Next generation wireless networks aim at providing substantial improvements in spectral efficiency (SE) and energy efficiency (EE). Massive MIMO has been proved to be a viable technology to achieve these goals by spatially multiplexing several users using many base station (BS) antennas. A potential limitation of Massive MIMO in multicell systems is pilot contamination, which arises in the channel estimation process from the interference caused by reusing pilots in neighboring cells. A standard method to reduce pilot contamination, known as regular pilot (RP), is to adjust the length of pilot sequences while transmitting data and pilot symbols disjointly. An alternative method, called superimposed pilot (SP), sends a superposition of pilot and data symbols. This allows to use longer pilots which, in turn, reduces pilot contamination. We consider the uplink of a multicell Massive MIMO network using maximum ratio combining detection and compare RP and SP in terms of SE and EE. To this end, we derive rigorous closed-form achievable rates with SP under a practical random BS deployment. We prove that the reduction of pilot contamination with SP is outweighed by the additional coherent and non-coherent interference. Numerical results show that when both methods are optimized, RP achieves comparable SE and EE to SP in practical scenarios.A preliminary version [1] of this work will be presented at IEEE GLOBECOM 2017. rates. Then, in order to properly study the effect associated with intercell interference in a large practical network with an irregular BS deployment, we adopt the stochastic geometry framework developed in [8] wherein BSs are spatially distributed according to a homogeneous Poisson point process (PPP). Within this setting, we calculate closed-form lower bounds of the achievable rates averaged over the UEs' spatial distribution. This provides powerful insights into the interplay of system parameters without requiring the use of heavy numerical simulations. Such lower bounds are then used to compute the EE of the network with both RP and SP taking into account the power consumed by transmission and circuitry. Numerical results are used to show that, when both methods are optimized, RP provides comparable SE and EE to SP in practical scenarios.The remainder of this paper is organized as follows. Section II introduces the network model.In Section III, the channel estimation process with RP and SP is detailed whereas the achievable rates with MRC are computed in Section IV. Section V presents detailed analytical comparisons between RP and SP. In Section VI, the average achievable rates are first computed for a random network deployment (based on stochastic geometry) and then used for computing the EE.Section VII illustrates numerical results while Section VIII concludes our work.Notation: We denote vectors by lower-case bold-face letters (e.g., x) 1 and matrices by boldface capital letters (e.g., X). 2 The operators E{·} and E{·|y} represent expected value and expected value conditioned on a realization of th...

This work focuses on the hardware design for the efficient operation of Massive multiple-input multiple-output (MIMO) systems. A closed-form uplink achievable data rate expression is derived considering imperfect channel state information (CSI) and hardware impairments. We formulate an optimization problem to maximize the sum data rate subject to a constraint on the total power consumption. A general power consumption model accounting for the level of hardware impairments is utilized. The optimization variables are the number of base station (BS) antennas and the level of impairments per BS antenna. The resolution of the analog-to-digital converter (ADC) is a primary source of such impairments. The results show the trade-off between the number of BS antennas and the level of hardware impairments, which is important for practical hardware design. Moreover, the maximum power consumption can be tuned to achieve maximum energy efficiency (EE). Numerical results suggest that the optimal level of hardware impairments yields ADCs of 4 to 5 quantization bits.

This work aims to design the uplink (UL) of a cellular network for maximal energy efficiency (EE).Each base station (BS) is randomly deployed within a given area and is equipped with M antennas to serve K user equipments (UEs). A multislope (distance-dependent) path loss model is considered and linear processing is used, under the assumption that channel state information is acquired by using pilot sequences (reused across the network). Within this setting, a lower bound on the UL spectral efficiency and a realistic circuit power consumption model are used to evaluate the network EE. Numerical results are first used to compute the optimal BS density and pilot reuse factor for a Massive MIMO network with three different detection schemes, namely, maximum ratio combining, zero-forcing (ZF) and multicell minimum mean-squared error. The numerical analysis shows that the EE is a unimodal function of BS density and achieves its maximum for a relatively small density of BS, irrespective of the employed detection scheme. This is in contrast to the single-slope (distance-independent) path loss model, for which the EE is a monotonic non-decreasing function of BS density. Then, we concentrate on ZF and use stochastic geometry to compute a new lower bound on the spectral efficiency, which is then used to optimize, for a given BS density, the pilot reuse factor, number of BS antennas and UEs. Closedform expressions are computed from which valuable insights into the interplay between optimization variables, hardware characteristics, and propagation environment are obtained.

In Massive MIMO base stations (BSs), the hardware design needs to balance high spectral efficiency (SE) with low complexity. The level of hardware impairments (HWIs) indicates how strong the signal distortion introduced by hardware imperfections is. In particular, the analog-to-digital converters (ADCs) have an important impact on signal distortion and power consumption. This article addresses the fundamental problem of selecting the optimal hardware quality in the Massive MIMO space. In particular, we examine the optimal HWI and ADC bit allocation per BS antenna to maximize the SE. The results show that in co-located arrays with low channel gain variations across antennas, equal ADC bit allocation is optimal. In contrast, cellfree Massive MIMO systems benefit the most from optimizing the ADC bit allocation achieving improvements in the order of 2 [bit-per-channel-use] per user equipment when using regularized zero-forcing (RZF). In addition, when including the impact of power consumption in cell-free Massive MIMO with RZF, allocating low values of mixed ADC bit resolutions across the BS antennas can increase the energy efficiency up to 30% compared to equal ADC bit allocation.

No abstract