Performance Evaluation of a Cellular Millimetrewave Mobile Broadband System Demonstrator José Garcia 1, Manuel Dinis 2 and José Fernandes 1,3 1 Universidade de Aveiro, Instituto de Telecomunicações, 3810 Aveiro Portugal 2 Portugal Telecom Inovação S.A., Rua Eng. J. Ferreira Pinto Basto, 3800 Aveiro Portugal Phone: +351 34 377900, Fax: +351 34 377901, e-mail: garcia@av.it.pt Abstract In this paper a radio propagation simulation tool is used to estimate the power distribution and channel time dispersion throughout the cell in typical MBS scenarios and the measurements performed during the SAMBA field trials campaign are used to validate the results. Moreover, it is shown how good the simulated match the measured results in complex propagation scenarios. Also in this paper, experimental data collected in the measurements campaign performed in a large indoor environment, using a two-cell configuration, is analysed in order to evaluate the system performance in presence of large channel time dispersion values and the handover algorithm implemented in an MBS Demonstrator. I. INTRODUCTION Broadband services are very demanding in terms of bandwidth and due to the saturation of the spectrum at lower frequency bands, the allocated spectrum for MBS (Mobile Broadband Systems) services lies in the millimetrewave frequency band. At these frequencies, the propagation impairments are considerably different from those faced by systems operating in the UHF band. The propagation environment becomes much more adverse, due mainly to high signal attenuation caused by the obstacles within the scenario, being the cell shapes highly dependent of the scenario geometry and the propagation environment. Cellular planning is one important issue from the point of view of the optimisation of the available resources, as a way to not only minimise the costs for the telecommunications operators in the system deployment, but also to maximise the traffic capacity. The high data rates supported by a broadband radio link are very demanding also in terms of power budget and channel impulse response characteristics. The antenna configuration strongly impacts the system performance mainly due to the inverse proportionality between the achievable bit rate per carrier and the channel time dispersion and the direct proportionality with the required transmitted power. Therefore, a correct choice of the mobile terminal (MT) and base station transceiver (BST) antennas, its location within the coverage area and techniques to adapt its radiation pattern to the specific characteristics of the scenario becomes mandatory for a reliable millimetrewave mobile broadband system. The work presented in this paper focus on the problem of cell coverage for a given type of propagation environment, capable to cope with the above mentioned requirements. Emphasis is given to the coverage of a large indoor scenario in order to evaluate the systems performance in presence of large channel time dispersion. This was done by evaluating the impact of downward tilting and the height variation of the BST antenna as well as the use of multiple-coverage, which requires de implementation of handover, to keep the channel time dispersion within controlled levels. II. ANTENNA CONFIGURATIONS The antennas are one key RF-component in mobile radio communication systems in general, but its role becomes critical for millimetrewave systems. The right antenna combination at the BST and MT may contribute significantly to improve the system performance by providing an adequate signal level over the cell and favour some multipath discrimination. The main antenna configuration used in this work is the wide cell developed in the ACTS SAMBA project framework which was used in the field trial measurements [1][2]. With this antenna configuration good performance characteristics can be achieved, such as fairly uniform power distribution throughout the cell and the possibility to control the time dispersion of the channel impulse response and, furthermore, it allows movement freedom of the MT. The radiation pattern of the BST antenna has a secant square shape in the elevation plane to compensate the free space loss, so the MT can have an hemispherical radiation pattern while keeping the power distribution fairly uniform throughout the cell [3]. III. CELL COVERAGE For the Aveiro trials several typical MBS application scenarios were chosen. Although, in this paper we restrict ourselves to a large indoor environment, the "Aristides Hall" pavilion, and to a wide city street, the "Nova" street. Nova Street It is a typical urban street in a residential area with 36m wide and 370m long. There are buildings on both sides of the street with heights ranging from four to eight floors and a grass area with several trees along the street's axis (see Fig. 1). There are also several parking zones and during some short periods of the day, the traffic flow can be quite intense. 3 He is now with European Commission, DG Information Society - Unit E4
One of the main problems of millimetrewave propagation in outdoor scenarios is shadow fading caused by obstacles, demanding the use of multiple-coverage to overcome this fact [4]. As in the field trials the MT was mounted in a van and its antenna placed on top of it, the obstruction by most obstacles was not a problem. Nevertheless the attenuation of the trees revelled itself as a serious problem, as it was expected from previous studies [5]. cars have to be considered, because the multipath components they originate became more significant. Even though, for the situation when the BST antenna height is bellow the top of the trees the simulation model does not need to be very complex by including all objects, only the cars parked, and still a quite accurate power distribution estimate can be obtained. Return Path BST Ongoing Path Tree Fig. 1 - Paths considered in "Nova" street Both trees and cars parked in the street were included in the simulation models. The trees are 8m high and were defined as two cylinders, one modelling the body and the other the foliage. The foliage density was modelled in such a way that the signal attenuation would be about 2~3dB. The number of cars included in the model was defined according to the time of the day when the measurements were performed. Fig. 1 sketches the ongoing and the return paths and also the BST location considered in both the simulations and in the field trials measurements. Fig. 2 and Fig. 3 show a quite good agreement between the simulated and measured power distributions for two different BST antenna heights. A reduction of the cell length and width is evident with the lower BST antenna height. The reduction of the cell width has a direct impact on the fading depth, due to the confinement of the multipath components by the lateral buildings, leading to the lower fading depth observed in Fig. 3. Nevertheless the fading depth can be further reduced for both BST antenna heights by resorting to space diversity. The measured results allow to confirm that, when using shaped lens antennas, the cell's dimensions and the energy at its boundaries can be controlled by adjusting the BST antenna height [1]. The simulated power distribution exhibits a deeper fading than the measured one. The main reasons for this fact are the difficulty of accurate characterisation of the electromagnetic properties of the materials within the scenario and the fact that the reflections are considered to be especular in the simulator. Nevertheless a quite accurate estimation of the mean received power distribution was achieved using a raytracing simulation tool with a low computational effort. In conditions of strong line of sight (LOS) component the simulation model used can be simplified, reducing the simulation time and computational effort, due to the fact that cars parked along the street do not need to be considered. However when the trees attenuate the LOS component the Fig. 2 - Simulated and measured received power for a BST antenna height of 11.2m: ongoing path; return path
during the SAMBA trials. The BST is located at the point (1.9, 33.3)m, with the antenna at 6.5m height, rotated -45º in the horizontal plane. The MT antenna has a constant height of 1.5m. Fig. 5 depicts the results obtained for a path along the X direction. The simulated and measured results are very similar, except for the positions closer to the BST, where the fading is deeper in the simulated results. For these positions, due to the radiation pattern and the orientation of the BST antenna, the LOS component is weaker and the received power is due mainly to the multipath components. Due to the difficulty of modelling the electromagnetic properties of all the different materials within the scenario and to the fact that the reflections are considered to be especular in the simulator, the fading depth is higher. For the other positions, the LOS component is very strong and there is almost a perfect mach between both power distributions. 45 m BST 25 m BST1 Central Path BST2 10 m Y Rows of seats made of concrete X Fig. 4 - Aristides Hall pavilion Fig. 3 - Simulated and measured received power for a BST antenna height of 7m: ongoing path; return path Aristides Hall Pavilion Aristides Hall is a large indoor scenario, dedicated to sport events and it is therefore one possible MBS application scenario, see Fig. 4. Results for the wide cell antenna configuration, highlighting the problems concerning the channel time dispersion, due to multipath propagation phenomena are presented in [4]. Several cellular configurations were studied by extensive simulation, in order to meet the system s requirements. It was concluded that the requirements could be met only by resorting to multiple-coverage, and two different cellular configurations were proposed. One of these configurations uses two BSTs located in the upper corners of Fig. 4, with the antennas tilted 9º downwards. With this antenna configuration, the time dispersion constraints are met, at the expense of a little penalty in the NRP uniformity [4]. The simulation results obtained when the BST is located in the upper left corner of Fig. 4 will be compared with the measurements performed Fig. 5 - Simulated and measured power distributions for a path along the X direction, at 8.3m from the seating rows The results obtained allow to conclude that resorting to a ray-tracing simulation tool, the power distribution along a wide indoor scenario can be predicted with a good accuracy and low computational effort. Furthermore, it is pointless to increase the complexity of the simulation model, by including all the objects within the pavilion.
IV. HANDOVER STUDIES Handover studies were performed also in "Aristides Hall" pavilion, considering one of the two solutions proposed in [4]. One BST is located at (0.5,20,6.5)m and the other at (44.5,20,6.5)m, with the antennas tilted 7º downwards. This is the best solution to cover the whole pavilion s surface, except the seating rows, both from the point of view of the power distribution and the channel time dispersion constraints, here represented by the sliding delay window (SDW) parameter. In the area of interest, the SDW is bellow 30 ns allowing the transmission of very high bit rates [4]. These handover studies were carried out in order to evaluate the performance of the chosen cellular configuration and of the implemented handover algorithm. The received power distributions obtained for the central path shown in Fig. 4 and for two different movement directions, named B and C, are depicted in Fig. 6. Movement direction B means that the MT is moving along the positive X direction, from BTS1 to BST2, and movement direction C means the opposite. The received power distributions in Fig. 6 are basically a concatenation of the two individual cells. The maximum received power levels are obtained for two different positions, one at a distance of 17 m from BST1 and the other at 15 m from BST2, corresponding to the maximum received power positions of the two individual cells. These results allow to verify that the cells are asymmetric the cell of BST1 is longer. Analysing Fig. 6, the handover occurs at approximately 27 m from BST1, while Fig. 6 shows that it occurs at 22 m from BST2. The antenna configuration used in both BSTs was such that each BST covered half of the scenario, assuring a sufficient overlapping zone so that the handover could be performed successfully. Nevertheless, due to the larger front-end gain of BST2 it was necessary to increase its antenna rotation angle by 2 degrees. As the rotation angle of BST2 antenna is larger than the one of BST1, the power within its cell decreases faster, and when the MT moves towards BST1, handover occurs at a distance of 22 m from BST2. For the opposite movement direction, handover occurs only at a distance of 28 m from BST1, because the power within its cell has a smother decay. Another important aspect to notice from Fig. 6 is that the histeresis mechanism implemented in the handover algorithm avoids the effect of multiple and successive handovers ( pingpong effect), which introduces an unnecessary system overhead. Close to BST2 the link breaks down due to the lack of power, but the system is able to recover nevertheless about 2 m after the link falls definitely, and the system is unable to recover once again. This fact is a consequence of the BST2 antenna rotation angle, which leads to the decrease of the power levels and the reduction of its uniformity in regions near the BST. Regarding the BER, Fig. 7 shows the distribution of the BER and MRC power levels as a function of distance, being the BER levels very low within each cell and only increasing in regions at the cell's limits and near the BSTs, due to the reduction of the received power. Fig. 6 - Received power as a function of distance: movement direction B; movement direction C The statistical behaviours of the received power and the BER are shown in Fig. 8 and Fig. 9, only for movement direction B, being quite similar for direction C. Fig. 8 allows to conclude that for 90 % of the cases the received power levels obtained using MRC combining are above 70 db. From Fig. 9, in 90 % of the cases the BER values are below 5 x 10-4, which complies quite well with the expected behaviour of the equaliser. Given the number of samples collected during the measurements performed in this path and to the sampling rate used in the control and monitoring software, it is possible to determine the average velocity of the MT and the handover elapsed time. The handover elapsed time corresponds to the time interval between the instant when the decision to make the handover is taken and the instant when it is performed successfully. In the figures of the received power distributions it corresponds to the time interval between the instant when the power falls below 90 dbm and the instant when it rises again to a level imposed by the carrier frequency to which the MT has switched. The average velocity of the MT in this path was 3.22 km/h for movement direction B and 3.24 km/h for direction C, which leads to a handover time of 0.62 seconds
and 1.01 seconds, respectively. The handover elapsed times are quite high and, therefore, the handover algorithm or some of its parameters need to be adjusted for this specific scenario. It is important to notice in Fig. 6 that, although in a deep fading region, the MT performs handover successfully requiring an handover time lower than the one required for the opposite movement direction. This fact is due to the high power levels of the carrier frequency of BST2, which makes the synchronisation process easier between the MT and that BST. Fig. 9 - CDF of the BER Fig. 7 - MRC received power and BER as a function of distance: movement direction B; movement direction C Fig. 8 - CDF of the received power level V. SUMMARY AND CONCLUSIONS In this paper it was shown, by comparing simulated and measured results, that very accurate power distribution predictions can be made with low computational effort, even for complex scenarios with several objects and LOS obstruction conditions, using a ray-tracing simulation tool. For "Aristides Hall" pavilion and using a two BST antenna configuration the system shows a very good performance, due to fact that the multipath components caused by the reflections on the walls are controlled using downward tilting of the BST antennas. Therefore, the BER levels are quite below the 0.5 x 10-3 maximum value specified for the system's equaliser. The performance evaluation of the handover algorithm implemented in the system leads to the conclusion that it is quite robust to the presence of fading, due to its hysteresis mechanism. The handover execution time depends strongly on the power levels of the destination BST. The higher the power levels, the lower the handover time, because the synchronisation process between the MT and the BST becomes easier. In summary, the Trial Platform has shown a good performance, proving that the technology is ready for MBS deployment, and many of the technical options are suitable for a future system. VI. REFERENCES [1] J. Fernandes and C. Fernandes, Impact of Shaped Lens Antennas on MBS Systems, PIMRC 98, Boston, USA, pp. 744-748, 1998. [2] M. Dinis, J. Fernandes, M. Prögler and W. Herzig, The SAMBA Trial Platform in the Field, ACTS Mobile Communications Summit 99, Sorrento, Italy, pp. 1013-1018, June 1999. [3] C. Fernandes, M. Filipe, L. Anunciada; "Lens Antennas for the SAMBA mobile terminal", Proc. ACTS Summit 97, Aalborg, pp. 635-640, Oct 1997. [4] J. Fernandes, A. Marques and J. Garcia, Cellular Coverage for MBS Using the Millimetre-Wave Band, ACTS Mobile Summit, Sorrento- Italy, June 1999 [5] J. Fernandes and J. Garcia, Multiple Coverage for MBS Environments, PIMRC 2000, London, UK, September 2000