RESEARCH POSTER PRESENTATION DESIGN © 2015 www.PosterPresentations.com Mobile data traffic volume keeps growing. Available radio spectrum is limited. Coordinated MIMO (Co-MIMO) increases the system-level spectral efficiency (SE). Performance can be further increased by more antennas per cooperating node (e.g. RRH). However, cost and power constraints limit the number of elements on conventional arrays. (E.g. each antenna is typically fed by its own radio frequency (RF) module.) Load-controlled parasitic antenna arrays (LC-PAA) offer an attractive alternative! Introduction Coordinated MIMO and Cloud RAN / RRHs Transmission Protocol Learning phase: For each beam combination, each BS transmits a pilot signal. Then, the UTs estimate the direct and cross channels and report to their BS the channel estimates. Beam selection phase: Based on this information, the BSs select jointly the optimum beam combination, in terms of the achieved sum-rate (SR) throughput. Precoding / transmission phase: The BSs transmit precoded symbols over the selected beams. Degrees of Freedom of LC-PAAs We consider single-fed LC-PAAs with M elements (i.e., 1 active and M 1 passive antennas), were each parasitic element terminates to a load with purely imaginary impedance. The degrees-of-freedom (DoF) provided by this antenna system are given by: DoF = M 1 2 + 1 Sum-Rate Capacity Bounds of LC-PAAs Low-SNR: C DoF ×SNR×log 2 e. High-SNR: C K ×log 2 DoF ×SNR . A THENS INFORMATION TECHNOLOGY (AIT) K ONSTANTINOS NTOUGIAS, DIMITRIOS NTAIKOS, CONSTANTINOS B. P APADIAS R OBUST MULTI-CELL PRECODING OVER L OAD-C ONTROLLED BEAMS WITH SINGLE-FED ANTENNA ARRAYS Using load-controlled antennas for Cooperative MIMO LC-PAAs use less RF chains than antennas. One or a few active elements are surrounded by closely-spaced passive elements. Parasitic elements are terminated via tunable loads. By adjusting load impedances, we control the mutual coupling between the antennas. Thus, we can shape and steer radiation beam patterns as desired. Have been shown to achieve beamforming (BF), transmit diversity (Tx Div), open-loop / closed-loop MIMO (OL-/CL-MIMO), and multi-user (MU) precoding via load control. However, attaining arbitrary solutions for the loads is not always possible. For instance, the spatial multiplexing (SM) solutions depend on the input signal. Also, the desired solutions may require loads that result in system instability. Finally, channel-based multi-user precoding with LC-PAAs is virtually unexplored. Our divide-and-conquer approach: “First form the beams, than precode over them. Yet load circuit / tuning limitations discourages the shaping of arbitrary / optimal beams. Beam switching: “Switch between different sets of fixed loads (predetermined beams). Figure 1: A beam generated by a single-RF printed Yagi array with 7 parasitic elements. System Setup We assume a single cooperation cluster comprising K cells (RRHs) with one active user each. Each RRH is equipped with a single-fed LC-PAA of 1 active and M−1 passive elements. Each user utilizes a terminal with a single, omni-directional antenna. At each timeslot, each RRH can generate 1 out of L predetermined beams. K beams selected out of L possible combinations in total. CSI is shared between the BSs and single-user decoding is used at the user terminals (UT). Figure 2: System setup for K = 2, L = 4 and M = 5. Coordinated MIMO relies on the cooperation between neighboring nodes. The cooperating base stations (BS) exchange control information and channel state information (CSI) and /or user data to control the inter-cell interference (ICI). The various flavors of Co-MIMO differ in terms of their performance vs. backhaul capacity and delay requirements tradeoff. The cloud radio access network (Cloud-RAN) architecture facilitates the application of Co- MIMO. Centralized pools of virtual base-band units (BBU) are connected to remote radio heads (RRH) that are located at the cell sites through optical transport systems. Beam Selection Criterion BS computes for each beam combination l =1,…, L the channel vector h ∈ℂ ×1 which is comprised of the direct channel UT -BS , h  , and the cross channels UT -BS , h  , k,m = 1, , K, m k. Then, the BSs exchange these channel vectors and form the composite channel matrix H × whose columns are h . Next, the performance metric T = tr H H −1 is calculated. Finally, the beam combination l that results in the minimum value of T is selected. System and Signal Models System model: y = Hx + n = HWP 12 s + n Multi-cell zero-forcing (ZF) precoding: Low complexity and good performance at high signal-to-noise-ratio (SNR): h w 2 =0⇒ F = H + = H HH −1 , W = F :,k F :,k . Power Allocation SR throughput maximization problem: max >0 R = =1 R = =1 log 2 1 + SINR = =1 log 2 1 + h w 2 p s.t. p P (per-BS power constraints) Water-filling solution: p = v 1 h + Water level Noise power Effective channel after precoding Performance Evaluation Figure 3: Numerical simulation results for various beamwidths over a channel with 5 scatterers. Conclusion Precoding over switched LC-PAAs is a highly promising approach for Co-MIMO networks with small access points (such as RRHs) that need to employ more antennas than RF chains. [1] D. Ntaikos, B. Gizas, C. B. Papadias, L. Roullet, F. Taburet, “Over -the-air demonstration for RRH-based LTE access with the use of parasitic antenna arrays: Results from the FP7 project HARP ,” Eur. Conf. on Netw. and Comm. (EuCNC), Paris, France, June 29 July 2, 2015. [2] K. Ntougias, D. Ntaikos, C. B. Papadias, “Robust Low-Complexity Arbitrary User- and Symbol-Level Multi-Cell Precoding with Single-Fed Load-Controlled Parasitic Antenna Arrays,” 23 rd IEEE Int. Conf. on Telecomm. (ICT), Thessaloniki, Greece, May 16 18, 2016. [3] K. Ntougias, D. Ntaikos, C. B. Papadias, “Coordinated MIMO with Single-Fed Load-Controlled Parasitic Antenna Arrays”, 17 th IEEE Int. Workshop on Sig. Proc. Adv. in Wireless Comm. (SPAWC), Edinburgh, UK, July 3 6, 2016. {kontou, dint, cpap}@ait.gr Coordinated MIMO and Cloud RAN / RRHs Objectives In previous works, we have studied various multi-cell user- and symbol-level precoding schemes and low-feedback techniques and compared their performance with the one accomplished in equivalent setups that utilize conventional antennas (see [1-3]). Here we are interested in the analysis of the approach, especially in asymptotic regimes. References