Energy Efficiency and Performance in mobile networks deployments with femtocells

Luis M. del Apio, Ignacio Berberana, Luis Cucala, Emilio Mino Daz, Oscar Moreno.
Telefnica Investigacin y Desarrollo, Madrid, Spain e-mail: {apio, emino, lcucala, omj}@tid.es

EstherTorrecilla
Telefnica Espaa Madrid, Spain e-mail esther.torrecillamarquez@telefonica.es

Abstract The new generations of cellular technologies and the current trend towards small cells or femtocells, will offer an improved spectral efficiency per area, but they also offer an opportunity to improve the energy efficiency, measured as the power consumption needed to provide a certain throughput in a given area. In this paper, two network performance indicators, aggregated throughput and energy efficiency, have been analyzed and compared in two network architectures; a traditional deployment based on outdoor macro base stations, for the provision of outdoor and indoor coverage, and a deployment where some of the indoor traffic is supported by femtocells. Performance simulations and energy calculations have shown that the introduction of a femtocell layer, complementing the macrocell layer is a great leap forward in system performance and energy efficiency, when compared with current indoor coverage based on outdoor macro and micro base stations. Additionally, this paper present a study on how much could be reduced the mobile network energy requirements of macrocell deployments in high density urban deployments, in low traffic conditions, based on the application of two Self Organizing Networks (SON) techniques for energy saving, selective disconnection and power reduction of eNodeBs and HeNBs. Keywords. performance LTE, femtocells, energy, efficiency, system

each other, with minimal radio transmission losses that translate in a higher spectral efficiency [1]. Currently around of 70% of 3G traffic is originated indoor, as it is stated by equipment vendors and telecom operators [2], and in the future with the increasing use of smartphones, indoor traffic volume will increase. As an answer the mobile telecom industry is working on the femtonode concept to provide indoor coverage from base stations placed at the customers premises, as a complement of the macro/micro cell layers [3]. In this article femtonode is referred to the physical equipment and femtocell is applied to the radio coverage produced by the femonode. This paper is organized as follows; section II analyzes a typical macrocell urban scenario, section III studies an equivalent urban area served by a hybrid macro and femto network and shows the improved energy efficiency, section IV proposes other techniques to reduce the mobile network energy requirements in macrocells, and finally section V is devoted to conclusion. II. MACROCELL COVERAGE ANALYSIS FROM THE ENERGY EFFICIENCY POINT OF VIEW

I. INTRODUCTION Current mobile networks provide radio coverage by means of layers of base stations, based on a first layer of high power macrocells, and a second layer of medium power microcells. In general these layers are intended to provide simultaneous outdoor and indoor coverage, and thus are installed in outdoor locations. This standard procedure has worked fine as long as the traffic has been mostly originated from outdoors, but with the introduction of data-related services the scenario has reversed and the main traffic demand comes now from indoors. In this indoor traffic demand scenario, a from outdoors to indoors coverage solution is obviously not optimum, a from indoors to indoors approach becoming preferable. The former solution incurs in high external walls penetration losses, and the latter is a natural evolution of the cellular concept, reusing resources in ever smaller cells, and can make a better use of the radio power as the femtonode and the user are located near

A. Urban macro network deployments characteristics The macrocell urban scenario is usually intended to provide a good outdoor coverage and a sufficient indoor service, and its performance in dense urban deployments is usually interference-limited, because the unwanted power from neighboring cells becomes noticeable, as there is a high degree of overlapping between small urban neighboring macrocells. As a reference, 3GPP uses for the dense urban reference scenarios an inter site distance (ISD) of only 500 m (3GPP Case 1) [4]. B. LTE macrocell urban scenario simulation A macrocell reference scenario with 18 LTE eNodeBs (54 cells) has been selected as an example of a real urban deployment, where all the traffic is served with macrocells. This reference scenario is a dense urban area in Madrids city center (Salamanca district), with an area of 2.86 km2 and an average inter-site distance of 250 meters.

The eNodeBs transmit a power of 46 dBm (40 watts) per sector, the operation frequency is 2 GHz, the bandwidth is 10 MHz, and diversity is used in the eNodeB receivers. The overall capacity of this scenario is calculated considering 10 active users per cell, i.e. 540 active users are randomly distributed over the whole scenario, outdoor and indoor. For the energy requirements analysis, the most favorable situation for the macro deployment has been selected; outdoor base stations that do not need any shelter nor air conditioning, thus the maximum consumed power per site, with three cells, is 790 W. On the other hand, an eNodeB has always a minimum power level consumption, even when no traffic is served, that is approximately a 40% of the full traffic load situation. In intermediate traffic situations, the power consumption is interpolated. These data, due to confidentiality constraints, is an average of the Telefnicas eNodeBs suppliers in Spain. The macrocell reference scenario has been evaluated by means of a radio planning tool [5] that calculates the signal level by means of ray tracing and real cartography of Madrids city centre. The signal level is calculated in a grid of points in the reference scenario, and in a subsequent step the throughput in each point is calculated using a LTE look up table (SINR throughput), obtained from a Telefnicas proprietary system level simulator. Finally the scenario average capacity is computed. C. Total throughput, total radiated power and total consumed power Figure 1 present the available throughput map. The aggregated average capacity for the 18 sites is 437.4 Mbps.

D. Figures of Merit calculation In order to give a clear insight of the energy efficiency of this scenario, some good analogies are the efficiency figures used in cars in America (miles per gallon) or in Europe (liters per 100 Km), where the throughput is equivalent to distance, and the consumed watts are equivalent to gallons or liters of fuel. Let us call these figures of merit Mbps per watt, Mb/W, and its inverse, watts per Mbps, W/Mbps. In the proposed reference scenario, when only the radiated power is considered, Mb/W equals 0.203, and this value decreases to 0.031 when the total consumed power is taken into account. The Europe style figures, W/Mb, are perhaps more intuitive in this macro reference scenario; they mean that it is necessary to consume 32.51 watts and to radiate 4.94 watts to provide an average bit rate of 1 Mbps per user. III. COMBINED MACRO AND FEMTO COVERAGE ANALYSIS FROM THE ENERGY EFFICIENCY POINT OF VIEW Section II has described the current network architecture used worldwide, but in the next future, as the customers request better indoor coverage and a higher throughput, an important fraction of the houses in a given area will install a femtonode. This section analyzes this future hybrid macro and femto network, from the throughput and energy points of view A. Multilayer macro and femto urban scenario simulation We are going to analyze the same geographical area described in II.B, but considering that a fraction of the houses will install a femtonode. For this study it was used a 8 m 10 m reference apartment (Figure 2), with brick exterior walls and plasterboard partitions, which is typical of Madrids Salamanca district simulated area.

Figure 2. Reference 80 m2 apartment scenario with a central femtonode

Figure 1. Available throughput map in the reference scenario

The total radiated power is 2,160 watts (40 watts/cell multiplied by 54 cells) and the total consumed power is 14,220 watts (790 watts/site multiplied by 18 sites).

In this scenario, the in-building area has a ground floor extension of 975,188 m2, and taking into account that the average number of stories per building in the district is five; the total indoor area is 4,875,942 m2, and there should be 60,949 reference apartments in this area. The indoor section of the reference scenario has been evaluated by means of a Telefnicas proprietary indoor

propagation tool that calculates the SINR level using statistical and semi-empirical procedures, considering the layout and characteristics of the materials that compose the reference apartment. Afterwards, it is computed the outdoor section of the scenario, following the process explained in section II.B for macrocells. In this exercise, it is proposed a future situation of high femtocell acceptation ratio, where an arbitrary 18% of the apartments, 10,971, have installed a femtocell. Most of the apartments will enjoy a higher throughput than what could be provided from a macro, improving the aggregated throughput in the reference scenario, and at the same time they will offload the traffic and coverage requirements on the macro layer. For example, if an 18% of the indoor area is served by femtonodes, the operator can afford to reduce every macro site power and tolerate an indoor coverage gap (from the macro) of the order of an 18%. In our reference scenario, the simulations performed indicate that macrocell radiated power per eNodeB can be reduced from 46 dBm to 38 dBm (an equivalent result could be obtained reducing the number of sites), which renders a site power reduction from 790 W to 391 W. Given this new reference scenario, let us proceed to analyze the aggregated throughput and power consumption. It is assumed that femtocells and macrocells are using different frequencies; therefore there are not significant interferences amongst the macro cell and femtocell layers. This exercise is deliberately pessimistic and therefore it is not considering femtocells without interferences from neighbor femtocells, and thus most of the femtonodes are highly interfered. The femtonodes are divided into four categories: Type 0 Without an interferer femtocell. Not considered in this study. Type I Serving Femtocell surrounded by one interferer femtocell. The simulation scenario sets that a 10% of femtocells fall into this class (1,097). Type II Serving Femtocell surrounded by two interferer femtos. 40% of the scenario femtocells are classified Type II (4,388). Type III Serving Femtocell surrounded by five interferer femtos. 50% of the femtocells are considered to be Type III (5,486). The percentages proposed for the distribution of the different types of femtonodes are an educated guess, based on the homes distribution in a densely populated area as Madrids Salamanca District. Every femtocell is located at a height of 1 meter and radiates 15 dBm from a 6 dB gain omnidirectional antenna. The consumed power per femtocell is 8 watts (data obtained from commercial 3G femtos). The operation frequency is 2.6 GHz and the bandwidth is 10 MHz. Figure 3 depicts the capacity results for a Type III femtocell. The average cell throughput for each femtocell type is: 27.5 Mbps for Type 0, 19.1 Mbps for Type I, 17.9 Mbps for Type II and 17.5 Mbps for Type III.

There is not a noticeable performance degradation increase between Type II (2 neighbor interferes) and Type III (5 interferers), because in Type III two of the interferers are located at the other side of the corridor, separated with additional walls). This is the reason for not considering some intermediate Types, for example a serving femtocell surrounded by 3 or 4 interfering femtocells. Figure 3 depicts the capacity results for a Type III femtocell.

Figure 3. Femtocell Type III throughput map

B. Total throughput, total radiated power and total consumed power The simulations that have been done to evaluate the aggregated capacity of the 10,971 femtocells and the 54 macrocells in the reference scenario, taking into account the type of femtocells, provide a total aggregated average throughput calculated for the whole scenario of 195,884 Mbps, from which 195,534 Mbps come from the femtonodes, and 350 Mbps from the macrocells. Regarding energy, the total radiated power is 687 watts (346 W from the femtonodes, and 341 W from the macros) and the total consumed power is 94,806 watts, the femtonodes being responsible of 87,768 W, and the macros consuming 7,038 W. C. Figures of Merit calculation In the macro plus femtonodes scenario, the figure Mb/W equals 285 Mbps/watt when only the radiated power is considered and 2 Mbps/watt when all the power consumption concepts are taken into account. If the inverse parameter, W/Mb, is used, the scenario consumes 0.484 watts and radiates 0.0035 watts to provide a bit rate of 1 Mbps D. Comparison with the current macrocell approach The energy efficiency comparison is clearly favorable to the macro plus femto approach. Table 1 summarizes and compares the figures of merit of both deployment solutions.

TABLE 1. Macrocell vs. Femtocell energy efficiency comparison Figures of Merit Macro-only Macro & Femto deployment deployment Mb/W of 0.031 2 total power Mb/W of 0.203 285 radiated power

It can be observed that the hybrid macro and femto approach is about 65 times more efficient in terms of total consumed power than the macrocell approach and about 1,400 times more efficient in terms of radiated power. These figures reflect that femtocells can improve not only indoor coverage and throughput, but also greatly improve indoor mobile service energy efficiency. However, it must be taken into account that the hybrid scenario absolute power consumption has been multiplied by 6.6 with respect to the reference one, thus making it mandatory to implement some femtonode switching-off procedures when they are not providing service, as it is described in Section IV-E. IV. STRATEGIES TO REDUCE ENERGY CONSUMPTION IN THE MOBILE NETWORK A. Selective power reduction or Base Station disconnection 3GPP LTE (Long Term Evolution) standards introduces self-optimization functions (SON) [6] that can be used to provide energy savings without impairing system performance when there is a low use of the mobile network (e.g. night time, weekends). Two strategies, reduction of the radiated power and selective disconnection of cells, can be applied in dense urban zones, where a big number of base stations with overlapping coverage are used. It is not on the scope of this paper to define the SON algorithm on energy saving but to analyze its effects on terms of energy saving and scenario throughput. Simulation setup In the case of a selective shutdown of some cells, it is assumed that the disconnected cells only switch off the radio section, keeping active other elements, like the backhaul with the core network, in order to facilitate a quick power up of the base station. As it was described in Section III.B, a base station with three sectors when switched off in this way will still consume about a 40% of the full-power situation (i.e. 280 W). The same city area and number of eNBs that were depicted in Figure 1 have been simulated for the analysis of the effects of these energy saving techniques. The eNodeBs radiate the same maximum power of 46 dBm per sector, but in this case the bandwidth is reduced to 5 MHz and a 2x2 MIMO configuration has been used. B. Selective disconnection of cells It has been simulated the effect of switching off some macrocells. These cells were selected with the criteria of not creating coverage gaps in the scenario; therefore neither the

cell border cells nor the ones with scarce overlap were disconnected. Since this scenario is interference-limited, shutting down some cells does not lead to a significant degradation of the average SINR distribution, compared with the reference scenario, and thus there is only a reduction of the total available throughput that can be offered to the terminals. The explanation is straightforward; switching off cells decreases the signal level in some zones, but in most areas the coverage is assumed by neighbor cells that now operate with a lower interference. The initial reference scenario provides an accumulated throughput of 226 Mbps. When 6, 15 or 24 cells are switchedoff, the total available throughput decreases 10%, 22% or 40% respectively. But as switching-off is performed when the total demanded traffic is low, a reduction of the total throughput is not important, and what must be analyzed is the area where 1, 3, 5 and 10 Mbps throughputs can be offered to the user terminals, in order to asses if a good service can be offered to those customers still active. As it is shown in Figure 4, a 1 Mbps service can be supported in all the area when switching off 15 cells. Even though the available throughput is partially affected, energy savings compared with the reference scenario are important. For example, switching-off 6, 15 or 24 cells renders an energy reduction of 7%, 17% or 27% respectively.

Fig. 4. Throughput distribution, switching off 15 cells

C. Reduction of radiated power In this case the energy optimization strategy is to reduce the transmitted power in all the base stations, as it was done in Section III when part of the indoor traffic was served by femtonodes. The impact of a transmitted power reduction from 46 dBm to 30 dBm, in 4 dB steps, applied to all the cells in the reference scenario, has been evaluated. For example, figure 5 presents the available throughput when the transmitted power is 38 dBm, where some white

spots can be observed, meaning that there are areas where only a service below 1 Mbps can be offered. When the transmitted powers are 46, 42, 38, 34 and 30 dBm, the accumulated throughputs are 226, 202, 179, 157 and 136 Mbps respectively. The energy savings for these values of power are 0%, 36%, 50.5%, 56% and 59%.

predefined macrocell it is assumed to be far away from home and the femtonode is switched off. A more powerful implementation is to detect the customer presence by means of a low-power radio interface activated in the User Equipment, for example a Bluetooth Low Energy, which is always active in the UE but whose low power characteristic does not degrade significantly the battery lifetime. When the user arrives at home, a short range radio connection between the UE and the femtonode is established, and the latter can switchon its radio section accordingly (the opposite is done when the user leaves the home and the short range connection is lost). Currently, this strategy is under evaluation in 3GPP [8] [9]. V. CONCLUSION

Fig. 5. Throughput distribution, with a reduced transmitted radio power of 38 dBm (applied to all cells).

D. Analysis of selective disconnection vs power reduction The transmission power reduction approach provides a better energy efficiency than the switching off strategy, because the power reduction can be applied to all the cells, with the exception of the non-overlapping cells, and thus reducing interference more efficiently. For example, a 4 dB power reduction leads to a throughput decrease of only 10%, but enjoys an energy saving of 36% On the other hand, when switching off 15 cells, the throughput reduction is more severe (40%) and the energy saving is less rewarding (26%). E. Selective disconnection of femtonodes Even though the simulations that have been presented in this section have been done for a macro network, all the procedures and conclusions can be applied to the hybrid macro and femto scenario, in particular if it is taken into account that the femtonode users will not be at home during an important fraction of the day, and then the femtonode can be switched off, rendering a potential sharp reduction of the aggregated power consumption . The most efficient approach is to actually detect when the users are not at home, and thus switch off the radio section of the femtonode. Some implementations have been proposed for switching off the radio section of a femtonode when the user is not in the neighborhood of his/her femtonode [7]; they detect when the User Equipment is camped in the nearest macro cell to the femtonode, in order to decide when switching on or off the femtonode. When the User Equipment is not camped in a

This study has demonstrated that the best strategy to provide indoor wireless mobile connectivity, from both the service and the energy points of view, is to place the transmitting station indoors, and this can be done by means of a femtocell layer that complements the traditional macrocell deployments. Energy calculations and throughput simulations show that the energy efficiency, measured as the required energy for a given throughput, is much better in a hybrid macro and femto network than in the traditional macro outdoors deployments. However, even though the energy figures of merit are very good for femtonodes, their large number can add up to a very high aggregated power consumption, so it will become important to implement strategies to reduce their power during operation and even switching them off, when they are not serving traffic. Another solution to save energy that can be applied to both macrocells and femtocells deployments in interference-limited dense urban scenarios, is to switch-off or reduce the radiated power of some stations during low traffic demand periods. In a macro deployment simulated scenario, a reduction of 4 dB in the eNBs transmitted power, leads to a 36% energy saving and only penalizes throughput in a 10 %, demonstrating a huge energy saving potential. REFERENCES
[1] [2] [3] Arraycom , Coopers law http://www.arraycomm.com/serve.php?page=Cooper G. Mansfield, Femtocells in the US Market - Business Drivers and Consumer Propositions, FemtoCells Europe 2008. June 2008. D. Calin et al. On femto deployment architectures and macrocell offloading benefits in joint macro-femto deployments. IEEE Communications Magazine, vol. 48, Issue 1. January. 2010, Page 26 3GPP: Requirements for further advancements for Evolved Universal Terrestrial Radio Access (E-UTRA) (Release 7), 3GPP TR 25.912 V9.0.0,Tables 13.4 e/g/i/j, October 2009 Sidarel, Volcano propagation ray tracing tool, http://www.siradel.com/1/volcano-software.aspx 3GPP TR 36.902, Evolved Universal Terrestrial Radio Access Network (E-UTRAN); Self-configuring and self-optimizing network (SON) use cases and solutions, V9.3.0, December 2010. Patent application US 2009/0285143, Apparatus and method for saving power of femto base station in wireless communication system R3-110030, Dynamic H(e)NB Switching by Means of a Low Power Radio Interface for Energy Savings and Interference Reduction. 3GPP TSG RAN WG3 Meeting, Dublin, Ireland ,17 21 January 2011 R3-101619, Energy Saving for HeNBs, 3GPP TSG RAN WG3 Meeting, Montreal, Canada, 10th-14th May 2010

[4]

[5] [6]

[7] [8]

[9]