Uplink Energy Efficiency in LTE Systems Xavier Ponsa,c , Raul Palaciosb , Christophe Grueta , Lirida Navinerc , Hugo Marquesd , Fabrizio Granellib and Jonathan Rodriguezd a
EADS-Cassidian, France; b University of Trento, Italy; c Institut Mines - Télécom, Télécom - ParisTech, CNRS, LTCI, France; d Instituto de Telecomunicações, Portugal
[email protected]
Abstract—This paper proposes a new method to model the instantaneous uplink (UL) energy efficiency (EE) of a mobile terminal when it is transmitting to a base station using the Long Term Evolution (LTE) technology. It is known that the transmitted power can vary significantly, mainly depending on the used modulation code scheme (MCS), or the path loss between the user equipment (UE) and the eNodeB (eNB), or even the number of resource blocks (RBs) used to send the data. Even though this transmitted power, also known as irradiated power, does not correspond to the final power consumption during an UL transmission, its variability has an impact on it. Unlike existing models, this paper considers the power consumed in the radio frequency chain and the power consumed by the baseband processing. The proposed model is expected to improve the accuracy of the theoretic measurements of both the power consumption and the EE of UL transmissions in LTE systems. Index Terms—Long Term Evolution (LTE), uplink, energy efficiency, power consumption model.
I. I NTRODUCTION With the evolution of mobile radio communication networks the data rate has grown from few kilobits per second in the Global System for Mobile Communications (GSM) to few hundreds of megabits per second in the fourth generation (4G) Long Term Evolution (LTE). At the same time, due to the new functionalities empowered by current terminal processing capabilities, the power consumption of mobile terminals has also increased. In GSM, mobile terminals could operate several days before running out of battery whereas nowadays, using the third generation (3G) technology, terminals have to be recharged every 1-2 days, even considering that current batteries have considerable higher capacity than before. In LTE, this trend is expected to be maintained or even increased [1]. In this paper, we propose a new accurate method to calculate the user equipment (UE) power consumption, and also the energy efficiency (EE) when a mobile terminal is transmitting data to an eNodeB (eNB) using the LTE technology. The EE has been introduced as an effective performance metric to evaluate and compare the energy consumption of different systems and networks. When compared with the power consumption, the EE provides the relative value of the useful bits of information that can be transmitted with one Joule (bit/Joule). On the other hand, the power consumed is an absolute magnitude and provides the number of Joules consumed during a second without taking into account the information transmitted.
To deal with EE is very important to know the UE power consumption as well as the throughput that a mobile terminal can achieve. Once the transmitted power is known the consumed power can be obtained. Most of the UE power models developed since now considered constant power transmission when a mobile terminal is sending data, either using an average power transmission or the lowest power transmission value. In order to model the UE power consumption three different aspects are taken into account: (i) the modulation and the code rate used to send data (modulation code scheme - MCS); (ii) the path loss and the channel conditions between the UE and the eNB; and (iii) the number of resource blocks (RBs) per transmission time interval (TTI) - bandwidth - used to transmit all the data. In addition, we calculate the average throughput that a mobile terminal can achieve depending on the number of RBs available in UL and the MCS used. These results allow us to calculate a more accurate power consumption to obtain the uplink (UL) EE in an LTE system. Next sections describe in more detail this concept, the structure is as follows: in Section II we present the related work and the motivations of this work; Section III provides the description and decomposition of the UE power consumption model explaining the concepts of the transmitted power, the radio frequency (RF) chain consumed power and baseband (BB) processing consumed power; Section IV presents our achieved results for the energy efficiency of a UE when it is transmitting and finally, Section V concludes the paper. II. R ELATED W ORK In the last years the power consumption has been one of the most researched topics in mobile radio networks. During this period, several techniques aiming to reduce the power consumption of the eNB, such as switching off Base Stations (BS) [2], cell zooming [3], and other approaches appeared. This extensive research in the BS side is due to the high power consumption of a BS, which can reach up to the 55% of the total power consumed in a mobile radio communication network [4]. In this work, we aim to target the UE power consumption and its EE only during the UL transmission periods. We basically focus in UL transmission since it is the most energy consuming activity of a mobile terminal even if it takes short time. It is widely known that an increased amount of power is required to assure that the transmitted data reaches the eNB.
Most of the previous UE power consumption models do not consider important parameters such as the modulation used to transmit, the path loss and the channel conditions between the eNB and the UE, the number of RBs used per millisecond (ms), or the bandwidth used for data transmission. In [5], the authors present the impact of the discontinuous reception (DRX) and Discontinuous Transmission (DTX) techniques in terms of power saving by using the UE power consumption model proposed in [6]. This model assumes constant power consumption during a TTI of 3W. This constant power consumption value is also used in [7] to develop an EE Vertical Handover algorithm. In [8] the authors propose an LTE power model that includes all the LTE radio resource control (RRC) states, which is later tested and compared with real measurements. In this paper, the power consumed during all the transmission procedure can be clearly seen, since the UE is in RRC Idle state before transmitting until it comes back to RRC Idle state. During the transmission procedure, the instantaneous power level varies significantly due to different data rates, which accounts for our research motivations. Another LTE UE power consumption model is presented in [9]. The authors suggest that the global power consumed in the UE during the transmission procedure should be splited in two stages. The first part of the power consumption refers to the BB processing, which depends on the transmission rate. The second part refers to the power consumed in the RF chain. In this part the power consumption depends on the efficiency of the Power Amplifiers (PA) and it is proportional to the power transmitted by the antenna. In this paper we clearly see that the power consumed in the UE is not constant and can vary according to different factors that we will consider in this paper. Another fundamental paper for our work to obtain the UE power transmitted and the UE throughput is [10]. In this paper, the authors analyze the capacity of the LTE downlink (DL) compared to the Shannon capacity bound. Since the Shannon bound cannot be achieved, the authors propose a modified Shannon bound formula for LTE: This formula introduces two parameters: the bandwidth efficiency and the signal-to-noise ratio (SNR) efficiency. These parameters allow considering all the restrictions that reduce the LTE bandwidth efficiency, such as the adjacent leakage ratio (ACLR), the cyclic prefix (CP), and the Reference Signals (RS). This Shannon bound formula has also been used in [11] and [12], to calculate the spectral efficiency (SE), the throughput, and the coverage for different scenarios. III. UE P OWER C ONSUMPTION M ODEL This section details the UE power consumption model at the physical layer. As described in [9], the power consumption in a UE occurs in the RF chain and in the BB processing. In our case, we go a step further, and we present a model to determine the exact amount of transmitted power at every TTI depending on different aspects like the modulation used, the distance between the BS and the eNB, and the number of RBs used. Once the transmitted power is known, and following the
work presented in [9], we obtain the UE power consumed in the physical layer during UL transmission with better accuracy. A. Transmit Power The transmit power is the power measured in the antenna port when the UE is transmitting data to an eNB. At the reception site, the SNR should be sufficient to allow the eNB to properly receive and decode the signal. We define this value of the SNR as the target SNR, which will depend on the MCS used for the UE to transmit the signal. In addition, the power transmitted should partially or completely compensate the path loss that the signal experiences between the UE and the eNB [13]. Another important aspect to be considered is the thermal noise. This noise is frequency dependent and can vary according to the number of RBs used per TTI. Therefore, it can affect the SNR computed at the receiver side, hence making impossible the signal decoding. 1) Spectral Efficiency Based on [10], the average SE for all the available MCSs is calculated in order to know the target SNR associated with each MCS. To that end, the number of useful resource elements (REs) that are inside all of the RBs belonging to the Physical Uplink Shared Channel (PUSCH) need to be known. This value depends on the system bandwidth, the CP used and the number of antennas used to transmit. To count all the REs available for useful data, all the RBs used for the Physical Uplink Control Channel (PUCCH) and the RBs used for the Physical Random Access Channel (PRACH) are discarded. These REs are only used for control data. In addition the REs that are not dedicated to transmit useful data have to be eliminated too. These REs are used to send Reference Signals (RS). The SE value will depend on the MCS used to transmit data. Depending on the MCS, the number of useful bits that can fit in a RE can be very different. In LTE there are 3 different types of modulation schemes with several code rates, which are combined in a total of 15 available MCS in LTE UL [14]. Each of these 15 MCS has an associated SNR target. Once this choice is made, the number of useful bits transmitted is established. Thus, dividing the number of total useful bits with the total bandwidth the average SE is obtained. 2) Shannon Bound Based on the Shannon bounds and once the SE is known, the minimum SNR needed at the reception side, hereafter target SNR, can be calculated. In [10], the modified Shannon formula (1) is presented: SN Rtarget bps = BWef f · η · log2 1 + (1) SE Hz SN Ref f Where BWef f is the bandwidth efficiency of the used technology (i.e., LTE), SN Ref f is the efficiency of the SNR required in LTE, and η is a correction factor. Once all the parameters of the modified Shannon bound are computed, the target SNR can be calculated knowing the SE previously calculated. 3) Path Loss: Channel Model
In order to obtain the transmitted power in the UE, the path loss has to be considered to determine the target SNR at the receiver. In LTE the UE power transmission is regulated by the UL Power Control mechanism [13], [15] with which the path loss can be fully or partially compensated. We suggest performing a complete compensation of the path loss, since we consider that this mechanism, known as open loop power control, is the most power consuming algorithm when compared with partial path loss compensation. This makes the results presented in this paper to correspond to the lowest EE that a UE can achieve in the LTE UL transmission. To obtain the path loss a channel model is selected. There are several channel models for different scenarios, and even some projects try to classify some of them depending on the scenario [16]. In this UE power consumption model the macro cell propagation model in rural environment, which follows the Okumura-Hata model [17], is used. This model provides a path loss value depending on the distance between the UE and the eNB. This is a very simple and well-known channel model, which is realistic enough to give us acceptable results. This Okumura-Hata channel model has been considered in the standardization for rural areas [18] and in other relevant references works such as [12] and [19]. 4) Noise To achieve the SNR target in the receiver side, the noise has to be considered in order to compute the transmitted power. The received noise depends on the receiver noise figure (N FRX ) as well as the bandwidth. The noise is frequency dependence due to the thermal noise (N0 = 174dBm/Hz). Therefore, depending on the number of RBs used for transmitting in one TTI, the frequency band can increase, increasing the received noise. The expression for the noise behaviour at the receiver is given by: N [dB]
= N0 + N FRx + +10 · log10 (#RB · 180KHz/RB)
(2)
Once the target SNR, the path loss, and the noise at the receiver are obtained, it is possible to calculate the transmitted power that is needed in the UE to reach the eNB with sufficient signal strength for successful data reception. In order to obtain the transmitted power, we compute all the gains and losses (3): PT x [dB]
= SN Rtarget − GAT x + LCableT x − GARx + +LCableRx + P athLoss + N oise
(3)
Where GAT x and GARx are the antenna gains in the UE and eNB, respectively and LCableT x and LCableRx are the cable losses at the UE and eNB, respectively. All in all, the transmitted power basically depends on the MCS used, the distance between the eNB and UE - which allows us to determine the path loss - and the number of RBs used - frequency band. B. RF Power Consumption This subsection explains how the power consumption in the RF chain is obtained. As explained, the power consumption
in the RF chain depends on the transmitted power. Depending on the efficiency of the whole RF chain, where the PA efficiency has a major influence, the power consumed can be substantially different in different devices. In [9] the authors use the transmitted power to measure the power consumed by the RF chain. The power consumption is limited within -40dBm and 23dBm. According to our research, the RF chain power consumption behaviour is quite similar amongst all the devices [9]. Minor differences can be observed among the devices due to the Power Added Efficiency (PAE) which is defined for a specific PA. There are 3 different working zones. The first one occurs when power consumption is constant and the lowest possible, for a very small transmitted power -in our case between 40dBm and -30dBm. The second zone is when the PA is working in its optimum efficiency zone. Even if the power consumption increases linearly with the transmitted power, it is still under 2.4W where the third zone starts. The third zone is when the PA is operating inefficiently and the power consumption starts increasing considerably. There is a big gap between the second and the third working zone, where the RF chain power consumption increases from 2.4W to 3.8W for a power transmission bigger than 10dBm. As such the power consumed in the RF chain can be calculated depending on the following parameters: MCS, path loss and number of resource blocks used for transmission. C. BB Processing Power Consumption Another part of the global power consumption of a UE terminal when transmitting to an eNB is the power consumed during the BB processing. The power consumed during this step depends on the data rate. In [9] the authors report the BB processing power consumption for different UL data rates used to transmit in UL. The power consumed due to the BB processing is quite constant around 2.12W. Having calculated the power consumed in both the RF chain and BB processing, we can compute the total amount of power consumed instantaneously in the UE depending on the path loss between the UE and the eNB, the MCS used, and the number of RBs. IV. N UMERICAL R ESULTS In this section we present the results obtained by following the proposed UE power consumption model. We use in MATLAB to plot the UE consumed power and the UE EE as a function of the distance. We will focus on a 5MHz single input single output (SISO) system using a carrier frequency of 1.8GHz. First of all, the UE transmitted power is calculated by using the model explained in Section III. For that the SE has to be calculated. The SE will depend on the number of available REs and the number of useful bits per RE. Hence, we can say that the SE depends on the bandwidth and the MCS. In our 5MHz LTE system, every 1ms there are 8 RBs dedicated to the PUCCH and, every 10ms there are 12 RBs dedicated to PRACH. We consider a PRACH with a configuration of 0
TABLE I SE AND TARGET SNR FOR ALL THE MCS MCS SE[bps/Hz] SNR (dB)
1 0.11 -2.69
2 0.17 -0.52
3 0.27 2.06
4 0.43 4.95
5 0.63 7.70
6 0.84 10.25
7 1.05 12.59
preamble, which has a duration of 1ms and occupies 6 RBs of bandwidth [20]. We obtain 408 useful RBs in 10ms. Each RB has 12 different subcarriers, which contain different REs depending on the CP. Using a normal CP (4.7µs, [20]) there are 7 REs per slot (0.5ms). In a subframe (1ms) there are 14 REs, where two of them are used to send Demodulations RS (DM-RS) and another one is used to send Sounding RS (SRS). Therefore, in one subframe there are 11 REs dedicated to transmit data [20]. We obtain an average value of 2692.8 REs per ms. These REs can contain more or less bits depending on the MCS. In Table I the average SE of our system is shown for all the available MCSs. Based on [10], we performed some simulations to obtain the best modified Shannon parameters that better fit the link adaptation curve. In our case, we focus on the modified Shannon bound parameters for a SISO configuration using 5MHz bandwidth in the UL, as shown in Table II: We calculate these parameters considering different implementation issues that exist in UL. The BWef f was calculated as in [21], and according to the following restrictions: •
•
•
•
Random Access Channel (RACH): 488/500 At each radio frame (10ms), there is a subframe (1ms) reserved to request access to eNB using only 12 RBs. Normal Cyclic Prefix (CP): 14/15 In 1ms subframe there are 14 symbols of 66.67µs, instead of the theoretical 15 symbols that can fit in 1ms without CP. Reference Symbols (RS): 11/14 At each subframe (1ms), there are 3 RS: 1 SRS and 2 DM-RS. Therefore, using Normal CP, only 11 symbols are available for useful data out of 14, considering the RS. Physical Uplink Control Channel (PUCCH): 21/25 In a 5MHz system every subframe contains 8 RBs dedicated to PUCCH, while all the rest 42 RBs are dedicated to useful data, PUSCH.
With these considerations, we obtain BWef f = 0.6. The modified Shannon bound can be obtained, once the parameters are calculated, and then the taregt SNR can be calculated. In Table I, we present the target SNR values for each MCS available in LTE UL. In order to overcome the noise, it has to be calculated. The
TABLE II M ODIFIED S HANNON B OUND PARAMETERS Configuration SISO
(BWef f · η, SN Ref f ) (0.346, 2)
8 1.36 15.79
9 1.71 19.26
10 1.95 21.50
11 2.37 25.56
12 2.78 29.51
13 3.22 33.74
14 3.64 37.76
15 3.96 40.75
Fig. 1. LTE UL power consumption.
typical values that we use to calculate the noise with (2) are: N FRX = 4dB, GAT x = 0dB, GARx = 7dB), LCableT x = 0dB and LCableT x = 2dB. Once the UE transmitted power is calculated following (3), the UE power consumption considering the RF and BB parts can be calculated. Thus, we reproduce the curves from [9] in order to calculate the total amount of power consumed depending on the MCS and the number of RBs used. Fig. 1 illustrate the power consumed for all the MCSs versus the distance between the UE and the eNB and considering that only 1 RB was used for transmission. As it can be seen the MCSs availability depends on the distance due to the transmitted power limitation, 23dBm. At some distances more than 23dBm are required to properly transmit with a specific MCS, but because of the transmitted power limitation, this MCS cannot be used. In order to obtain the EE that a UE has in UL when transmitting the UE power consumption and the average throughput need to be calculated. The average throughput that a mobile terminal can achieve depends on the MCS and the number of RBs. In order to calculate it, the number of useful REs available in the system is calculated. We obtain an average number of 2692.8 REs per 1ms using the whole band. Bringing that number to a RB level, we obtain 64.11 average REs per RB, considering that there are 42 RBs in 1ms available for PUSCH in a 5MHz bandwidth system and a normal CP. In Fig. 2 we can see the throughput per RB in our system for all the available MCS. As expected, the throughput increases as the MCS increases. Finally, the EE can be calculated as the relation between the throughput and the consumed power. As it was expected a higher MCS achieves a higher EE. In Fig. 3 the EE is showed for all the MCSs depending on the distance. As closer the UE is from the eNB, the higher the EE
under different rules. The distribution can depend on several parameters, such as the service and the distance. Taking into account the scheduling policy we can better obtain the EE depending, for instance, on the service required for each mobile terminal or the distance between the UE and the eNB. R EFERENCES
Fig. 2. LTE UL throughput.
Fig. 3. LTE UL energy efficiency.
is. This is because the UE transmits with a very low power. Again we confirm that not all the MCSs are available at all possible distances. V. C ONCLUSIONS In this paper we have proposed an accurate method to calculate the EE of a UE when it is transmitting. Most of the existing works we found dealing with UE LTE EE, consider constant power consumption. This assumption does not reflect a real scenario. The new model addresses this issue and takes into consideration parameters such as the used MCS, the distance between the UE and the eNB and the number of RB used to obtain the UE transmitted power and the corresponding EE. We expect that this work will help the research community obtaining more accurate results on the evaluation of LTE systems that deal with UL transmissions and possibly trigger a healthy discussion on the topic. In future work, we will consider a scheduling policy. This technique will allow us to know the distribution of the RBs
[1] Nokia Siemens Network, “LTE Smartphone Measurements,” Sep. 2011. [2] A. Bousia, E. Kartsakli, L. Alonso, and C. Verikoukis, “Energy efficient base station maximization switch off scheme for lte-advanced,” in Computer Aided Modeling and Design of Communication Links and Networks (CAMAD), 2012, pp. 256–260. [3] Z. Niu, Y. Wu, J. Gong, and Z. Yang, “Cell zooming for cost-efficient green cellular networks,” Communications Magazine, IEEE, vol. 48, no. 11, pp. 74–79, 2010. [4] Z. Hasan, H. Boostanimehr, and V. Bhargava, “Green cellular networks: A survey, some research issues and challenges,” Communications Surveys Tutorials, IEEE, vol. 13, no. 4, pp. 524–540, 2011. [5] M. Polignano, D. Vinella, D. Laselva, J. Wigard, and T. Sørensen, “Power Savings and QoS Impact for VoIP Application with DRX/DTX Feature in LTE,” in IEEE Vehicular Technology Conference (VTC), May 2011, pp. 1–5. [6] Nokia, R2-071285, “DRX Parameters in LTE,” Mar. 2007. [7] ICT-248577 C2POWER consortium, “D6.1: Energy Efficient Vertical Handover algorithms. Intermediate Specification,” 2010. [8] J. Huang, F. Qian, A. Gerber, Z. M. Mao, S. Sen, and O. Spatscheck, “A Close Examination of Performance and Power Characteristics of 4G LTE Networks,” in ACM International Conference on Mobile Systems, Applications, and Services (MobiSys), 2012, pp. 225–238. [9] A. R. Jensen, M. Lauridsen, P. Mogensen, T. B. Sørensen, and P. Jensen, “LTE UE Power Consumption Model: For System Level Energy and Performance Optimization,” in IEEE Vehicular Technology Conference (VTC), Sep. 2012, pp. 1–5. [10] P. Mogensen, N. Wei, I. Z. Kovacs, F. Frederiksen, A. Pokhariyal, K. I. Pedersen, T. Kolding, K. Hugl, and M. Kuusela, “LTE Capacity Compared to the Shannon Bound,” in IEEE Vehicular Technology Conference (VTC), Apr. 2007, pp. 1234–1238. [11] A. B. Saleh, S. Redana, J. Hämäläinen, and B. Raaf, “On the coverage extension and capacity enhancement of inband relay deployments in lteadvanced networks,” Journal of Electrical and Computer Engineering (JECE), vol. 2010, pp. 1–10, Jan. 2010. [12] H. Holma and A. Toskala, LTE for UMTS - OFDMA and SC-FDMA Based Radio Access. John Wiley & Sons, Ltd, 2009. [13] A. Simonsson and A. Furuskar, “Uplink Power Control in LTE Overview and Performance, Subtitle: Principles and Benefits of Utilizing rather than Compensating for SINR Variations,” in IEEE Vehicular Technology Conference (VTC), Sep. 2008, pp. 1–5. [14] 3GPP TS 36.213 V11.2.0, Evolved Universal Terrestrial Radio Access (E-UTRA) - Physical Layer Procedures, 3rd Generation Partnership Project Std., Feb. 2012. [15] C. U. Castellanos, D. L. Villa, C. Rosa, K. I. Pedersen, F. D. Calabrese, P.-H. Michaelsen, and J. Michel, “Performance of Uplink Fractional Power Control in UTRAN LTE,” in IEEE Vehicular Technology Conference (VTC), May 2008, pp. 2517–2521. [16] IST-4-027756 WINNER II, “D1.1.2 V1.2 WINNER II Channel Models,” 2008. [17] M. Hatay, “Empirical formula for propagation loss in land mobile radio services,” IEEE Transactions on Vehicular Technology, vol. 29, no. 3, pp. 317–325, Aug. 1980. [18] 3GPP TR 36.942 V11.0.0, Evolved Universal Terrestrial Radio Access (E-UTRA) - Radio Frequency (RF) system scenarios, 3rd Generation Partnership Project Std., Feb. 2012. [19] A. Mishra, Advanced Cellular Network Planning and Optimization. John Wiley & Sons, Ltd, 2007. [20] S. Sesia, I. Toufik, and M. Baker, LTE - The UMTS Long Term Evolution: From Theory to Practice. Chichester, UK: John Wiley & Sons, Ltd, Feb. 2009. [21] L. Zhang, “Network Capacity, Coverage Estimation and Frequency Planning of 3GPP Long Term Evolution,” Master’s thesis, Linköping University, 2010.
