1. Introduction
A very relevant use case in wireless communication systems presently is the Internet of Things (IoT) [
1]. Moreover, with the fifth-generation (5G) technology, the market is expanding towards massive IoT deployments connected to complex smart sensor networks, which require low-cost devices with low power consumption [
2]. Since IoT devices typically consist of battery-powered nodes, energy efficiency (EE) is a critical issue, so that long-term operation without battery replacement is feasible. This demand makes EE a key design goal for 5G IoT applications [
3].
Multiple antenna (MIMO) systems have been used to mitigate the effects of fading, which allows improving the link reliability so that wireless nodes transmit with reduced power. However, each antenna should be connected to a radio frequency (RF) chain and, therefore, multiple antennas may increase the power consumption at the circuit level. In that sense, antenna selection (AS) is an important technique [
4], yielding the same diversity gains as MIMO, but with lower power consumption due to the reduced number of active RF chains. For instance, the EE optimization of some MIMO techniques, when considering the effect of reconfigurable RF transceivers, is discussed in [
5]. Along with AS, singular value decomposition (SVD) beamforming and single antenna (SISO) are considered. Results show that EE can be improved considerably with the AS technique for short and moderate distances. On the other hand, the spatial diversity gains provided by SVD are important to extend the range. In addition, the combination of antenna selection and low-resolution phase shifters for device-to-device communications is investigated by [
6]. The proposed antenna selection algorithm is shown to increase the energy efficiency by suppressing interference from other devices in the network, as well as improving the received signal power.
Another important venue to improve the EE is to reduce the power consumption at the circuit level, e.g., through battery management techniques [
7] or receiver architecture redesigns [
8]. CMOS technology downscaling imposes huge challenges for RF designers, which need to meet severe system requirements while keeping power-consumption as low as possible. Seeking for low-power front-end receivers compliant with the required specifications of 5G-based IoT scenarios, both receiver architecture and circuits have to be optimized. To support a huge number of applications in different communication standards, usually multiple RF front-ends are designed to operate in a single band/standard each, thus potentially consuming considerable power overall [
9]. Hence, a low-power wideband RF front-end receiver is essential for multi-standard communications.
Figure 1 illustrates two different wideband receiver architectures. The conventional receiver chain in
Figure 1a has a band-pass filter (BPF), followed by a low-noise amplifier (LNA) and an active mixer. In wideband applications, BPF are generally surface acoustic wave (SAW)-filters organized in banks, which are switched on/off depending on the band or channel required. These SAW-filters make the receiver bulky and costly. Several strategies on the LNA designs can be found in the literature for wideband receivers, depending on the metric to be optimized. To save die area, wideband inductorless LNAs with tunable active shunt-feedback architecture have been proposed in [
10]. Also employing an active shunt-feedback, but focusing on reducing noise figure, a wideband noise-canceling CMOS LNA with enhanced linearity was proposed by [
11]. In addition, aiming at reducing the power-consumption and improving the linearity, the circuit designed in [
12] is a dual-path noise and nonlinearity canceling LNA. Furthermore, instead of designing a single wideband LNA, a multi-standard beamforming RF front-end is presented in [
13], which consists of four independent receiving paths. This circuit covers a wideband operation, yielding better noise figure and linearity at the expense of increasing the overall power-consumption.
Common to the previously mentioned low-power wideband LNAs is their focus on providing low noise figure at the expense of reduced gain and linearity. Alternatively,
Figure 1b presents the low-noise transconductance amplifier (LNTA) architecture. The multiple off-chip bulky SAW-filters are replaced by on-chip reconfigurable filters, denoted as N-path filters [
14]. Then, the combination of the high output impedance of LNTA with the low impedance of the N-path circuitry enables significant attenuation of frequencies outside the desired band, thus providing high RF selectivity on-chip. Therefore, the association of the LNTA with the multiple reject-band filters (RBF) in parallel allow the reception and amplification of the desired signal in wideband, through multiple smaller bands, allowing reduced noise figure at the front-end. In addition, to reduce power-consumption, a passive mixer is placed right-after the LNTA. The position of the RBF could change around the LNTA, depending on the desired behavior. LNTAs have recently drawn considerable interest due to their greater linearity performance [
15].
For instance, inductorless designs implementing noise and distortion cancellation techniques are presented in [
16,
17]. Moreover, a high-linearity RF receiver architecture is shown in [
14], adopting Miller band-pass filters for channel selection. To reduce the front-end receiver power consumption, in [
18] the LNTA is followed by a current mode passive mixer that provides sufficient linearity to permit coexistence with large out-of-band interference arising from other transceivers. Nevertheless, despite the enhanced linearity of LNTAs, these circuits face severe requirements for impedance matching networks and bandpass filters. As a result, occupied area enlarges significantly. In addition, to improve receiver sensitivity, LNTAs are designed to provide a sufficient gain to the weak incoming RF signals, increasing the power consumption [
19]. In summary, the above mentioned LNTA-based designs aim to improve linearity and reduce the noise figure, at the expense of a significant increase in the power consumption.
Given all the challenges imposed by energy-constrained applications, in this paper we present an energy efficiency analysis that merges aspects of front-end receiver architectures with the spatial diversity improvements of using multiple antennas. Then, we take into account the power consumption of the required circuit blocks, for both LNA and LNTA architectures, in two MIMO scenarios: one based on AS scheme, in which only one pair of antennas remains active in each transmission; and other based on SVD beamforming, which uses all transmitting and receiving antennas to increase robustness against channel fading. Moreover, as a reference, we also consider the SISO case. Results show that low power consumption is not enough to guarantee the best performance in terms of energy efficiency. There is a tradeoff between link distance, noise figure and power consumption that can be optimized through the proper combination of receiver architectures and spatial diversity techniques.
The remainder of this paper is organized as follows.
Section 2 details the RF receiver designs and the communication model. The considered transmission schemes are mathematically detailed in
Section 3. Then, some numerical results are discussed in
Section 4, while
Section 5 concludes the paper.
4. Results and Discussions
In this section we provide a few numerical results with the two considered RF front-end architectures. The system parameters are listed in
Table 2. In addition, we consider nodes from a wireless sensor network with the same number of antennas to transmit and receive,
. The transmission system is evaluated using the closed-form expressions for the outage probabilities. In other words, given the target outage probability
we find the required transmission power for each scheme, using (
10), (
12) and (
14), respectively for the SISO, AS and SVD schemes. In addition, these expressions also depend on the noise figure of the receiver, so that each RF front end architecture impacts the required minimum transmission power. With that in hand, the energy efficiency of each scheme is computed using (
16).
Table 3 shows the energy efficiency of the LNA designs, selected from [
10,
11,
12], as a function of the transmission distance. Then, for each LNA we consider SISO, AS and SVD schemes, with the architecture that yields the best performance highlighted in orange, cyan and green shadings, respectively. Please note that the front-end designs with the lowest power consumption are not necessarily the ones with the best energy efficiency. The association of low noise figure and low power consumption clearly plays a very important role in designing an energy efficient system. Then, in
Table 3 we observe that the LNA in [
12] yields the highest energy efficiency for SISO, AS and SVD. Next,
Table 4 considers the LNTA designs selected from [
14,
17,
18]. As we observe, the LNTA in [
14] yields the highest energy efficiency for SISO when
m, being outperformed by the LNTA in [
17] when
m. Using the AS scheme, the LNTA in [
18] performs best when
m, the LNTA in [
14] performs best when
m and the LNTA in [
17] performs best when
m. Therefore, the blue shading shows the highest energy efficiency across architectures. Finally, the SVD scheme follows the same idea, so that the best architecture starts with the LNTA in [
18] when
m, shifting to the LNTA in [
14] when
m and to the LNTA in [
17] when
m. As a consequence, in the sequel we pick the LNA/LNTA architectures with the best performance among their counterparts. According to
Table 3, the LNA in [
12] is the best choice, while the LNTA in [
17] outperforms the other LNTAs most of the time according to
Table 4.
Figure 4a plots the energy efficiency as a function of the distance between transmitter and receiver considering the LNA in [
12] and the LNTA in [
17], for the SISO, AS and SVD schemes with
antennas. As we can observe, the LNTA architecture at the receiver usually achieves higher energy efficiency, except for short transmission ranges with the MIMO schemes. In this example, the LNA outperforms the LNTA in terms of EE with AS when
m, while this distance increases up to
m with the SVD scheme. This indicates that the very low power consumption of the LNA design in [
12] plays an important role to maximize the EE in short transmission distances, while the lower noise figure of the LNTA from [
17] becomes more important when the distance increases, since it allows alleviating the transmitted power of the PA, at the transmitter side. Complementing the analysis, the importance of the spatial diversity brought by the multiple antennas becomes more evident in
Figure 4b, which increases the number of antennas to
. As we observe, similar conclusions can be drawn for the AS scheme, where the LNA becomes more energy efficiency for up to
m. However, when submitted to the SVD scheme, the LNA presents better performance than the LNTA. The performance difference decreases when the transmission distance increases.
Next,
Figure 5 shows the energy efficiency of the AS scheme as a function of the number of antennas, considering
and
m, for the LNA and LNTA architectures. As we observe, when
m the LNA outperforms the LNTA in terms of energy efficiency, while this conclusion inverts for
m. As previously mentioned, the noise figure of the selected LNTA is lower than that of the LNA counterpart, which becomes more important when the distance increases, regardless of the number of antennas. In addition, we also observe that the energy efficiency increases with the number of antennas up to a saturation level.
For the SVD scheme, on the other hand, there is an optimal number of antennas to maximize the energy efficiency, which depends on the transmission distance, as depicted by
Figure 6 for
m. Similarly as for the AS scheme, the LNA outperforms the LNTA for short transmission distances. When
d increases, the LNTA performs better for a reduced number of antennas (less than 4). Improving the spacial diversity allows the LNA to surpass the LNTA, even presenting a worst noise figure. For the SVD scheme, the number of antennas determines the number of receivers. Hence, power consumption becomes a key metric in the energy efficiency performance.
Finally,
Figure 7 plots the energy efficiency of the AS and SVD schemes as a function of
, comparing the LNTAs from [
17] and [
18], for
m. As we observe, the LNTA designed in [
17] is always more energy efficient with the AS scheme, in which a single pair of RF chains is active for communication. However, there is a trade-off between [
17] and [
18] for the SVD scheme when the number of antennas increases. With a few antennas (less than 3), the energy efficiency using [
17] is higher, while it is outperformed by [
18] with more antennas. By taking the parameters of
Table 1 into account, we observe that the LNTA in [
17] has very low noise figure and high power consumption, with
dB and
mW. On the other hand, the design in [
18] has very low power consumption at the expense of a higher noise figure, with
dB and
mW in this case. Therefore, we observe that low noise figure is more important when the number of active antennas is low (e.g., with AS and SVD with a few antennas), while low power consumption becomes crucial when the number of active RF chains at the receiver increases.
5. Conclusions
Energy efficiency behavior modeling has been presented in this paper. The proposed model sought to place different front-end receivers in relation to different communication schemes. More precisely, two different receiver architectures composed by LNA and LNTA, respectively, were employed with SISO, AS, and SVD schemes, for different ranges.Nakagami-m fading distribution is used to characterize the wireless channel, in order to be better compliant with IoT scenarios, since the parameter m can be used to model scenarios where nodes have line-of-sight conditions. After an analysis of the energy efficiency between receivers based on the state of the art, the best candidate for each front-end architecture was selected. A comparative performance study of the different communication schemes was carried out, showing which distances and conditions stand out for the selected front-end designs. The results show that, for short-range scenarios, LNA presents increased EE performance, particularly due to its very low power consumption. On the other hand, when the communication distance increases the very low noise figure provided by N-path LNTA-based architectures outperforms the low power consumption of the LNA-based designs, yielding higher EE for SISO and AS transmission schemes. For SVD, the LNA always presents the better performance in terms of EE, exposing that IoT applications are strongly dependent on energy consumption of the RF circuits. Finally, our analysis also shows that the selected front-end architecture depends on the number of active antennas at the receiver, so that low noise figure is more important with a few active antennas at the receiver, while low power consumption becomes more important when the number of active RF chains at the receiver increases.