Analogue MIMO Detection
- PDF / 840,692 Bytes
- 10 Pages / 600.03 x 792 pts Page_size
- 32 Downloads / 185 Views
Analogue MIMO Detection Robert J. Piechocki,1 Jose Soler-Garrido,1 Darren McNamara,2 and Joe McGeehan1, 2 1 Centre
for Communications Research, University of Bristol, Merchant Venturers Building, Woodland Road, Bristol BS8 1UB, UK Research Laboratory, Toshiba Research Europe Ltd, 32 Queen Square, Bristol BS1 4ND, UK
2 Telecommunications
Received 1 December 2004; Revised 17 May 2005; Accepted 8 July 2005 In this contribution we propose an analogue receiver that can perform turbo detection in MIMO systems. We present the case for a receiver that is built from nonlinear analogue devices, which perform detection in a “free-flow” network (no notion of iterations). This contribution can be viewed as an extension of analogue turbo decoder concepts to include MIMO detection. These first analogue implementations report reductions of few orders of magnitude in the number of required transistors and in consumed energy, and the same order of improvement in processing speed. It is anticipated that such analogue MIMO decoder could bring about the same advantages, when compared to traditional digital implementations. Copyright © 2006 Hindawi Publishing Corporation. All rights reserved.
1.
INTRODUCTION
Turbo codes and more general turbo principles (turbo equalisation, turbo multiuser detection, etc.) are bound to have a substantial impact on the next-generation wireless systems. The turbo principle requires exchange of the so-called soft information, which is a probabilistic measure. In current implementations (e.g., turbo coding) this information is sampled then quantised (digitised) and handled by digital signal processors. The amount of digital information to be processed by DSPs and FPGAs is enormous and represent a “bottleneck” for high speed digital systems. However, the soft information, being analogue in nature, is best represented in analogue domain (e.g., electric currents or voltages). More interestingly, it can be processed in this form by analogue networks as well. The analogue decoding paradigm formulated in [1, 2] takes this stand. First analogue implementations of binary decoders were reported in the literature in [3–5]. Those implementations reported reductions of 1–3 orders of magnitude in number of required transistors and in consumed energy, and the same order of improvement in processing speed. More ambitious CMOS-only implementation of analogue decoders was recently reported in [6, 7]. In truth, it was the neural networks community that first used analogue VLSI circuits to build simple artificial neural networks [8]. Both neural networks and communications engineering are by and large examples of computation, and as a result the fundamental building blocks are the same in both cases. Some fundamentals of analogue computations stem directly from the universal Turing machine
paradigm worked out by Alan Turing nearly 70 years ago [9]. Subsequently, they were used in many versions of analogue and mixed-mode (micro-) processors built over the last few decades. In this contribution we extend the concept of ana
Data Loading...
