FADING DEPTH EVALUATION IN MOBILE COMMUNICATIONS FROM GSM TO FUTURE MOBILE BROADBAND SYSTEMS

Transcription

1 FADING DEPTH EVALUATION IN MOBILE COMMUNICATIONS FROM GSM TO FUTURE MOBILE BROADBAND SYSTEMS Filipe D. Cardoso 1,2, Luis M. Correia 2 1 Escola Superior de Tecnologia de Setúbal, Polytechnic Institute of Setúbal, Setúbal, Portugal, fcardoso@est.ips.pt 2 Instituto de Telecomunicações, Instituto Superior Técnico, Lisboa, Portugal, luis.correia@lx.it.pt Abstract A technique for fading depth characterisation in wideband mobile communications systems is described. The approach, based on the eigenvalue decomposition technique, allows deriving the cumulative distribution functions of the received power for Rayleigh and Ricean fading channels. Application examples for different systems, namely GSM, UMTS, HIPERLAN and MBS, working within different environments, as proposed by standard-setting bodies, are shown in detail, and the results are discussed, giving a general insight into the fading depth behaviour. In the considered environments, the fading depth experienced by GSM is between 9.4 and 18.4 db. The fading depth observed by UMTS is below the one for GSM by 2.1 to 10.3 db. Regarding HIPERLAN, the observed fading depth is 2.6 to 4.9 db below the one for UMTS. MBS experiences fading depths that are usually below 9.7 db (a worst case of 10.6 db is observed in City Streets under NLoS). Since the observed fading depth depends on the system and environment characteristics, different fading margins should be considered for each system and working environment. Keywords Mobile Communication Systems, Wideband, Time-domain, Short-term Fading, Fading Depth. I. INTRODUCTION The performance of a communications system operating in a fading environment depends on a huge set of parameters, namely the received power. Since the received signal fades, it is necessary to provide additional power (fading margin) in order to achieve the desired link quality. Moreover, the necessary short-term fading margin depends on the system bandwidth; hence, for different systems working within typical environments, different fading margins are needed in order to achieve the desired link quality. Until nowadays, much of the research was based on the assumption of analysing narrowband signals, whose bandwidth is much narrower compared to the coherence bandwidth of the propagation channel. Under these conditions, the signal at the receiver is said to be in a frequency flat fading environment, and varies usually with the Rayleigh or Ricean distributions, depending on the existence of a Line-of-Sight (LoS) component. However, when one considers wideband signals, whose bandwidth is greater than the coherence bandwidth, the signal at the receiver is distorted, but the fading depth is smaller than the one obtained from the Rayleigh or Ricean distributions. This decrease in the short-term fading depth should be taken into consideration, since it produces a more accurate evaluation of the power budget, hence, a better radio network planning. A frequency-domain theoretical model for the study of wideband signal transmission is presented in [1]. The model incorporates physical parameters, such as the number and amplitude of arriving waves, the propagation path length, the signal bandwidth, and the carrier frequency. An approach for the study of the short-term fading depth dependence on system bandwidth and environment specific features is proposed in [2]. An analytical expression for the evaluation of fading depth, accounting for the maximum difference in propagation path length, and having the Rice factor as a parameter, is derived from fitting simulated data from the model in [1]. A time-domain level variation analysis technique of wideband signals in Rayleigh fading environments is proposed in [3], the approach being based on the eigenvalue decomposition technique. In this paper, a time-domain technique for fading depth characterisation in wideband Rayleigh and Ricean environments is described [4]. The short-term fading depth is evaluated from the Cumulative Distribution Function (CDF) of the received power, and represented as a function of the Rice factor, and the product between the system bandwidth and the rms delay spread of the propagation channel, as suggested in [3]. Results for the fading depth observed by GSM, UMTS, HIPERLAN and future Mobile Broadband Systems (MBS) [5] are presented and discussed. Moreover, different fading margins are proposed for each system and working environment. This paper is organised as follows. The time-domain technique for fading depth characterisation in wideband Rayleigh and Ricean environments is presented in Section 2. The fading depth behaviour for GSM, UMTS, HIPERLAN and MBS, working within different environments, characterised by their discrete Power Delay Profiles (PDPs), is shown in Section 3. Conclusions are drawn in Section /02/$ IEEE PIMRC 2002

2 II. FADING DEPTH CHARACTERISATION Assuming that the PDP of the propagation channel can be described by a set of arriving waves from distinct paths, each wave is the sum of several ones arriving so close to one another that they are not distinguishable by the receiving system. Thus, the receiver s capability to discriminate waves arriving with different delays depends on its resolution, which is inversely proportional to the system bandwidth. If there is Non-Line-of-Sight (NLoS) between the Base Station and the Mobile Terminal, all amplitudes of the arriving waves are Rayleigh distributed, otherwise, the first arriving wave is usually modelled as Ricean [6]. As described in [4], starting from the continuous or discrete PDP of the propagation channel, a covariance matrix whose elements are given by the product among the correlation function between different frequency components, the frequency response of the filter used in the transmitting equipment, and an incremental bandwidth that depends on the system bandwidth, is generated. Performing the eigenvalue decomposition of the covariance matrix, the obtained eigenvalues, m, correspond to the decomposition of the signals that fade incoherently. The respective decomposed signals and their powers correspond to the eigenvectors and eigenvalues respectively. The PDF of the received power, s, is then obtained from the convolution of the PDF of the LoS component and the scattered ones, and represented as a function of the obtained eigenvalues [4]: 2K p( s) e 2 a d K 2x1 2 ad K x M M λm e I 0 a M d m2 λ λ m k2 km sx m where a d is the amplitude of the LoS component, K is usual Rice factor, and I 0 is the modified zero-th order Bessel function of the first kind. Since there is no closed-form expression for (1), the evaluation of the CDF of the received power has to be performed by means of numerical methods. The fading depth is evaluated as the difference in power corresponding to 50 and 1% of the CDF and it is represented as a function of K, and the product between the system bandwidth and the rms delay spread of the propagation channel, B. [4]. III. SIMULATION RESULTS A. Initial Considerations In practice, the PDP is a continuous function of the delay. There are several types of PDPs that have been used for modelling the propagation channel. These models are not intended to cover all possible operating environments. k dx (1) Instead, they are designed in order to span the overall possible range of environments. The exponential and two-stage exponential ones are commonly used for the evaluation of mobile communication systems [6]. However, since most of the channel simulators are usually based on a tapped-delay line structure, discrete propagation models, derived from continuous ones, are usually recommended by standard-setting bodies for simulating the propagation channel [6]. In the next sub-sections, one presents results on the fading depth observed by different systems working within different environments, as proposed by standard-setting bodies. One also evaluates the observed fading depth whenever the considered system bandwidth is below the one for which the models are designed for, hence, one also evaluates the fading depth observed by GSM with the models for UMTS and HIPERLAN. A similar approach is used for UMTS. In order to properly understand if for a given environment small variations on the propagation conditions (i.e., on the value of the rms delay spread) are significant regarding the observed fading depth, one evaluates B.. Globally, one can state that this dependence is significant when the system bandwidth is above the coherence bandwidth of the propagation channel, i.e., B. > 0.02 Hz. s (e.g., Fig. 1). In this situation, the propagation channel is usually referred as a wideband one. For lower values of B., narrowband case, the fading depth is practically constant and almost independent of B.. B. GSM Propagation Models There is a large set of channel models for GSM that has been proposed by the Joint Technical Committee for Personal Communication Systems (JTC/PCS) and the European Telecommunications Standards Institute (ETSI), the latter ones being the most commonly used for system evaluation purposes [7]. In to order to evaluate the fading depth observed within different environments, one has to evaluate B., for each environment, as presented in Table 1. One should recall that the GSM bandwidth is 200 khz. Table 1 Value of and B. for GSM environments. Environment [ns] B. [Hz. s] Rural Area Typical Urban Hilly Terrain As one can observe from Table 1, for Rural Area environments B. = 0.02 Hz. s, hence, it is expected that the observed fading depth does not depends significantly on small variations of the propagation conditions, which affects the value of. For Typical Urban and Hilly Terrain environments, the observed fading depth is more sensitive to

3 environment changes, since B. > 0.02 Hz. s. Globally, one can state that, for the same system bandwidth, the fading depth decreases with increasing values of. Since one has the value of B. for each environment, one can evaluate the observed fading depth, as illustrated in Fig. 1 and presented in Table 2. From Fig. 1, one observes that the fading depth remains constant for large values of B.. This is due from considering a discrete PDP; hence, the fading depth remains constant for a system bandwidth above the one that is larger enough in order to discriminate all arriving waves. One must remember that discrete PDPs are simple realisations of the real ones; therefore, this behaviour is an approximation error, rather than an inherent characteristic of the propagation channel. Fig. 1. Fading depth observed by GSM, Rural Area. Table 2 Fading depth observed in GSM environments. Environment NLoS K = 6 db Rural Area Typical Urban 12.0 Hilly Terrain 10.3 As one can see from Table 2, under NLoS, the fading depth observed by GSM ranges from 10.3 db in Hilly Terrain environments to 18.2 db in Rural Areas. However, if one considers the existence of LoS, with K = 6 db, the fading depth observed in Rural Area environments decreases to 11.2 db. For Typical Urban and Hilly Terrain environments only NLoS is considered, since the proposed channel models are not intended for LoS. C. UMTS Propagation Models Using the channel models for UMTS, as proposed by ETSI [8] and Third Generation Partnership Project (3GPP) [9], one can compare the observed fading depth, namely for GSM and UMTS (for the latter the system bandwidth is 5 MHz). The results for B. for each system and environment are presented in Table 3. Table 3 Value of and B. for UMTS environments. Environment [ns] B. [Hz. s] GSM UMTS Indoor A Pedestrian A Vehicular A Indoor B Pedestrian B Vehicular B Rural Area Typical Urban Hilly Terrain From Table 3, one observes that UMTS behaves as a wideband system in all kind of environments, since B. > 0.02 Hz. s, hence, the observed fading depth becomes significantly dependent on environment changes that affect the value of. Regarding GSM, one concludes that this effect is not significant in Indoor, Pedestrian A and Rural Area environments. For the remaining classes of environments, GSM behaves like a wideband system, thus, it experiences lower fading depths; however, it is more sensitive to environment changes that affect propagation. The fading depth observed in different environments is presented in Table 4. For the Vehicular B environment, only the results for NLoS are presented, since this model is not intended to be used if LoS is assumed. Table 4 Fading depth observed in UMTS environments. Environment GSM UMTS NLoS K = 6 db NLoS K = 6 db Indoor A Pedestrian A Vehicular A < 7.4 < 5.9 Indoor B Pedestrian B < 6.3 < 5.2 Vehicular B 10.3 < 7.7 Rural Area Typical Urban < 5.1 < 4.3 Hilly Terrain Under NLoS, GSM experiences fading depths between 10.3 (in Vehicular B environments) and 18.4 db (in Indoor A and Pedestrian A ones). One must note that these values are close to the ones obtained with the GSM channel models. Fading depths between 9.4 for Hilly Terrain and 11.2 db for Indoor, Pedestrian A and Rural Area environments are observed for K = 6 db. The difference in fading depth between LoS and NLoS is usually lower than 7.2 db.

4 The fading depth observed by UMTS, under NLoS, ranges from less than 5.1 in Typical Urban environments to 12.5 db in Indoor A ones. Under LoS, and K = 6 db, the fading depth observed by UMTS is between less than 4.3 db verified in Typical Urban environments and 9.1 db in Indoor A and Pedestrian A ones. The difference in fading depth between LoS and NLoS is usually below 3.4 db. From the difference in fading depth observed between the LoS and NLoS situations, one concludes that UMTS is less sensitive to the existence of LoS. Moreover, the fading depth experienced by UMTS is usually between 2.6 to 9.5 db and 2.1 to 5.9 db below the ones for GSM, for the NLoS and LoS cases, respectively. Therefore, one concludes that the fading margin for UMTS can be considerably reduced when compared to the one for GSM. D. HIPERLAN Propagation models Five channels models A, B, C, D and E were designed for HIPERLAN/2 simulations in different environments [10]. The product B. for each environment is presented Table 5. The considered system bandwidth is 23.5 MHz. Model A corresponds to a typical office environment. Model B corresponds to a typical large open space environment with NLoS conditions or an office environment with large delay spread. Models C and E correspond to typical large open space indoor and outdoor environments with large delay spread. Model D corresponds to LoS conditions in a large open space indoor or an outdoor environment. Table 5 Value of and B. for HIPERLAN environments. Environment [ns] B. [Hz. s] GSM UMTS HIPERLAN Model A Model B Model C Model D Model E Table 6 Fading depth observed in HIPERLAN environments. Environ. GSM UMTS HIPERLAN NLoS K = 6 db NLoS K = 6 db NLoS K = 6 db Model A Model B Model C Model D Model E The observed fading depth is presented in Table 6. As one can observe, under LoS, which corresponds to the situation for which HIPERLAN is mainly intended to work, it experiences a fading depth between 3.6 and 5.2 db. Under NLoS, the fading depth ranges from 3.6 to 6.3 db. Hence, in the considered environments, the difference in fading depth between LoS and NLoS is usually below 1.1 db. The fading depths observed by GSM and UMTS are usually lower than the ones obtained with the UMTS Indoor channel models. Concerning the fading margin for HIPERLAN, one concludes that it can be lower than the one for UMTS, since the experienced fading depth is usually between 2.6 to 4.9 db and 2.6 to 3.3 db below the one for UMTS, for the NLoS and LoS cases, respectively. E. MBS Propagation Models Besides some experimental PDPs obtained from experimental measurements in different environments, proposed channel models for MBS are not available yet. Moreover, a proper system specification is far from being proposed. From experimental measurements in [11] it seems that the PDP of the propagation channel in MBS systems can be modelled by exponential and two-stage exponential PDPs. Thus, in order to derive some preliminary results for the fading depth observed for MBS, one assumes an exponential PDP, as described in [4]. Two different system bandwidths are considered, 50 and 100 MHz for MBS1 and MBS2, respectively. Typical values of rms delay spread within different environments are extracted from measurements in [11]. The measurements were made in different environments, namely City Streets, City Squares, Small Rooms and Corridors. Carrier frequencies of 59 and 62.5 GHz were used for outdoor and indoor measurements, respectively. A complete description of the experimental measurements setup can be found in [11]. The variation range of the rms delay spread for each class of environments and the product B. are presented in Table 7. As one would expect, MBS is clearly a wideband system, since B. >> 0.02 Hz. s, independently of which environment is considered, hence, being sensitive to environment changes that affect propagation parameters. Table 7 Value of B. for MBS environments. Environment [ns] B. [Hz. s] MBS1 MBS2 City Street City Square Small Room Corridor Using the curves for the fading depth dependence on B. that correspond to the case of an exponential PDP [4], one derives the fading depth variation range, Table 8.

5 Table 8 Fading depth observed in MBS environments. Environment MBS1 MBS2 NLoS K = 6 db NLoS K = 6 db City Street City Square Small Room Corridor As one can observe from Table 8, under LoS, the fading depth observed by MBS1 ranges from 3.2 to 8.0 db in outdoors, and 5.1 to 7.4 db in indoors. If one considers MBS2 the fading depth is usually between 0.7 to 1.4 db below the ones for MBS1. Under NLoS, there is an increase in fading depth of about 0.2 to 2.6 db and 0.1 to 1.7 db for MBS1 and MBS2 respectively. IV. CONCLUSIONS A time-domain analysis technique of wideband signals, in Rayleigh and Ricean fading environments, is used to derive the cumulative distribution function of the received power for various fading channels, whose PDPs are expressed as continuous or discrete functions of the delay. There are some similarities regarding the fading depth observed in different environments. For values of B. below 0.02 Hz. s the observed fading depths are similar, hence, independent of the PDP of the propagation channel. This corresponds to a situation where the system bandwidth is below the coherence bandwidth of the propagation channel, thus, signals are in a frequency flat fading environment. For large values of B., the fading depth depends on the type of the PDP, namely the number, amplitude and delay of the arriving waves. The fading depth experienced by GSM is between 9.4 and 18.4 db. For UMTS fading depths range from less than 4.3 to 12.5 db. HIPERLAN experiences fading depths between 3.6 and 6.3 db. In the considered environments, and for the given system bandwidths, the fading depth observed by MBS is usually between 2.4 and 9.7 db (a worst case of 10.6 db is observed in City Streets under NLoS). As a final conclusion, one can state that, since the observed fading depth depends on the system bandwidth and environment specific features, one should consider different fading margins according to different system bandwidths and working environments when performing power budget calculations. Hence, one can say that the fading margins for UMTS, HIPERLAN and MBS can be significantly reduced, when compared to the ones for GSM, while achieving the desired link quality; this enables more efficient and less costly radio network planning. REFERENCES [1] S. Kozono, S. Seino and H. Nakabayashi, Received Signal-Level Characteristics of Wide-Band Transmission in Mobile Communications, in Proc. of PIMRC 96 7 th IEEE Intern. Symp. on Personal, Indoor and Mobile Radio Communications, Taipei, Taiwan, Oct [2] F.D Cardoso and L.M. Correia, An Analytical Approach to Fading Depth Dependence on Bandwidth for Mobile Communication Systems, in Proc. of WPMC 01 The 4 th Intern. Symp. on Wireless Personal Multimedia Communications, Aalborg, Denmark, Sep [3] T. Inoue and Y. Karasawa, Theoretical Analysis on the Level Variation Properties of Wideband Signals in the Rayleigh Fading Environment, in Proc. of PIMRC th IEEE Intern. Symp. on Personal, Indoor and Mobile Radio Communications, Osaka, Japan, Sep [4] F.D. Cardoso and L.M. Correia, A Time-domain Technique for Fading Depth Characterisation in Wideband Mobile Communication Systems, in Proc. of XXVIIth General Assembly URSI 2002, Maastricht, The Netherlands, Aug [5] R. Prasad, Universal Wireless Personal Communications, Artech House, London, UK, [6] L.M. Correia (ed.), Wireless Flexible Personalised Communications, John Wiley, London, UK, [7] K. Pahlavan and A.H. Levesque, Wireless Information Networks, John Wiley, London, UK, [8] ETSI, Selection procedures for the choice of radio transmission technologies of UMTS, TR v3.1.0, ETSI, Sophia Antipolis, France, [9] 3GPP, Deployment aspects, TSG RAN WG4, TR v0.1.0, Sophia Antipolis, France, [10] J. Medbo and P. Schramm, Channel Models for HIPERLAN/2 in Different Indoor Scenarios, EP BRAN 3ERI085B, ETSI, Sophia Antipolis, France, Mar [11] S.A. Mohamed, G. Lovnes, E. Antonsen, R.H. Raekken, B. Nigeon and J.J. Reis, Report on Propagation Measurements, RACE- MBS Deliverable, R2067/BTL/2.2.2/DS/P/031.a1, RACE Office, Brussels, Belgium, Nov