System Performance Challenges of IMT-Advanced Test Environments

Transcription

1 System Performance Challenges of IMT-Advanced Test Environments Per Burström, Anders Furuskär, Stefan Wänstedt, Sara Landström, Per Skillermark, Aram Antó Ericsson Research [per.burstrom, anders.furuskar, stefan.wanstedt, sara.landstrom, per.skillermark, aram.anto@ericsson.com] Abstract The test environments for system level evaluations of IMT-A (International Mobile Telecommunications-Advanced) candidate technologies are analyzed. Compared to the evaluation criteria for previous IMT standards, the IMT-A test environments provide significantly more realistic settings for evaluating system-level performance. The models include line-ofsight probability in all scenarios, different groups of users in terms of propagation and speed, as well as indoor and outdoor deployments. This contribution scrutinizes the evaluation conditions and presents resulting performance for a basic Long- Term Evolution (LTE) candidate system. Each environment contains challenging characteristics. Examples include a mix of indoor and outdoor users, a mix of line-of-sight and non-line-ofsight conditions, and terminal speeds making accurate tracking of channel quality difficult. Together, these characteristics suggest that a radio transmission technology fulfilling the ITU IMT-A requirements is well suited for meeting a variety of challenges of real deployment scenarios. Index Terms IMT-Advanced standard, ITU requirements, system simulations, radio network performance I. INTRODUCTION TU, the International Telecommunication Union, has Ispecified guidelines [1] to be used for evaluating the performance of future technologies trying to conform to the IMT-A (International Mobile Telecommunications-Advanced) standard. Noticeable features of IMT-A technologies are the required performance of 15 bits/hz/s peak spectral efficiency and support for 1 MHz bandwidth. The evaluation guidelines include a set of test environments, effectively models and assumptions on system level simulations, that the proponent technology should be evaluated with, as well as a set of requirements [] that should be fulfilled. The requirements to be evaluated by system simulations are on cell spectral efficiency, cell-edge user spectral efficiency, defined by the five percent worst users, and on the number of simultaneous VoIP (Voice over Internet Protocol) users in a cell. In this contribution, the test environments are analyzed in terms of coupling gain and geometry characteristics output from a system simulator. Also some general properties of the environments are discussed and basic performance results for an IMT-A candidate technology are presented. The purpose is to highlight the aspects of the environments that constitute the Test environment Deployment scenario TABLE 1 Overview of IMT-A Test Environments Indoor Indoor hotspot Microcellular Urban micro Base coverage urban Urban macro High speed Rural macro Channel model Carrier frequency (GHz) ISD (m) UT speed (km/h) User distribution indoors 5 % outdoors indoors, outdoors in 5 % in vehicles vehicles outdoors BS antenna heights 6 m, mounted on ceiling 1 m, below 5 m, above 35 m, above most significant challenges to reaching the performance requirements, focusing on cell and cell-edge user spectral efficiency. II. IMT-A TEST ENVIRONMENTS AND SCENARIOS Four test environments including five deployment scenarios, each with its own set of channel model parameters, are specified as environments where the technology seeking IMT-A-compliance should be evaluated in. In addition to the environments, parameters that describe system properties are also given, including carrier frequency, output power, antenna properties and noise level figures. See Table 1. The names of the environments capture the areas of applicability: Indoor hotspot, Microcellular, Base coverage urban and High speed. By passing the requirements in all test environments a candidate technology would perform well in a large span of real deployments. Four of the five scenarios have system performance requirements that the candidate technology must fulfill. The fifth scenario is optional and not treated here. Briefly described, the indoor hotspot deployment (InH) is assumed to provide very high bit rates to indoor, slow-moving users. It is set in a large single floor, having two access points with omni antennas at ceiling level and no outer interference sources. The microcellular (UMi) scenario considers a dense /9/ $6. 9 IEEE 8

2 urban setting, with antenna heads on building sides, below roof level, providing capacity to both indoor and outdoor pedestrian users. In this scenario, the propagation model covers aspects of building penetration and can optionally be described by either a Manhattan style square street-like layout or a traditional distance-based path loss. The latter is to be used for evaluation purposes. The third scenario is the urban macro (UMa) deployment. With antenna heads five meters above surrounding building roofs and an inter-site distance (ISD) of 5 meters, the associated channel model is designed to simulate that of a macro-layer network serving city users on street level. All users are here vehicular, moving at 3 km/h. Finally, the high speed case (RMa) is a macro-network with ISD 173 m, covering a rural area with users moving around in cars or trains, all at 1 km/h. The antennas are located high above surrounding building roofs. The three latter scenarios have associated penetration losses for users indoors or in cars. All deployments have rectilinear user motion patterns. More details are given in [1]. III. METHODOLOGY The models and assumptions were implemented in a semidynamic multicell one-reuse system simulator. Here, semidynamic refers to a structure where a user population in a radio network is evolved in discrete time steps. During a time step the channels are assumed constant. The simulator supports any type of Orthogonal Frequency Division Multiplexing (OFDM) based radio interface. No system specific assumptions, other than those specified by ITU in [1], were needed to create the descriptions of the environments. To simulate the system performance of an existing technology, parameters in line with 3GPP s Long-Term Evolution standard was used, see Table. Antennas Transmission scheme Scheduling Power control Link adaptation Uplink overhead Antennas Transmission schemes Scheduling Channel estimation Link adaptation Downlink overhead TABLE SIMULATION PARAMETERS Uplink 1 TX 4 RX, 4 wavelengths separation 1x4 Single Input, Multiple Output (SIMO) Frequency Division Multiple Access (FDMA), channel dependent Fractional power control, P -8dBm, α.8 [4] Error-free but delayed and quantized feedback, MCSs based on LTE transport formats [4] 14.9%, 4 RB for control channels and reference symbols Downlink InH 4 TX, 4 wavelengths separation UMi, UMa and RMa 4 TX,.5 wavelengths separation RX,.5 wavelengths separation InH codebook-based precoding adaptive rank Multiple Input, Multiple Output (MIMO) UMi, UMa and RMa coordinated beam forming (CBF) within site with multiuser (MU)-MIMO Proportional Fair in Time and Frequency Non-ideal channel estimation, Channel Quality Indication (CQI) error per resource block is N(,1) db Non-ideal, based on delayed feedback PUSCH CQI mode 3-1 wideband PMI, frequency selective CQI, according to [4] 3.95%, 3 TTIs per subframe and reference symbols for 4 TX (max overhead) System Models C.D.F. [%] Coupling gain distributions 3 Indoor hotspot Urban micro 1 Urban macro Rural macro Coupling gain (Prx - Ptx) [db] Fig. 1 Coupling gain distributions for the four deployments. The urban micro case stands out with its four user categories: line-of-sight/non-line-ofsight and indoors/outdoors. Indoor users have a db through-wall penetration loss as well as a loss associated with the distance from the outer wall. OFDM parameters According to [3] Modulation and coding Antenna downtilt Receiver scheme Path gain [db] P(LoS) -5-1 LoS and non-los properties of UMi, UMa and RMa UMi LoS 1 UMi non-los UMa LoS 1-1 UMa non-los RMa LoS RMa non-los 1 - QPSK, 16QAM & 64QAM Code rates accoding to [5] 1 degrees for UMi, UMa, 6 degrees for RMa MMSE with branched receiver diversity Distance normalized to cell radii Fig. Path gain and probability of line-of-sight as a function of the distance between antennas in the outdoor scenarios. 81

3 IV. RESULTS A. Coupling gain The distributions of coupling gain, defined as the ratio of received to transmitted power of the signal, are given in Fig. 1 for all four scenarios. The measure includes the effects of antenna gain, shadow fading, path loss and penetration losses. The urban micro deployment has a large dynamic range, over eight orders of magnitude, with its four categories of users, from outdoor line-of-sight to indoor users with severe through-wall losses. The urban macro scenario shows less spread but has a lower average. This captures the behavior of a macro layer addressing vehicular-bound users at street level in an environment with severe scattering and low probability of line-of-sight. B. Line-of-sight (LoS) conditions In Fig., the path gain and probability of LoS as a function of distance from the base station are plotted for the deployments that use traditional three-sector antennas. These scenarios have a non-zero probability of LoS at the cell edge (defined by ITU as ISD/sqrt(3)). Because of the large difference between LoS and non-los path loss for a given distance, if a terminal has LoS to any base station, this will almost always be the link with best gain. Thereby the serving cell is very often not at the site in closest proximity to the terminal. For UMi, the resulting overall probability of being a line-of-sight user is very high. This is visible in a scatter plot of the coupling gain as a function of distance for the two link types, see Fig. 3. Here the indoor users have been omitted for clarity. Of the outdoor users, 77 per cent have line-of-sight to their serving base station, which in 71 per cent of the cases lie outside their geometrical cell belonging. This constitutes a difference to networks with less diverse propagation characteristics than those described in Fig. 3. Average coupling loss [db] Urban micro coupling gain, outdoor users LoS Non-LoS Prediction from channel model Distance to serving base station [m] Fig. 3 Coupling gain as a function of distance to the serving base station for the three categories of users in the urban micro scenario. Black drawn lines represent the coupling gain predicted by channel model data produced by adding the maximum antenna gain of 17 db. Not all users lie in the antenna boresight, hence the realizations are somewhat lower. Of outdoor users, 77 per cent are in line-of-sight to their serving base station. Fig. 4 Combined spherical pattern of the IMT-A antenna model. Users under the vertical lobe will have the same gain to all sectors of the serving site. C. Antenna diagram and geometry The geometry, defined as the received power of serving cell divided by the sum of the received powers to all other cells plus noise, is a measure of wideband SINR. It is used here to assess cell isolation of the hexagonal deployments UMi, UMa and RMa. The InH scenario is omitted from the discussion due to the special characteristic of having only one interferer. In the outdoor scenarios, sectorized antennas are to be used for evaluations. The antenna element pattern is specified [1] by the following formulas θ A ( θ ) = min 1, Am, θ 3 db ϕ ϕ tilt A ( ) e φ = min 1, Am, ϕ 3 db where A is the relative antenna gain in the horizontal (θ) and vertical (φ) angular directions, respectively, A m = db represents maximal attenuation, and the numerators are the Average geometry [db] Geometry versus coupling gain, UMa LoS Non-LoS Average coupling gain [db] Fig. 5 Geometry versus coupling gain in the UMa scenario. The users close to the base station, at -3 db geometry, are highlighted. 8

4 C.D.F. [%] Geometry distributions Indoor hotspot Urban micro Urban macro Rural macro Wideband C/(I+N) [db] Fig. 6 Geometry distributions for the deployment scenarios. Comparing the hexagonal deployments, the RMa shows the best cell isolation, while the urban scenarios have users with severe inter-cell interference. half-power beam widths at 7 and 15. The antenna boresight gain is 17 dbi, and the combined antenna gain is to be computed as A tot = min[ ( A( θ ) + Ae ( φ) ), Am ]. The resulting antenna diagram is depicted in Fig. 4. Compared to measured antenna models [6], the ITU model has a wide effective vertical lobe. The purpose of using this wide a lobe is to model the effects of vertical scattering in the evaluation environments. The same antenna pattern is employed in all three scenarios. An antenna feature of interest is the resulting maximum attenuation sphere of db. Users not in the antenna lobe will be in an area where the geometry is -3 db, as the gains to the non-serving sectors of the site have the same gain as the serving sector. This is clearly visible in a scatter plot of geometry versus coupling gain, see Fig. 5. Measurements from real network deployments do not show this behavior [7]. Instead, the users at -3 db geometry are usually found at the site borders. The geometry distributions of the three hexagonal deployment scenarios, using the IMT-A antenna, are given in Fig. 6. In each scenario, an optimized antenna down tilt is used. The antenna model and the resulting cell isolation represent an important part of the challenges of all three outdoor deployments. D. Main system performance challenges A subset of the system performance requirements (c.f. []) which the proponent technology must meet or exceed are summarized in Table 3. Comparing the outdoor scenarios, the UMi requirements represent the minimum performance to be expected by the system when set in closely spaced, hexagonal cells in a dense urban area. The challenge here will be to address the large range of coupling losses and allowing multistream transmissions to LoS users while simultaneously controlling the severe inter-cell interference due to low cell isolation. The urban macro scenario requirement will have to be fulfilled with similar cell isolation as the micro case, but with the added complexity of vehicle-borne users. In all three TABLE 3 IMT-A Requirements on Spectral Efficiency System spectral efficiency (bits/s/hz/cell) Cell-edge spectral efficiency (bits/s/hz) DL UL DL UL Indoor hotspot Urban micro Urban macro Rural macro scenarios, the downlink requirements are the most challenging because the maximum allowed number of receive antennas on user terminals is limited to two. At the base station, up to eight antennas may be used, allowing for advanced receiver algorithms and interference cancellation schemes to improve uplink performance. This reduces the impact of inter-cell interference from low cell isolation. Conversely, in the downlink these gains are moderate, at best. Network-centric inter-cell interference coordination (ICIC) or more advanced coordination schemes could counteract, or exploit, the low cell isolation. The scenarios with vehicle-bound users (UMa and RMa) are especially challenging to the downlink of any system using user channel feedback reports, as the channel coherence time will be small compared to the time between measurement and next corrected transmission. In fact, due to better overall geometry as provided by the propagation formula, the Rural macro scenario provides better system performance than the Urban macro does in traditional, uncoordinated deployments. Unless the user has line-of-sight, assuming a reporting period in a typical IMT-A concept being in the vicinity of 5-1 ms, the UMa user speed of 3 km/h results in a completely uncorrelated channel before one reporting period is completed, minimizing the differences between the scenarios that concern link adaptation. This is supported by simulations of downlink performance as a function of terminal speed [8], showing that beyond km/h, fast fading can not be followed at a carrier frequency of GHz. Performance does not degrade significantly with increased speed once the reporting period exceeds the channel coherence time. The high probability of line-of-sight in the UMi case could necessitate the use of cross-polarized antennas to achieve a sufficient fraction of multi-stream transmissions. For UMa, tracking the angular velocity of users would be much easier than following the fast fading, implying the use of array antennas, spaced closely together to allow beam forming. A combination of array antennas and spatial or polarization diversity might be beneficial in both cases. As shown, both urban scenarios have low cell isolation. Theoretically, this lends itself well to exploitation by centralized signal processing schemes that form super-cells to coordinate simultaneous transmissions from multiple base stations, given sufficiently detailed channel knowledge of all links. This would create fewer borders of uncontrolled interference. 83

5 Cell-edge user tp [bps/hz] Avg cell tp [bps/hz/cell] ±.%.6.67 ±.8% p ±1.% 1.71 ±.1%.93 ±4.1%.75 ±3.4%.13 ±.8%.11 ±6.% Fig. 7 Uplink performance at 95% confidence interval of a basic LTE setup. The requirements of Table II, represented by the dashed line, are met in all environments. Due to the low level of interference in the indoor environment, the requirements are here easily reached by any technology able to handle the other environments. E. System performance of LTE To assess the system performance in the four test environments, spectral efficiency results of a basic LTE system are shown in Fig. 7 and Fig. 8. Detailed simulation parameters are given in Table. In the uplink, an LTE Release-8 setup with 1x4 SIMO meets the requirements in all environments. In the downlink, the InH results are obtained with uncorrelated transmit antennas and single-user MIMO. For the other environments, the system is set up with four correlated transmit antennas, MU-MIMO and intra-site coordinated beam forming. The performance of this scheme represents work in progress and refined algorithms and parameter tuning may produce different results. Nonetheless, it is clear that the performance in the urban environments has the least margins to the requirements. V. CONCLUSIONS Technology candidates for IMT-Advanced, the next generation of radio networks for cellular systems, will face many challenges from the test environments for system simulations. The models as specified by ITU allow for sophisticated evaluations of system-level performance due to multiple levels of variety, from high-speed to indoor stationary users, a large dynamic range of coupling gains, and channel models with very different fading properties on both large and small scale. As shown, the requirements put on the system performance will be especially challenging for the downlink, while the uplink will be easier. The large variety of models and assumptions of the IMT-Advanced evaluation framework suggests that a technology fulfilling the requirements is well positioned to provide excellent performance in real network deployments. REFERENCES [1] ITU, Guidelines for evaluations of radio interface technologies for IMT-Advanced, ITU-R M.135, [] ITU, Requirements related to technical performance for IMT-Advanced radio interface(s), ITU-R M.134, [3] 3GPP E-UTRA, Physical Channels and Modulation, TS V8.4. [4] 3GPP E-UTRA, Physical Layer Procedures, TS V8.4. [5] E. Dahlman, S. Parkvall, J. Sköld and P. Beming, 3G Evolution, HSPA and LTE for Mobile Broadband, nd edition, Academic Press, 8. [6] F. Gunnarsson, M. N. Johansson, A. Furuskär, M. Lundevall, A. Simonsson, C. Tidestav, M. Blomgren, Downtilted Base Station Antennas - A Simulation Model Proposal and Impact on HSPA and LTE Performance, IEEE VTC-Fall, September 8, Calgary, Canada [7] S. Tenorio, Y. Le Pezennec and M. Sierra, 3G HSDPA evolution: MIMO and 64QAM Performance in Macrocellular Deployments presented at European Wireless 8, Prague, Czech Republic [8] A. Simonsson, Q. Yu and J. Östergaard, LTE Downlink x MIMO With Realistic CSI: Overview and Performance Evaluation, submitted for publication to PIMRC 9 Avg cell tp [bps/hz/cell] ±.3%.8.36 ±.% ±4.9% ±1.9% Cell-edge user tp [bps/hz] ±5.4%.81 ±14.3%.7 ±13.3%.1 ±5.1% Fig. 8 Downlink performance at 95% confidence interval of a basic LTE setup. The dashed line represents the requirements of Table II. The outdoor area requirements are reached with MU-MIMO and coordinated beam forming. 84