Energy Efficiency Improvement Through Pico Base Stations For A Green Field Operator

Transcription

1 2012 IEEE Wireless Communications and Networking Conference: Mobile and Wireless Networks Energy Efficiency Improvement Through Pico Base Stations For A Green Field Operator Malik Wahaj Arshad, Anders Vastberg and Tomas Edler Royal Institute of Technology (KTH), Stockholm, Sweden; Huawei Technologies AB, Stockholm, Sweden {mwars, vastberg}@kth.se tomas.edler@huawei.com Abstract Mobile telecommunication operators are now focussing on emerging markets due to the current highly competitive mobile telecommunication sector in established markets. The deployment of wireless mobile infrastructure in these emerging markets or a green field scenario requires an innovative energy efficient approach, which is not feasible in an incumbent operator scenario. This paper describes a combined macro and pico cellular heterogeneous wireless network architecture, and analyses its energy efficiency with respect to variation in inter site distance. The increase in capacity and power saving through sparse network deployment is investigated in terms of area spectral efficiency and area power consumption respectively. The results suggest that the deployment of pico cells along with a traditional cellular network can improve the energy efficiency of the network, as well as provide gains in terms of increased inter site distance. Finally. the indifference curves of Energy Efficiency and number of pico nodes indicate the optimum deployment scheme for multiple area spectral efficiency targets. Index Terms Pico, Power Consumption, Heterogeneous Networks, Energy Efficiency, Macro Offloading, Quality of Service (QoS). I. INTRODUCTION The number of mobile phone subscribers worldwide increased from 700 million to 5 billion during the period 2000 to 2010 with the fastest growth occurring in developing regions including China and India [1]. Mobile phone penetration increased globally from 20% in 2003 to 67% in 2009 [2]. This phenomenal increase in mobile phone users increased the global energy consumption. Both the operators and equipment manufacturers are now focusing on methods to improve the energy efficiency of telecommunication networks. The equipment manufacturers are proposing new transceiver designs, improved cooling and power amplifier solutions for the access sites, and alternative power sources [3], while operators have focused on introducing architectural changes in their network deployment which provides both cost and energy efficiency improvements. The proposed solution is the introduction of heterogeneous networks (HETNET) where low power nodes would be deployed along with macro cellular base station (BS) within the coverage area of the cell [4]. The potential energy saving is expected to come from the fact that the low power nodes are much closer to the end user and thus face lower path losses and propagation distances as compared to the macro BS. Also, the shifting of some of the users to pico cellular BSs eventually leads to a higher macro cellular average signal to interference and noise ratio [5]. Moreover, the radio access network consumes around 60% of the power consumed by the mobile telecommunications industry [1].Therefore, if low power nodes such as pico and femto cellular BSs can offload the macro cellular BSs, then HETNET can reduce energy, increase capacity, and improve coverage of future cellular systems. Numerous studies have been made pertaining to HETNET deployment and the potential energy saving. As the users at the cell edge require the highest transmission (Tx) power of the macro BS, an investigation which involves positioning of micro BS at the cell edge to save the macro BS energy is performed in [6].The performance evaluation parameters for heterogeneous deployment strategies including area power consumption (APC), area spectral efficiency (ASE) and area throughput are introduced in [7] then used in [8], [9],and [10]. Linear power models are used to calculate the power consumption for both the micro and macro BS in [11]. The power consumption of system under varying load conditions is also analyzed in [10]. Network management techniques and different deployment strategies are introduced in [12]. An Investigation that evaluates the energy efficiency of deploying residential pico cells with open access is performed in [13]. In this paper, we investigate the energy saving achieved through traffic offloading and support for sparse network deployment by pico BS in joint macro pico BS deployment. The three parameters namely APC,ASE and Energy Efficiency are used to analyze these energy gains. The following scientific question is focused in this paper: What would be the optimum deployment strategy in terms of APC for a green field operator, given a target ASE? The remainder of the paper is organized as follows. Section II introduces the system model, performance metric, propagation model and the power consumption models. Section III provides the simulation setup and the results. Section IV concludes the paper. II. SYSTEM MODEL The scenario considered in this paper is an outdoor wide area cellular network. The simulation area involves 21 hexagon cells with wrap around which simulates an infinite hexagonal grid of macro BS. The inter site distance (ISD) between macro BSs is varied from 400 m to 500 m which is standard in sub urban scenario. Each macro site covers 3 cells so there are 7 macro sites. Each macro site uses 3 directional antennas to provide coverage in 3 cells. The pico nodes are placed along with the macro BS within the cell area to offload the traffic from the macro BS and the number of pico nodes in each cell /12/$ IEEE 2197

2 is kept constant. The pico BS use omni directional antennas for transmission. The pico (BS) nodes are assumed to be correlated with hotspot locations and are placed in the hotspot center to achieve maximum offloading gain. The simulation involves both uniform and non uniform traffic distribution. The non uniform traffic distribution involves creating hot spots with a fixed number of users present in a relatively small area, very close to each other. As outdoor propagation models are used, the hotspots simulate a public square in a sub urban area. The ratio of user distribution between hotspot and uniform distribution is maintained at 34:66 for all ISD. The user equipment (UE) selects a BS based upon the gain matrix. The problem with this approach is that the gain matrix does not address the power difference between macro and pico BS. This leads to unsatisfied UE as the pico BS cannot provide satisfactory power levels to the high number of connected users due to its limited range. Therefore, range extension is utilized to overcome this issue. A. Propagation Model There are three main factors contributing towards deterioration in signal quality due to channel propagation namely path loss, slow fading (shadowing), and fast (multipath) fading. In this study we neglect the fast fading to simplify the simulation scenario. The 3GPP case 1 propagation models for heterogeneous, outdoor macro pico network, listed in [14], are modeled. The path loss model and simulation parameters are given in Table 1. The propagation models involved both line of sight and non line of sight propagation. The minimum distance between a macro BS and a pico node is 130 m and the minimum distance between two pico BS nodes is maintained at 90 m. B. POWER CONSUMPTION MODEL In this investigation, we have considered a distributed BS with the following linear power model [15]:- P dbs = Load + P backhaul (1) The power consumption models for distributed BS have three parts. The first static part reflects the static power consumption of the respective BS. The static power consumption involves the power consumed by the power amplifier, baseband unit, feeder network, and site cooling system. The second dynamic load dependent part is controlled by the number of Physical Resource Blocks (PRB) used by each BS. The load on the macro BS is defined as: PRBs utilized by macro BS users Load = (2) PRBs available to the macro BS The last static part is the back haul power consumption of the BS. The back haul power consumption is based upon a micro wave link with a power consumption of 125 watts and a rectifier loss of 17% [15]. P backhaul = 125.Loss rectifier (3) Similarly, the pico BS power consumption in Active mode [15] is defined as P pico =14.9+Load.(2.1) + P pico backhaul (4) Similarly, pico BS power consumption consists of a static part, a load dependent part and the back haul power consumption. The back haul power involves a fast ethernet point to point (PtP) optical network terminal (ONT) and the fiber PtP WAN interface [16]. A rectifier loss of 17% is also considered for the backhaul power consumption. C. PERFORMANCE METRIC The performance metric for this investigation is physical layer user throughput based upon the Shannon formula. Thus, for a user x, the achievable data throughput is evaluated based upon Shannons law as:- Th channel (x) =B(x).log 2 (1 + SINR(x)) (5) Where B(x) and SINR(x) are the allocated bandwidth and signal to interference and noise ratio respectively. The outage in the network is maintained around 10% with a tolerance of 2%. The concept of APC is utilized to analyze the network energy consumption. APC is used to characterize the power consumption of a network independent of its size. It can be defined as the average power consumption per cell divided by the cells area [7]. APC can be expressed mathematically as:- P AP C = P Cell. (6) A cell where P Cell is the power consumption of the macro BS and A cell is the cell area. In HETNET scenario, the cell power is the sum of macro BS power as well as the Pico BS power. APC is measured in watts/km 2. The network performance in terms of capacity is analyzed with the concept of ASE. The ASE can be defined as the achievable rates in a network per unit bandwidth per unit Area [7]. Mathematically, it can be expressed as:- S = 1. S (X=x).dx (7) A cell Where A cell is the cell area, A is the total area of the system, S X=x is the spectral efficiency of user x in the area A. Thus, ASE is the average of the individual user spectral efficiency over the system area A. It is measured in bits/sec/hz/km 2. The ASE only provides information about the mean achievable throughput rates. The individual user throughput rates may vary from the mean. III. SIMULATION SETUP AND RESULTS In this investigation, an orthogonal frequency division multiple access (ODFMA) cellular network with hexagonal deployment is considered. The simulations for a green field operator involve analysis of the system performance while varying the ISD in HETNET scenario. The variation in a HETNET scenario involves deploying different pico BS densities at varying 2198

3 Area Spectral efficiency (bits/sec/hz/km 2 ) ASE at 0 pico BS ASE at BS ASE at 2 pico BS ASE at 3 pico BS ASE at 4 pico BS 0.1 Inter Site Distance (ISD) Fig. 1. Area Spectral Efficiency as a function of Intersite Distance. ISD. The number of hotpots increases with increasing ISD. As the simulated area is sub urban, it is realistic that the number of hotspots doubles as the coverage area doubles. Thus, if the reference area is A for an ISD of 400 m with 1 hotspot, at 4 times area A 4 hotspots are simulated. The number of pico BSs is iteratively increased to analyze the effect of each pico BSs deployment on the network energy consumption. Also, as propagation losses increase with increasing ISD, the number of users in the reference (macro BS) only scenario is reduced to meet the QoS threshold. The ISD is varied from 400 m to 800 m which increase the area per macro cell by 4 times. Thus, at the maximum ISD of 800 m, the number of simulated hotspot is 4 per cell. The ratio of user distribution between hotspots and uniform distribution is kept at 34:66 for all ISD. Thus, at an ISD of 800 m, 34% users are evenly distributed over the 4 hotspots. The antenna tilt angle is also reduced with increasing ISD. The reduction in antenna tilt angle helps to increase the coverage area of the macro BS. The pico BS is deployed iteratively in the hotspots. With each pico deployed per cell, hotspot users are offloaded from the macro BS. As the users are offloaded from the macro BS, the macro BS can serve more users. Thus, the number of users in the cell is increased until the outage increased to 10% with a standard deviation of 2%. The ratio of users between uniform and hotspot distribution is maintained while increasing the users in the cell. A reuse one system is considered such that the same frequency resources are used in all cells. Due to spectrum sharing, interference issues arise. The effect of this interference is incorporated into the results by scaling the received SIR of every user. The SIR scaling is done by calculating the number of PRBs which are utilized within the system and then scaling the interference matrix by the ratio of the calculated PRBs and total available PRBs. This results in optimizing the interference matrix to the interference every user is actually receiving based upon the number of PRBs it utilized. All the simulations are based upon fair share distribution. Thus, all the resources available are equally distributed among all the users and the users are utilizing all the allocated resources. As noted previously, the user offloading by pico BS is accounted for by introducing new users in the system untill the outage reaches 10%. The increase in users is performed while maintaining the ratio of the number of users between uniform and hotspot. The increase in the number of users increases the ASE of the network. The results for APC, ASE, and Energy Efficiency at different pico deployments for multiple ISDs are based upon 500 simulated realizations. The high number of realizations helps to average out the effect of variation in hotspot positions within the cell area. The resolution of the data points is increased through cubic spline interpolation. The change in backhaul power due to variation in ISD is neglected. The variation in ASE with Inter site distance is shown in Figure 1.The curve shapes are consistent with the simulation methodology. As the number of hotspots increase with increasing area, the deployment curve for high order pico BS deployment starts at relatively higher ISD values. The ASE decreases monotonously with ISD and it increases with the increasing number of pico BS per cell. The decrease in ASE with increasing ISD is expected as the propagation losses increases with increasing ISD. The increase in ASE due to the first pico BS at lower ISD values is highest as compared to other high order pico BS deployment scenarios. There is an increase of approximately 36% in the ASE at BS deployment for an ISD of 400 m. The increase in ASE reduces to 26% at an ISD of 500 m. The increase in ASE with pico BS is a result of offloading macro users to the pico BS. When users are offloaded from the macro BS, more resources become available for the remaining macro users and thus ASE increases. In this study, the number of PRBs allocated to pico BS users are limited to those needed to reach the threshold QoS requirement. If the pico BS users were allocated all the available resources, the increase in ASE would have been more substantial. The high gain in ASE at low ISD values can be explained by the high number of users in a single hotspot at lower ISD values. As more users are offloaded to the pico BS, hence more resources are available for distribution among the remaining macro BS users. Also, as the ISD increases, the hotspot users start to spread over multiple hotspots as compared to a single hotspot at lower ISD values. Thus, fewer users are offloaded by each individual pico BS at higher ISD values, leading to a reduction in offloading gain per added pico BS. As the ISD increases in the case of 0 pico BS (our reference scenario), the propagation losses increase and it becomes difficult to satisfy the threshold QoS. Therefore, the number of users are reduced with increasing ISD for the reference scenario, while the ratio between uniform and hotspot users is kept constant. As more users are offloaded to the pico BS, more users are introduced into the system until the outage reaches 10%. At ISD values around 700 m, the difference in ASE between the reference scenario relative to 1 and 2 pico BS deployment is negligible. There is still some gain at 3 pico BS deployment, but it is very small. This decreasing trend at high ISD values can be explained by the distribution of users in increasing number of hotspots. Even at 4 pico BS deployment for an ISD of 2199

4 Area Power Consumption (Watts/km 2 ) APC at 0 pico BS APC at BS APC at 2 pico BS APC at 3 pico BS APC at 4 pico BS 500 Inter Site Distance (m) Area Power Consumption (W/km 2 ) pico 0 pico 0 pico 0 pico 2 pico ASE 0.15 ASE 0.2 ASE 0.3 ASE 0.4 ASE 0.5 ASE 0.6 ASE 0.7 ASE 0.8 ASE pico 2 pico 3 pico ,4 pico 0 pico 2 pico 0 pico 2 pico 500 Intersite Distance (m) Fig. 2. Area Power Consumption as a function of Inter site Distance. Fig. 4. Area Power Consumption as a function of Inter site distance for multiple ASE targets Area Power Consumption (W/km 2 ) Fig APC at 0 pico BS APC at BS APC at 2 pico BS APC at 3 pico BS APC at 4 pico BS Target Area Spectral Efficiency (bits/sec/hz/km 2 ) Area Power Consumption for multiple Target Area Spectral Efficiency 800 m, the ASE is around 0.16 bits/sec/hz/km 2.This trend emphasizes that user distribution in hotspots and the number of hotspots per cell strongly affect the gain in ASE. At higher ISDs, the number of hotspots increases but the number of users per hotspot decreases so the offloading gain is small. Figure 2 depicts the APC as a function of ISD at multiple Pico BS deployments. The trends for APC are roughly similar to that of ASE. Again, the curve shapes are consistent with the simulation methodology. The APC decreases monotonously with increasing ISD and it increases with the number of pico BS deployed. The decrease in APC can be explained by the increasing area at larger ISD. Also, as discussed earlier for higher ISD, the number of users in the system is reduced to maintain the QoS threshold for 90% of the users. This results in less resource utilization at the macro BS for the same macro power, leading to a decrease in APC. The increasing APC trend with the increasing number of pico BSs deployed is due to power consumption of pico BS. This power offset is eventually compensated for the increasing ASE due to PRB offloading from the macro BS. In this study, we proposed an optimal pico BS deployment strategy for the green field operator which results in minimum APC for a given ASE target. For target ASE targets varying from 0.1 bits/sec/hz/km2 to 0.9 bits/sec/hz/km 2.Figure3isusedto determine the ISD that can be reached for the respective ASE target with multiple pico BS deployment scenario. The APC at the respective set of ISDs is then evaluated using the APC curve in Figure 2. Figure 3 depicts the APC as a function of the target ASE for multiple pico BS deployments. The trend shows that the APC increases with the increasing target ASE. It can also be observed that each pico BS deployment strategy can meet the target ASE to a certain degree. Each deployment has its upper and lower limits for achivable ASE, i.e. the macro only scenario can achieve ASE taragets from 0.15 to 0.7 bits/sec/hz/km 2. Another observation is that the APC reduce with greater pico BS deplopyment for a target ASE higher than 0.25 bits/sec/hz/km 2. The reverse trend is observed below the target ASE of 0.25 bits/sec/hz/km 2. This can be explained by relatively low power consumption in macro only scenario for low ASE targets. Another interesting trend is the relatively low target ASE achieved by greater pico BS deployment. As discussed earlier in the explanation of Figure 1,this low achievable ASE at greater pico BS deployment can be explained by the deployment of greater number of pico BSs at higher ISD where pico BS is not very effective in increasing the ASE due to high propagation losses and the distribution of users over increased number of hotspots. Thus, the 3 pico BS deployment can only achieve a maximum target ASE of 0.25 bits/sec/hz/km 2. Figure 4 depicts the APC as a function of ISD for multiple ASE targets. The respective pico BS deployment at each ISD is also marked. The trend indicates that as the target ASE increases, the respective curve shifts left and upward. The upward shift indicates that higher ASE leads to higher APC. The left shift in the curve shows that only lower ISDs can support higher ASE targets, thus the network becomes dense with increasing ASE targets. The left shift also indicates that the higher ASE can be achieved when user density in hotspots is high. The curves are also marked with the respective pico BS deployment for a target ASE. Thus, each target ASE can be achieved by multiple pico BS deployments at different ISDs. The target ASE up to 0.3 bits/sec/hz/km 2 can be achieved with different pico BS deployments with similar ASE. At a target ASE higher than 0.3 bits/sec/hz/km 2,the 2200

5 Energy Efficiency (kbits/joule) bits/km bits/km bits/km bits/km bits/km bits/km bits/km 2 0 Pico BS 1 Pico BS 2 Pico BS 3 Pico BS 4 Pico BS 0.15 bits/km 2 Energy per bit(mjoule/bit) Pico BS 1 Pico BS 2 Pico BS 3 Pico BS 4 Pico BS 0.25 bits/km bits/km bits/km 2 bits/km bits/km bits/km bits/km Intersite Distance (meter) Fig. 5. Energy Efficiency as a function of Inter site distance for Multiple Pico BS deployment Intersite Distance (meter) Fig. 6. Energy per bit as a function of Inter site distance for multiple Pico Deployments per cell APC reduces for pico BS deployment which are also deployed at relatively larger ISD. This suggest that it is more feasible to utilize pico BS deployment for higher ASE targets, as this strategy results in a more sparse network and eventually reduction in OPEX and CAPEX for the green field operator. The limiting factor for higher ASE target is the respective upper limit of the supporting ISD. For instance, the target ASE of 0.4 bits/sec/hz/km 2 can only be supported till an inter site distance of 570 m. Based upon the results discussed so far, we can now evaluate the energy efficiency achieved through pico BS deployment at different ISD. The energy efficiency of the network can be expressed as: Achievable Area Spectral Efficiency EE = BW. (8) Area Power Consumption The achievable ASE in (8)is the ASE that each Pico BS deployment can achieve near a target ASE. The bandwidth for the investigated LTE downlink scenario is 20 MHz. Figure 5 depicts the energy efficiency of the network as a function of ISD for multiple pico BS deployments. The respective achievable ASE with a spacing of 0.5 bits/km 2 is also shown in Figure 5.The trend shows that the energy efficiency along with the achievable ASE improves with the introduction of pico BS at lower ISD. For instance, the energy efficiency at BS per cell relative to a macro only scenario is improved by 26% at an ISD of 400 m, along with an improvement in ASE by 35%. At relatively higher ISD, similar gains can be achieved with greater pico BS deployment. At an ISD of 550 m, the energy efficiency gain with BS is reduced to around 9% and at 2 Pico BS, the energy efficiency gain is around 18%. The reduction in energy efficiency gain at 2 pico BS deployment can be explained by the distribution of users in the increased number of hotspots. Thus, a single pico BS is not able to offload the traffic required to achieve a substantial energy gain. The energy efficiency at ISD higher than 670 m starts to show a negative gain at greater pico BS deployment. This can be explained by the similar ASEs but different APCs for 0, 1, and 2 pico BS deployments at larger ISDs. This trend is quite evident for 2 pico BS deployment as its energy efficiency starts to fall below that of a 0 pico BS deployment at an ISD greater than 700 m. As at larger ISDs, the number of hotspots increases which leads to a decrease in the number of users per hotspot. Thus, each pico BS is not able to offload as many users which leads to a decrease in ASE and eventually a decrease in energy efficiency as well. If the energy efficiency gain is compared for the same ASE with greater pico BS deployment then this shows positive gains in energy efficiency as well as ISD gains. For instance, at a target ASE of 0.45 bits/km 2, the energy efficiency improved by 12.42% for 2 Pico BS deployment as compared to 0 Pico BS deployment. As discussed earlier, the ASE with greater pico BS deployment starts to equal the macro only scenario for ISDs greater than 670 m. Thus, there are no energy gains for 1 and 2 pico BS deployments at ISDs greater than 670 m. The 3 and 4 pico BS deployment still give a marginal gain at greater ISDs. The reason for energy gains at 3 and 4 pico BS deployment is that almost all the hotspot users are now offloaded which results in an increase in achievable ASE and eventually energy efficiency. The energy per bit for different pico BS deployments at multiple ISDs can be evaluated by taking the inverse of the energy efficiency already evaluated. Figure 6 depicts the energy per bit required to meet the target ASE for different ISDs. The trend depicts that at ISDs up to 670 m, the energy per bit decreases with greater pico BS deployment for the same ASE target. Also, there is a gain in terms of increased ISD for the same target ASE with greater pico BS deployment. Above an ISD of 670 m, the energy per bit of greater pico BS deployment start to surpass the energy per bit requirement of macro only scenario. The change in the energy per bit trend at larger ISD can be explained by the increased energy consumption with greater pico BS deployments accompanied by relatively lower off loading of resources from macro BS with the addition of each pico BS. These factors collectively increase the energy per bit requirements of the network at higher ISD. 2201

6 TABLE I LTE SYSTEM PARAMETERS Carrier F requency 2.0 GHz Bandwidth 20 MHz FFT size 2048 Number of Subcarriers 1200 Subcarrier spacing 15 khz Noise per subcarrier 127 dbm T hermal noise 174 dbm/hz Macro antenna gain 15 dbi P ico antenna gain 5 dbi Macro T X power 40 W Macro antenna height 32 m P ico T X power 250 mw Macro sectors 3 P ico sectors 1 IV. SUMMARY AND CONCLUSIONS In this paper, the energy efficiency gains of combined macro and pico cell wireless network architecture was investigated as a function of ISD. The results show that, for a given ISD, both the Energy Efficiency and ASE of the heterogeneous network improves with the deployment of pico cells. For instance, at an ISD of 400 m, the introduction of BS per cell improves the energy efficiency by 26% along with improving ASE by 35%. With regard to ISD for a given ASE target, pico cell deployment provides gains in terms of reduction in APC as well as an increase in ISD. Also, as the ASE target increases, the achievable deployment scenario shifts towards lower ISD. The user distribution in the pico cells as well as the number of pico cells per macro cell also strongly impacts the energy efficiency. The results are based on snap shot analysis at different ISDs. More meaningful results can be achieved by using a dynamic system simulator which considers scheduling of macro BS users along with user mobility. Also, an investigation of the impact of advanced antenna techniques such as MIMO and Advance Interference management schemes such as CRS-IC and ICIC on the energy efficiency of the LTE wireless network would be an interesting research direction. [3] G. Koutitas and P. Demestichas, A Review of Energy Efficiency in Telecommunication Networks, Telfor, vol. 2, no. 1, pp , Nov [4] Avneesh Agrawal, Trends in Wireless Communications. available at keynote.pdf. [5] F. R. A. J. Fehske and G. P. Fettweis, Energy efficiency improvements through micro sites in cellular mobile radio networks, in Proceedings of the 2nd Workshop of Green Communications in conjuntion with GLOBECOM 2009, pp [6] G. P. Fettweis and E. Zimmermann, ICT energy consumption trends and challenges, in Proceedings of the 11th International Symposium on Wireless Personal Multimedia Communications, Lapland, Finland, September 2008., pp [7] A. J. F. F. Richter and G. P. Fettweis, Energy efficiency aspects of base station deployment strategies in cellular networks, in Proceedings of the 70th Vehicular Technology Conference (VTC Fall) 2009, pp [8] F. Richter and G. P. Fettweis, Cellular mobile network densification utilizing micro base stations, in Proceedings of IEEE International Conference on Communications (ICC) 2010, pp [9] P. M. F. Richter, A. J. Fehske and G. P. Fettweis, Traffic demand and energy efficiency in heterogeneous cellular mobile radio networks, in Proceedings of the 71th Vehicular Technology Conference (VTC Spring) 2010, pp [10] M. G. O. B. F. Richter, G. P. Fettweis, Micro Base Stations in Load constrained cellular mobile radion networks, in Proceedings of the IEEE 21st International Symposium on personal, indoor and mobile radio communication s, Sept 2010, pp [11] G. F. O. B. Oliver Arnold, Fred Richter, Power Consumption Modeling of Different Base Station Types in Heterogeneous Cellular Networks, in Proceedings of the Future Network and Mobile Summit, June 2010, pp [12] Thomas Bohn, Dieter Ferling, Patrick Jschke, Anton Ambros, Most Promising Tracks of Green Radio Technologies, INFSO-ICT EARTH, Deliverable 3.1, available at pub/bscw.cgi/d29584/earth WP4 D4.1.pdf. [13] L. H. H.Claussen and F. Pivit, Effects of Joint Macrocell and Residential Picocell Deployment on The Network Energy Efficiency, in Proceedings of IEEE 19th International Symposium on Personal, Indoor and Mobile Radio Communications 2008, pp [14] 3GPP, 3GPP v9.0, available at html-info/36814.htm. [15] Huawei, Huawei contributions towards Mobile VCE, 2010, available at [16] ITU-T SG15, Code of Conduct on Energy Consumption of Broadband Equipment, available at T09-SG TD-GEN-0212/en. V. ACKNOWLEDGEMENT The work presented in this paper is the result of a joint collaborative project between the Royal Institute of Technology (KTH) and Huawei Technologies Sweden AB. The authors would like to express their gratitude to Huawei Technologies Sweden AB for funding this research and to Mr. Christer Qvarfordt for his valuable suggestions towards the simulation methodology. REFERENCES [1] W. Tuttlebee, S. Fletcher, D. Lister, T. OFarrell, and J. Thompson, Saving the planet The Rationale, realities and research of Green Radio, The Journal of the Institute of Telecommunications Professionals, vol.4, no. 3, pp. 8 20, Sep [2] Susan Teltscher, Vanessa Gray,d Desire van Welsum, Philippa Biggs, World Telecommunication /ICT Development Report 2010: Monitoring the WSIS Targets a mid-term review. available at unesco.org/communication/documents/wtdr2010 e.pdf. 2202