REALISTIC scenarios have not been analyzed so far for. Raising Coverage and Capacity using Fixed Relays in a Realistic Scenario

Transcription

1 Raising Coverage and Capacit using Fied Relas in a Realistic Scenario Rainer Schoenen, Wolfgang Zirwas and Bernhard H. Walke Abstract Multihop techniques are known as a practical solution for covering huge radio cell areas when there are onl ver few base stations (BS). This is the case when fiber access is limited and BS CAPEX and OPEX are ver epensive. For WiMAX and GPP-LTE it is possible to operate rela stations which are onl fed over the air link. While having some inherent overhead due to increased radio resource usage, there are nevertheless impressive gains in the coverage compared to a BS alone and also in the capacit of the whole radio cell. Near rela nodes (RN) there is not onl better SINR to the user, which is obvious, but it is often more efficient to associate to a RN instead of the BS, taking into account all resources used for the first and second hop. Therefore this is a low cost measure to increase the sstem efficienc. This has been shown in the literature for artificial scenarios so far. In this paper we stud a realistic scenario using data from topolog information and ratracing. The cit of Jerse was taken as eample. The data is analzed numericall, with all laer- and laer- performance models specified analticall. We stud the case of one base station onl, one BS with four RNs, and the latter plus another ring of nine RNs. The BS has fiber access, while the first hop of Relas (H) is fed over the air from BS and the second hop H is fed b the RNs of group H. We present results for the gains in coverage and capacit that are obtained b these Multihop techniques. Inde Terms Relaing, Multihop, LTE, Coverage, Capacit I. INTRODUCTION REALISTIC scenarios have not been analzed so far for the use of fied rela nodes in future OFDMA-based transmission schemes like GPP-LTE [] or WiMAX []. The benefit for the coverage and capacit in rela enhanced radio cells was shown in the past with free-space propagation or Manhattan Grid []. Other related work in this area mostl analzes regular cellular geometries without considering realistic pathloss due to obstructions [] []. In this paper we build upon D models of the cit of Jerse [] with ratracing tools [8], stud the scenario for D and scalar performance measures and etend the analsis for realistic areas around the BS. Compared to abstract and regular cellular scenarios [9], this provides a good proof-of-concept. The benefit of Multihop is the cost-efficient provisioning of radio access over the area, no matter if the BS is in lineof-sight or not. Using RNs is viable at much lower cost than providing more base stations due to their need for high-rate fied network access at ever location. Due to the limited power of transmitters and the high path loss in non-line-ofsight conditions the received signal strength is not sufficient Wolfgang Zirwas is with Siemens, Munich, German The author is with the Chair of Communication Networks at RWTH Aachen Universit, Facult, German This work is funded b the BMBF ScaleNet project in German in man urban areas. To improve the coverage therefore more sending stations are needed, which can be full BSs or rela nodes (RN s of store-and-forward tpe). OFDMA-based transmission allows to coordinate radio resources in time and frequenc domain. In an rela enhanced cell we assume a full coordination, so that there is no intra-cell interference. Therefore, a resource block can onl be used b one of the actors, either BS or one of the four RN H or one of the nine RN H. For this reason, an traffic PDU that goes from BS to a user terminal (UT ) associated to RN consumes three resource blocks R, R, R, one on each hop,,. The size of R i in bits shall be the same, but the size in terms of time and frequenc bandwidth (T F ) depends on the modulation&coding scheme (P hm ode) on this subchannel. Close to the sender, the higher received SINR value allows the highest PhMode, i.e. the highest data rate. At the cell border the offered data rate is one order of magnitude lower (QPSK / compared to QAM / for LTE []). A terminal operating at the lowest PhMode occupies a ten times higher part of the spectrum than a terminal operating at the highest PhMode. That means the average cell capacit is overproportionall determined b the maimum possible rate at the outer regions. The paper is structured as follows. Section II defines the scenario. Net, the used laer (PHY) and laer (DLC) models are eplained. Numerical results presented in the net section, which also provides D result maps of the area. The last section ends with a concluding summar. II. SCENARIO DEFINITION Jerse has been chosen for this scenario because the area of appro..9km is a tpical cell size. B using a radio network planning tool, base stations and all rela nodes are placed at best possible locations, see Fig.. The topolog (building placement) was known in advance and ratracing tools have been used to obtain the received signal power P R,i at each location from each possible transmitter site. The determination of the BS and RN locations (above roof top) was performed iterativel with a radio planning and placement tool. The phsical parameters and calculation steps are given in section III. In the end a coverage map was obtained which shows for each point on the map which serving station a terminal is best associated to (best server). This can be the base station, if this is the rate optimal association, but it can also be one of the relas RN H or RN H. The maimum achievable rate at a certain point is then determined. From

2 PhMode modulation code rate SINR min in [db] QPSK /.9 QPSK /. QPSK /.8 QAM /. QAM / 9.8 QAM /. QAM /. 8 QAM / 8. TABLE I LTE PhModes AND THEIR REQUIRED SINR min Fig.. The Scenario map of Jerse showing the BS (middle) and RN placement as well as the polgon of interest (green) MI [bit/s/hz] QPSK / QPSK / QPSK / QAM / QAM / QAM / QAM / QAM / QAM / shannon these matri (D) results we calculate scalar performance measures like the overall coverage and capacit, which are increased compared to using onl one BS. Therefore we compare the scenario with ) one BS onl, ) the BS plus a ring of four RN H and ) the BS plus a ring of four RN H plus a second ring of the nine RN H. III. PHY AND DLC MODELS From link level to MAC throughput, the performance of the eample sstem is evaluated b calculating the following steps. Transmit Power: dbm at the BS, dbm at the RN, Bandwidth: b = 8MHz net (MHz sstem), Frequenc:.GHz apropriate for LTE or WiMAX, Pathloss I: D model of the cit scenario (walls of buildings), Pathloss II: ra tracing to capture multi-path propagation, Noise: Thermal noise power is N = 9.dBm, SINR: the first performance measure below PHY laer, MI: mutual information determined from SIN R and modulation (Eq. ), BER: bit error ratio, the PHY performance result, PER: packet error ratio, the result after channel decoding, Throughput: determined b bandwidth, PhMode (modulation and code rate), ARQ overhead (Eq. ), Second Hop Throughput: reduced b resources required on first hop (Eq. ). Third Hop Throughput: reduced b resources required on first and second hop (Eq. ). A. PHY laer The received power P R,i on ever location is the output of software tools for ratracing. Indoor coverage was modeled b assuming wall thinkness and attenuation. The net SINR [db] Fig.. Link level curves for the LTE modulation&coding schemes (PhMode). QAM is not used et. steps were analzed analticall-numericall using Matlab. With SINR = P R,i /(N + I) the signal to noise ratio is easil determined. For each SIN R level between around and there in another P hm ode chosen, depending on the estimated performance of this P hm ode in terms of bis/s/hz. For determining the required link level results we build upon the mutual information (MI) method []. We appl the steps SINR MI, MI BER and BER PER to get the packet error probabilit. For the SINR MI approimation. was used [9]: MI shannon (SINR) = log ( + SINR/dB ) () MI(SINR, m) = ([s MI shannon (SINR)] w + m w ) /w () s = s(m) =.9.8 (m mod ) () w = w(m) = m + () m is the modulation inde, i.e. the number of bits per smbol (=QPSK,...8=QAM). Figure shows the outcome, taking also the coding rate into account to get the PHY throughput. LTE coders have rates /, /, / and / []. The PhModes in this figure are given in Table I (QAM was not used). Within this cell, all RNs are coordinated b the BS, so there is no intra-cell interference.

3 Scenario coverage [%] capacit [Mbit/s] spec.eff.[bit/s/hz] BS onl...9 BS+H BS+H+H TABLE II SCALAR RESULTS FOR THE RELAY SCENARIO WITHIN THE SQUARE AREA Scenario coverage [%] capacit [Mbit/s] spec.eff.[bit/s/hz] BS onl BS+H BS+H+H TABLE III SCALAR RESULTS FOR THE RELAY SCENARIO WITHIN THE POLYGON AREA Fig.. The percentage of each station group used as a function of the cell radius from..8m B. DLC laer On DLC (MAC) laer, there is an overhead due to framing, signaling and ARQ retransmissions. The latter depends on P ER, which can be taken into account when assuming selective repeat ARQ b equation. r abovearq = r belowarq ( PER) () In total we obtain a MAC overhead of M AC/P HY =.%. The PHY overhead of P HY/RAW =.% comes from OFDM cclic prefi duration. Under multihop operation there are individual resources needed on ever hop. The constant packet length requires a different resource share depending on the used PhMode which determines the maimum rate r i,ma usable on each hop. Therefore we can get the maimum rate on the second hop to be and on the third hop r = (r,ma + r,ma ) () r = (r,ma + r,ma + r,ma ) () For ever location (, ) we can now determine the best rate out of r, r, r which gives us the result in Figure. One of the three rates is maimum and the inde i of the maimum r i determines the best server, i.e. it shows which station the UT at that location should be associated with. The performance results have been obtained over different areas. First, the full area was used, including less densel populated parts ( square ). Second, a circular area with a radius of 8m around the BS was defined, which gives a realistic cell size ( circle ). Third, the area within a polgon (Fig. ) was studied (the urban populated area). IV. ANALYSIS RESULTS The analsis has been carried out to generate the twodimensional data in fig. and. Scalar results were onl counted in the polgon area defined in Fig.. Here we show the downlink onl that also applies to the uplink if the uplink pathloss is the same (FDD). But the benefit in terms of capacit reveals if we derive scalar performance measures from it. The coverage (in % of the area) of the scenarios differing b the number of relas involved is determined b counting all locations with SINR > SINR min. For LTE SINR min =.9dB holds. Figure (a) shows the coverage of each scenario. The sstem capacit is determined b assuming equal traffic load for each user terminal and a homogeneous user densit over the area. This means that a UT far outside, having a low PhMode, requires more share of the resources than a UT close to the BS. The following equation [] for the capacit C considers this: C = polgonarea dd (8) Capacit(, ) Figure (b) shows the capacit of each scenario. Figure shows that in a multihop scenario, more and more of the coverage area of BS is taken over b RNs. The capacit C in bit/s can be used to calculate the spectral efficienc e = C/b using the used bandwidth b. The performance metrics are shown in Table II for the full area and in Table III for the densel populated polgon. According to this, two tiers of relas, compared to the BS onl scenario, increase the coverage b a factor of.8 overall and. within the cit, and both the capacit and spectral efficienc b a factor of approimatel. both overall and within the cit. Figure shows the partition of area which is served b either BS, H or H dependent on the radius of the cell. We observe, that a first ring of relas comes in earl, and a second ring is worth to use beginning from m. These numbers show the benefit of using relas. The radio power sum is of course more here, but the epensive fiber access to the BS is saved, which is the limitation in earl rollout phases. V. CONCLUSION In this paper a realistic urban scenario was treated offering numeric results for the coverage and capacit of a multihop radio cell. We analzed the topologic structure of the cit to get the path loss, SINR and data rate values for ever location in that area. Net the coverage (in %) and capacit (in Mbit/s) as well as the spectral efficienc were derived from that, calculated for the whole sstem, but counted onl within a bounded polgon area. Doing this analsis for three scenarios, one with a base station onl, one with two hops (one tier of relas) and one with three hops (two tiers of relas),

4 (a) coverage using relas (b) capacit using relas Fig.. Coverage and capacit compared for three scenarios: BS (one BS onl), BS+H (with one tier of relas), BS+H+H (BS with two rela hops H+H). Three different integration areas are the parameter. Fig.. The fraction of area served b each group of servers, determined within the polgon area remarkable gains were visible when using relas. There is onl one interface to the fied network in all three scenarios. All the relas help to cover areas behind obstructions (buildings) where the path loss from the BS would be too large. The multihop techniques also improve the spectrum efficienc, because for all covered areas there is onl one radio channel needed. The alternative was to place several BSs to serve the same area. But this requires more channels, since a cluster order of one suffers from too much interference. One outlook is an economic comparison of the CAPEX and OPEX related to the scenarions, e.g. []. Another outlook is the analsis of multiple cells with interference effects at the border. REFERENCES [] [] [] R. Pabst, B. Walke, D. C. Schultz, and et al, Rela-Based Deploment Concepts for Wireless and Mobile Broadband Radio, IEEE Communications Magazine, pp. 8 89, Sep. [] V. Sreng, H. Yanikomeroglu, and D. Falconer, Coverage enhancement through two-hop relaing in cellular radio sstems, in Proc. IEEE Wireless Communications and Networking Conference(WCNC ),. [] H. Yanikomeroglu, Fied and mobile relaing technologies for cellular networks, in nd Workshop on Applications and Services in Wireless Networks (ASWN ), - Jul, pp. 8. [] R. Schoenen, J. Eichinger, and B. Walke, On the OFDMA FDD mode in G-LTE, in Proceedings of the th International OFDM Workshop (InOWo ), Hamburg, German, Aug. [Online]. Available: [] R. Schoenen, W. Zirwas, and B. Walke, Capacit and Coverage Analsis of a GPP-LTE Deploment Scenario, in Proceedings of the Broadband Wireless Access Workshop BWAW 8 colocated with ICC 8, Beijing, China, Ma 8. [Online]. Available: [8] M. Lott, M. Weckerle, and et al, Hierarchical cellular multihop networks, in Proceedings of the IEEE EPMCC,. [9] R. Schoenen and B. Walke, On PHY and MAC performance of G-LTE in a multi-hop cellular environment, in Proceedings of the rd IEEE International Conference on Wireless Communications, Networking and Mobile Computing (WiCOM), Shanghai, China, Sep. [Online]. Available: [] K. Brueninghaus and D. e. a. Astel, Link performance models for sstem level simulations of broadband radio access sstems, in Proceedings of the th Annual IEEE International Smposium on Personal, Indoor and Mobile Radio Communications, Sep, pp.. [] K. J. e. a. Ekstrom H., Furuskar A., Technical solutions for the G long-term evolution, IEEE Communications Magazine, pp. 8, Mar. [] C. Homann and S. Goebbels, Dimensioning cellular wima part i: Singlehop networks, in Proceedings of European Wireless, Paris, France, Apr, p.. [Online]. Available: http: // [] Bogdan Timus, Deploment Cost Efficienc in Broadband Deliver with Fied Wireless Relas, Ph.D. dissertation, KTH Stockholm,.

5 8 (a) SINR of BS onl (in db) (b) PhMode of BS onl 8 (c) SINR of BS with H (in db) (d) PhMode of BS with H 8 (e) SINR of BS with H and H (in db) (f) PhMode of BS with H and H Fig.. On the area map of Jerse, these figures show the SINR [db] and PhMode [..8] for a scenario with BS onl, with one tier of relas H, and with two rela hops H+H

6 (a) Rate of BS onl (b) Best server (coverage) of BS onl (c) Rate of BS with H (d) Best server (coverage) of BS with H (e) Rate of BS with H and H (f) Best server (coverage) BS with H and H Fig.. On the area map of Jerse, these figures show the available rate capacit [bit/s] and best server (middle==bs, =H, =H) for a scenario with BS onl, with one tier of relas H, and with two rela hops H+H