ANALYSIS OF LTE-ADVANCED CAPACITY AND
DATA RATES PERFORMANCE UNDER HIGHBROADBAND DEMAND
Vladimir Nikolikj, MSc, MBA & Prof. Dr. Toni Janevski2
Vip Operator DOOEL Skopje, FilipVtoriMakedonski 3, Skopje, R. Macedonia
(
2Faculty of Electrical Engineering and Information Technologies, RugjerBoshkovik bb, Skopje,
R. Macedonia (
ABSTRACT
In the recent years followed by strongly growth, mobile broadband starts governingthe traffic. For the purpose of satisfying user it is from high importance mobile network operators to assess the needed network capacity, when facing the boosting demand and shortage of the available spectrum. The deployment of 3G and nowadays 4G has resulted in increased capacity, but due to theessentialshift of the traffic balance of the mobile networks caused by broadbanddata load, the support of the much higher peak rates, higher throughput and coverage will become necessity soon. LTE-Advanced (or LTE Release 10) is expected more and more to significantly enhance the LTE Release 8 resulting in abetter user experience. Altogether, this paperaddresses the high level analysis of LTE-Advanced capacity and data rates dimensioningas key issue being considered for the future.
Keywords: LTE-Advanced, Capacity, Spectral efficiency, Peak data rate, Mobile broadband
- INTRODUCTION
Growing mobile traffic demands and their forecasts for the coming years (as estimated e.g. by Qualcomm in [1]) are indicating, that the traffic increase will not be possible to be secured by currently deployed network architectures. Thus, the key aspect of this paper is to analyze downlink capacity and data rates performance of forthcoming radio access technology (RAT)LTE-Advanced by combining the mobile broadband demand and amount of available spectrum. Since the spectrum is a finite, non-exhaustible common resource, our ultimate goal is to assess the impact of the FDD spectrum and its value over the LTE-Advanced RAT in the multi-access environment. In our analysis we use different combinations of system bandwidths, from20 to 100 MHz. Consequently we evaluate the higher spectral efficiency brought with LTE-Advanced and its impact on the achieved area capacity shown as Mbps per km². The unique contribution of this paper comes from analysis of the capacity and data rates performance when using 100MHz bandwidth spectrum as key attribute of LTE-Advanced obtained by the carrier aggregation functionality. We focus on the evaluation of the load impact on the network performance by assumingvery high usage, going even up to 90 GB per user and month.
This paper is organized as follows. In the next section we introduce the mobile user broadband demand. This is followed by an overview of the area and cell types of the populated areas to be covered with certain RAT. Next, a discussion on spectrum and it efficiency for the LTE Release 8 and 10 is provided.In the following two chapters we show the LTE-Advanced network analysis by introducing urban area network modeling for various mobile broadband demands. Here we compare the performance of LTE Release 8 and LTE-A in the context of needed network deployment in order to satisfy demand. We then providedownlink peak data rates evaluation of the bothreleases for various bandwidths available. Finally, conclusions are strained in the last section of the paper.
- MOBILE USER BROADBAND DEMAND
Mobile data usage and mobile broadband penetration continue with the rapid growth in the recent years in Europe. Starting from UMTS, the networks are upgraded to LTE and in the future to LTE-Advanced in order to offer high capacity mobile broadband services. According to [3] the mobile broadband penetration of all active users has reached 54% in EU countries (Denmark, Sweden and UK have around 100%). Based on the same document, the volume of mobile data traffic is expected to grow more than tenfold in the period from 2010 to 2015, reaching almost 8.000 million GB of data in Europe. Global mobile traffic in Q3-2012 hasreached around 900 Petabytes of data and around 190 petabytes of voice. In this paper we consider two levels of high usage of 15 GB and 40 GB and extreme demand of 90 GB per user and month. Also, we consider that the usage is spread out over six hours per day, translating into a busy hour rate of 16.7%, despite that current industry standard considers 12.5% busy hour rate [6].
- CELL CLASSES AND COVERAGE
According to [5] due to take up rate of mobile broadband and traffic pattern, the utilization rates will unavoidable increase and potentially reach critical levels, what will inevitable require upgrade of backhaul and network capacity, by increasing the number of sites. Alternatively, if possible, MNOs could increase the number of carriers by adding additional spectrum, which could replace the deployment of new sites. This brings spectrum to aessential asset as it could be a substitute for new sites.Currently, MNOs have licensed spectrum at different bands and the carriers in between are set at 800 MHz, 900 MHz, 1800 MHz, 2100 MHz and 2600 MHz bands. All parts from the available bandwidth provide different performance for coverage and capacity. For instance, as the low frequency propagates better than a higher frequency, the cell ranges differ based on the used band and the environment (rural, suburban, urban, and super-urban) and its population (inhabitants/km²). Depending upon the aim to establish indoor or outdoor coverage the ranges of the utilized cells (cell ranges or radio ranges) within the particular radio access network (RAN) could vary widely. According to [7], the urban cell range varies from 0.6 km to 1.4 km and suburban from 1.5 km to 3.4 km, with Okumura–Hata propagation model. The rural case shows clearly higher cell ranges that could reach up to 26 km for the outdoor mobile coverage. In[5], authors indicate that the modeling in the 800 MHz for rural area is done with cell ranges with maximum 5.65 km. Also, author in [13] describes the cells with different cell radius as different base station. With the combination of these inputs and the input coming from [2], [4], and [6], within the Table 1 we summarize the RAN characteristics related to different surface of the sample covered areas, populated with various number of inhabitants per km² and covered with various type of cell classes each having different radio range emphasized in km. Thus, for instance if we have certain sub-urban area with 20.000 km², it could be covered with 995 sites with cell range of 2.53 km. Here,we consider that all sites are set up into three sectors and the coverage calculated as follows:
Coverage = π* cell range (km) ²
- BANDWIDTH AND SPECTRAL EFFICIENCY
Going further, we would once again address that the size of the bandwidth owned by the particular MNO plays key role as well. The more bandwidth that can be used at one site the higher the capacity. In this paper we analyze different spectrum allocation alternatives, varying from 20, 50 and 100 MHz and in the different bands (we consider only the downlink transmission or one-way representation i.e. 50 MHz out of the both way transmission represented as 2*50 MHz). According to [10], carrier aggregation allows combining lower and higher bands— leveraging better coverage of the former with higher availability of the latter (up to 5 carriers and up to 100MHz supported in standards). In this paper weassume utilization of carrier aggregation functionality as LTE-Advanced attribute for the spectrum of 100 MHz in downlink. More on impact of carrier aggregation and network sharing could be found in [6].It is common understanding that spectrum could be used as a substitute to increase the number of sites in order to enhance the capacity. Also it is wieldy accepted that the MNOs could use different bands of spectrum in the different propagation environments. Consequently, the spectral efficiency is a key concept that determines the efficiency of spectrum (bps/Hz). The values of the spectral efficiency depend from the value of the RAT used. For instance, the LTE guarantees higher spectral efficiency in comparison with HSPA+, and LTE-Advanced higher than the LTE. It should be noted that the spectral efficiency varies with the distance from the base station.
Table 1.Range, coverage and number of sites for different types of areas.
Area type / Covered area (km²) / Inhabitants/km² / Radio or Cell range (km) / Cell coverage area (km²) / Sites / Sectors / Ratio (Sectors / Coverage Area)Rural / 200,000 / 20 / 5.65 / 100.24 / 1995 / 3 / 0.03
Sub-urban / 20,000 / 200 / 2.53 / 20.10 / 995 / 3 / 0.15
Urban 1 / 2,000 / 2,000 / 1.79 / 10.06 / 199 / 3 / 0.30
Urban 2 / 2,000 / 2,000 / 0.70 / 1.54 / 1300 / 3 / 1.95
Urban 3 / 2,000 / 2,000 / 0.57 / 1.02 / 1960 / 3 / 2.94
According to [8] three types of spectral efficiencies are identified: peak spectral efficiency, the average spectralefficiency, and cell edge spectralefficiency. The peak spectralefficiency is the highest data rate normalized by overall cell bandwidth assuming error-free conditions, when all available radio resources for the corresponding link direction are assigned to a single UE. For the LTE-Advanced, the system target to support downlink peak spectralefficiency of around 30 bps/Hz and uplink peak spectralefficiency of 15 bps/Hz.Nevertheless, rather than elaborating with peak rates (the theoretical maximum) and bit rates at cell boarders, the researchers calculate with average throughput over the entire cell area. As indicated in [5], this implies that it should be assumed that users are receiving similar bit rates close to the site as well as at the cell boarder. Thus, more applicable is the use of the average spectralefficiency that is defined as the aggregate throughput of all users (the number of correctly received bits over a certain period of time) normalized by the overall cell bandwidth divided by the number of cells. The average spectralefficiency is measured in b/s/Hz/cell.Author in [4] uses average spectral efficiency of 0.67 for HSPA evolution technology and 1.67 bps per Hz for future releases of LTE radio access technology. Authors in [6] considers average spectral efficiency of 2.0 for LTE for the urban environment scenario (in line with what equipment manufacturers presents), also indicating that the spectral efficiency varies with the distance from the base station, typical target values are > 10 close to the base station, < 0,01 at the cell border and 1-2 as average over the whole cell . According to [9] there is identified couple of downlink cell spectral efficiencies for the LTE-Advanced (FDD) (based on the propagation environment). The highest values of each are summarized within the next Table 2 (cell spectral efficiency is evaluated by assuming overhead corresponding to downlink control channels that spans L OFDM symbols with L = 1.). In this paper we focus on urban area analysis in order to compare the LTE and LTE-Advanced RATs with the following average downlink cell spectral efficiencies:
a)2.00 bps/Hz/cell, relevant for LTE; and
b)3.80 bps/Hz/cell, relevant for LTE-Advanced.
Table 2.Downlink spectral efficiency with the use of scheme and antenna configuration relevant for LTE-Advanced Releases (FDD).
Environment / Scheme and antenna configuration / Cell average [b/s/Hz/cell]Indoor / MU-MIMO 4 x 2 (C) / 6.6
Microcellular / MU-MIMO 8 x 2 (C/E) / 4.2
Base coverage urban / CS/CB-CoMP 8 x 2 (C) / 3.8
High speed / MU-MIMO 8 x 2 (C) / 4.1
- NETWORK CAPACITY
The capacity is the most vital key performance indicator based on which the performance of certain RAT should be evaluated. It is a function of the total number of sites, the amount of spectrum that is available and the spectral efficiency of the radio access technology. Based on [5] the capacity is calculated according to the following formula:
Capacity = Bandwidths (MHz) * number of sites * sectors * spectralefficiency (bps/Hz)
In order to express the capacity as throughput per area unit (Mbps per km²), we introduce the more convenient for calculation multiplier derived by division of the number of sectors and coverage area with single site, by what we adopted version of the capacity formula as follows:
Figure 1.Cell capacities in Mbps per km²as function of spectral efficiency of different RAT(LTE (2.0 bps/Hz) and LTE-Advanced (3.8 bps/Hz)) and cell radio coverage in km²and amount of used spectrum in MHz.
Capacity = Bandwidths (MHz) * (number of sectors / coverage area) * spectralefficiency (bps/Hz).
Based on this formula and ratios (number of sectors / coverage area) given in the Table 1 above, the Figure 1 summarizes the capacity values (in Mbps per km²) for LTE and LTE-Advanced considering different chunks of bandwidth (cell radio range is in km). As per Table 1, here we evaluatethe 3 types of urban scenarios, each with cells having the following radio ranges:
a)Urban 1: radio range of 1.79 km;
b)Urban 2: radio range of 0.70 km; and
c)Urban 3: radio range of 0.57 km.
Figure 1 clearly indicates that LTE-Advanced RAT outperforms the LTE RAT in terms of available average capacity per cell and km². Even the usage of cell classes with radio range of 0.7km and equipped with the LTE-Advanced technology shows better capacity performance than the LTE RAT implemented within the cells having smaller radio range of 0.57km. This could be interpreted in a way that for the same covered area, with smaller number of sites (higher radio range) and utilization of LTE-Advanced the better capacity performance will be achieved than more sites (lower radio range) and utilization of LTE. Also, it could be noted that with usage of LTE –Advanced and sites with radio range of 0.57km the cell capacity to be achieved is above 1.1 Gbps.
- LTE-ADVANCED NETWORK MODELING
This chapter approaches the evaluation of the capacity solely for LTE-Advanced RAT by combining the user’s demand for mobile broadband and amount of spectrum. We perform the modeling and analysis of network capacity for urban area having 2.500 km²densely populated with 2.200 inhabitants per km². We consider relatively high penetration rate for mobile broadband of 55%.According to [13], the spatial distribution of the traffic can be uniform or non-uniform.In this paper we consider uniform spatial traffic density within the simulated test area [12]. Further, we assume market with three MNOs all having equal market share. We analyze the Urban 2 scenario, assuming that each of the three MNOs covered the area initially with 1625 sites, yielding that each site is covering area of 1.54 km²(cell area). We assume that MNOs have the following spectrum status:
a)MNO 1 using 2x20 MHz in both the 1.8 GHz and the 2.6 GHz bands
b)MNO 2 using 2x50 MHz in both the 1.8 GHz and the 2.6 GHz bands
c)MNO 3 using 2x100 MHz in the 800MHz, 1.8 GHz and the 2.6 GHz bands (carrier aggregation in place)
Consequently, the analyzed spectrum is distributed in ranges from 20 MHz, 50 MHz and 100 MHz applicable for LTE-Advanced FDD services.Based on chapter 2, in the modeling analysis we assume three levels of demand shown through the following usage: 15, 40 and 90 GB per user and month. The usage busy hour rate is 16.7%. The download demand is transferred into the data rate as shown in the Table 3.
Table 3.Transformation of download to the Mbps/km²per MNO.
Load (GB) / Mbit/s/user / Mbit/s/MNO/km²15 / 0.185 / 75
40 / 0.494 / 199
90 / 1.111 / 448
According to [6], we than compare the capacity with the demand in order to calculate the utilization rate and after we determine if the network needs to be upgraded or if the installed capacity within the area is sufficient. In order to maintain the appropriate quality of service levels of multimedia data traffic with burst nature, in this paper we evaluate the LTE-Advanced network behavior with 60% the utilization rate, meaning there is thoughtful(but needed) spare capacity left of 67% on top of the needed capacity. Saying different we model future proved mobile network, since as indicated within [1], the increased capacity will help operators to address bursty smartphone traffic as well as increased use of social media applications, instant messaging and others. More on the analysis of LTE-Advancedforbursty traffic model can be found in [11].All the above inputs indicates that the capacity, expressed as Mbps per km², is 5x times higher for MNO 3 compared to operator MNO 1, ranging from 148 to 741 Mbps per km², as Figure 2 shows.Additionally, at Figure 3 we depict the utilization rate of the three networks for all three amounts of user demand.
It is clear that all of the three networks have necessary capacity to manage data volumes produced from an average usage of 15 GB per month and user.
Figure 2.Required capacity per km²for the three MNOs.
Figure 3.Utilization rate of the three MNOs for various user’smonthly download.
But with an enlarged usage to 40 GB per user and month MNO 1 is obligatory to upgrade its network, while the two other operators have abundant capacity to manage the requested user’s demand. Furthermore, an ultimately increased usage to 90 GB creates capacity problem for two out of the three MNOs, as illustrated in the Figure 3. Consequently, MNO 1 and MNO 2 are in the situation to upgrade their networks with more sites. Figure 4 shows the total amount of sites that are required to meet the demand for the three usage levels, implying that the initial 1625 sites are counted in.MNO 1, which only has 20 MHz to its disposal, needs to deploy about 2015 additional sites to handle the increased traffic generated by a usage of 40 GB per user per month.With an increased usage to 90 GB, MNO 1 has to install an additional 6565 sites in order to encounter the demand from its subscriber base, reaching a total of 8190 sites. MNO 2 has to deploy an additional 1651 sites, reaching 3276 sites. Nonetheless, MNO 3 is able to accomplish the increased demand from the customers without being required to make any interchanges to its originally deployed network.
Figure 4.Number of sites needed to manage increased traffic.