Spatial Diversity and Correlation for MIMO in BANs: Parametric Simulation Scheme

Transcription

1 Spatial Diversity and Correlation for MIMO in BANs: Parametric Simulation Scheme K. LUOSTARINEN, M. A. JADOON 2, J. SILTANEN 3, and T. HÄMÄLÄINEN 2 Metso Paper, Jyväskylä, FINLAND, kari.luostarinen@metso.com 2 University of Jyväskylä, Jyväskylä, FINLAND, aliman8@hotmail.com, and timoh@mit.jyu.fi 3 JAMK University of Applied Sciences, Jyväskylä, FINLAND, jarmo.siltanen@jamk.fi Abstract Use of MIMO for capacity increase has been practically utilized in personal area networks (PANs), indoor wireless channels and on Body networks [6]. Spatial diversity for on body channels is explored using parametric simulation scheme. The multiple input multiple output (MIMO) channel fading characteristics and spatial correlation results are given for virtual antennas. The 3-D E-field distribution at various regions is shown along with probability distribution functions graphs. Diversity and correlation results are discussed which give results of achieving same correlation behavior at different spacing. The strategy and results of spatial correlation will be discussed. Key-words: MIMO, Spatial Correlation, Diversity parametric simulation scheme. Introduction Body area networks deals with communications on and around human body. Using MIMO in body worn wireless communications has been the topic of interest in [],[2]. Most of the work has been done with measurement setup involving Tx and Rx antennas on human body while measuring capacities/correlation for different on-body-channels [3]. Considering lot of effort involved in experimental measurement domain so our theme was to work on simulation technique that can give us same results as of or very close to the results obtained while practically using body worn antennas to analyze the fading characteristics along with spatial correlation and diversity behaviour. Secondly, various scenarios were devised in simulated environment. After Parametric simulation study without effect of antennas, visualization of fading characteristics on cylindrical shaped body with all the properties close to human body using MIMO was done. Based on results of fading characteristics, spatial correlation and diversity analysis was performed which gave us new value of spatial distance between two transmitting and receiving antennas without losing the properties of the signal. Robustness in communication demands the best use of diversity. In our case we used spatial diversity, where spatial diversity is the use of two or more antennas on both transmitter and receiver sides. The use of diversity will be decided on the basis of correlation [4]. 2 Theoretical Background Fading due to body movement and channel variation is a major problem in Bans. If two or more independent and uncorrelated channels are available between transmit and receive nodes, received signals are obtained which are statistically independent of each other [5]. Diversity can be achieved in many ways, but space diversity is most important in body area networks. Two or more antennas can be obtained at the transmitter (MISO), at the receiver (SIMO) or at the both ends (MIMO) for this type of diversity. Antenna coupling can be a problem by placing two antennas close to each other, but for most applications spacing of half a wave length is suggested to mitigate this problem. [4],[5]. Theoretical Rayleigh model is used as a bench mark to characterize the indoor or outdoor channel. Various studies related to On-Body-Network shows that better performance is achieved in NLOS case as compared to LOS case. During these studies human posture was taken, and receives diversity was more effective when transmit and receive antennas are opposite sides of the body. [5],[6]. 3 Simulation Setup ISBN:

2 As shown in the flow chart below the scheme to be followed was very simple and concise. As shown in the flow chart (see Fig ) the scheme to be followed was very simple and concise. For our simulation scheme we decided to use cylindrical shaped body. Dielectric constant of the material to be used was which closely resemble to dielectric constant of human flush muscle. Software to be used was decided to be CST simulation software. Frequency to be used was 2.45 GHz. The setup consists of four transmit antenna positions (, 2, 3, 4) and three antenna orientations (x, y, z) (see Fig 2). Four positions were designed on the front portion (upper side) of the body. The positions were maintained at spacing of λ/2 in order to avoid the coupling effect of antennas on each other. Fig : FLOW CHART FOR OVERALL SCHEME The propagation behaviour was visualized and analyzed around the body in three different regions: Front, curved and back. Front portion was one, where we placed the antennas, the curved portion was along curved region of the body and back region was the one at 8 degree from front side antennas. The propagation characteristics were studied in all these regions individually. Visualization and analysis of fading characteristics were done for all four positions individually. The analysis was done by putting antennas in x, y and z orientations respectively and for various heights from the body. The visualization of the electric field around the body shows that when both the transmit and receive antennas are in line-of-sight (LOS) or semi-los, the field decays at an exponential rate with separation distance. In non-los (NLOS) situations, standing wave patterns were observed in the received fields (see Fig 3 (c)). Fig 2: DIFFERENT POSITONS OF TRANSMIT ANTENNAS IN Y ORIENTATION The fading of the NLOS conditions was determined to be log-normally distributed, which agrees with existing studies [7],[8]. 4 Fading Characteristics of On-Body Propagation To investigate the channel behaviour thoroughly, the propagation was analyzed in three different separate regions: The front portion, the curved portion, the back portion (see Fig 2). E-field variations at different heights from the antenna were also tested. Each case was analyzed for three orientations (x, y, and z). Body dimensions were in the x direction (-25 to 25).mm, y direction (-5 to 5 mm) while in z direction (-2 to 2 mm). The measurement was done for Y orientations at each of the four transmit positions. All transmitting antennas transmit simultaneously and virtual antennas were assumed at the back portion in the Y orientation. Different positions are configured to deploy MIMO at the spacing of λ/ Position The antennas in the y orientation and four transmit positions are shown in Figure 2. Simulations were done by placing the transmit antennas at positions, 2, 3 and 4, respectively. Position is at the center of the body, while position 3 was 6 mm away from Position in the positive z direction. In fact, 6 mm corresponds to half-a-wavelength (lambda/2 or λ/2) spacing for the frequency of 2.45 GHz. Similarly, Position 2 was 6 mm away from Position in the positive x direction and ISBN:

3 Position 4 was 6 mm away from Position 2 in the positive z direction. Symmetry in the results for Position is clearly shown in Figure 3 (a). With the transmit antenna being at the center of the front region, the E-field decays smoothly on both sides of the front region. This is followed by the smooth decay at the curved portion, which before entering the back region starts to exhibit large variations (i.e., a standing wave pattern). This trend of large E-field variations is continued into the entire back region. Therefore, we can deduce that there is one region on the body (from the front region to the most part of curved region) where the E-field decays smoothly and there exists another region (called the interference region) where large variations in the E-field are observed (from the fringe of the curved region to the back region). The reason for different behaviour is that creeping waves of comparable strengths arrive at the back region of the body via different propagation paths to create constructive and destructive interference, as shown in Figure 3 (c), whereas the propagation at the front portion, as can be seen in Figure 3 (a), is more like the LOS case. 4.2 Position 2 E-field variations due to this transmit position are asymmetrical in the front and back regions (and likewise asymmetrical between the two curved regions of the body, if compared to Position. The magnitude of the E-field along the curved region for Position 2 was stronger than Position, as it decays less sharply. The E- field peak was moved along with the movement of antenna from the center Position to the offcenter Position 2. This decay continues along the curved portion of the body. A smooth decay was observed along the curved portion and some fluctuations were observed only at the end of the curved portion. Standing wave fluctuations then continued into the back region. So based on propagation characteristics around the body two regions were observed: one was the smooth decaying region, which was from the front to the curved region and second was the interference region, which is found mainly at the back side of the body. 3(a) (a) 3(b) 3(c) Fig 3: 3-D VIEW OF E-FIELD PROPAGATION IN THE 3 (a) FRONT, 3 (b) CURVED, AND3(c) BACK REGIONS OF THE BODY FOR TRANSMIT 4.3 Position 3 and 4 By moving the antenna in the z direction to Position 3, the peak was shifted in the z direction to where the antenna was moved. In the curved region, the E-field was varied in accordance with the antenna movement, i.e., the field becomes asymmetrical (along the z direction) over the displayed region. When the antenna was further moved from Position 3 in the x direction to Position 4 (i.e., nearer to the curved region), the E-field peak shifted towards the curved region. This causes a peak just before the start of the curved region and the ISBN:

4 magnitude of the E-field were observed to be higher and a smoother in this case. In the back region, the E-field has similar behaviour as Position when the antenna was moved along the z direction to Position 3. But if the antenna is moved from Position 3 in the x direction to Position 4, there is more variation in the strength of the E-field that is due to constructive and destructive interference. This is due to the antenna at Position 4 being closer to one side of the back region. When the E-field values were compared at the back region for different antenna positions, it was found that higher E- field values were obtained when the antennas are placed closer to the back region (i.e., Positions 2 and 4. However, the fluctuations of E-field of different positions seemed to be independent of one another) for a given spatial location on the back region. This gives an indication that multiple antennas placed at different locations at the back region can be used to improve the robustness of on-body propagation in NLOS. In the next section comparative analysis for different antenna orientations (x, y, z) for all 4 transmit positions are given to investigate any difference that might arise from the change of antenna orientation to z and x orientations, respectively. The comparisons between E-field variations have been done to understand clearly the effect of using different orientations. 5 Comparative Analysis For Different Orientations 5. Antenna at Position The E-field distributions at the front side were almost the same for the different orientations (see Fig 4), with a difference of several dbs 4(a) 4(b) Fig 4: X 4(a) AND Z 4(b) ORIENTATIONS OF ANTENNAS PLACED ON FRONT SIDE OF BODY In the back region, the magnitude of E-field was higher for the y oriented antenna than the x and z oriented antenna, and a difference of 5 db was observed at some points. The E-field distribution showed more fluctuations for the z oriented antenna and was smoother for the x and y oriented antennas. Along the curve side, the E-field magnitude was highest for the y oriented antenna and it was smoother i.e. decayed slower than other orientations. The E- field magnitude for the y oriented antenna was almost 5 db and 2 db higher than the z and x oriented antennas, respectively. For the z oriented antenna, the E-field decayed more rapidly as it advanced along the curved portion, as compared to x oriented antenna. 5.2 Antenna at Position 2 At the front side strong peaks were observed for the z oriented and y oriented antennas. In the back region, for the y oriented antenna, the magnitude of the E-field was the highest among the different orientations, and a difference of around 5 db was observed at some points as compared to x and z oriented antenna. The E- field distribution was fluctuating more frequently for the z and y oriented antennas than for the x oriented antenna. Along the curve side, the E-field magnitude was highest for the y oriented antenna and decayed slower and smoother than the other two orientations. Similar to Position, the E-field magnitude for the y oriented antenna is almost 5 db higher than the z oriented antenna and 2 db higher than the x oriented antenna 5.3 Antenna at Position 3 ISBN:

5 In the front region, The E-field magnitude decays smoothly from the peak value. For the y oriented antenna, the peak was sharper at the position where the antenna was placed and it decayed rapidly away from that position. On the other hand, for the z and x oriented antennas, the E-field peak at the antenna position was not as sharper as Position, but it decayed more slowly at positions away from the antennas. The E-field magnitude at the back side for y oriented antenna was highest, it was almost 2 db higher than x and z oriented antennas, and significant variations in the E-field distribution were noticed over these orientations. Along the curve side, the trend observed for the different orientations is similar as Positions and Antenna at Position 4 For the y and z oriented antennas, the E-field peak was sharper than for the x oriented antenna at the position where the antennas was placed. For the y oriented antenna, the E-field decayed rapidly away from the antenna position, whereas the decay is slower for the other two orientations. The E-field magnitude at the back side for y oriented antenna was the highest among the three orientations and the highest peaks were almost 2 db higher than those of the x and z oriented antennas. As in Position 3, significant variations in the E-field distribution were noticed over these orientations. Along the curve side, strong E-field amplitudes were observed at the edge between the curved and front regions for all orientations. For the y oriented antenna, the E-field decayed more rapidly and smoothly as the receive antenna moved along curved side towards the back region. For the x oriented antenna, the E-field decayed slower than for the z oriented antenna. In order to quantify it more explicitly, we plotted the distribution of E-field magnitude in the back portion for different orientations of the transmit Hertzian dipole. The distribution was found to be lognormal see Fig 5) in all three orientations, which is consistent to previous findings [7],[8]. Receive antennas were positioned in the Y direction. The phase distribution of the backside is shown in Fig 6 and it is found to be almost uniform within the range from -pi/2 to pi/2. Density xyzeii2(:,4) data rayleigh lognormal Data FIG 5A: DISTRIBUTION OF E-FIELD MAGNITUDE IN THE BACK REGION FOR Y-ORIENTATION 6 Spatial Correlation and Diversity Space diversity is a technique in which more than one antenna is used at the transmit and/or receive sides. Antenna spacing is a very important issue for this technique. FIG5B: DISTRIBUTION OF E-FIELD MAGNITUDE IN THE BACK REGION FOR Z ORIENTATION. When two or more than two antennas close to each other, then mutual coupling between antennas should be low enough to minimize the interaction of one on the other (which reduces efficiency and affects the signal correlation) [4],[8]. For effective use of spatial diversity, the correlation characteristics for the on-body propagation must be evaluated. After extracting the fading characteristics of the on-body channel, we investigated whether MIMO techniques can improve the robustness for onbody communications. The effectiveness of MIMO techniques can be examined by calculating either the capacity or the diversity ISBN:

6 performance. FIG5C: DISTRIBUTION OF E-FIELD MAGNITUDE IN THE BACK REGION FOR X ORIENTATION. Desnity Phase distribution at the backside Phase FIG 6: PHASE DISTRIBUTION ON BACKSIDE OF BODY For the diversity performance, a key indicator is spatial correlation (low correlation is a necessary condition for good diversity performance). The other two indicators are total efficiency and the efficiency of individual antennas. investigated). Since we were interested in normal (or y) component of E-field, its correlation characteristics were shown in these plots. One obvious feature in the plots is that when the transmit antenna was placed at the center of body (Position ), then spatial correlation on back side was lower than Position 2 (compare Figures 7 and 8). The spatial correlation at two transmit positions spaced 6 mm (or λ/2) apart was found to be significantly lower than that observed for the receive positions, which was between.3 and.4 [3] (see Figures 7-2). Moreover, if transmit and receive antennas are swapped in this setup, then receive and transmit correlation are interchangeable concepts. Since correlation values of.5 is adequate to provide good diversity gain, this means that spacing greater than λ/4 (see Figures 7 to ) is sufficient for achieving good diversity gain. From this we can conclude that multiple antennas can be employed at both the transmit and receive sides to obtain independent copies of the signals [4], [8],[9] when the spacing between the antennas is greater than λ/4. To obtain the correlation at the transmit side, we calculated the correlation between transmit antennas at different positions and orientations by using equation (). ρ = () Where ρ = correlation between two transmitting elements at the front side of the body. is FIG 7: SHOWING ANTENNA ORIENTATIONS IN Y DIRECTIONS TO DETERMINE AND CALCULATION SPATIAL DIVERSITY 6. Analysis at different transmit positions We analyzed the correlation characteristics between the E-field vectors (representing fictitious received antennas or perfect E-field sensors) at the back portion of the body due to transmit antenna at different positions and orientations on the top side of the body (see Fig 7-). They show clearly an initially decreasing trend as distance increases from 5mm onward (5 mm is the smallest correlation distance correlation sample at position and is correlation at position 2(position+lamda/2). Correlation Correlation against Electrical length L/lamda FIG 8: TWO RECEIVE POSITIONS AT DIFFERENT SEPARATION DISTANCES ON THE BACK REGIONS OF THE BODY DUE TO TRANSMIT ANTENNA AT POSITION. ISBN:

7 Configuration diagram Fig-9 for correlation at the transmitter and receiver position is shown in figure.where c is the antenna of interest and rest are the samples around it at lambda/2 spacing. Correlation between all the elements is found as shown in the figure and is well below the threshold limit of.5.same configuration is adopted at the back side and correlation is calculated using equation () and various results are shown in graphical form in Fig -3. A smaller correlation value was obtained for the y oriented antenna as compared to other orientations. Since y orientation best supports the propagation of creeping waves. Thus, the lower correlation value for this orientation goes well with the propagation characteristics. Several diversity schemes, like selection combining and maximum ratio combining, can be used to observe the best performance for MIMO, but they are left for future work. Here we focused on creeping waves around a body by analyzing results extracted from simulations of on-body propagation, and to verify from correlation characteristics the potential use of MIMO to improve robustness for on-body communications. Visualization of simulation results gave interesting insights into the propagation. 6.2 Characteristics of on-body communications On-body propagation is best supported by creeping wave component given by the y orientation (i.e., normal to the body), as opposed to other orientations. Any existing path loss model can be used to model the path loss at the front region. The path loss increases smoothly with increasing distance (from the transmit antenna). The geometry of the curved region influences the propagation behaviour and the strength of the creeping wave component. Correlation against Electrical length Correlation against Electrical length Correlation Correlation L/lamda FIGURE 9: TWO RECEIVE POSITIONS AT DIFFERENT SEPARATION DISTANCES ON THE BACK REGIONS OF THE BODY DUE TO TRANSMIT ANTENNA AT POSITION L/lamda FIG : TWO RECEIVE POSITIONS AT DIFFERENT SEPARATION DISTANCES ON THE BACK REGIONS OF THE BODY DUE TO TRANSMIT ANTENNA AT POSITION 3. Correlation against Electrical length Correlation FIG : ALL POSSIBLE CORRELATION CONFIGURATION FOR EACH OF 4X4 MIMO ON TRANSMIT SIDE L/lamda FIG 2: TWO RECEIVE POSITIONS AT DIFFERENT SEPARATION DISTANCES ON THE BACK REGIONS OF THE BODY DUE TO TRANSMIT ANTENNA AT POSITION 4. ISBN:

8 For bodies with smooth curved surface, the height of the antenna from the body surface is a significant factor that influences the path gain, since a loss of a few dbs is observed when analysis was done for antenna height ranging from 5 to 35 mm. The polarization of the antenna should preferably be normal to the body surface, in order to take advantage of creeping wave propagation. 6.3 Analysis at different heights To analyze the behaviour of E-field for the same antenna position (i.e., Position ) but at different heights, we varied the heights in steps of 5mm. One obvious trend we observed was that, the strength of E-field decreases when the antenna height is increased. Secondly as we move away from the body, the decrease in correlation with separation distance is more than the case when the antenna was at 5 mm(see Fig 3). For all these heights, the correlation value was below.5 for spacing of λ/2. FIG3: TWO RECEIVE POSITIONS AT DIFFERENT SEPARATION DISTANCES ON THE BACK REGIONS OF THE BODY DUE TO ANTENNA AT POSITION AT THE HEIGHTS OF 5 MM AND 25 MM. 6 Conclusions The correlation characteristics of on-body propagation provide us with important evidence in the effectiveness of using MIMO to achieve robust and reliable on-body communications. The use of diversity is proposed to overcome fading and to improve communications performance. For space diversity, multiple antennas with different orientations were tested and the y orientation was found to give the lowest correlation on the receive side. Antenna spacing is a very important issue for this technique at both transmit and receive sides, and we have found that reliable connections can be obtained by using more than one antenna at spacing greater than λ/4. Correlation decreases as we move further away from body at heights ranging from 5 to 35 mm. References: [] I. Khan, Hall, P.S. Hall, Antennas and Propagation IEEE Transactions.vol. 58, Issue., pp.95 22, 2.' [2] A.A. Serra, P. Nepa,G. Manaral, and P.S.Hall, "Diversity for body area networks" in Proc. of XXIX URSI General Assembly, Chicago, IL, Aug. 7-6,28.Available: Cp2.pdf [3] M.R. Kamrudin, Y.I. Nechayev, and P.S. Hall, Performance of antennas in on body environment, in Proc. IEEE Int. Symp. Antennas Propagation.vol.3A, Washington DC, July 3-8,25,pp [4] A. A. Serra, P. Nepa, G. Manara, and P. S. Hall, Diversity for body area networks, in Proc. XXIX URSI General Assembly, Chicago, IL, Aug. 7-6, 28. Available: 8/papers/Cp2.pdf. [5] P.S. Hall, Y. Hao, V. I. Nechayev, A. Alomain, C. C. Constantinou, C. Parini,M. R. Kamarudin, T.Z. Salim, D.T.M. Heel, R. Dubrovka, A. S. Owadall, W. Song, A. Serra, P. Nepa, M. Gallo, and M. Bozzetti, Antennas and propagation for on-body communication systems, IEEE Antennas propagation Mag., vol.49, no.3, pp.4-58, Jun.27 [6] D. Neirynck, C. Williams, A. Nix and M. Beach, Channel characterization for personal area networks, in COST273 TD (5) 5, Lisbon, Portugal, Nov. -, 29. Available: 983/893//TD-5-5.pdf. [7] I. Khan, P. S. Hall, Y. I. Nechayev,L. Akhoondzadeh-asl, Antenna Technology I. Khan, Hall, P.S. Hall, Antennas and Propagation, IEEE Transactions. vol.58, Issue., pp.95 22, 2. ISBN: