IMEC has developed a multiple input multiple output (MIMO) near-maximum likelihood (ML) detection technique that is scalable in terms of performance and power consumption. The scalable MIMO functionality has been designed to meet the high spectral efficiency requirements in next-generation wireless standards such as 3GPP LTE. The scalable MIMO detector will be implemented on a software defined radio (SDR) platform that is based on parallel architectures. This implementation approach supports different communication standards and meets today's demand for high spectral efficiency, low power consumption, low cost and reduced time to market.
In order to address high spectrum costs, emerging standards such as 3GPP LTE provide high spectral efficiency. These emerging standards achieve high spectral efficiency through a combination of orthogonal frequency division multiplexing (OFDM) and spatial multiplexing (SM). To fully exploit these techniques, LTE terminals require high-complexity and power-consuming detection techniques such as multiple input multiple output (MIMO) maximum likelihood (ML) detection. Furthermore, wireless user terminals require support for multiple standards in order to operate in different parts of the world. This flexibility comes at the penalty of increasing design complexity and silicon area, since more functionality has to be implemented in the baseband platform. These requirements add up to a major challenge, because LTE terminals must be developed at low cost, short time to market and low power consumption in order to be competitive.
IMEC provides a solution for the majority of these challenges with an SDR implementation of a scalable near-ML detector. This solution adaptively and efficiently scales the ML search space. Firstly, the detection technique is scalable in terms of performance and power consumption. This scalability tailors the power consumption to the required high spectral efficiency given certain channel propagation conditions. Secondly, the scalable LTE MIMO functionality has been designed to be implemented on an SDR platform based on parallel architectures. This approach supports reduced time to market and power consumption, and it supports different communication standards. Thirdly, in combination with a flexible and scalable forward error correction (FEC) architecture, it allows a cross-layer controller to tune the detector and the FEC performance/complexity to obtain minimal power consumption given a required data throughput.
In this article, the novel MIMO near-ML detection technique is presented and its scalability in both performance and power consumption is evaluated. Next, the co-optimization of the near-ML detector algorithm and implementation for parallel architectures is discussed. Finally, the benefits of these approaches are summarized in view of the challenges that today's wireless terminals impose,.
Multiple antenna detection
MIMO wireless systems use spatial multiplexing (SM) as a way of increasing the data rate for 4G systems. In spatial multiplexing, a high rate signal is split into multiple lower rate streams and each stream is transmitted from a different transmit antenna in the same frequency channel. Although SM MIMO is a very powerful technique, its maximum benefits in terms of spectral efficiency can only be reached when advanced detection techniques such as maximum likelihood (ML) are applied. ML is a well-known non-linear technique that allows an optimal detection of signals transmitted over MIMO channels. It offers a higher performance than linear detectors in SM MIMO systems.
However, ML detectors are highly complex, with their complexity depending on the size of the likelihood search space. This in turn depends on the number of antenna-parallel streams, constellation order and excess delay. The size of this search space can be significantly reduced by combining ML detection with OFDM, which efficiently handles the problems created by multi-path channels. OFDM simplifies the multi-path channel propagation into orthogonal frequency domain signal tap propagation channels. OFDM modulation and MIMO are therefore two key components for supporting most of the emerging (and on-going) wireless communication standards such as 3GPP LTE. Still, in most cases, the complexity requirements of ML detection when applied to SM MIMO OFDM prohibit its practical implementation.
The complexity of the likelihood search space can be reduced by applying so-called sub-optimal techniques. In recent years, several scalable, low-complexity sub-optimal ML likelihood detectors have been proposed. Among the proposed methods, tree search detection algorithms are the most promising ones. Most contributions in this field constrain the complexity in a uniform and sub-optimal way by performing a uniform tree search which does not take into account the channel propagation conditions and signal-to-noise ratio. Although most of the proposed solutions result in a simple implementation, they suffer from severe performance degradation and require a large number of operations for achieving near-ML performance.