Practical Introduction to LTE Radio Planning

Practical Introduction to LTE Radio PlanningJ. Salo, M. Nur-Alam, K. ChangThis paper reviews basics of radio planning for 3GPP LTE. Both coverage-limited and interference-limitedscenarios are considered. For the coverage-limited scenario LTE link budget is compared to that of 3GPP Release 8HSPA+ with 2x2 MIMO. It is shown that, given the same 5MHz bandwidth, both systems have similar cell rangesbut, for a given target bit rate, there exists an optimum LTE system bandwidth that maximizes cell range in bothuplink and downlink. For the interference-limited scenario (with random uncoordinated interference) we illustratethe relationship between average network load, cell edge target throughput and cell range, as well as the notionof interference margin for cell range dimensioning. Impact of base station antenna configurations on dual-streamMultiple-Input Multiple-Output (MIMO) performance is demonstrated by means of a real-world measurementexample. The impact of advanced LTE radio resource management features are briefly reviewed. Finally, the mostimportant radio parameter planning tasks are introduced.1. INTRODUCTIONThis paper introduces LTE from the perspectiveof radio network planning. The paper ismainly targeted for readers with earlier experiencein radio planning and mobile communications.Some prior knowledge of radio engineeringand LTE is assumed as principles of OFDMAand SC-FDMA, as described in 3GPP LTE specifications,will not be reviewed in this paper. Insteadthe reader is referred to well-known literaturereferences [1–3]. The most important LTEradio interface parameters are summarized inTable 1 for the convenience of the reader. Ourfocus is on the FDD variant of LTE, althoughmost of the discussion is also applicable to TDD.AbbreviationsBCCH Broadcast ChannelCQI Channel Quality IndicatorFDD Frequency Division MultiplexingHARQ Hybrid Automatic Repeat RequestHS-DSCH HSDPA Downlink Shared ChannelLTE Long Term EvolutionPCI Physical Cell IdentityPDCCH Physical Downlink Control ChannelPDSCH Physical Downlink Shared ChannelPMI Precoder Matrix IndicatorPRACH Physical Random Access ChannelPRB Physical Resource BlockPUSCH Physical Uplink Shared ChannelRS Reference SignalSINR Signal-to-Interference-Noise-RatioTDD Time Division MultiplexingTTI Transmission Time Interval1

2 J. Salo et alTable 1Summary of main LTE radio interface parametersQuantity LTE DL LTE ULSystem bandwidth 1.4, 3, 5, 10, 15, 20MHz 1.4, 3, 5, 10, 15, 20MHzMultiple access OFDMA single carrier FDMACyclic prefix 4.7 microsec / 16.7 microsec 4.7 microsec / 16.7 microsecModulation QPSK, 16QAM, 64 QAM QPSK, 16QAM (64 QAM for Cat5 UE)Channel coding ⋆ turbo coding turbo codingHARQ 8 processes, up to 7 retransm./proc 8 processesPower control none ♮ cell-specific and UE-specificHandover hard, network-triggered hard, network-triggeredNum of Tx antennas 1, 2, 4 ‡ 1 ♯Num of Rx antennas arbitrary ≥ 2Transmit diversity Space-Frequency Block Code transmit antenna selection diversityBeamforming 3GPP codebook or proprietary noneSpatial multiplexing open-loop, closed-loop none †PRB allocation per TTI distributed or contiguous contiguous ♭⋆ For PDSCH and PUSCH. For L1/L2 control and common channels other coding schemes can be used.♮ Downlink Only static Reference Signal power boosting has been specified by 3GPP, vendor-specific implementationspossible.‡ With UE-specific reference signals (vendor-specific beamforming) arbitrary number of Tx antennas.♯ For simultaneous transmission. Transmit antenna selection is possible with RF switching.† Multiuser MIMO possible in UL in 3GPP Release 8.♭ Intra-TTI and inter-TTI frequency hopping is possible in UL.2. LTE COVERAGE IN NOISE-LIMITEDSCENARIO2.1. Definition of average SINRProbably the most useful performance metricfor LTE radio planner is the average signal-tointerference-and-noiseratio, SINR ave , defined asSINR ave = SI+N , (1)where S is the average received signal power, Iis the average interference power, and N is thenoise power. In measurement and simulationanalysis, the sample averages should be takenover small-scale fading of S and I and over alarge number of HARQ retransmissions (severalhundreds of TTIs, preferably). The average interferencepower can be further decomposed asI = I own + I other , (2)where I own and I other are the average own-celland other-cell interference power. In case ofHSPA, I own = (1−α)P own with α ∈ [0, 1] denotingthe average channel multipath orthogonalityfactor and P own denoting the own-cell transmitpower. In LTE, orthogonality is often assumedunity for simplicity, even though in realitythe following nonidealities may result innon-negligible own-signal interference in LTE:• Inter-symbol interference due to multipathpower exceeding cyclic prefix length• Inter-carrier interference due to Dopplerspread (large UE speed)

Practical Introduction to LTE Radio Planning 3• Transmit signal waveform distortion dueto transmitter nonlinearities, measured interms of Error Vector MagnitudeIn both LTE and HSPA the effective value ofα depends on the multipath characteristics andreceiver implementation. For HSDPA with receiverequalizer, α > 0.9 can be assumed in mostcases. In this paper, α = 1 is assumed for LTEand hence I own = 0. The impact of α < 1 can beseen only at high SINR ave ; at low SINR ave noisepower dominates the denominator of (1) over interferencepower, for both HSDPA and LTE.In the remainder of the paper, we dropthe subscript and denote average signal-tointerference-ratiosimply with SINR.2.2. LTE versus HSPA+ coverage in noiselimitedscenarioIn Table 2 link budgets of LTE and HSPA+ arecompared. Both systems have 5 MHz systembandwidth, 2Tx×2Rx MIMO antenna system,and equal antenna and RF characteristics for faircomparison. Note that the link budget is carrierfrequencyindependent, as it is given in terms ofmaximum path loss, not cell range. A singlecell in isolation is assumed, i.e., no interference(I = 0).The following differences in link budgets canbe seen:• In HSDPA, L1/L2 control and pilot channeloverhead is a fraction of the totaldownlink transmit power (typically about20%). In LTE, L1/L2 control channels consumea fraction of DL OFDM symbols,30% overhead in time/frequency resourceelement usage is assumed in Table 2, correspondingto three symbols wide L1/L2control channel region.• For a given bit rate target, required SINRis slightly different for LTE and HSPA+.However, in the noise-limited regime thedifference in target SINR is typically small,less than 2 dBs.• The noise bandwidth is about 5MHz forHSPA+ UL and DL, as well as LTE DL.For LTE UL, however, the noise bandwidthdepends on the number of allocated physicalresource blocks (frequency allocation),since symbols are detected in time-domainand noise is accumulated only over the actualoccupied bandwidth.While HSPA+ and LTE have similar performancein terms of coverage, the same is not truefor system-level capacity under network interference,where LTE has advantage over HSPA+due to more advanced radio features, such asmultiuser-MIMO, frequency-domain schedulingand inter-cell interference coordination.2.3. Optimizing LTE system bandwidth forcoverageIn this section it is shown that, for a given targetbit rate, there is an optimum system bandwidthin DL and UL in terms of maximizingcoverage. Of course, typically other decisioncriteria besides coverage come into play whenchoosing system bandwidth for practical networkdeployment. The purpose of the followingdiscussion is mainly to illustrate trade-offbetween bandwidth and coverage in LTE uplinkand downlink.Downlink (Fig. 1): Consider a fixed rate ofinformation transmission for a fixed transmitpower. As the number of allocated Physical ResourceBlocks (PRBs) becomes larger, code ratedecreases and channel coding gain in turn increases.Therefore, for a fixed DL system bandwidthit pays off to use as many PRBs as possibleto minimize required SINR 1 . On the otherhand, if one is allowed to use the system bandwidthas a design parameter, there exists an optimumsystem bandwidth for a given target bitrate. This is due to the fact that, assuming fixedtotal transmit power, transmit power per PRB(power density) increases as system bandwidthis reduced. This is illustrated in Figure 1. For bit1 One PRB is a contiguous chunk of 12 subcarriers, or180kHz, in frequency domain.

4 J. Salo et alTable 2Link budget comparison in noise-limited scenario, a single user at cell edge for 2x2 LTE and 2x2HSPA+, 5 MHz LTE bandwidth. Target bit rate: 1Mbps DL, 128Mbps UL.Quantity LTE DL HSDPA LTE UL HSUPATransmit power ⋆ 46 dBm 45 dBm ♭ 23 dBm † 24 dBmAntenna + feeder gain 15 dB 15 dB −3 dB −3 dBMHA gain - - 2 dB 2 dBRx Noise Figure 7 dB 7 dB 2 dB 2 dBSNR target −4 dB ‡ −5 dB ‡ −12 dB ‡ −13 dB ♦RX sensitivity −105 dBm ♯ −105 dBm −117 dBm −118 dBmMax path loss 164 dB 163 dB 154 dB 156 dB⋆ Sum power of two transmit antennas♭ max HS-DSCH power 42 dBm per antenna† Category 3 LTE terminal‡ All bandwidth allocated for the single user in DL and UL.♦ E b /N 0 target of 2dB for 128 kbps, processing gain of 15 dB .♯ UL and DL noise bandwidth is 5MHz.rate target of 128 kbps, the optimum bandwidthis 1.4 MHz, while for 512 kbps and 1Mbps targetbit rate, it is 3 and 5MHz, respectively.Uplink (Fig. 2): There is an optimum systembandwidth in the uplink, too, as illustratedin Figure 2. The difference to downlink isthat the noise bandwidth equals actual instantaneousscheduled UL bandwidth, which may beless than the system bandwidth. From Fig. 2, itcan be seen that for bit rates ≤ 256kbps the optimumnumber of PRBs is about 3−5, meaningthat 1.4MHz system bandwidth (6 PRBs) wouldbe sufficient for optimal coverage. For bit ratesabove 2Mbps, system bandwidth of 5MHz ormore is needed in order for system bandwidthnot to limit receive sensitivity (UL coverage).In practice, control and common channel coverageshould be taken into account, too. Simulationresults in [1] indicate that for low bitrate services uplink random access channel coveragemay be the limiting factor, instead of thePDSCH/PUSCH considered here.RX downlink sensitivity [dBm]−94−96−98−100−102−104−106−108−110LTE DL rx sensitivity vs system bandwidth for different bit rates4Mbps2Mbps1Mbps512kbps−112256kbps128kbps−1141.4 3 5 10 15 20System bandwidth [MHz]Figure 1. LTE downlink sensitivity versus systembandwidth for different bitrates, channelunaware scheduling. Total system bandwidthis allocated to a single user.

Practical Introduction to LTE Radio Planning 5RX sensitivity [dBm]−98−100−102−104−106−108−110−112LTE UL rx sensitivity vs allocated UL bandwidth for different bit rates2Mbps1Mbps512kbps256kbps128kbpswhere I max = ∑ K k=1 I max,k is the maximum interferencepower at cell edge, and γ k is the subcarrieractivity factor of the kth cell. We assumethat the average cell load is equal for all cells, inother words, γ k = γ, for all k.Signal-to-Interference-and-Noise ratio can bewritten as a function of SNR, SIR and cell loadγ as−114−116−118−1200 2 4 6 8 10 12 14 16 18Allocated UL bandwidth [MHz]SINR ===SI+NSγI max + N1γSIR min+ 1SNR(5)(6), (7)Figure 2. LTE uplink sensitivity versus allocatedbandwidth for different bitrates, channelunaware scheduling.3. LTE IN INTERFERENCE-LIMITED SCE-NARIOA somewhat simplified engineering view onSINR under non-negligible other cell interferenceis presented in this section. In particular,interference is here characterized only by itsaverage power, whereas (as in HSDPA) it is infact very bursty when examined at TTI level.Therefore, the interplay of HARQ and interference(traffic) time correlation is completely neglectedin the sequel, for the sake of potentialradio planning insights.3.1. SINR under random (uncoordinated) interferenceAt cell edge, the interference from K neighbouringcells can be written as (I = I other )I =K∑ γ k I max,k (3)k=1= γI max , (4)where SIR min =I Smaxand SNR =N S . We againemphasize that all quantities are average values,where the averaging is over small-scale fadingof S and I. The value of SIR min dependson network geometry and antenna configuration(but not on cell range) and is in practice determinedfrom system simulations or networkmeasurements. Typical values are SIR min =−4 . . . − 1dB. The interference is here implicitlymodelled as Gaussian noise, as is the commonpractice. It always applies that SINR

6 J. Salo et al20SNR versus SINR for different average load γγ=1To derive formula for the noise rise, we firstdefine in linear scaleSINR [dB]151050−5γ=0.5γ=0.25γ=0.1γ=0.05γ=0NR= I + NN(9)= SNRSINR . (10)Substituting, SNR = NR·SINR in Eq. (7) andsolving for NR results in−10−10 −5 0 5 10 15 20SNR [dB]NR =11−γ SINRSIR min. (11)Figure 3. Cell edge SINR versus SNR for differentnetwork average loads (γ), SIR min = −3dB.where NR denotes noise power rise due to interference,and SINR is determined by the throughputtarget. Cell range can then be solved in theusual manner from SNR intf where the receivedsignal power, S, is a function of path loss (cellrange) and effective isotropic transmit power.What remains to be done is to obtain an expressionfor the noise rise. Before doing so, it isuseful to note, again, that SINR can at cell edgenever exceed SIR minγin Eq. (7). As SINR → SIR minγ,SNR must in turn increase in order to keep SINRless than its upper limit. Therefore, one sees thatthe noise rise in Eq. (8) will be huge when thetarget SINR approaches SIR minγ, hence drasticallyshrinking the cell size.Graphically, the same concept can be seen inFig. 3 where the noise rise is the difference betweenthe noise-limited case (γ = 0) and loadedcase, see Eq. (8). For example, with 100% loadthe noise rise is 23dB at SNR = 20dB. As SINRapproaches SIR min = −3 dB, large increase inSNR is required for diminishing gains in SINR,and therefore cell edge throughput.Here SINR depends on the cell edge throughputtarget and SIR min on the network geometry.One can see that noise rise is very steep nearthe pole SINR = SIR minγ. Therefore, it is increasinglydifficult to reach the upper limit cell edgethroughput.3.3. Trade-off between cell range, networkload and cell edge throughputIn this section we illustrate the relationshipbetween the three basic network planning variables:cell range, average network load and celledge throughput. Towards this end, we writeSNR= S N(12)= EIRPLN , (13)where EIRP is the effective isotropic radiatedpower, and L is the path loss, given in linearscale asL = MAd B . (14)Here A is the median path loss at one kilometerdistance (depends on carrier frequency andpropagation environment), B is the path loss exponent,d is the distance in kilometers and M isthe shadow-fading margin.In linear scale, SINR as a function of cell ranged and network load γ is

Practical Introduction to LTE Radio Planning 7SINR =1γSIR min+ MAdB NEIRP(15)There is a one-to-one mapping between SINRand average throughput. Therefore, (15) describesthe tradeoff between γ, d and throughput.In the sequel, the three two-variable specialcases are considered.In the sequel, path loss model is based on3GPP system LTE system simulation specificationaccording to L[dB] = 128+ 37 log 10(d)[4]. Shadowing standard deviation is 8dB withshadowing margin of M = 8.7dB (95% singlecellarea probability) and building penetrationloss of 20 dB. Other parameters are as givenin Table 2, except that system bandwidth is 10MHz. The SINR-throughput mapping was simulatedfor a 2×2 transmit diversity system.3.3.1. Cell range vs network load, fixed celledge throughputIn Fig. 4, the relation of network load and cellrange is shown for various values of cell edgethroughput. As network load increases, the cellrange decreases for fixed cell edge throughput.3.3.2. Network load vs cell edge throughput,fixed cell rangeIn Fig. 5, the relation of network load and celledge throughput is shown for various values ofcell range. For our later discussion concerninginter-cell interference coordination, it is instructiveto note that when cell range is small, celledge throughput is sensitive to network load.For cell range of 0.5km, if one was able to reduceeffective network load from 0.4 to 0.1 celledge throughput would double from 6 Mbps to12 Mbps. Such a reduction could be possibleby applying interference coordination betweenneighboring cells, hence directing interferenceto other part of the frequency spectrum. Thiswill be discussed further later in this paper.network load10.90.80.70.60.50.40.30.20.1network load versus cell range, for fixed DL throughput1Mbps2Mbps3Mbps4Mbps6Mbps8Mbps00 0.5 1 1.5 2DL cell range [km]Figure 4. Network load versus cell range forfixed downlink cell edge throughput.Cell edge throughput [Megabits/sec]2018161412108642network load versus throughput, for fixed cell range0.5km0.75km1km1.25km1.5km00 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1network loadFigure 5. Network load versus cell edgethroughput for fixed cell range.

8 J. Salo et alCell edge throughput [Megabits/sec]24222018161412108642Cell edge throughput versus cell range, for fixed network loadγ = 0, 0.1, ..., 0.9, 100 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2Cell range [km]4. MULTIPLE-INPUT MULTIPLE-OUTPUT(MIMO) CONFIGURATIONS4.1. MIMO Transmission Schemes in LTEWhile in common engineer talk there are appearsto be several interpretations for the term,in this paper MIMO is taken to mean any radiointerface that has at least two antennas at bothends, i.e., with this definition smallest MIMOantenna configuration would then be 2Tx×2Rx.Note that in LTE all UEs are required to havetwo receive antennas, while the number of basestation antennas can be 2 1, 2, or 4. In principle,there are three ways to utilize MIMO antennas:Figure 6. Cell range versus cell edge throughputfor fixed network load.3.3.3. Cell range versus cell edge throughput,fixed network loadIn Fig. 6, the relation of cell range and celledge throughput is shown for various valuesof network load. As the cell range increases,the cell edge throughput approaches the samelimit regardless of cell load. Even the low networkload of 10% limits throughput to less than14Mbps by creating a noise floor from othercellinterference. It should be noted that downlinkcommon channels and signals alone causeat least > 10% network load over the ≈ 1 MHzbandwidth over which they are transmitted. Ontop of this, reference signals are transmitted overthe entire system bandwidth. Sites in early FDDnetwork implementations are not expected to beframe-synchronized, so that it is not possible toavoid inter-site interference by cleverly coordinatingtransmission in time and frequency.• Combined transmit/receive diversity: Normaltwo-branch diversity reception or transmissionhas the benefit of producing twocopies of the same signal for reception,which with a suitable signal combiningtechnique reduces fading variation. Similarly,a 2Tx×2Rx antenna transmissionresults in reception of four signal replicas,with the corresponding additional reductionin fading. This version of MIMOis hence simply an enhanced diversityscheme, and as such is not by many considered"true" MIMO, since it does not actuallyincrease transmission data rate. InLTE, space-frequency block coding baseddiversity scheme is used; this is an openloopscheme where UE precoder feedback(PMI) is not required at the base station.• Beamforming: This is very similar to the diversitytransmission, the main nomenclaturaldifference being that in beamformingone typically considers a physical antennabeam being constructed towards the UE.2 With precoding using UE-specific reference signals ("beamforming")arbitrary number of antennas can be used. However,this transmission scheme is not expected to be supportedin first vendor releases, due to hesitance from operatorsto deploy antenna arrays. The cell edge throughput gainfrom such configuration in rural/road environments wouldbe substantial however, e.g., 9 dB from an 8-antenna panelarray.

Practical Introduction to LTE Radio Planning 9This requires closely spaced antennas, unlikethe diversity schemes which requireat least a few wavelength antenna spacing.Beamforming using 3GPP codebook alsorequires PMI feedback from UE. Mathematically,both beamforming and transmitdiversity are both special cases of so-calledrank-1 precoding 3 . LTE Rel 8 supportsrank-1 precoding (or, "beamforming") usingpre-defined 3GPP codebook (for 2and 4 antennas), and any vendor-specificbeamforming when using UE-specific referencesignals (arbitrary number of basestation antennas).• Spatial Multiplexing: This is what manyconsider to be a true MIMO transmissionscheme. With beamforming and diversity,base station transmits a single stream ofinformation but uses the multiple antennasto either reduce fading (diversity) orincrease signal power (beamforming). Onthe other hand, with 2Tx × 2Rx spatialmultiplexing the idea is to transmit twoparallel information streams over the samebandwidth, hence theoretically doublingthe data rate and spectral efficiency. Bothopen-loop (only channel rank and CQIreported by UE) and closed-loop spatialmultiplexing are supported (also precodingmatrix information reported by UE).Delving slightly in implementation details,in LTE open-loop spatial multiplexinghave been engineered in such a waythat symbols in information streams areinterleaved between the MIMO subchannels.Thus, the average SINR experiencedby the transmitted symbols is the averageof the two MIMO subchannels’ SINRs.In uplink, spatial multiplexing is not supportedfor a single UE, but two differentUEs are allowed to transmit at the sametime; this is called multiuser-MIMO.3 In this paper, rank-1 precoding and beamforming are consideredsynonymous.In practice, the choice of transmission schemedepends on instantaneous radio channel conditionsand is adapted continuously. The main differenceto non-MIMO transmission that, in additionto SINR, now also the channel rank hasto be considered, since spatial multiplexing isfeasible only if the instantaneous radio channel(during 1ms TTI) supports transmission ofmore than one information stream, or in termsof matrices, the 2Tx×2Rx matrix channel’s conditionnumber and SINR are good enough. As arule of thumb, in a spatially uncorrelated channelspatial multiplexing becomes useful whenSINR > 10 dB. Looking at Fig. 3 one can seethat this requires very low network load for spatialmultiplexing to work at cell edge, assumingthat other-cell interference is uncoordinated.4.2. Benefit of MIMOThe benefits of different MIMO schemes indownlink are roughly as follows:Transmit/receive diversity: Compared to 1Tx×1Rx the gain of 2Tx × 2Rx is about 6 − 7 dBdue to following factors: i) transmit power isdoubled by adding another amplifier (3dB); ii)average received signal power is doubled becauseof two-antenna reception (3dB); iii) diversityfrom four signal paths brings additionalgain which however strongly depends on thepropagation environment (here pessimisticallyassumed 0−1dB, higher gains are possible).Rank-1 precoding: For slow-moving UEs thegain from closed-loop transmit diversity is 1−2 dB higher than with the open-loop transmit/receivediversity. For fast-moving mobilesthis improvement over diversity diminishes, orgoes negative, since the CQI feedback cannottrack the channel fast enough.Spatial multiplexing: This scheme does not improvelink budget, but increases data rate. Withtwo antennas and an ideal MIMO radio channel,the data rate would be doubled. In practice,gains are considerably smaller due to the factthat for ideal operation the SINR of the two parallelsubchannels should be high enough to supportthe same modulation and coding scheme.

10 J. Salo et alThis is very rarely the case, and the secondstream must be transmitted at lower informationbit rate. This point is further illustrated inthe following subsection.4.3. Example of Measured MIMO Radio ChannelSpatial multiplexing transmission is feasibleonly at those time instants when the radio channelconditions are favorable. There exist awealth of literature on the theory of MIMO communicationsystems, and the interested readeris referred to the references. From radio planningpoint of view, the 2Tx×2Rx MIMO channelpower response λ at any given time instantcan be written, on a subcarrier, as 4 λ = λ 1 + λ 2 ,where λ 1 and λ 2 are the power responses ofthe first and second MIMO subchannels, respectively.Each of the subchannels carries one informationstream. As with a normal non-MIMOchannel, the powers of the subchannels experiencesignal fading. If the power of the weakersubchannel, say λ 2 , is very weak, the second informationstream should not be transmitted atthat time instant, since it will either be buriedin noise, or alternatively, from link adaptationpoint of view, it is more spectrally efficient totransmit one stream with higher-order modulationand/or less channel coding. In eithercase, usage of spatial multiplexing depends onthe instantaneous channel conditions which arereported to, and tracked, by the base station.When the second subchannel is weak, the channelis said to have "rank one", and spatial multiplexingis not used 5 .Fig. 7 presents an example of a narrowbandmeasurement of a 2Tx×2Rx channel inan urban environment at 2.1 GHz carrier frequrency.Two base station antenna configurationsare shown: i) two vertically polarized4 The symbol λ is used here for historical reasons, since theMIMO subchannel power is related to the eigenvalues of theMIMO radio channel.5 Mathematically, the rank of the 2Tx×2Rx channel is practicallyalways two, so this terminology is strictly speaking amisnomer.transmit antennas at 3 wavelength separation(||); ii) two cross-polarized transmit antennasat 0 deg and 90 deg angles (+). The dual-branchUE antenna is a realistic terminal microstrip antennaintegrated in the terminal chassis, whosebranch polarizations are not well-defined, as istypically the case for small integrated antennas.The figure shows the powers of the two MIMOsubchannels as a function of travelled distance.Roughly in the midpoint of the measurementroute the UE enters line-of-sight and the signalpower of the stronger stream (λ 1 ) increases.The second stream power behaves differentlyfor the two antenna setups. With the crosspolarizedantenna base station setup the secondstream (λ 2 ) is about 15 dB weaker than thestronger one. In contrast with the vertically polarizedantennas the second stream is about 30dB weaker. Thus, if the channel SINR is 20 dB,with the cross-polarized configuration the firststream SINR would be almost 20dB while thesecond stream would experience about SINR ≈5 dB which would still allow fairly high bit ratetransmission over the second MIMO subchannel.However, with the vertically polarized antennasthe second stream would be SINR ≈ −10dB, which would not support high-rate transmissionof the second stream.Fig. 8 illustrates the same measurementroutes using two different figures of merit,namely the total MIMO channel power responseand ratio of MIMO subchannel powers,1 λλ 2. From the upper subplot it can be seenthat the total power responses of the two antennasetups are slightly different. Verticallypolarized configuration has almost negligiblyhigher received power in the line-of-sight, whilethe cross-polarized configuration, in turn, hasslightly better power response in non-line-ofsight.From the lower subplot we note that thecross-polarized antenna setup has advantage interms of the second subchannel power in line-ofsightconditions; the second subchannel is about10 dB weaker than the first, compared to the 30or so dBs for the vertically polarized case.

Practical Introduction to LTE Radio Planning 11Summarizing the main radio planning messagefrom the two figures:• Dual-stream transmission is only feasiblewhen the SINR of the second subchannelis high enough. This depends on the totalchannel SINR and the ratio of the subchannelpowers.• Base station antenna configuration has impacton the spatial multiplexing performancein two respects: total received signaland spatial multiplexing gain.• In terms of total received signal power, thetwo compared antenna configurations areroughly equal with slight advantage for v-pol in line-of-sight, and vice versa in nonline-of-sight.• In terms of spatial multiplexing gain, x-polhas large advantage in line-of-sight. Thereis no noticeable difference in non-line-ofsight.The results indicate that cross-polarized basestation antenna configuration should be favoredwhen usage of spatial multiplexing is desired, asit offers more robust performance in all channelconditions. While these conclusions are in thispaper illustrated by means of a single measurementexample, similar conclusions have beendrawn from large channel measurement campaignsreported in literature [5].5. RADIO RESOURCE MANAGEMENT FEA-TURESSo far, we have discussed a simple "baseline"LTE radio configuration, assuming fully random(uncoordinated) interference and no specialradio resource management features. TheLTE radio interface specification Release 8 supportsseveral advanced air interface functionalitiesthat potentially improve network performancein certain environments. These are explainedin the following.subchannel power, dBsubchannel power, dB0−10−20−30−40two vertically polarized BTS antennas(||), three wavelength separation−500 2 4 6 8 10 12 14 16 18 20distance, meters0−10−20−30−40two cross−polarized BTS antennas (+)−500 2 4 6 8 10 12 14 16 18 20distance, metersFigure 7. Example of measured MIMO radiochannel, power response of the MIMO subchannelsfor two base station antenna configurations.MIMO channel power response, dBλ 1/ λ 2, dB0−5−10−15−20−25channel power response, v−pol versus x−pol−305 10 15 20distance, meters5040302010ratio of MIMO subchannel powers, v−pol versus x−pol05 10 15 20distance, metersλ 1λ 2λ 1λ 2v−pol (||)x−pol (+)Figure 8. Example of measured MIMO radiochannel, total channel power response and subchannelpower ratio for two base station antennaconfigurations.

12 J. Salo et al5.1. Frequency-Aware UL/DL SchedulingBaseline scheduler assumes no knowledge ofthe instantaneous frequency response of the radiochannel. For slow-moving UEs such knowledgeis available at the BTS in the form of subbandCQI feedback from UE. In case there isonly a single user to be scheduled in downlink,it is optimal to allocate whole bandwidth, sincethis offers highest coding gain. In UL, the optimalbandwidth for a single simultaneous usermay be less than the system bandwidth, seeFig. 2 for an example. Focusing on the downlink,in the case there are multiple simultaneousUEs to be scheduled during a TTI, the basestation can schedule each UE in the frequencysubband with the strongest channel power response,hence avoiding fading dips in frequencydomain. This results in frequency-domain multiuserscheduling gain, which can be quantifiedin terms of network-level throughput gain andcell edge throughput gain.Practical (including system imperfections)achievable downlink multiuser schedulingnetwork-level throughput gains have been simulatedto be about < 30−40% at network level,for 3−7 users [6]. For the cell edge, throughputcan be almost doubled with a suitable schedulingalgorithm (always vendor-dependent). Herethe gain should be interpreted relative to thethroughput of a frequency-unaware schedulerwith the same instantaneous number of usersper TTI. Of course, even if the network-levelthroughput increases, actual per-user throughputwill decrease when the number of simultaneoususers increases.Another benefit of the frequency-awarescheduling emerges in case the time correlationof other-cell interference power is long enoughso that the scheduler has time to learn the interferencespectrum and divert transmission toa cleaner part of the system bandwidth. Thisis only possible if UE is moving slowly andthe interference is persistent in time over, say,a few dozens of TTIs. This interference rejectiongain can be seen similar to what is obtainedfrom inter-cell interference coordination, to bediscussed next.5.2. Inter-Cell Interference CoordinationThere are two basic methods to coordinatefrequency usage between cells in a network:static and dynamic. Static coordination meansfixed allocation of frequency resources per cellover extended periods of time while dynamicfrequency allocation means fast coordinationwithin a time frame of seconds or even less,without need for manual operator intervention.In LTE, dynamic inter-cell interference coordinationis inherently supported by 3GPP-specifiedsignalling between base stations.The network-level throughput gain from frequencyuse coordination is most visible withlow-to-medium network load; when the networkload is high no coordination will helpsince all frequency resources of all cells tendto be used most of the time. Hence, interferencecoordination is most useful for noncongestednetworks. Looking at Fig. 5, considerablecell edge throughput gains are possibleif one can reduce the effective network loadseen by the UE by allocating neighbor cell frequencyresource on a different part of the systembandwidth. From Fig. 5, if the cell radiusis small the throughput gain can be considerable.At network-level, throughput gainssimilar to channel-aware scheduling have beenreported in literature for low-to-medium networkloads (γ < 50%). Summarizing, intercellinterference coordination is mostly useful insmall-range cells during off-peak hours or duringearly phases of network life cycle.5.3. Uplink Fractional Power ControlLTE supports both cell-specific and UEspecificuplink power control. In LTE downlink,there is no power control in the conventionalsense. With LTE power control schemeit is possible to compensate only a fraction ofthe path loss, reducing transmit power of celledge UEs that generate most of the interferenceto other cells. This is in contrast to conventional

Practical Introduction to LTE Radio Planning 13power control schemes that attempt to compensatepath loss to the full extent allowed by theUE transmit power. The benefit of the fractionalscheme is that, due to reduction in UE transmitpower, network-level interference is reducedand overall UL throughput is increased. Thisimprovement is at the expense of reduction incell edge UL throughput. For each optimizationgoal - maximizing cell edge throughput or networkthroughput - there exist an optimum optimumpower control parameter set. For earlysystem simulation examples, see [7].6. INTRODUCTION TO RADIO PARAME-TER PLANNINGThe three most important tasks in LTE prelaunchradio parameter planning are: i) PhysicalCell Identity (PCI) allocation, ii) PhysicalRandom Access Channel (PRACH) parameterplanning, and iii) uplink reference signal sequenceplanning. Obviously, there are numerousother types of air interface parameters thatcan be tuned in LTE, such as UL power controland handover thresholds. However, the threetasks picked out above are the ones that mustbe planned before network launch, while the remainingones can arguably be optimized also afterlaunch, provided that reasonable default valuesare available. In the following, we briefly reviewthe basics of the three main pre-launch parameterdesign tasks. More detailed treatmentof parameter optimization and planning will bea topic of a separate exposition.6.1. PCI planningIn LTE, Physical Cell Identity (PCI) allocationis a task somewhat similar to scrambling codeallocation in WCDMA. PCI is encoded in thephysical layer synchronization signal transmissionand is used by the UE for neighbour cellhandover measurement reports. Hence, as inWCDMA, the PCI should uniquely identify theneighbouring cell to the serving eNB, within acertain geographical area. Consequently, PCIreuse distance should be large enough so thatUE cannot measure and report two cells withthe same PCI. This part of the PCI allocationshould not pose a problem since there are 504PCIs defined.The second, perhaps less obvious, purpose ofthe PCI is to serve as a resource allocator parameterfor downlink and uplink Reference Signals(RS).DL Reference Signal: Downlink reference symbols("LTE pilot signal") are allocated in a timefrequencygrid as shown in Fig. 9. The figureillustrates three frame-synchronized cells withPCIs 9, 10, and 11. For example, these couldrepresent different cells of the same eNB site 6 .In time domain the RS are always transmittedin the same OFDM symbol. However, in frequencydomain each cell has a different shiftdetermined by modulo-3 of the PCI, denotedPCI mod3. In Fig. 9, cell #1 has PCI mod3 =0, cell #2 has PCI mod3 = 1, while cell #3has PCI mod3 = 2. With this PCI allocation,the RS of different cells do not overlap in frequency,which results in less interference onUE channel estimation. If neighbouring sitesare frame-synchronized (e.g., using GPS), or theframe timing offset between sites is known, thePCI allocation should also be coordinated betweenneighbouring sites, which brings additionalcomplexity in the allocation process. Onthe other hand, if the neighbouring sites are notframe-synchronized or the frame timing offset isnot known (i.e., random), one cannot coordinatePCI allocation between sites.It should be noted that assigning PCIs asshown in Fig. 9 results in RS overlapping withthe control and data Resource Elements of theneighbouring cells. Therefore, a design choicemust be made between RS↔RS interference andRS↔PDSCH/PDCCH intererence. The currentengineering consensus seems to be that the latterchoice is favoured, due to the fact that iftwo frame-synchronized cells have the same6 Note that Fig. 9 shows transmission of both transmit antennasin the same grid, to save space. Frame synchronizationin this context means that the DL transmission of a radioframe starts at the same time instant in all three cells.

14 J. Salo et alFigure 9. Location of Reference Symbols within one PRB for different PCIs. Frame-synchronized cellsshown.PCImod3, the additional drawback is that theirPrimary Synchronization Signals will interferewith each other, causing problems in cell searchand handover measurements.UL Reference Signal: LTE uplink shared datachannel (PUSCH) carries Demodulation ReferenceSignal (DM RS). Optionally also SoundingReference Signal is transmitted in the uplink.The uplink DM RS are constructed fromZadoff-Chu sequences which are divided into30 groups. Roughly, this means that for a givennumber of PRBs allocated in the uplink thereare 30 different base sequences that can be usedas the reference signal 7 . The cross-correlationbetween the base sequences is on average low,which is beneficial from inter-cell interferencepoint of view. It follows that the planning requirementis that the neighbouring cells shouldbe allocated different base sequences. The simplestmethod is to ensure that PCI mod30 of po-7 There are actually two groups of 30 sequences definedfor PRB allocations of 6 or more. These can be pseudorandomlyalternated. However, this option is not consideredin this paper.tentially interfering cells is different. This is becausein the simplest scheme the DM RS base sequenceindex is equal to u =PCI mod30, whereu = 0 . . . 29 is the base sequence index. An exampleof this simple PCI-based DM RS sequenceallocation scheme is shown in Fig. 10. In practicalnetwork deployments this simple planningcriterion cannot be always fulfilled. As a remedyto such a case, there are additional sequence allocationschemes built-in in the 3GPP specification,most notably a static base sequence indexoffset parameter, DM RS cyclic shift planning,and pseudo-random base sequence hopping ("uhopping").These will be discussed next.6.2. UL Reference Signal sequence planningThe simple PCI-based based sequence uplinkDM RS allocation scheme from the previous sectionmay be difficult to plan since the same basesequence is reused in every 30th cell, which maylead to insufficient cell separation.There are three options to reduce uplink intercellinterference in such cases:

Practical Introduction to LTE Radio Planning 15Figure 10. Simple PCI-based uplink DM RS allocation scheme. Different base sequence u allocated toevery cell based on PCImod30, every cell has ∆ ss = 0.Figure 11. Example of how the parameter ∆ ss can a base sequence collision between two cells. Theindoor cell has been allocated PCI=30 which would result in u = 0 if ∆ ss = 0, hence creating ULinterference with cell PCI=0. Setting ∆ ss = 29 gives u = (30+29) mod 30 = 29 instead.

16 J. Salo et alFigure 12. Example of u-hopping. Cluster #1 has hopping-pattern σ = 0 and cluster #2 has hoppingpatternσ = 1. Sequence offset ∆ ss = 0 for all cells. Systematic base sequence collisions within a groupare avoided within a group. Random base sequence collisions are possible in the cluster border.Figure 13. Example of cyclic shift planning. Cells of the same site are allocated the same base sequenceu using the offset parameter ∆ ss . Inter-cell interference between cells of the same site is mitigated bysetting different cell-specific cyclic shift (cs) for cells of a site.

Practical Introduction to LTE Radio Planning 17• Bypass the simple PCI-based base allocationscheme by explicitly defining the basesequence used in the cell. This brings additionalflexibility to base sequence allocation,and effectively decouples the PCIplanning from uplink DM RS base sequenceplanning.• The base sequence u can change pseudorandomlyfor every 0.5ms time slot.This planning option randomizes base sequencecollisions and averages inter-cellinterference.• Different cyclic shifts of a ZC sequence areorthogonal. This can be utilized by assigninga different cyclic shift on two cells thatuse the same base sequence u. The cellspecificstatic cyclic shift is broadcasted onBCCH.Defining u independently from PCI: The simplestscheme assigns the base sequence index u to acell as modulo-30 of PCI. Optionally, the basesequence can be assigned to a cell asu = (PCI+ ∆ ss ) mod 30 ,where ∆ ss = 0 . . . 29 is an offset parametersignalled on BCCH. In the simple PCI-basedscheme ∆ ss = 0. With ∆ ss it is possible to avoidcollisions in cells that would otherwise use thesame u due to PCI allocation. An example of ∆ ssplanning is shown in Fig. 11.Pseudo-random u-hopping: If u-hopping is activated,the base sequence used in the cell changesat every time slot in a pseudo-random fashion.The index of base sequence in time slot n 8u n = (ν n + PCI+ ∆ ss ) mod 30 ,where ν n = 0 . . . 29 is pseudo-random integerdefined by the hopping-pattern. The hoppingpattern defined used in a cell is defined by theindex σ = ⌊ PCI30⌋. Thus, there are 17 u-hopping8 There are 20 time slots in a radio frame. The hopping patternis re-initialized at the beginning of every radio frame.patterns defined since there are 504 PCIs defined,i.e., σ = 0 . . . 16. A foreseen methodof PCI planning, where near-by cells are assignednear-by PCI values, results in groupingof cells into "clusters-of-30", where within eachcell cluster the same hopping-pattern is used. Toprevent systematic collisions, static part of thesequence group assignment, (PCI + ∆ ss ) mod30, should be different, especially for framesynchronizedcells. At the border of two cellclusters having different σ, random sequencegroup collisions are possible since two differenthopping patterns are utilized. An example ofthis planning scheme is shown in Fig. 12.Cyclic shift planning: So far we have consideredonly how to reduce inter-cell interferenceby assigning different base sequences to everyneighbouring cell. Another interference reductionmethod follows from the fact that ZC sequenceshave the useful property that two differentcyclic shifts of the same base sequence areorthogonal 9 . Therefore, if a pair of cells havebeen allocated the same base sequence u, intercellinterference can be reduced by assigning adifferent cyclic shift to the cells. This scheme canbe applied to cells of one site as shown in Fig.13. Other applications, tangential to our presentdiscussion, are uplink multi-user MIMO (whereeach UE uses a different cyclic shift) and, in LTE-A, uplink single-user MIMO (where each UEtransmit antenna uses a different cyclic shift), ora combination of the two. It should be notedthat cyclic shift planning can also be combinedwith the other two schemes discussed in thissection.6.3. PRACH parameter planningThe random access procedure in LTE uplinkbegins when UE transmits a preamble to theeNB. The specific preamble is selected randomlyby UE from a pre-defined set of 64 Zadoff-Chusequences. To avoid call setup anomalies, eachcell (within a reuse distance) has its own uniqueset of 64 preambles, and the information of the9 Cyclic shifts of extended ZC sequences used in uplink DMRS are not orthogonal, however.

18 J. Salo et alFigure 14. Example of root sequence allocation for different cells. Four sequences per cell are allocated,PRACH configuration index = 8, see Table 3.specific set to use in the cell is broadcasted onBCCH. The preamble sequence allocation dependson cell range: for example, if one (unrealistically)requires that all cells must be accessiblefrom 100km away then there would beonly a total of 839 preamble sequences available.As each cell must have exactly 64 preambles,the total number of cells within a reuse distancewould then be only 83964≈ 13, and in this caseeach cell consumes 64 ZC sequences (one fullZC sequence per preamble) out of the total of839 ZC sequences available for PRACH. Moresequences can be generated if the "random accessradius" of cells is dimensioned more realistically;continuing the example, if maximumrandom access cell range is chosen as 1km, oneZC sequence generates 64 preambles, insteadof the one preamble in the 100km case. Summarizing,the PRACH parameter planning taskconsists of deciding i) the maximum cell rangefor random access, and ii) allocating the ZC sequencesto cells.Table 3 lists the number of ZC sequences requiredper cell, for a given random access radius.One can see that a cell requires five ZCsequences per cell for up to 7.3km radius, whichis typically sufficient for urban and suburbanmacro cells. This results in sequence reuse factorof at least 839/5 ≈ 167 cells, hence allowsfor easy planning process. Example of root sequenceallocation is shown in Fig. 14. Thepoint to notice here is that root sequence indicesof cells must not overlap within the reuse distance10 . As for typical scenarios there are plentyof ZC sequences available, for safety margin andease of planning one can overdimension the randomaccess radius of the cells in a given planningregion, to simplify allocation. In the exampleof Fig. 14 each cell consumes four sequences,corresponding to maximum radius of 5.4km forevery cell.If root sequences used in neighbouring cellsoverlap, the transmitted preamble may be de-10 To reduce signalling load, only the index of the first rootsequence used in the cell and the PRACH configuration indexis transmitted on BCCH.

Practical Introduction to LTE Radio Planning 19tected in multiple cells. The drawback are the"ghost" preambles which consequently resultin unnecessary PDCCH and PUSCH resourcereservation in those cells that whose random accessresponses the UE chooses to neglect 11 .Table 3PRACH ZC sequence parameters from the 3GPPspecification.PRACH configurationNumber Random acdequencesin-of root secessradiusper [km] ⋆cell1 1 0.72 2 13 2 1.44 2 25 2 2.56 3 3.47 3 4.38 4 5.49 5 7.310 6 9.711 8 12.112 10 15.813 13 22.714 22 38.715 32 58.70 64 118.8⋆ Includes 1.2µsec implementation margin for delayestimation7. CONCLUSIONIn this paper topics related to radio planningin LTE were discussed. Besides the link budgetfor coverage and interference-limited cases, thetradeoff between cell range, cell edge throughputand network load was discussed under uncoordinatedinterference. Impact of MIMO andantenna configurations was illustrated using a11 Similar problem exists in GSM networks where ghost randomaccesses result in unnecessary SDCCH reservation.measurement example. Finally, some advancedradio resource management features and detailedradio parameter planning tasks were introduced.REFERENCES1. H. Holma et al (eds.), LTE for UMTS, Wiley,2009.2. F. Khan, LTE for 4G Mobile Broadband,Cambridge University Press, 2009.3. S. Sesia et al (eds.), LTE, The UMTS LongTerm Evolution: From Theory to Practice,Wiley, 2009.4. 3GPP TS 25.814, v7.1.0, 2006.5. P. Suvikunnas et al, Comparison of MIMOAntenna Configurations: Methods and ExperimentalResults, IEEE Trans VT, 20086. A. Pokhariyal et al, Frequency DomainPacket Scheduling Under Fractional Loadfor the UTRAN LTE Downlink, IEEE VTC,20077. C. U. Castellanos et al, Performance of UplinkFractional Power Control in UTRANLTE, IEEE VTC, 2008.Jari Salo received the degrees of Master ofScience in Technology and Doctor of Science inTechnology from Helsinki University of Technology(TKK) in 2000 and 2006, respectively.Since 2006, he has been with European CommunicationsEngineering Ltd, where he worksas a consultant for mobile wireless industry andoperators. His present interest is in radio andIP transmission planning and optimization forwireless networks. He has written or co-written14 IEE/IEEE journal papers, 35 conference papers,and is the recipient of the Neal ShepherdMemorial Best Propagation Paper Awardfrom the IEEE Vehicular Technology Society,for the best radio propagation paper publishedin IEEE Transactions on Vehicular Technologyduring 2007. He is also a co-creator of thewide-band spatial channel model adopted by3GPP for LTE system-level simulations. E-mail:jari.salo@eceltd.com

20 J. Salo et alMohammad Nur-A-Alam obtained his Masterof Science degree from Tampere Universityof Technology (TUT) in 2009 and is currentlyworking towards Ph.D in the same department.After attaining his BSc. in Electricaland Electronic Engineering from Islamic Universityof Technology in 2003, he worked in TelecomMalaysia International as a Network Planningand Optimization Engineer. From 2008 to2010, he was with ECE Ltd as LTE Consultant.He is currently working for Nokia Siemens Networksas system solution manager and his interestare in LTE/HSPA+ radio network planning.Kwangrok Chang received the degrees ofMaster and Ph.D in Electronic and Electrical Engineeringfrom POSTECH (Pohang Universityof Science and Technology) in 1995 and 1998,respectively. In July of 1998 he joined LG ElectronicsLtd for two years, after which he hasbeen with Nokia Siemens Networks Ltd., wherehe is currently heading the Network Planningand Optimization Group of Japan and Korea.His present interests are network capacity managementand enhancement of end user performancesin high speed data networks, such asHSPA, and network planning/dimensioning ofLTE system.