Servicenavigation

Project Details

Back

Projekt Print View

Massive MIMO Systems for Communication and Localization

Subject Area Electronic Semiconductors, Components and Circuits, Integrated Systems, Sensor Technology, Theoretical Electrical Engineering
Term from 2014 to 2019
Project identifier Deutsche Forschungsgemeinschaft (DFG) - Project number 259038323
 
Final Report Year 2018

Final Report Abstract

In this Massive MIMO project, innovative communication and localization techniques have been researched both in terms of theoretical approaches and practical realizations and verifications while particularly considering the mitigation of hardware impairments. In order to address this aspect, a fully-functional massive MIMO testbed, suitable for efficient and online self-calibration, has been designed and built in order to demonstrate the approaches. A Min-Max-MSE HL-THP scheme was proposed for DL massive MIMO systems with imperfect CSI. The proposed scheme comprised an inner linear BF that is designed based on statistical CSI and outer nonlinear THP modules, which exploit the instantaneous overall CSI of the cascade of the actual channel and the inner BF. Thereby, UTs are divided into groups, where in each group, a THP module successively cancels the intra-group interference, whereas the inter-group interference is suppressed by the inner BFs. Since the per-group channel matrices have a much smaller size than the actual channel matrix, the proposed Min-Max-MSE HL-THP has a substantially lower computational complexity than the purely nonlinear Min-Max-MSE THP, while achieving a similar performance. The simulation results showed that the proposed Min-Max-MSE HL-THP achieves a considerably lower Max-MSE than purely linear RZF and PGP-RZF precoding, especially in scenarios, where the number of BS antennas is not much larger than the number of UTs and/or the channel vectors are highly correlated. In the second part, we investigated the impact of IQI, which is one of the most important HWIs, on the achievable rate of UL multi-cell massive MIMO systems, and proposed an IQA-WLMMSE channel estimator and an IQA-WLMMSE detector. The proposed IQA-WLMMSE detection scheme processed the real and the imaginary parts of the received signal separately, and hence, was able to jointly mitigate the IQI and the multiuser interference. We used techniques from random matrix theory and derived an analytical expression for the asymptotic achievable sum rate of the proposed IQA-WLMMSE receiver in the large system limit. The analytical results can be used for performance evaluation without the need for performing lengthy Monte-Carlo simulations. Our simulation results revealed that if left unattended, the IQI leads to severe achievable rate losses, even in scenarios, where the number of BS antennas is much larger than the number of UTs. Moreover, our analytical and simulation results showed that the proposed IQA-WLMMSE receiver achieves substantially higher sum rates than the conventional IQU-MMSE receiver, and performs close to the MMSE receiver in an ideal system, i.e., without IQI. For 5G communications, TDD Massive MIMO has been considered due to its simple implementation and very high performance. The communications model for TDD Massive MIMO has, therefore, been investigated in the project and enables channel estimation in the uplink using the known orthogonal pilot sequences. By employing the principle of channel reciprocity within the same channel coherence time and assuming hardware impairments calibration, the downlink precoding is performed. The performance of different channel estimators has been evaluated with simulations and it has been established that the minimum mean square error (MMSE) estimator performs better than the least square (LS) estimator. Yet, the LS is computationally most efficient and its performance is very close to the element-wise MMSE under high SNR conditions. For evaluating the downlink performance, the conjugate beamforming (CBF) and Zero-forcing (ZF) precoders have been simulated in terms of achievable sum-rate (bits/s/Hz). It has been concluded that CBF performs better in low SNR regime with an additional advantage of decentralized antenna processing. Whereas, ZF provides very high performance from medium to high range SNR scenarios. Afterwards, the proposed theoretical communication model has been implemented on the real testbed in an indoor environment with strong multipath propagation. The measurements have been recorded for uplink where a single user transmits four successive OFDM subframes (each of 1𝑚𝑠) that include 1 pilot symbol along with the 9 payload data symbols (QAM mapped). The estimated channel response between terminal and the antenna array shows deep multipath fading ≥ 40dB at some of the antenna elements. The maximal ratio combining (MRC) technique has been employed to cater this problem and decoding the user data. The results show promising values in range of 3.2% – 4.1% for rms EVM (lower to high order modulation schemes). As the most challenging task in practical massive MIMO operation, we addressed the problem of performing the calibration of the array in a way that does not rely on hardware changes, nor requires excessive extra hardware. We proposed a novel calibration technique only relying on state-of-the-art transceiver hardware (i.e. power amplifiers with an integrated monotonic power detector) and a few semi-passive backscatters without any electrical connection to the actual MIMO array. The innovative algorithm covers all relevant linear hardware impairments of generic RF transceivers, which are mixer I/Q phase and amplitude imbalance, carrier leakage, as well as amplitude and phase of the transmit and receive transfer function, including time delay. The algorithm has been successfully demonstrated using a setup of our proposed Massive MIMO module design, resulting in a LO leakage suppression of -57 dBc and image rejection of -60 dBc. The calibration scheme was then reviewed and published in the most relevant journal for microwave research, the IEEE Transactions on Microwave Theory and Techniques. Based on these results we investigated highly efficient implementations of massive MIMO base stations suited for maximal flexibility required for hardware-oriented research based on the novel theoretical findings in the area of communications, calibration and localization. An entirely modular approach for a 120 antenna MIMO system wholly covering the 5.8 GHz ISM band in one channel has been designed and built resulting in a highly flexible massive MIMO testbed implementing the presented calibration approach and capable of transmitting or receiving arbitrary signals on all antennas simultaneously in the specified band.

Publications

 
 

Additional Information

© 2024 DFG Disclaimer / Copyright / Privacy Policy

Textvergrößerung und Kontrastanpassung