Revised version 4-9-21 1 Experimental Investigation of the Joint Spatial and Polarisation Diversity for MIMO Radio Channel Jean Philippe Kermoal 1, Laurent Schumacher 1, Frank Frederiksen 2 Preben E. Mogensen 1;2 Center for PersonKommunikation 1 Niels Jernesvej 12, DK-922 Aalborg, Denmark fjpk,schum,pmg@cpk.auc.dk Nokia Networks 2 Niels Jernesvej 1, DK-922 Aalborg, Denmark frank.frederiksen@nokia.com Abstract This paper presents analysis of MIMO radio channel measurements. A description of both the measurement set-up and the picocell environment is given. The correlation properties of the MIMO radio channel are investigated considering different diversity techniques such as spatial, polarisation and joint spatial and polarisation diversity techniques. It is shown that a combination of space and polarisation diversity is an attractive solution for achieving compact MIMO implementation especially when a large number of antenna ports is considered. The issue of unbalanced Branch Power Ratio (BPR) is addressed and its influence on the power gain of the potential subchannel generated by the MIMO concept is analysed. Keywords MIMO radio channel measurement, joint spatial and polarisation diversity, eigenanalysis. 1. Introduction The inherent demand for higher data rates is a topic of high attention in the research community. For some time now, the concept of Multi-Input-Multi-Output (MIMO) has been under investigation as reported in the literature by [1],[2] among others and modelled by [3], [4] among others. Briefly, the MIMO structure of the radio channel shows that a large capacity enhancement can be achieved via parallel channelling. Only recently, experimental results have been presented in [5] [6] [7] [8] [9]. This paper presents empirical results derived from MIMO radio channel measurements performed as part of the European IST research project METRA (Multi Element Transmit and Receive antennas) [1]. The correlation properties of the MIMO radio channel is investigated considering different diversity techniques such as spatial, polarisation and joint spatial and polarisation. The issue of unbalanced Branch Power Ratio (BPR) is addressed and its influence on the power gain of the potential subchannel generated by the MIMO concept is analysed. These results are derived from the measured complex narrowband information using the eigenvalue decomposition. The use of collocated polarised antenna is presented as a solution to avoid the use of a large cumbersome antenna array for picocell environment. Two case studies where a 4 4 MIMO setup using single polarised antennas and a 4 4 set-up using a combination of single polarised and dual polarised antennas are compared. The rest of the paper is organised as follows. The measurement set-up is described in Section 2. The environments where the measurement campaign was conducted are outlined in Section 3. Empirical results will be presented in Section 4 along with concluding remarks in Section 5. 2. The measurement set-up In general, MIMO is understood as a M N concept, where M and N are the number of antenna ports at the base station (BS) and the mobile station (MS) respectively. A simplified sketch of a MIMO set-up is presented in Fig. 1, where the transmitter (Tx) at the MS is on the left side and the stationary receiver (Rx) located at the BS is on the right side. In this study, the measurement set-up provides experimental results where M;N 2 [1; 4]. Figure 1: Illustration of the MIMO measurement configuration. At the MS, an interleaved antenna array with 4 vertically polarised sleeve dipoles, as one can see in Fig. 2, was moved along a linear slide covered with a microwave absorber shield over a distance of 11.8 over a duration of 5 s. Such antenna array arrangement was used to reduce the mutual coupling effect between the antenna ports. After post-processing, a linear array 1 is derived from it with a spacing of.4. The Tx uses a 1-to-4 switch with a switch interval of 5 μs between each antenna port of the antenna array, implementing a pseudo parallel 1 Note that a mechanical artifact enabled the implementation of two arrays at the MS. This eventually increases the statistical figure of the propagation channel measurement campaign. Consequently the MS consists of (4 2) vertical sleeve dipoles
Revised version 4-9-21 2 transmission 2 within 2 μs. Channel sounding measurements were performed every 2 ms at a carrier frequency of 2.5 GHz (UMTS band) and a chip rate of 4.96 Mcps. The complex narrowband information was extracted from the wideband channel data. More detailed descriptions of the MS and the testbed (i.e. Tx and Rx) are documented in [6] and [11] respectively. The BS consisted of 2 sets of four parallel Rx channels, each connected to one specific antenna configuration. The first configuration consisted of a uniform linear array with four vertically polarised sleeve dipoles with a spacing of 1.5. The second set-up was composed of two dual polarised ±45 ffi patch antennas. Depending on the environment investigated, the antenna patches were orientated in a ±45 ffi or a +9 ffi = ffi position. Fig. 3 summarises the antenna configuration used in this paper with respect to each measured environment. Base Station The two main locations were the campus of Aalborg University (AAU) and the Aalborg International Airport (Denmark). For each environment several MS locations were selected to provide a set of measurements where both line-of-sight (LOS) and Non-LOS (NLOS) were present. Moreover, several BS locations were selected in addition to the MS locations in order to increase the statistical information of the environment. This is shown in Fig. 4 and Fig. 5 with the arrows representing the positioning of the MS and the shaded area indicating the different locations of the BS. Three of the measured environments characterised as picocell are presented in this study. The first measured environment referred to as NOVI2, provides an example of a building with several small offices on the same floor. The second environment which is of the same size as NOVI2, denoted NOKIA illustrates a typical modern open office environment. The last measured environment was the Aalborg International Airport, denoted in this paper Airport, which provides a very large indoor open area. Environment NOKIA and Airport are illustrated in Fig. 4 and Fig. 5 respectively. microwave absorbers interleaved antenna array Mobile Station Figure 2: MS with a microwave absorber shield covering the linear slide during a measurement. The two antenna arrays at the BS can be seen at the background. A zoom on the MS presents the interleaved antenna array used for MIMO radio channel measurements. Measured environment Novi2 Nokia Airport MS BS (dipole) BS (patch) Antenna set-up 3 Figure 3: Summary of the different antenna configurations. 3. Description of the investigated environments The results presented in this paper are extracted from measurement campaigns undertaken in different picocell environments. 2 In this way, only one transmit antenna is active at a time, thus providing isolation between the transmit antennas. Since the switching is relatively fast and the environment stationary, it approximates a parallel transmission for low mobile speeds. 3 3 Figure 4: Example of the different positions of the BS and the MS for an open picocell environment (i.e. NOKIA). The arrows represent the displacement of the MS (1 to 6). The 3 shaded areas represent the different positions of the BS. In total 18 pairs (MS,BS) of measurement positions were investigated for this environment. Table 1 summarised the number of measurement positions which have been made during the measurement campaigns as well as the mean signal-to-noise ratio (SNR) related to each environment. Here the mean SNR is computed over all the measured positions for each environment. Number Number Total mean Environment of BS of MS number of SNR positions positions positions in [db] NOVI2 3 7 21 31 NOKIA 3 6 18 31 Aiport 2 8 16 3 Table 1: Summary of the number of investigated environments and the mean SNR associated to each environment.
Revised version 4-9-21 3 between two antennas of different polarisation and separated by a certain spatial distance. Fig. 6 introduces the scatter diagram of jρ spacej versus jρ xpol j for the environment NOVI2. These correlation coefficient results are extracted from all the 21 measurement positions performed in this environment. This correlation coefficient is computed between two antenna ports at the BS and for all the 8 dipoles of the MS to increase the statistical amount of data.jρ spacej is derived for two antenna ports separated with 1.5 and jρ xpol j considers two collocated ±45 ffi dual polarised antenna ports. It can be seen that for the spatial domain, jρ spacej exhibits very low values whereas for jρ xpol j they are significantly higher. 1.9.8.7 ρ xpol.6.5.4.3.2.1 Figure 5: Example of the different positions of the BS and the MS for a very large open picocell environment (i.e. Airport). In total 16 pairs (MS,BS) of measurement positions were investigated for this environment. 4. Experimental results In this section, the correlation properties of the MIMO radio channels will be investigated for different BS antenna configurations. Direct comparison between single polarised and dual polarised antenna configurations will be presented. The power correlation coefficient between antenna ports using different diversity techniques such as space, polarisation and joint spatial and polarisation diversity will be analysed. In addition, issues related to the BPR when dual polarised antennas are used will be addressed. This will be presented first in a Single-Input- Multi-Output (SIMO) approach and then extended to a MIMO configuration to present the impact of the BPR. Following this, eigenvalue results will be presented when a 4 4 antenna configuration is considered. Finally, the power level of the different subchannels obtained in a MIMO perspective will be presented for all the picocell environments. 4.1. Power correlation coefficient The power correlation coefficient jρj has previously been defined in [6]. In the rest of this paper, to keep the discussion simple, jρj will be mentioned as the correlation coefficient. In that respect, when considering different diversity techniques, then jρ spacej is the spatial correlation coefficient associated to the space diversity, jρ xpol j is related to the cross polarisation correlation coefficient of a dual polarised antenna. Then finally, jρ j refers to the correlation coefficient which exists.1.2.3.4.5.6.7.8.9 1 ρ space Figure 6: Scatter diagram of jρ spacej versus jρ xpol j for NOVI2. Decorrelation exists in the spatial domain. Higher values of correlation are noticeable in the polarisation domain. Fig. 7 represents the scatter diagram where jρ j is plotted from both a theoretical and empirical perspective. The environment is the same as for Fig. 6. The theoretical values are obtained from the equation defined in [12] such that jρ j = jρ spacej jρ xpol j (1) The diagonal line represents the case where the equation (1) is valid. It can be seen that for very low correlation values this is very sensitive and therefore values lower than.5, for instance, will be disregarded. It can be seen that for higher correlation values, the measurement data confirm the work of [12] for the investigated picocell environment. 4.2. Branch Power Ratio impact It is well known that the correlation coefficient drastically impacts the performance of a MIMO scenario. It is known that when the antenna branches are decorrelated, the MIMO configuration provides significant power gain for each parallel subchannel. However, it is only true for equal branch power. The present study investigates the impact of the BPR on the evaluation of the parallel subchannel of the MIMO radio channel. A SIMO configuration will be used as a first study case to understand the propagation mechanism involved when the BPR varies from a low value to a high value. Then a second study case will be presented using a 2 2 MIMO configuration for the two BPRs.
Revised version 4-9-21 4 Theoretical ρ 1.9.8.7.6.5.4.3.2.1 High sensitivity.1.2.3.4.5.6.7.8.9 1 Measured ρ Fig. 9 and Fig. 1 present the cumulative distribution (cdf) of the eigenvalue computed for a 1 2 antenna configuration. The analysis is performed using one vertically polarised dipole at the MS and a dual polarised patch antenna at the BS. In the first case, the patch antenna is orientated such that it exhibits a ±45 ffi polarisation (Fig. 9). This is the situation for the measurement undertaken in NOKIA. In the second case, for the Airport environment, the patch is tilted of 45 ffi in the elevation plane such that a +9 ffi (vertical) and ffi (horizontal) polarisation (Fig. 1) is achieved. The reason for selecting these two measurement positions is that they are both open picocells exhibiting a similar 1 1 channel cumulative distribution (almost Rayleigh). The cdfs are computed over a full slide run of the Tx for one measurement position of picocell NOKIA and Airport respectively. In order to increase the statistical figures, the cdfs were computed over all the 8 antenna ports of the MS. The cdfs are normalized to the strongest 1 1 radio channel. Fig. 9 illustrates clearly that the power gain from using two antennas spatially separated but with a low BPR is the same whether the ffi or the 45 ffi antenna port is used. Obviously, there is no gain added to the space diversity technique by using polarisation diversity since jρ spacej is low. Fig. 1, on the other hand, demonstrates an imbalance in the power gain available depending on which branch used. On the lowest branch ( ffi ), a gain is noticeable due to the use of polarisation diversity but remains negligible compared to the power gain of the 9 ffi branch. Figure 7: Scatter diagram of the measured jρ j versus the theoretical jρ j for NOVI2. Sensitiveness very high for low correlation value. Good matching for higher correlation value. Fig. 8 summarises the antenna configuration with their associated BPR and correlation coefficient. The value of jρ spacej, jρ xpol j, jρ j and the BPR for the two examples illustrated in Fig. 9 and Fig. 1 are averaged over the 8 antenna ports of the MS. It can be seen that the correlation values are very low. Environment NOKIA Antenna set-up at the BS BPR [db].9.22 1.4.23.7.22.2.4.6.8.2.4.6 space: o.8 space: 45 o xpol: +/ 45 o 5 5 5 1 15 Figure 9: Illustration of the influence of the BPR on the power gain for a 1 2 SIMO configuration: ±45 ffi. Airport +9 +9 +9 +9-8.16-8.12.3.11 Figure 8: Summary of the antenna configuration with their associated BPR and correlation coefficient for a 1 2 scenario where the antenna port at the MS is vertically polarised..2.4.6.8.2.4.6 space: +9 o.8 space: o xpol: +/ 9 o 5 5 5 1 15 Figure 1: Illustration of the influence of the BPR on the power gain for a 1 2 SIMO configuration: +9 ffi and ffi. Note the imbalance in the power gain compared to Fig. 9 Fig. 11 and Fig. 12 present a MIMO approach to the pre-
Revised version 4-9-21 5 vious SIMO analysis. A 2 2 antenna set-up is considered. Here, two vertically polarised dipoles are considered at the MS. The same antenna topology is used at the BS as for the SIMO analysis. The eigenvalues are normalized to the strongest 1 1 channel. One can see in Fig. 11 that the two eigenvalues are optimized since the BPR is low while when the BPR is much larger (-8 db) the imbalance exhibited in the SIMO case is still present as shown in Fig. 12. This clearly highlights the fact that although the correlation value exhibits strong decorrelation no optimum MIMO configuration can be achieved when the branch power is unbalanced. Therefore, it is recommended to use an antenna arrangement such that both the BPR and the correlation coefficient are as low as possible when considering implementing a MIMO configuration..2.4.6.8.2.4.6 space: o.8 space: 45 o xpol: +/ 45 o 5 5 5 1 15 Figure 11: Illustration of the influence of the BPR on the power gain for a 2 2 MIMO configuration: ±45 ffi. tool when implementing MIMO set-up in picocell environment. It will be shown that the use of collocated dual polarised antenna could solve this issue. Two 4 4 MIMO scenarios are presented here. These two set-ups differ in the type of antenna configuration used at the BS. The first scenario considers 2 dual polarised ±45 ffi patch antennas (with a low BPR) and the second scenario is implemented using 4 vertically polarised sleeve dipoles. At the MS, 4 vertically polarised dipoles are considered. The environment is the same as the one used in Fig. 11 (i.e. NOKIA)..2.4.6.8.2.4.6.8 λ 4 5 5 5 1 15 λ 3 Figure 13: Eigenvalue MIMO result for a 4 4 antenna configuration using 4 dipoles at the MS and 2 dual polarised ±45 ffi patches. 4 eigenvalues can be identified, 2 of them being significant (P> db).2 λ 4.2.4.6.8.2.4.6 space: +9 o.8 space: o xpol: +/ 9 o 5 5 5 1 15 Figure 12: Illustration of the influence of the BPR on the eigenvalue for a 2 2 MIMO configuration: +9 ffi and ffi. 4.3. Impact of the joint spatial and polarisation diversity A major drawback in employing space diversity is usually the actual physical size of the antenna array implemented. Consequently, one may see the use of space diversity as a cumbersome.4.6.8.2.4.6.8 5 5 5 1 15 Figure 14: Eigenvalue MIMO result for a 4 4 antenna configuration using 4 dipoles at the MS and 4 dipoles at the BS. 4 eigenvalues can be identified, 2 of them being significant (P> db). Very similar behaviour to Fig. 13. The first and the second MIMO scenario performances are illustrated in terms of the cdf of their eigenvalues in Fig. 13 and Fig. 14 respectively. It can be seen when comparing both figures, that they exhibit very similar behaviour. Four eigenvalues can be identified, two of them being significant (P> db). This enables to conclude that using four dipoles or two dual λ 3
Revised version 4-9-21 6 polarised patch antennas will provide the equivalent results in terms of subchannel gain in the MIMO radio channel. The advantage of using the joint spatial and polarisation solution is the possibility of using a more compact antenna set-up when large a number of antenna ports is considered. An outage level of 1% is typically a good measure for system level performance. The power gain of each subchannel is extracted from this threshold and cumulated over all the measurement positions of the two picocell environments NOVI2 and NOKIA. The cdf of the power gain of each subchannel is presented in Fig. 15. It can be seen that the first two subchannels offer little variance in the power gain for all the various measurement positions for the two investigated picocells. For the weaker eigenvalues, the slope is less steep, however they have practically no significant impact on the general power gain. This indicates that wherever a BS is considered to be placed, one can expect to have almost identical power gain. log P(power gain<abscissa).1.2.3.4.5.6.7.8.9 λ 4 λ 3 5 5 5 1 15 Power gain of the subchannel [db] Figure 15: Power gain of the parallel subchannels for 4(dipoles) 2(±45 ffi patch) MIMO configuration. The power gain is extracted from the 1% outage level of the eigenvalue (Fig. 13) for each measurement position for the two picocell environments using the ±45 ffi patch orientation. Note the little variance of the 2 significant parallel subchannels (P> db) 5. Conclusions This paper presented MIMO radio channel measurements as part of the European IST project METRA. The measurement set-up and the investigated picocell environments were described. The investigation of space, polarisation and joint polarisation diversity is presented. It was shown that a combination of space and polarisation diversity is an attractive solution for compact MIMO implementation especially when a large number of antenna ports is considered. This was shown to be a very robust solution too, since the power gain of the significant parallel subchannels exhibited a little variance. This indicates that for the investigated picocell environments wherever the BS is considered to be placed the power gains remain almost identical. However, it was outlined that it is necessary to have equal BPR and decorrelated antenna branches to achieve an optimum implementation of the MIMO concept. 6. References [1] G.J. Foschini, Layered Space-Time Architecture for Wireless Communication in Fading Environment When Using Multi-Element Antennas, Bell Labs Technical Journal, pp. 41-59, Autumn 1996. [2] J. Bach Andersen, Array Gain and Capacity for Known Random Channels with Multiple Element Arrays at Both Ends, IEEE Journal on Selected Areas in Communications-Wireless Communication Series, vol. 18, no. 11, pp. 2172-2178, November 2. [3] K.I. Pedersen, J.B. Andersen, J.P. Kermoal and P.E. Mogensen, A Stochastic Multiple-Input Multiple-Output Radio Channel Model for Evaluation of Space-Time Coding Algorithms, Proceedings of VTC 2 Fall, pp. 893-897, Boston, United States, September 2. [4] Lucent Technologies, Further link level results for HS- DPA using multiple antennas, 3GPP TSG RAN WG1, TSGR1#17()1386, Stockholm, Sweden, November 2. [5] J.P. Kermoal, P.E. Mogensen, S.H. Jensen, J.B. Andersen, F. Frederiksen, T. B. Sørensen, K.I. Pedersen, Experimental Investigation of Multipath Richness for Multi- Element Transmit and Receive Antenna Arrays, IEEE Proc. Vehicular Technology Conference, pp. 24-28, Tokyo, Japan, May 2. [6] J.P. Kermoal, L.Schumacher, P.E. Mogensen, K.I. Pedersen, Experimental Investigation of Correlation Properties of MIMO Radio Channels for Indoor Picocell Scenarios, IEEE Proc. Vehicular Technology Conference, pp. 14-21, Boston, Massachusetts, September 2. [7] L. Schumacher, J.P. Kermoal, F. Frederiksen, K.I. Pedersen, A. Algans, P.E. Mogensen, MIMO Channel Characterisation,IST Project IST-1999-11729 METRA Deliverable 2, February 21. [8] M. Steinbauer, A. F. Molisch, A. Burr, R. Thomae, MIMO channel capacity based on measurement results, Proc. ECWT 2, pp. 52-55, Paris, France, 5-6 October 2. [9] D.P. McNamara, M. A. Beach, P. Karlsson, P. N. Fletcher, Initial characterisation of Multiple-Input Multiple-Output (MIMO) Channels for space-time Communication, IEEE Proc. Vehicular Technology Conference, pp. 1193-1197, Boston, Massachusetts, September 2. [1] http://www.ist-metra.org [11] F. Frederiksen, P. Mogensen, K.I. Pedersen, P. Leth- Espensen, A Software Testbed for Performance Evaluation of Adaptive Antennas in FH GSM and Wideband- CDMA, Conference Proceeding of the 3rd ACTS Mobile Communication Summit, Vol.2, pp. 43-435, Rhodes, Greece, June 8-11 1998. [12] P.C.F. Eggers and J. Toftgaard and A.M. Oprea, Antenna Systems for Base Station Diversity in Urban Small and Micro Cells, IEEE Journal on Select Areas in Communications, vol. 11, no. 7, pp. 146-157, September 1993.