Analyzing and evaluating the energy efficiency based on multi-5G small cells with a mm-waves in the next generation cellular networks

ABSTRACT


INTRODUCTION
The mobile subscriptions grew around 4% between Q1 2016 and Q1 2017, to reach 7.6 billion in Q1 2017; while mobile data traffic grew around 70% in the same period, to reach around 9.7 ExaBytes (EB) per month in Q1 2017; the reason behind this increase is a continuous development in cellular networks and its ability to meet mobile subscriber needs and provide high data speed rates. Accordingly, it is expected that the mobile subscriptions are growing to be 9 billion by 2022, with mobile data traffic approximately up to 71 EB/month, where that 75% will be for mobile video traffic [1]. These challenges have been a crucial step towards the fifth generation of cellular networks, which it is expected to be deployed by 2020 [2].
5G will have to fulfil many requirements, and a critical one is to deliver high network energy performance [3]. High energy performance requires a fundamental change of design principles and implementation practices within the mobile telecom industry. Thus, 5G technology adopts a set of new technologies; a detailed survey on 5G networks has been presented in [4]. The intensive deployment of 5G small cells with mm-waves along with large MIMO will play a significant role in developing 5G cellular network, to overcome the explosive increase in mobile traffic and improve EE of the network, as well as  In the EE mathematical model given in (1), the path loss, fading, shadowing, and the interference effects from adjacent cells are considered. However, for ease of understanding, this EE model is simplified and explained in the following paragraphs. In a communication system, the EE, measured in bits per joule, and expressed mathematically [10], where R is the total data delivered to the UE, and BS tot P is the total power consumed. BS tot P can be calculated by the following formula [16], where DC  is the DC-DC regulator losses, MS  is the main power supply losses, and cool  is the air-conditioner losses. However, by introducing RRHs, feeder losses are avoided by mounting the power amplifier (PA) close to the transmit antenna and using an optical fiber cable. In addition, cooling (air-conditioner) losses are omitted in small BS [16].
is the output power per transmit antenna, RF P is the radio frequency power, and BB P is the baseband power. Since that the C-RAN architecture is considered in this study, PBB is omitted, because a baseband unit is installed in the central office. PA  is the PA power efficiency, and N TRX is the total transceiver antennas in the site.

Data rate
As known of Shannon theory, the data rate is directly proportional with the antennas and bandwidth, and logarithmically proportional with SINR [10], where BW is the bandwidth; and SINR is the signal to interference plus noise ratio, and can be calculated by the following formula, where P I is the interference power of the neighbour BS 2 and BS 3 (which are denoted as P I2 and P I3 , based on Figure 1, N o is the noise in wireless channels; and P rx is the received power at UE

5G propagation model
The basic propagation model of the received power (P rx ) as stated below [17], where G1 is the antenna gain at BS1, and PL_mod_SUI is the path loss model for 5G small cell. The transmission power of BS1 (P tx_BS1 ), depends on two main factors: (ii) the required coverage area, and (ii) the surrounding environment (propagation fading). To simplify the model derivation, the macro BS transmission power is normalized as P o = 40 W with the coverage radius R o = 1 km. Similarly, the P tx_BS1 with coverage radius Rc1 is denoted by [18], The mm-wave spectrum is very sensitive to losses due to reflection and diffraction, which greatly depend on the material and the surface. Moreover, reflection and diffraction reduce the range of mm-waves and decrease the ability of a mm-wave signal to penetrate buildings (e.g., brick and concrete) and most solid materials (the attenuations for different materials are given in [19,20]. In this study, the modified Stanford University Interim (SUI) path loss model (P L_mod_SUI ) for a frequency of 28 GHz for Non-Line of Sight (NLOS) is proposed, as given in (8). This mathematical model (the modified SUI path loss model) is constructed based on extensive empirical measurements [8], where is the mean slope correction factor obtained directly from the NLOS empirical results, ( 1 ) is an original SUI model at a distance 1 , ( ) is an original SUI model at a reference distance do, P L (do) is the free space path loss in dB at do, and Xσ is a typical lognormal random shadowing variable with 0 dB mean and standard deviation σ. The original SUI model P L_SUI (Rc1) for a frequency above 2 GHz is [8], where,  (13) where λ, Xfc, and XRX denote the carrier wavelength, the correction factors for frequency, and receiver heights, respectively. f MHz is the carrier frequency. h TX and h RX is the transmitter and receiver antenna heights, respectively. The parameters a, b, and c are constants used to model the terrain types encountered in the service area. The SUI terrain type A is considered, with parameters given as a = 4.6, b = 0.0075, and c = 12.6 [8].
The interference power of the adjacent cells that affect the received signal at the UE can be expressed mathematically as, The transmission power of the BS2 and BS3, and the modified-SUI NLOS path loss model are expressed similarly to BS1, as given in the following equations, where ∆ represents the distance between the UE and the edge of the cell1 within cell1. For example, if the UE is being determined at the edge of cell1, ∆ will equal zero

SIMULATION SETUP
The simulation layout is given in Figure 1. The modified SUI model at 28 GHz is considered in this study; additional details about the simulation parameters that are used in the present study are given in Table 1. Note that the simulation parameters for both cell 2 and cell 3 are similar to cell 1 .

RESULTS AND DISCUSSION
The received power level is commonly used in wireless communication as a measure of the quality of wireless connections because each radio receiver can only detect and decode signals with strengths greater than the minimum receiver sensitivity power. This section first discusses the effects of the path loss, fading and shadowing on the received power level at the UE within cell 1 , as well as the interference effects from adjacent cells (cell 2 and cell 3 ) on the received power level at the UE over different cell radii. Because the EE is a function of both the data rate and the total BS power consumption, the second part of the discussion focuses on the data rate that will be delivered to the UEs under the effects of propagation path loss, multi-path (small-scale) fading, shadow (large-scale) fading and the interference power of the adjacent cells. The final part of this section evaluates the EE based on the proposed multi-LTE macro BSs system model. The received signal power decreases rapidly as the transmit-receive distance (radius of cell) increases ( ( ) = − . This power decrease is because the path loss, fading and shadowing within the cell, as well as the interference of adjacent cells and noise, are increasing. Figure 2 shows the effect of both the path loss, fading and shadowing within cell 1 and of the interference of adjacent cells (cell 2 and cell 3 ) on the received power level at the UE versus cell radii.
It is clear that when the radius increases, the received power level of BS 1 decreases and the interference power of BS 2 and BS 3 increases, reaching a maximum at the edge of cell 1 . This can be expressed as the SINR given in (5) and defined as the ratio between the received power level of BS 1 and the interference power of BS 2 and BS 3 plus noise. Figure 3 shows the SINR over different radii. Where that the mm-wave spectrum (28 GHz) is very sensitive to the losses, SINR at edge of cell1 drops to -10.6 dB.  (4), the data rate is directly proportional with the antennas and bandwidth, and logarithmically proportional with SINR. Hence, the key 5G considerations are focused on the physical layer (PHY), intensive deployment of 5G small cells with mm-waves along with large MIMO: -5G Small cells, radio coverage of 5G small cell BSs is 200 m. However, a simulation of network parameters for different cell radii is performed. As shown in Figure 3, when the distance between the BS and UEs has increased, SINR is degraded, which considered as the metric for quality of wireless connections. Thus, the worst received wireless signal of 5G small cell BS is -10.6dB at the edge of the cell1; which will reflect on the data rate as shown in (4) as well as on the amount of data transferred in the joule, as shown in (2); due to using a low order of modulation technique. -Millimeter wave (mm-wave): a three BW values have investigated to cover the most potential standards, which are 0.25, 0.5, and 1 GHz. Based on (4), Figure 4 shows the relationship between the data rate and the SINR values for different bandwidths and for 64 transmit antennas.
The following analysis discusses the data rate at the edge of cell 1 over different of the BW values. As shown in Figure 5, the data rate up to 1.93, 3.86, and 7.72 Gbps with BW 0.25, 0.5, and 1 GHz, respectively. Which obtained based on (4). It is clear that at the largest BW the data rate amounted four times at the lowest BW. These results mean that the 5G small cell can support a new high-speed data application, including multimedia communications, online gaming, and high-quality video streaming. Anyway, since that the EE is a function of the data rate and BS power consumption. The EE performance versus the SINR for different BWs is shown in Figure 5, based on the multi-5G small cells network system proposed in Figure 3. The total power consumption in this type of system grows proportionally with transmission power of the base station P tx_BS as is shown in (3). As shown in Figure 5, the EE achieved at a 0.25 GHz BW and 64 antennas at the edge of the cell 1 , where the SINR is -10.6 dB, can be up to 5.18 Mbits /J. This result computed using (2), a data rate of 1.93 Gbps taken from Figure 4, and by dividing by the total power consumption of 372.86 W (according to as shown in (3) 64× [(4.12+0.7) / (1-0.08) (1-0.1)]). However, with a 1 GHz BW and 64 antennas, the EE can be as high as 20.70 Mbits/J, resulting in data rates of up to 7.72 Gbps, as shown in Figure 5. -Large MIMO, is another method that can increase the capacity and EE [21][22][23]. Using Large MIMO technology, providing high multiplexing as well as array gain at the same time. Large MIMO technology is not only spectrum efficient but energy efficient as well. As the number of BS antennas increases, the received SINR increases linearly [24][25][26]. Figure 6 summarizes the data rate that can be achieved with various numbers of antennas for different BWs at the edge of cell 1 .
At the edge of cell 1 (radius tual 200 m) where the SINR is the lowest (-10.6 dB), the total data rate with a 0.25 GHz BW and 32 antennas can be up to 0.96 Gbps. With a 1 GHz BW the total data rate can be up to 3.86 Gbps with the same number of antennas. However, the multiple signal paths due to multiple antennas at the transmitter are responsible for the large throughput. It is clear that when using 64 antennas, the data rate (1.93 Gbps with 0.25 GHz BW, and 7.72 Gbps with 1 GHz BW, respectively) is 2 times greater compared with using 32 antennas.
The data rate increases with an increasing number of antennas, and the EE is a function of the data rate (as shown by (2)). Therefore, the EE improves with a larger BW size and a larger number of antennas.

CONCLUSION
This paper presented a model and an investigation of the EE of multi-5G small cell system cellular networks. In addition, the concepts of new technologies of the large MIMO in the millimeter range communication, are considered. Since that the EE is a function of both the data rate and the total BS power consumption. Therefore, the EE improves with a larger BW size, a larger number of antennas, and decrease the total BS power consumption. The simulation results show that the large BW and the number of antennas combined with a high SINR increases the throughput, which increases the number of bits transferred per joule of energy and thus improves the EE of the cellular network.