high quality reference magnetic dipole. The spatial separation was introduced to minimize coupling between antennas, and the resulting disturbance of the omni-directional characteristics was therefore considered to be minimal. A separation of about 1 m (roughly 6λ) was used, except for the measurements with the reference magnetic dipole where the separation due to a different mounting was 30 cm (~2λ).The electric dipole and the reference magnetic dipole had very similar radiation patterns, with less than 0.5 dB difference over the range of azimuth and elevation angles of interest and an XPD in excess of 20 dB. In contrast, the cloverleaf antenna had an XPD of about 6 dB and also some ripple in the radiation pattern. Still, with the help of careful calibration it was possible to extract useful results also from this antenna.
VV
-90
-100
-110
-120 0
50
100
150 Time [s]
200
250
300
Figure 1. Average path gain (relative received power) per combination of transmit and receive polarization, from a representative measurement route where the reference magnetic dipole was used.
840
The behavior is very similar on all outdoor and indoor routes, and can be summarized as follows:
Locally there can be significant differences, as at about 180s in Figure 1 where pHH is almost 6 dB stronger than pVV
•
The two cross-polarized components, pVH and pHV, have the same power on average.
•
The cross-polarized components, pHV and pVH, are heavily suppressed, usually by 5-15 dB, compared to the two co-polarized components Results regarding the component pVH were only available in outdoor measurements where the reference magnetic dipole was used. While it is well-known that parallel and perpendicular polarizations have different coefficients for transmission, reflection, and diffraction, the average received power in the cellular scenario seems to be independent of whether a vertically or a horizontally polarized wave was transmitted. It is not obvious why, but it can be speculated that the differences tend to average out on multipath channels. This lack of difference between vertically and horizontally polarized waves is in contrast to some earlier results [3][5] but confirms others [1][2]. On the other hand, the strong attenuation of the crosspolarization scattering is easier to understand. Most man-made surfaces tend to be either horizontal or vertical, and thus will preserve the polarization of an impinging horizontally or vertically polarized wave that is propagating in either the vertical or the horizontal plane. Slanted or irregular surfaces cause cross-polar scattering, but such surfaces are relatively rare in the considered deployment. Table I lists the measured power differences, averaged over all measurement routes. For comparison, the values reported in earlier work has also been listed. The grey values are from indirect measurements. Further analysis of both the indoor and the outdoor data showed that the suppression of the cross-polarization component depends on the total received power, such that a larger path loss leads to a higher cross-polarization scattering. This is exemplified in Figure 2, where it can be seen how the ratio between the cross-polarized component pHV and pVV varies from -12 dB at higher total received powers down to about -4 dB at the lower power range. In contrast, the ratio between pHH and pVV remains close to zero for all values of pVV. TABLE I.
Outdoor routes Indoor routes Lee [1] Turkmani [2] Lotse [3] Kaunilainen [5] Knudsen [6]
AVERAGE PATH GAIN DIFFERENCES BETWEEN CO- AND CROSS -POLARIZED COMPONENTS Average of pHH/ pVV [dB] 0.6 0.1 0 ~0 [-6,-11] [-1.6,2.2]
Average of pHV/ pVV [dB] -8.9 -6.9 -6 [-10.8,-13.3] [-7,-13] [-2.3,-9.5] -5.5
[dB]
•
5
VV
The two co-polar channel components, pVV and pHH, have the same power on average
Path gain relative to p
•
0 -5
-10 -15
p /p HV VV p /p HH
-20 -150
VV
-140 -130 -120 -110 -100 -90 Path gain for co-polar vertical, p [dB] VV
Figure 2. Co- and cross-polar powers as a function of total received power (here given by pVV). Error bars denote +- one standard deviation per 5 dB pVV interval. Results are from all indoor measurements.
As can also be seen in Fig. 2, the ratio between the local average powers of two polarization components has a standard deviation of about 3 dB. This quantifies the observation of local differences in power from Fig. 1. These local power differences remain for several seconds even at driving speeds, corresponding to tens of meters of travel. Since the duration of an imbalance between e.g. pHH and pVV can be quite long, even co-polarized vertically or horizontally polarized antennas at transmitter and receiver is not a guarantee for optimal power transfer. The ability to adapt to whichever of pHH and pVV is currently strongest would be necessary, requiring polarization adaptability at both transmitter and receiver. B. Fast fading characteritics The fast fading variations of the elements of H(t) were found to follow the same statistics – in most cases Rayleigh fading was observed. Fig. 3 shows an example of the fast fading processes over a limited segment, where the four coefficients appear to undergo independent Rayleigh fading. The independence is confirmed by a correlation analysis, where the estimated correlation coefficients are summarized in Table II. Mean envelope correlations are invariably low, however both the spatial separation of the two terminal antennas as well as the orthogonal polarizations used at transmitter and receiver are possible explanations. The spatial separation of the terminal antennas is in itself sufficient to result in uncorrelated fast fading. Only in the case where coefficients correspond to the same terminal antenna can the lack of correlation uniquely be attributed to the independence of the polarization channels. Such correlation coefficients are indicated by bold-faced values in the table. In these cases it is evident that changing the base station polarization results in an uncorrelated fast fading process.
841
-80
0
Relative power [dB]
-90 Path gain [dB]
pVV(τ) pHH(τ) pHV(τ) pVH(τ)
-5
-100 -110
s VV sHH s VH s
-120
HV
-130 0
0.05
0.1 Time [s]
0.15
-10 -15 -20 -25 -30 -35
0.2
-40 -45
Figure 3. Example of fast fading variations on a short segment of an outdoor route.
TABLE II.
ρ(|s1|,|s2|) shh svh shv svv
MEAN FAST FADING ENVELOPE CORRELATIONS (OUTDOOR ROUTES)
shh 1 0.02 0.05 0.07
svh 0.02 1 0 0.07
shv 0.05 0 1 0.02
svv 0.07 0.07 0.02 1
C. Wideband polarization characteristics Wideband polarimetric channel measurements such as those reported in [8] provide further insight into the propagation mechanisms that give rise to the observed signal strength differences among various polarization combinations at transmitter and receiver. An example of how the power of the co- and cross-polarized components vary as a function of the propagation delay is given by the power delay profiles shown in Fig. 4. For shorter delays, where the waves carrying most power are received, there are quite rapid fluctuations, both in the total power as well as in the relation of co-polarized to cross-polarized power. For some peaks it is even so that the cross-polarized components are stronger than one of the copolarized, as for e.g. the first arriving peak seen near 200 ns in Fig. 4. Many of these peaks can be attributed to a few or only a single interaction along the propagation path. Some of these interactions have been identified as scattering from irregular objects and diffraction at large angles. In contrast, the behavior at larger delays shows an almost constant offset between coand cross-polarized components and minimal difference between the two possible co- and cross-polarized components respectively. These components are more likely a result of multiple specular reflections. The total average power per polarization component, integrated over all delays, is similar to that reported in the narrowband measurements described above. For a further analysis, including the identification of individual propagation paths and corresponding scattering objects, the reader is referred to [8].
200
400
600 800 1000 1200 1400 Time delay, τ [ns]
Figure 4. Power delay profiles from a 200 MHz bandwidth macrocell outdoor measurement.
IV. DISCUSSION The results presented above show that vertical and horizontal polarizations are mainly preserved during propagation from the transmitter to the receiver. This poses the question if this property is unique to vertical and horizontal polarization, or if any polarization will be preserved. To answer this question, first consider how the received polarization r = [rv rh]T can be expressed as
r = H ⋅t
(3)
where H is the polarization transfer matrix defined in (1), and t = [tv th]T is the transmit polarization vector. It has been shown that the off-diagonal elements of H are quite weak, and that each element experiences uncorrelated fading and therefore will have a random phase. It is therefore possible to approximate H as
1 0 H≈ 0 e jϕ
(4)
where ϕ is a random angle. Only the eigenvectors of (4) will be preserved, and these are simply vertical, tv = [1 0]T, and horizontal, th = [0 1]T, polarization. Other polarizations, such as slant 45°, t45 = [1 1]T, or circular, tRHCP = [1 i]T, will, depending on the random phase shift ϕ, be transformed into other polarizations including an orthogonal polarization when ϕ is equal to π. This is true also if amplitude fading of the components of H is included in the analysis, as the eigenvectors are unchanged. Therefore, it can be concluded from the analysis of these measurements that horizontal and vertical polarizations are the only polarizations that are preserved during propagation in a cellular environment.
842
V. CONCLUSIONS A series of narrowband and wideband measurements have been performed to characterize how polarized radio waves propagate through an environment typical of a cellular system deployment. The subsequent analysis supports the following observations: •
There is no fast fading correlation between signals received on (or transmitted from) vertically and horizontally polarized BS antennas, irrespective of the MS polarization.
•
Vertically and horizontally polarized radio waves experience the same path loss on average.
•
The amount of power that is scattered from vertical to horizontal polarization or vice versa is suppressed by on average 5-15 dB compared to co-polar propagation. The cross-polar scattering increases with increasing path loss.
•
Local variations of the co- and cross-polar power are much slower than the fast fading variations, and have a standard deviation of around 3 dB.
•
Vertical and horizontal polarizations are the only polarizations that are mainly preserved during propagation in the cellular environment.
REFERENCES [1]
[2]
[3]
[4]
[5]
[6]
[7] [8]
W. C. Y. Lee and Y. S. Yeh, “Polarization diversity system for mobile radio”, IEEE Transactions on Communications, Vol. COM-20, No. 5, pp. 912-923, Oct 1972. A. M. D. Turkmani, A. A. Arowojolu, P. A. Jefford and C. J. Kellet, “An experimental evaluation of the performance of two-branch space and polarization diversity schemes at 1800 MHz”, IEEE Transactions on Vehicular Technology, Vol. 44, No. 2, pp. 318-326, May 1995. F. Lotse, J-E. Berg, U. Forssén and P. Idahl, “Base station polarization diversity reception in macrocellular systems at 1800 MHz”, Proc. IEEE Vehicular Technology Conference (VTC’96), Atlanta, pp. 1643-1646, May 1996. T. B. Sørensen, A. Ø. Nielsen, P. E. Mogensen, M. Tolstrup and K. Steffensen, “Performance of two-branch polarisation antenna diversity in an operational GSM network”, Proc. IEEE Vehicular Technology Conference (VTC’98), Ottawa, Canada, pp. 741-746, May 1998. A. Kaunilainen, L. Vuokko and P. Vainikainen, “Polarization behaviour in different urban radio environments at 5.3 GHz”, COST 273 TD(05)018, Bologna, Italy, Jan 2005. M. B. Knudsen and G. F. Pedersen, “Spherical outdoor to indoor power spectrum model at the mobile terminal”, IEEE Journal on Selected Areas in Communications, Vol. 20, No. 6, pp. 1156-1169, Aug. 2002. P. H. Smith, ““Cloverleaf” antenna for F.M. Broadcasting”, Proc. of the IRE, Vol. 35, Issue 12, pp. 1556-1563, Dec. 1947. J. Medbo, M. Riback, H. Asplund and J-E. Berg, “MIMO channel characteristics in a small macrocell”, Proc. IEEE Vehicular Technology Conference (VTC’2005Fall), Dallas, Sept.2005.
843
