Long Term Evolution Part 5: Radio Link Budget J. Hämäläinen, 2016 Department of Communications and Networking Contents 1 Principles of RLB 2 LTE downlink RLB 3 Some details on RLB a) eNodeB powers and antennas b) Path loss and shadowing margin c) Interference, SINR and data rate 4 DL Link Budget Examples 5 LTE Uplink Radio Link Budget Network Planning Our focus area Network planning consists of 3 phases: - Dimensioning, detailed planning and optimization Dimensioning Note: We omit core network 150 140 Path Loss [dB] 130 + 120 110 100 90 0 0.2 0.4 0.6 0.8 Distance from BS [km] 1 1.2 EIRP 58dB Margins 23dB Sensitivity -100dB Allowed PL 135 dB 1000 x 5000 x 1.4 Area and propagation information Radio Link budget # Network elements Detailed planning TX power 43dBi Antennas 2 Antenna tilt 5o Parameter x, y, z Input from dimensioning Network planning tools System simulations BS + RS Configurations and topology plan Optimization + Operating network (drive tests, monitoring) Optimized system 1. Principles of RLB Background • Radio Link Budget (RLB) is a basic tool in radio engineering. It is used to compute estimates for e.g. Can be used to estimate – Received power in terminal/eNodeB user rates – Allowed propagation loss – Connection range between transmitter and receiver • RLB take into account the gains and losses from the transmitter, the communication medium (wireless channel in our case) and the receiver. • Typical parameters are related to the propagation model (radio environment), antennas (antenna directivity/gains), feedlines (cable attenuation) and the receiver properties (product specific sensitivity, noise figure) Example: RLB in LTE cell coverage estimation LTE eNodeB with 3-sector transmission User on the distance where minimum required rate can be provided – assuming a certain load. Example: We require that user should reach 1Mbit/s data rate when 10% of the cell resources are allocated for him/her. Question: How far from eNodeB user can be? This distance gives cell range under above constraint Simplified RLB EIRP = Effective Isotropic Radiated Power, contains transmitted power and antenna gain Transmitter characteristics BS Transmitter BS to MS Total transmission power 43 dBm (20 W) Transmitter antenna gain 15 dBi EIRP 58 dBm Margins Channel characteristics Receiver characteristics Number that is used to estimate the cell range Shadow fading margin 7 dB Interference margin 4 dB Penetration loss 10 dB Total Margin 21 dB UE Receiver (Max coverage) Receiver sensitivity -100.7 dBm System gain 158.7 dB Allowed Propagation Loss 137.7 dB Simplified RLB: Terminology • dBi = dB(isotropic). It is the forward gain of a certain antenna compared to the ideal isotropic antenna which uniformly distributes energy to all directions. • dBm = dB(1 mW) is a measured power relative to 1 mW (e.g. 20W is 10*log(1000*20)= 43 dBm • Effective isotropic radiated power is the amount of power that would have to be emitted by an isotropic antenna (that evenly distributes power in all directions and is a theoretical construct) to produce the peak power density observed in the direction of maximum antenna gain. – EIRP can take into account the losses in transmission line and connectors and includes the gain of the antenna. RLB through equations • In this case the Allowed Propagation Loss (APL) can be calculated as follows: APL = EIRP - min { PRX } - M Total = PTX + GA - min { PRX } - M SF - M I - M Penetration Here min { PRX } = Receiver sensitivity [dBm] PTX = Transmission power in BS [dBm] GA = BS antenna gain [dBi] M SF = Shadow fading margin [dB] M I = Interference Margin [dB] M Penetration = Indoor penetration loss [dB] Penetration loss simply depends on the expected building wall losses. 10 The RLB principle TX/RX parameters Data rate requirement User resource allocation • TX/RX parameters depend on the network deployment – Equipments (eNodeB, UE, antennas), site properties • Data rate requirement Link budget Allowed propagation loss – Depends on the service – Data rate can be mapped to required signal to interference and noise ratio (SINR) • User resource allocation Path loss model – Traffic expectations • Path loss model System range – Environment/clutter type 2. LTE Downlink Radio Link Budget 12 LTE Downlink RLB • In the following we go through this LTE downlink radio link budget in details • This is a snapshot from excel tool that is given for participants • There will be some solved examples discussed later. Resource allocation and rate requirements Parameter Comment Number of PRBs This is estimated by assuming the operation bandwidth and number of users served at the same time. In case of 10MHz band we have 50 resource blocks (48 for data). Then 10PRB takes 10/48 of all resources Data rate In this case we assume 2Mbits/s target rate Remark on rate requirement • In case of constant bit rate service (like real time video) the 2Mbits/s requirement defines how much resources user continuously employs • In usual case (e.g. web browsing, file downloading, streaming video) the data transfer happens in bursts so instantaneous rate can be high while there are time gaps between transmissions for user (time multiplexing of users) • Example: If user on cell edge download 1 Mbit file – (s)he needs round 0.1 seconds for all 48 PRBs OR – (s)he is given 5PRBs for 1 second time period – Other options – of course – are possible as well Transmission characteristics Parameter eNodeB TX power Comment Typical value is 20W-60W (43dBm-48dBm) 20W on 5MHz band (as in WCDMA/HSPA) 40W on 10MHz band (most usual test case for Rel.8 LTE) 60W on 20MHz band Antenna gain Discussed later in more details. Typical 1.3 m high panel antenna at 2 GHz band gives 18 dBi gain in main direction Cable loss Loss between the eNodeB antenna and the low noise amplifier. The cable loss value depends on the cable length, cable type and frequency band. UE receiver (1/2) Parameter Comment UE Noise Figure (NF) NF measures of degradation of the SNR by the components in the RF receiver chain, product specific. Typical values: 6-11dB Thermal noise = Boltzmann constant x T (Kelvin) x Effective bandwidth. Here Boltzmann constant = 1.38 x 10^(-23) J/K (J = Ws) Reference temperature 20 Celsius = 290 Kelvin Effective bandwidth = Number of PRB’s x 180kHz Thermal Noise Receiver Noise Floor Receiver noise floor = UE NF + Thermal noise UE receiver (2/2) PHICH = Physical HARQ Indicator Channel PBCH = Physical Broadcast Channel PDCCH = Physical DL Control Channel Parameter Comment SINR Required Signal to Interference and Noise Ratio depends on the data rate, number of PRBs and link efficiency. We consider this in more details later in this slide set Minimum required power in receiver required to detect the signal. Receiver sensitivity = Receiver Noise Floor + SINR Receiver sensitivity Control channel overhead Control channel overhead includes the overhead from reference signals, PBCH, PDCCH and PHICH. 5%-25% leads to 1dB-4dB overhead. RX antenna gain Depends on the receiver antenna, usually 0dBi for Margins and losses Parameter Comment Body loss Body loss is typically included for voice link budget where the terminal is held close to the user’s head. 3-5dB for voice. Depends on the propagation environment. Typical values: 4-7dB. Will be discussed later in more details. Shadowing loss Interference margin Interference margin accounts for the increase in the terminal noise level caused by the other cell interference. If we assume a minimum G-factor of −4 dB, that corresponds to 5.5dB IM (10*log10(1+10^(4/10)) = 5.5 dB). Typical values for IM: 3dB – 8dB. Indoor penetration loss Depends on the building types. In urban area 20dB, in suburban/rural area with light buildings 10dB. Allowed propagation loss APL = PTX + GA( NodeB) - LCable + GA(UE) - min { PRX } - M SF - M I - LC - Lbody - LPenetration 3. Some details on RLB RLB elements: a) eNodeB powers and antennas eNodeB transmission power • For macro eNodeB typical value is 20W-60W (43dBm-48dBm) – 20W on 5MHz band (as in WCDMA/HSPA) – 40W on 10MHz band (most usual test case for Rel.8 LTE) – 60W on 20MHz band • For micro eNodeB typical value is 5-10W (37dBm-40dBm) • For pico eNodeB typical value is 100mW-2W (20dBm-33dBm) – 3GPP limit for pico eNodeB TX power is 250mW (24dBm) – There are many ‘pico’ products with TX power 250mW-2W • For femto eNodeB TX power is limited to 100mW (20dBm) – Typical values are some tens of milliwatts Example eNodeB products Ericsson 6201 multi-standard macro base station, indoor installation Nokia macro base station, outdoor wall installation Macro eNodeB antennas: 3-Sector site solutions • Site = location for base station, antennas, cables, etc. • The use of 3 sectors in each site is the most common approach • Omnidirectional antennas can be used in cells with low load • Here color code refers to coverage areas of different antennas (frequencies can be same or different in different sectors) 1 2 3 1 1 2 3 2 3 1 1 2 24 3 2 3 Typical macro eNodeB site antenna • Dual band X-polarization antenna for each sector – 2 transmission branches in downlink and uplink in each sector antenna – Dual band: 800MHz and 2100 MHz – Design contains only 3 antennas but still 6 feeder cables needed 25 Panel antenna example: Kathrein 742 215 • Kathrein multi-band dual-polarization panel antenna (model 742 215) • Typical macrocell eNodeB antenna – Round 18dBi antenna gain – Support 2100MHz and 800MHz bands • Let us look this antenna in more details and compare antenna measurements with the general 3GPP modeling used in simulations (see antenna slides for more details) 26 Horizontal and vertical gain patterns 27 Comparison with Gmain 18dBi 3GPP model q3dB(V ) q3dB(H ) GFB 65 deg 30dB 6.2 deg SLL -18dB Kathrein 742215 Simple gain model Source: F. Gunnarsson et al: “Downtilted Base Station Antennas – A Simulation Model Proposal and Impact on HSPA and LTE Performance”, IEEE VTC 2008 q tilt = 10 degrees 29 ‘High power’ pico eNodeB • Ericsson indoor pico (6401), multistandard • TX output power: 1W • http://www.ericsson.com/ourportfolio/products/rbs-6401 30 Model for pico eNodeB antenna gain pattern 1/2 • Horizontal gain pattern GH ( ) 5dBi 0dBi -180deg -10dBi -100deg-70deg -15deg 15deg 70deg 100deg 180deg Model for pico eNodeB antenna gain pattern 2/2 • Vertical gain pattern GV ( ) 0dBi -3dBi -180deg -100deg -70deg -15deg 15deg • Total gain pattern G ( , ) GH ( ) GV ( ) 70deg 100deg 180deg 32 Example • Two 2W indoor nodes with given antenna gain patterns • Received wideband powers simulated by WinProp tool • 2GHz carrier • Excellent indoor coverage obtained ‘Low power’ Femto base stations Nokia 3G femto base station Vodafone 3G femto base station RLB elements: b) Path loss and shadowing margin b1) Average path loss Single slope model • The most commonly used average path loss model is the so-called single-slope model where – “L0” is the average path loss at reference distance “r0” – “n” is the path loss exponent (which depends on antenna heights, carrier frequency, and propagation environment) – In free space n = 2 ➢ The single-slope model is valid, e.g., when dealing with homogeneous environments Free space model • Free space propagation is an example of a single slope model, i.e., In this situation, we have that Dual-slope model • Another path loss modeling approach is provided by the dualslope model, i.e., where “r0” is known as the break-point distance Linear loss model • In waveguide propagation and absorbing propagation environments, the linear loss model can be sometimes used • Linear model is suitable for tunnels, and indoor propagation through walls (i.e., excellent signal propagation) Environment types (1) – Outdoor propagation environments: Subdivision is done into: • Base station antenna located above roof tops, and • Base station antenna located below roof tops Environment types (2) – Outdoor to indoor propagation environments Subdivision is done again into same categories as outdoor case: • Base station antenna located above roof tops • Base station antenna located below roof tops Environment types (3) – Indoor propagation environments Subdivision is done, e.g., into: • Number of floors in a building to be covered, • Landscape of the (office) building, and so on – Offices or flats with many rooms – Corridors, tunnels, etc. Okumura-Hata Model (1) • (Okumura-) Hata model is one of most common models for signal prediction in large macrocells – This model exists in many version, and is defined for limited ranges of parameters • Originally, this model is valid for: – Distances: 1-100 km – Frequency ranges: 150-1500 MHz (it was extended later) 43 Okumura-Hata Model (2) • Okumura used extensive measurements of base station-tomobile signal attenuation in the city of Tokio (Japan) – He developed a set of curves that gives the median attenuation (relative to free space) of signal propagation in irregular terrain – The base station heights for these measurements were 30-100 m ➢ The Hata model is an empirical formulation of the graphical path-loss data provided by Okumura (model is isotropic) – Closed-form formulas provided by Hata simplify path loss calculations (four different environments were defined) 44 Okumura-Hata Model (3) • The original Hata model is given by where the parameters (and their corresponding units) are 150 and 1500 MHz Okumura-Hata Model (4) • The correction factor for the mobile antenna height “ai(hMS)” depends on the size of the coverage area: – Large/dense city (i.e., “i = 1”), – Medium/small size city (i.e., “i = 2”), – Suburban area (i.e., “i = 3) and rural/open area (i.e., “i = 4”) Okumura-Hata Model (5) ➢ Correction factor for the mobile antenna height (cont’d) Okumura-Hata Model: PL vs. Range (1) Carrier frequency 48 Okumura-Hata Model: PL vs. Range (2) Carrier frequency 49 Okumura-Hata Model: PL vs. BS Antenna Height Carrier frequency 50 Okumura-Hata Model (6) • In mobile communication systems (like GSM and 3G), base station antennas are rarely placed on locations over 40 meters in height • Systems like television and radio broadcasting may use towers that higher than 100 meters – The height of the FM- and TV-mast (Helsinki-Espoo) located in Latokaski (Espoo), has a current height of 326 meters (third highest structure in Finland) • In the next slide, we illustrate the impact of the environment type in the path loss attenuation – It is found that difference between large and medium size cities is small, while – Path loss attenuation in suburban and open areas is far more smaller than in city environments Okumura-Hata Model: PL vs. Range (3) Environment Type 52 Okumura-Hata Model (7) • Later on, Okumura-Hata model was extended to the 15002000 MHz frequencies, in the COST 231 research program – The distance interval was also extended • ITU-R sector adopted this model in Recommendation P.529 where MS antenna height correction factor is the same as in the previous model, and the additional term is given by Okumura-Hata Model (8) • Finally, an extension to the Okumura-Hata model for distances between 20-100km is given by the expression where the “β” parameter is given by Note: The COST 231 extension of the Okumura-Hata model is a single slope model for distances in the range of 1-20 km Remark on path loss models • Okumura-Hata model represents the most well-know macrocell model • Walfisch-Ikegami model is an other well-known model for urban area • In network planning tools above-mentioned models are used with additional clutter corrections – Several different clutter types can be defined with different additional dB/m loss factors – Planning tool design companies are usually not publishing all their models • There are also several models for small outdoor and indoor cells – Yet, the smaller are the cells, the worse are the ‘isotropic models’ that don’t take into account the environment (e.g. building) structure. – When high carrier frequency communication take place in cellular systems (5G) ray tracing models become more important. 55 b2) Shadow fading 56 Example: Path loss measurements Shadow fading (1) • Obstacles with a size from tens to hundreds of wavelengths (on the different propagation paths) cause a variation of the path loss around the average path loss “L” • This variation is random but, however, it is correlated when measured in nearby locations • Shadowing, caused for e.g. by big buildings, can be critical for mobile users located in cell edge areas Shadowing Tx Rx – The effect of shadow fading may create large coverage holes ➢ In many measurements, it has been observed that shadow fading “Ls” can be described with a log-normal distribution*, and the probability density function is given by 1 f ( Ls ) e 2 s L2s 2 s2 * Loss measured in logarithmic scale (i.e., [dB]) is normally distributed Shadow fading (2) • Values obtained in shadow fading measurements* – Carrier frequency: 2.0 GHz Network area/ Parameter Standard deviation Dense urban / Urban 8,5 dB Sub-urban 7,2 dB Rural 6,5 dB * Values reported in different sources vary generally from 6-10 dB Shadow fading (3) 1 Scaled probability distribution function (PDF) 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 -30 -20 -10 0 10 Shadow fading value [dB] 20 30 Probability distribution function for shadow fading when SF standard deviation “σs” is 4 dB (o), 8 dB (*), and 12dB (x) Shadow fading (4) • The cumulative distribution function (CDF) is given by PLs L0 L0 0 L0 f (t ) dt 1 Q , s where “Q(x)” is the Marcum Q-function, Q( x) 1 2 e x t2 2 dt 1 x erfc 2 2 erfc(x) = complementary error function Shadow fading (5) 1 Cumulative distribution function (CDF) 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 -30 -20 10 0 -10 Shadow fading value [dB] 20 30 Cumulative distribution function for shadow fading when SF standard deviation “σs” is 4 dB (o), 8 dB (*), and 12dB (x) Combined path loss and shadowing (1) • In network planning, a Shadow Fading Margin (SFM) is added on top of average path loss • This margin is based on the required target power level (i.e., maximum allowed path loss in the link budget) to guaranteed a given outage probability in the system • Let us denote – “Lmax” = maximum allowed attenuation (link budget parameter) – “Pout” = allowed outage probability at cell edge (QoS of network) – “Ltot” = Total loss including shadowing and average path loss • Then we set (*) PLtot Lmax Pout S-72.3216 RC Systems I (5 Ltot L Ls cr) Lecturesattenuation 3-5, Autumn ”Lmax” = maximum allowed (system specific parameter) 2013 63 ”Ltot” = Total loss at cell edge (network design parameter) Combined path loss and shadowing (2) • From formula (*) we obtain 1 PLs Lmax L Pout • Then, using the Marcum function representation we get Lmax L Pout Q s • It is common to use the inverse Marcum’s function to obtain the (average) path loss that fulfill target requirements, i.e., L Lmax sQ1 Pout • Yet, note that the inverse Marcum’s function does not exist in closed-form S-72.3216 RC Systems I (5 cr) Lecturesattenuation 3-5, Autumn ”Lmax” = maximum allowed (system specific parameter) 2013 64 ”Ltot” = Total loss at cell edge (network design parameter) Inverse Marcum’s function 0 10 -1 10 -2 Q(x) Q(x) 10 -3 10 -4 10 -5 10 0 0.5 1 1.5 2 x x 2.5 3 3.5 4 Combined path loss and shadowing (3) • Revisiting the derived formula that combines path loss and shadowing, L Lmax sQ 1 Pout • The (average) path loss on the left-hand side depends on: – Distance between transmitter and receiver, – Antenna heights of both, transmitter and receiver, – Carrier frequency, environment type, … OkumuraHata model • The value on the right-hand side depends on: – Standard deviation of shadow fading (environment type), – Allowed attenuation (system specific parameter: link budget) – Outage probability (network target performance: guaranteed QoS) In link budget calculations, so-called SF margin is given by M SF s Q 1 Pout Combined path loss and shadowing (4) Lmax Path Loss L Ltot 50% Edge reliability (i.e., “Pout”) depends on assumed “σSF” and shadow fading margin Path Loss + shadowing 90% Lmax Ltot Shadow fading standard devidation (σSF) grows L ➢ As “σSF” grows for same fading margin, the edge reliability is reduced ➢ As “σSF” grows, a larger fading margin is required to maintain the same edge reliability 67 RLB elements: c) Interference, SINR and data rate Co-channel interference: general formulation (1) • General formulation for the received signal: Interfering TX #1 K r S 0 S k nW d1 k 1 where “S0” is the desired signal, “Sk” refers to the interfering signal originated in the k-th (co-channel) transmitter, and “nW” is additive white Gaussian noise (i.e., Thermal noise) Receiver d2 Interfering TX #2 d0 dK Interfering TX #K Source of desired signal Interference in LTE, illustration Desired signal from serving eNodeB Interfering signal from other eNodeB Interfering signal from adjacent sector S-72.3216 RC Systems I (5 cr) Lectures 11-12, Autumn 2013 70 UE eNodeB Co-channel interference: general formulation (2) • Let us compute the expected powers: P E S0 2 RX 0 P TX 0 P / L0 E S k 2 RX k P TX k / Lk P E nW 2 N Here “L0” refers to the path loss attenuation between the receiver and the source of desired signal, “Lk” corresponds to the path loss attenuation between the “k-th” interfering transmitter and the receiver, “PN” the AWGN power • The instantaneous path loss can be written as follows: Antenna gain x Shadow fading Lk = Lk (dk ) Gk (j k )× L × hk SF k Average path loss 2 Fast fading Co-channel interference: general formulation (3) • The shadow fading follows a lognormal distribution, i.e., k 10 linear scale LSF 10 k where “ξk” is a sample from a zero mean Gaussian process, with standard deviation “σk” – If it is assumed that all TX-RX pairs admit the same shadow fading characteristics, then k k • Shadow fading in different links is usually correlated – In most studies of cellular systems, the correlation between shadow fading in different links is assumed to be the same – A typical value for correlation, which corresponds to BS antennas located above the roof-top of buildings, is 0.5 – Correlation between adjacent sector antennas is 1.0 LTE: General formulation of SINR • Now, the general formula for the SINR is given by SINR= P L × h0 G0 (j 0 ) L0 (d0 ) TX 0 K åP 2 SF 0 L × hk × dk ×Gk (j k ) / Lk (dk ) + PN TX SF k k 2 k=1 • This formula can be used in simulations. We have PkTX =transmission power of kth eNB (for certain PRB) LSF k =SF towards kth eNB hk = kth eNB signal fast fading channel power response (for certain PRB) 2 Gk = Antenna gain pattern towards kth eNB dk = 1 if kth eNB is transmitting on this PRB, otherwise 0 Lk (dk ) = Distance dependent path loss towards kth eNB PN = AWGN power General formulation of SINR • If we ignore shadow and fast fading, and assume transmission on all PRB’s, then we obtain wideband SINR that is also called as Geometry-factor (G-factor) Gfactor = P0TX LSF 0 G0 (j 0 ) L0 (d0 ) K TX SF P å k Lk dkGk (j k ) / Lk (dk ) + PN k=1 • G-factor can be used to describe the network deployment, see the next slide – 30% of UEs with ITU modiﬁed Vehicular A at 30 km/h and – 10% of UEs with ITU modiﬁed Vehicular A at 120 km/h Cell selection Best cell selected with 0 dB margin Transmission power Uplink: Max 24 dBm (latest 3GPP speciﬁcations use 23 dBm Antenna conﬁguration Downlink: 1 × 2, 2 × 2 Uplink: 1 × 2, 1 × 4 Receiver algorithms Downlink: LMMSE (Linear Minimum Mean Square Error) Geometry factor • G-factor distribution depends on the network G-factor Uplink: deployment. LMMSE receiver with Maximal Ratio Combining depends on – Antennas, sectorization – Applied average path loss model 1 0.9 0.8 – Frequency reuse (e.g. δ ) 0.7 • Example: – 3GPP macro 1 (ISD = 500m) – 3GPP macro 3 (ISD = 1700m) – 3GPP micro CDF 0.6 0.5 0.4 0.3 0.2 Cell edge: 5%-ile level Interference margin for 3GPP Macro 1: M I =10× log10 (1+10 -G ) 3GPP, macro 1 3GPP, macro 3 3GPP, micro indoor:outdoor = 50:50 0.1 0 -10 -5 0 5 10 15 Geometry (dB) 20 25 30 G-factor has value -4dB =Figure 5.5dB 9.9 Distribution the average wide-band channel SINR (geometry factor) macro ca onof cell edge 3 and micro The corresponding downlink system performance numbers for LTE SIMO and 2 × Source (picture): transmission Holma, Toskala: for inUMTS, Wiley schemes areLTE presented Figure 9.10 and Figure 9.11, respecti vely. Notes on the G-factor distribution • Distribution shows that G-factor varies in macrocell deployment between -5dB and 15dB – Thus, in fully loaded macro network the (average) SINR is limited to 15dB and data rates are limited accordingly – The 50% level of G-factor CDF indicates that (average) SINR in fully loaded macro network is close to 4dB • If microcells are added to the network much higher SINR values can be reached • In e.g. indoor small cells very high SINR values are possible Mapping between SINR/G-factor and throughput in LTE • When creating RLB we require a certain bit rate on the cell edge. – We may, for example, require that rate on cell edge is 1Mbps when user can be allocated 10% of the radio resources. • We can use the following very simple approximation for a bit rate (throughput): (*) TP = BW× M × A× log2 (1+ SINR/ B) = NPRB × BWPRB × SE where SE is the Spectral Efficiency and – BW is the allocated bandwidth (in terms of PRBs) – A and B are factors that are selected such that SE approximates the LTE link spectral efficiency. – Factors A and B depend on the number of antennas and physical layer performance. – M = number of data streams Mapping between SINR and throughput • Previous spectral efficiency formula provides an approximation for the Adaptive Modulation and Coding (AMC) schemes (also called as Modulation and Coding Schemes (MCS)) applied in the system. Resulted parameters of the configurations MIMO M A B SIMO 1x2 1 0.62 1.8 MIMO 2x2 2 0.42 0.85 MIMO 4x4 4 0.40 1.1 MIMO 8x8 8 0.33 1.4 • We note that these parameters represent approximations based on certain simulations. Thus, different assumptions on channel estimation and detection algorithms may lead to slightly different results. Spectral efficiency 30 Max SE Max SE Spectral efficiency [bits/s/Hz] 25 20 SIMO (1x2) 15 MIMO (2x2) MIMO (4x4) MIMO (8x8) 10 5 Max SE 0 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 SINR [dB] 81 LTE Rel.8 peak bit rate examples Modulatio n Stream s Ideal SE [bits/s/Hz] Practical SE [bits/s/Hz] LTE Max bit rate (20MHz) 64QAM 1 6.0 3.75 75Mbps 64QAM 2 12.0 7.5 150Mbps 64QAM 4 24.0 15.0 300Mbps • If carrier aggregation approach is used (LTE-Advanced), then rate can be multiplied by number of carriers. • Maximum spectral efficiencies may take place only when SINR is extremely large => reachable in practical macrocellular networks only in special cases. • Note the difference between ideal and practical SE • These spectral efficiencies set upper limit for SE growth in previous slide 4. DL Link Budget Examples Urban area example (DL) • Assume the link budget parameters below, 10MHz band, 2GHz carrier, 35 meter Radio antenna Communication base station height andSystems 1.5 meterII, UEExercise height. 3, 2014 • Compute the coverage in case of large city for 2Mbps service when eNodeB allocates 4 PRBs for the user (12 users/cell served simultaneously). – – Problem downlink RLB (excelcan in Noppa): thePRBs following link b What happens1. forLTE the service coverage if eNodeB allocate allAssume available 48 for this user (target rate being the same 2 Mbps)? 2.1GHz carrier, 25 meter base station antenna height and 1.5 meter UE height: Increase the user rate 5Mbps and solve problem again Parameter BS TX power BS antenna gain BS cable loss UE noise figure Interference margin RX antenna gain RX body loss Control channel overhead Indoor penetration loss Shadow fading margin BS antenna configuration Value 40W 18dBi 2dB 7dB 4dB 0dBi 0dB 1dB 20dB 7dB 2x2/4x4 MIMO Results • Case 2Mbit/s and 4 PRB’s: – Range in large city = 300 meters – Range in suburban area = 680 meters • Case 2Mbit/s and 48 PRB’s: • Case 5Mbit/s and 4 PRB’s: – Range in large city = 880 meters – Range in suburban area = 2.0 km – Range in large city = 110 meters – Range in suburban area = 240 meters • Case 5Mbit/s and 48 PRB’s: – Range in large city = 650 meters – Range in suburban area = 1.5 km Remark on range (1/3) • Question: We increase the amount of radio resources 12 times (4 PRB ->48 PRB) but range is not increasing directly proportionally. Why range increase is so small? • Answer: If more PRB’s are used, less data needs to be loaded per PRB => SE and accordingly SINR requirement decreases – In 2Mbit/s case SINR requirement decreases from 8.8dB to -7.5dB, the difference being 16.3dB (see next slide) – In 5Mbit/s case SINR requirement decreases from 24.2dB to -2.8dB, the difference being 27.0dB (see next slide) – This increases allowed propagation loss with same amounts. Yet, if distance from/to eNodeB is short, then the path loss increases fast as a function of distance (see next slides) Remark on range (2/3) 10 LTE 2x2 MIMO maximum spectral efficiency (7.5bits/s/Hz) 9 Spectral efficiency [bits/s/Hz] 8 SISO Spectral Efficiency 7 6 5 MIMO Spectral efficiency (2x2) 4 3 Shannon AWGN bound (SISO) 2 1 0 -10 -8 -6 -4 -2 0 2 4 6 8 10 SINR [dB] 12 14 16 18 20 22 24 26 Remark on range (3/3) 16.3dB 27.0dB Other remarks • In 4 PRB case we can serve 12 users at the same time while in 48 PRB case we can serve only single user. – Range extension by using more resources per user can take place only when cell load is low. • User usually needs 2-5Mbit/s rates just during very short time periods – In e.g. web browsing/streaming video data is transferred in bursts. – Thus, if instantaneous rate is high, user will have good use experience. Rural area example • In suburban area LTE is used on 800MHz to provide mobile broadband for single houses – Note: interference margin is decreased to 2dB. • Assume three cases: 1. User is inside a light single house with 10dB indoor penetration loss 2. User is outside the house 3. User has a LTE based fixed wireless system containing directive antenna with 10dBi gain (2dB cable loss) on the house rooftop (7 meters height). LTE receiver is connected to indoor WiFi (through cable) that provides indoor connectivity. • What is the maximum distance from eNodeB to receiver (= rooftop antenna) for 2Mbit/s and 5Mbit/s services if receiver can apply 4/8 PRBs? Illustration of connection options eNodeB Yagi antenna (10dBi gain) Antenna cable (2dB loss) LTE receiver and WiFi router Outdoor UE See, e.g.: http://www.smartcoverage.eu/4G-antenna/4g-lte-aerial-antenna.html Results 4 PRB case: 8 PRB case: • • Indoor user with 10dB penetration loss – – • – – • • 2Mbit/s, suburban area, range = 25.0km 5Mbit/s, suburban area, range = 9.0km • 2Mbit/s, suburban area, range = 4.6km 5Mbit/s, suburban area, range = 2.6km Outdoor user with 0dB penetration loss – – 2Mbit/s, suburban area, range = 5.9km 5Mbit/s, suburban area, range = 2.2km Rooftop directive antenna + indoor distribution – – – – 2Mbit/s, suburban area, range = 3.1km 5Mbit/s, suburban area, range = 1.1km Outdoor user with 0dB penetration loss Indoor user with 10dB penetration loss 2Mbit/s, suburban area, range = 8.9km 5Mbit/s, suburban area, range = 5.0km Rooftop directive antenna + indoor distribution – – 2Mbit/s, suburban area, range = 37.6km 5Mbit/s, suburban area, range = 20.9km Remarks • If cell dimensioning is done based on outdoor coverage, then indoor coverage can be achieved using additional directive rooftop antenna – Costs falling on the user/subsidized by operator or government? – Shared directive antennas + local distribution one good option (village case) • If only fixed wireless based connectivity is assumed very large cells can be used – There will be large coverage holes outdoors and especially for direct indoor coverage – Doubling the cell range decreases the required number of eNodeB’s to ¼ from original case. 5. LTE Uplink Radio Link Budget LTE Uplink RLB • Note that besides data rate the same parameters are used as in DL • The UL data rate with 10PRB’s is 1Mbit/s while it was 2Mbit/s in DL. • Yet the ALP is almost the same as in DL. UE resource allocation and rate Parameter Comment Number of PRBs This is decided by eNodeB after estimating the required bandwidth and scheduling uplink users that are served at the same time. Thus, in this case 1 Mbit/s takes 10/50 of all resources. In (Rel.8/9) uplink PRB’s are given continuously in frequency. Data rate In uplink we have 1Mbit/s while in DL we assumed 2Mbits/s target rate (APL will be almost the same) UE transmission characteristics Parameter UE TX power Comment In uplink maximum TX power is 23dBm. It is assumed here since this RLB consider cell edge user. Power control is used in uplink => TX power can be less than 23dBm as well. UE antenna gain UE antenna gain depends on the device type. Typical value is 0dBi while fixed wireless LTE transceivers may have even 10dBi antenna gain. Body loss Not visible in this UL RLB but 3-5dB body loss can be subtracted. EIRP EIRP = TX power + antenna gain (-body loss) TX power UE power is divided between PRB’s. Thus, with larger eNodeB receiver (1/2) Parameter Comment eNodeB Noise Figure NF measures of degradation of the SNR by the components in the RF receiver chain, product specific. The minimum performance requirement is approximately 5 dB but the practical performance can be better like 2 dB. Thermal noise = Boltzmann constant x T (Kelvin) x Effective bandwidth. The bandwidth depends on the number of allocated resource blocks. With 10 PRB’s we have 121dBm. Thermal Noise Receiver Noise Floor Receiver noise floor = eNodeB NF + Thermal noise eNodeB receiver (2/2) Parameter Comment SINR Required Signal to Interference and Noise Ratio depends on the data rate, number of PRBs and link efficiency. Minimum power in receiver required to detect the signal. Receiver sensitivity = Receiver Noise Floor + SINR (in UL this is given per PRB) eNodeB antenna gain, same as in DL Receiver sensitivity RX antenna gain Margins and losses Parameter Comment eNodeB cable loss Same as in DL. Shadowing loss As in downlink. Values: 4-7dB. Interference margin Interference margin reflects the increase in the eNodeB receiver noise level caused by the interference from (other cell) users. Since LTE uplink is orthogonal, there is no intracell interference but we still need a margin for the other cell interference. This margin depends on the UL target capacity. That is, there is a tradeoff between capacity and coverage. Penetration loss As in downlink Allowed propagation loss APL = PTX + GA(UE) - LBody + GA(eNodeB) - min { PRX } - M SF - M I - LCable - LPenetration Example: Recall the first DL example • Assume the previous link budget parameters (from DL example) 10MHz band, 2GHz carrier, 35 meter base station antenna height and 1.5 meter UE height – – eNodeB Noise Figure = 2dB (In DL for UE it was 7dB) Antenna configuration is now 1x2 SIMO (2 eNodeB antennas, 1 UE TX antenna) • Compute the coverage in case of large city for 1.3Mbps service when eNodeB allocates 5 PRBs for the user (10 users/cell served simultaneously) • Ranges when assuming 1.3Mbit/s and 5 PRB’s: – Range in large city = 300 meters – Range in suburban area = 680 meters Rural area example (UL) • In suburban area LTE is used on 800MHz to provide mobile broadband for single houses – Note: interference margin is decreased to 2dB. • Assume three cases: 1. User is inside a light single house with 10dB indoor penetration loss 2. User is outside the house 3. User has a LTE based fixed wireless system containing directive antenna with 10dBi gain (2dB cable loss) on the house rooftop (7 meters height). LTE transceiver is connected to indoor WiFi (through cable) that provides indoor connectivity. • What is the maximum distance from transmitter to eNodeB for 1.3Mbit/s service if receiver can apply 5/10 PRBs? Illustration of connection options eNodeB Yagi antenna (10dBi gain) Antenna cable (2dB loss) LTE receiver and WiFi router Outdoor UE See, e.g.: http://www.smartcoverage.eu/4G-antenna/4g-lte-aerial-antenna.html Results 5 PRB case: 10 PRB case: • • Indoor user with 10dB penetration loss – • – • • 1.3Mbit/s, suburban area, range = 25.3km • 1.3Mbit/s, suburban area, range = 4.35km Outdoor user with 0dB penetration loss – 1.3Mbit/s, suburban area, range = 6.0km Rooftop directive antenna + indoor distribution – – 1.3Mbit/s, suburban area, range = 3.1km Outdoor user with 0dB penetration loss Indoor user with 10dB penetration loss 1.3Mbit/s, suburban area, range = 8.4km Rooftop directive antenna + indoor distribution – 1.3Mbit/s, suburban area, range = 35.5km