Energy Consumption In Wireless Networks Progress Report

Transcription

1 Energy Consumption In Wireless Networks Progress Report Niclas Jacobsson #1, Mickael Lopes #2, Shohidul Islam #2 #1 Department of Signals and Systems,Chalmers Chalmers University of Technology #2 Erasmus Student, Chalmers University of Technology {jacobssn, lopesm, Abstract Since the mobile phone relies on a limited power source and provides more demanding services today, minimizing the energy consumption is one of most important factors to consider. In this context, the focus will be set for two areas: modulation formats in the physical layer and the RRC-protocol in the network layer. The modulation formats section will be a literature study only, to show the energy advantages by using higher modulation combined with channel coding. Main focus will be set on the RRC-protocol and its functionality. The simulation chapter will show the importance of precise timer settings to gain energy efficiency as well as a comparison between the original RRC-protocol and a RRC-protocol proposed by us. Keywords energy consumption, mobile phone, RRC-protocol, RRC-timer, modulation formats account. Since the network layer is where information about network status and functions for establishing and reconfiguring a connection exists, this is where direct energy consumption regulation can be performed. The network layer is able to reconfigure the user, putting it in different power states, when no data transmissions occur the network layer may put the user in a state that consumes less energy. This is all regulated by the Radio-Resource- Control-protocol (RRC-protocol) and will be mainly investigated by simulations in this report. The physical layer affects power consumption in an indirect way. By utilizing the air interface in a manner that is as efficient as possible. With the use of proper modulation and coding, a lower cost in terms of power per transmitted bit can be achieved. DELIMITATIONS It will not be given any detailed information of the mobile CPU, which indeed has an mayor role due to the issue of energy consumption in mobile units. This component is considered to be an own area of study. The report will not discuss channel coding. I. INTRODUCTION Mobile communication has taken a big leap forward in its evolution during the last decade. A phone today is not only used for traditional tasks such as voice conversation and texting but now additional services have become standard such as; video conversations, file downloading, web browsing, streaming, Voice-over-IP and much more. All of these tasks proves to be very energy consuming in different ways and demands higher requirements on the phone, such as battery life and capability of operating at high data-rates. One can wonder how mobile phones can operate in such high bandwidths without draining all energy of the mobile power source. One of the most important aspects of wireless communication is efficient use of the air interface. This report will investigate implementations done in the physical layer and network layer to reach a high level of efficiency there are considerations at different levels that needs to be taken into II. MOBILE DESIGN AND ENERGY STATES The UMTS network is one of the most common used mobile communication technologies today. It consists of three subsystems: User Equipements (UE), UMTS Terrestrial Radio Access Network (UTRAN), and the Core Network (CN). In this context, there are two important factors which determine the energy consumption due to network activity. The first one is the transmission energy, it is proportional for length and transmit power level of the signal (physical layer). The second factor is the UMTS Radio Resources Control (RRC) protocol. Due to limited radio resources this protocol has been implemented in each UE to efficiently use it. The main task of RRC protocol is to introduce a state machine in the UE and to control it. There are four standards states used: Cell_DCH ( cell dedicated channel ) Cell_FACH (cell forward access channel) URA_PCH/Cell_PCH (URA/cell paging channel) IDLE The Cell_DCH state is used when the cell phone need to ensure a high throughput and low delay for transmission, it is typically used for high volume data or voice like real-time traffic such as streaming video. In this state, the mobile phone

2 has a dedicated channel which demands high power consumption. The Cell-FACH state shares the channel with other users and is typically used when there is low transmission activity (such as web browsing) where a delay is allowed and uses about half of the power compared to the DCH state. The URA_PCH/ Cell_PCH states are only used to listen to the channel but uplink data transfer is not allowed. This state costs one percent power of the DCH state. The IDLE state is typically used in the absence of any network activity, in this state the RRC connection is released. This state costs minimal energy. between two successive packet will be very short (<T1) but if the UE receives low volumes of data the time between two successive packet will increase, if the time between two successive packets becomes higher than T1, the RRC will put the UE in the Cell_FACH state. It s the same process for the transition between Cell_FACH and IDLE state with the inactivity timer T2, if the UE does not receive data after the time T2, it will change states. The transitions between IDLE and DCH state and the transition between IDLE and FACH state are triggered by user activity, the state that the UE will be transitioned into depends on the Buffer Occupancy (i.e. data load) level in the Radio Link Control (RLC) layer. RRC state Cell_PCH/URA_PCH Cell_FACH Cell_DCH Power Consumption(mA) < 5 ma ma ma Table I: Power Consumption in RRC States [3] In order to simplify the problem, it will be assumed that the PCH- and IDLE- states are the same, and we will denote them by the IDLE state. We can make this simplification because we are analyzing this process from an energy point of view. Even if the PCH cost energy, it is negligible compared to the additional states (We will assume a transition scheme as illustrated in figure Fig. 2). Fig. 2: Proposed inactivity timer not triggered[8] Fig. 3: Proposed inactivity timer triggered[8] Fig 1: The radio resource control state machine[7] The way this RRC-protocol works [3] is pretty simple, we need two inactivity timers T1 and T2. For instance, if the UE is in the Cell_DCH state and receives data, the end of each received packet will trigger an inactivity timer (T1), if the UE receives high volumes of data (e.g. streaming video), the time The challenge is to configure the inactivity timers T1 and T2. [1] A Short time length will cause frequent reconnections that is a issue that needs to be avoided. Even if the IDLE state does not cost a lot of energy, the transition from IDLE to DCH causes non-negligible power consumption and frequent reconnections would also cause a great deal of unnecessary latency for the end-user. On the other hand, prolonging the length of the inactivity timer will keep the UE in a state it does not need and will decrease the efficiency of radio resources and consumes more energy. To conclude, the RRC protocol permits the UE to increase its efficiency and to economize its power consumption, but this protocol is controlled by two important parameters, the inactivity timers T1 and T2. The configuration of these parameters are very important because an inappropriate time setting for these timers will decrease the UE efficiency and cause unnecessary energy consumption. II. HIGH ORDER OF MODULATION This chapter will try investigate if there is energy benefits by using higher order of modulation.

3 A. EVOLUTION OF 3G To achieve transmission of digital data by air the data bits needs to be processed into symbols and then be carried by a sinusoidal wave carrier, this is done by using a so called modulation. A basic explanation of modulation is to map a certain amount of bits into symbols to increase the transmission rate and gain spectral efficiency [4]. Because the mobile operators are limited to operate in a very restricted spectrum, the spectral efficiency becomes exceptionally important factor to constantly improve. There are different modulation formats used depending on service and country. The most common formats used for mobile telephony are the Quadrature Amplitude Modulation (QAM) and Phase Shift Keying (PSK) formats. QPSK was the first modulation format used in first 3G release Wide Band Code Division Multiple Access (W-CDMA) and carries 2 bits per symbol and enables a theoretical downlink speeds up to 2 Mbit/s. Later on, a updated release of WCDMA, named High-Speed Downlink Packet Access (HSDPA) was released with the modulation 16QAM. This modulation carries 4 bits per symbol and enables a increased downlink speed up to 7,2 Mbit/s. The lastest release of HSPDA that use 64QAM-modulation, which progress carries 8-bit per symbol and supports a theoretical rate up to 84 Mbit/s [5]. High order of modulations (HOM) has contributed huge improvement regarding the data-rates of the WCDMA technology and this chapter will discuss the basics of HOM. A. HIGHER ORDER OF MODULATION In theory with perfect conditions the transmission rates could reach almost unlimited values with higher order of modulation (HOM), though in reality this is much more complicated because of unwanted interferences caused by the surroundings which are commonly referred to as noise. The noise in wireless channels is caused by adjacent signals transmitted in the channel and this is one of the most significant factors that will be taken in to account here. As seen in figure [5] below, the constellation points is significantly more narrow in higher order of modulations (HOM) [10]. Fig. 5: Constellation and decision regions of QPSK, 16- and 64 -QAM [10] Narrow decision regions in a noisy environment will make symbols harder for the receiver to decode correctly and this generates higher probability of having bit errors (BER). A common term of measurement digital communications is the Signal to Noise Ratio per bit (SNR) or Eb/N0 (where Eb is energy per bit and N0 is the total noise density [6]) This is used to measure the signal power compared to the noise in the channel. By study figure [6] below it can be seen that for a fixed BER the corresponding SNR increases for HOM which also proving that HOM are significantly more sensitive to noise [11]. Figure 6: The figure shows error probability (BER) of uncoded bits and Eb/N0 (SNR per bit) in a Additive White Gaussian Noise channel.[11] In conclusion, higher order of modulation needs more energy per bit to achieve the same BER as a lower order of modulation, therefore it will need other implementations, such as channel codes, to achieve energy efficiency. Channel codes will not be investigated here because of time constraints. III. SIMULATION OF THE RRC-PROTOCOL A. RRC-TIMER SIMULATION The purpose of this simulation is not only to show that the two standard timers defined in 3G standard phones have been carefully chosen to reduce the power consumption but it is also to show that our proposed RRC-protocol can be even more efficient during web browsing utilization. The simulation results will confirm that the energy consumption is at lowest at T1 = 5 seconds and T2 = 12 seconds. All programs below have been implemented in C. The first step of this simulation is to implement a program which will return an array of random data request in order to simulate 10 minutes of web browsing utilization (due to the random aspect of the program, we had to generate long time- user utilization to have coherent result ). The returning array will give two parameters, the time when the data request is asked (T) and

4 the size of the data requested (M), this two parameters follow a Gaussian probability distribution (GM). The Gaussian probability distribution, with Mean μ and Standard Deviation σ, is a Gaussian function[9] of the form: 1 2 ² ² Where P(x)*dx gives the probability that a variable Y with a Gaussian Distribution takes on a value in the range [x,x+dx].the notation is Y ~ N (μ,σ²) which means that the variable Y follows Gaussian Distribution with the mean μ and the standard deviation σ. Fig 6 : Standard deviation diagram of the Gaussian distribution [9] Thus, T ~ N(μ 1,σ 1 ²) and M ~ N(μ 2,σ 2 ²) with μ 1 = 15 s and σ 1 =5 s (this parameters have been chosen in an arbitrary way ) and μ 2 = 100 kb and σ 2 = 50 kb (this values have been computed by sampling 20 daily used websites ). The algorithm of the program is as following: DataRequest() A : array Int T = 0 Int M = 0 Int i = 0 While T < 600 do T = T + GD(μ1,σ1) M = GD(μ2,σ2) A[i]=T A[i+1]=M i=i+2 Return A Fig 7 : DataRequest Algorithm The second step of this simulation is to compute the RRC protocol [8] according to its description given before. This program takes in argument the array returned by the DataRequest program, and it computes the energy consumption. The table II below lists the different parameters of the simulation. Promotion time IDLE -> DCH 2 s FACH -> DCH 1.5 s DCH -> FACH 0.5 s FACH -> IDLE 0.5 s State Radio Power DCH/FACH/IDLE 800/400/0 mw Promotion Radio Power IDLE -> DCH 1.1 J FACH -> DCH 1.05 J DCH -> FACH 0.3 J FACH -> IDLE 0.3 J Data Rate DCH 32 kbps FACH 10 kbps Transfer Energy Power consumption to R(x) =0.025*x+2.4 J download x kb of data Table II : Parameters of the RRC function[7][2] The transition from FACH state to DCH state is controlled by the Buffer Occupancy level in the RLC : when the buffer is full, the RRC move to the DCH state. Due to the complexity of this condition, we had chosen to trigger this transition if and only if a data request is received before the end of the last download, this is to keep the program as realistic as possible. As the programs are implemented, the simulation can begin. In the first simulation, T2 is set to 12 s and T1 varies, and in the second simulation, T2 varies while T1 is set to 5 s. The result is displayed in figure 9a below and shows the overall average of 500 runs for each timer values (due to the random aspect). Figure 8a: Result of the first simulation: T2 = 12 s and T1 varies

5 In the proposed RRC-protocol, the transition from FACH- to DCH-state is triggered when the Buffer Occupancy is full (like the standard RRC protocol) and if the user wants to download a data of size larger then M. When the download is completed the RRC moves back to the FACH state. A simulation with computed power consumption of the proposed idea has been done. In the figure 10 below, a comparison is shown between the original RRC-protocol versus the proposed RRC-protocol during web browsing (when T2 = 12 s). Figure 8b: Result of the second simulation: T1 = 5 s and T2 varies In the first graph we can see shorter T1 is, the better. So the question is why don t we choose a shorter T1? It s because the choice of T1 is a compromise between energy consumption and delay. The second graph shows that a shorter T2 is more power consuming. This fact can be explain by the simulation condition. Indeed the FACH state is the more appropriate state for a web browsing utilization, if we short T2, the DCH state will be more used than the FACH state and so the power consumption will increase. The simulation has proved the energy consumption drops as T2 increases and then stabilizes for T2 > 10s. The simulation has proved the standard timers values are coherent with the most restricting constraints of mobile application, which is power consumption. A. PROPOSAL RRC-PROTOCOL The purpose of this part is to propose a better RRC protocol with a main objective to limit the DCH state. To achieve this goal, the RRC protocol below is proposed. Figure 10 : Energy consumption of standard RRC protocol(blue) and proposal RRC (green) for M = 200 kb. The result shows decrease of 2.4 % for the proposal RRC and therefore seems to be more effective during web browsing utilization. Furthermore, the effect of the proposed RRC has to be confirmed in other utilization aspects as well as its impact on the transmission delay. We will not go deeper in to this elaboration to test these variables because of time constraints. Fig 9: Proposed RRC-protocol

6 REFERENCES [1] C. Lee, J. Yef, J. Chen Impact of Inactivity Timer on Energy Consumption in WCDMA and cdma2000 [2] N. Balasubramanian, A. Balasubramanian, A. Benkataramani Energy Consumption in Mobile Phones: A Measurement Study and Implications for Network Applications [3] P. Perälä, A. Barbuzzi, G. Boggia, K. Pentikousis Theory and Practice of RRC State Transistions in UMTS Networks [4] Agilent Techologies. (2001) Agilent, Digital Modulation in Communication systems An introduction. [Online]. Available: [5] E. Dahlman, S. Parkvall 3G Evolution HSP And LTE For Mobile Broadband 2 nd Edition [6] Dr. Robert A. Nelson, Modulation, Power And Bandwidth [Online]. Available: h.pdf [7] F. Qian, Z. Wang, A. Gerber, Z. Morley Mao, S. Sen, O. Spatscheck Characterizing Radio Resource Allocation for 3G networks [8] T. Yamamoto, S. Saruwatari, H. Morikawa Energy-Efficient Upload Engine for Participatory Sensing [9] Wikipedia normal distribution [Online] Available: [10] Andrea Goldsmith, Wireless Communications, Cambridge University pess, [11] (2002) The Agilent Technologies website. [Online]. Available: 60&aid=61&cid=W-CDMA/HSDPA/HSUPA

7 Review question: group 12 Can you name and describe the three different states of the RRC protocol? The three standard states are Cell_DCH state, the Cell_FACH state and the Cell_IDLE state. The Cell_DCH state is used when the cell phone need to ensure a high throughput and low delay for transmission, it is typically used for high volume data or voice like real-time traffic such as streaming video. In this state, the mobile phone has a dedicated channel which demands high power consumption. Your data The Cell-FACH state shares the channel with other users and is typically used when there is low transmission activity (such as web browsing) where a delay is allowed and uses about 50% of the power compared to the DCH state. Your data Data for other users The IDLE state is typically used in the absence of any network activity,in this state the RRC connection is released. This state costs minimal energy (1% of the DCH state).