Design of indoor communication infrastructure for ultra-high capacity next generation wireless services

The proliferation of data hungry wireless devices, such as smart phones and intelligent sensing networks, is pushing modern wireless networks to their limits. A significant shortfall in the ability of networks to meet demand for data is imminent. This thesis addresses this problem through examining the design of distributed antenna systems (DAS) to support next generation high speed wireless services that require high densities of access points and must support multiple-input multiple-output (MIMO) protocols. First, it is shown that fibre links in DAS can be replaced with low-cost, broadband free-space optical links, termed radio over free-space optics (RoFSO) links. RoFSO links enable the implementation of very high density DAS without the need for prohibitively expensive cabling infrastructure. A 16m RoFSO link requiring only manual alignment is experimentally demonstrated to provide a spurious-free dynamic range (SFDR) of > 100dB/Hz^2/3 over a frequency range from 300MHz- 3.1GHz. The link is measured to have an 802.11g EVM dynamic range of 36dB. This is the first such demonstration of a low-cost broadband RoFSO system. Following this, the linearity performance of RoFSO links is examined. Because of the high loss nature of RoFSO links, the directly-modulated semiconductor lasers they use are susceptible to high-order nonlinear behaviour, which abruptly limits performance at high powers. Existing measures of dynamic range, such as SFDR, assume only third-order nonlinearity and so become inaccurate in the presence of dominant high-order effects. An alternative measure of dynamic range called dynamic-distortion-free dynamic range (DDFDR) is then proposed. For two different wireless services it is observed experimentally that on average the DDFDR upper limit predicts the EVM knee point to within 1dB, while the third-order SFDR predicts it to within 6dB. This is the first detailed analysis of high-order distortion effects in lossy analogue optical links and DDFDR is the first metric able to usefully quantify such behaviour. Next, the combination of emerging MIMO wireless protocols with existing DAS is examined. It is demonstrated for the first time that for small numbers of MIMO streams (up to ~4), the capacity benefits of MIMO can be attained in existing DAS installations simply by sending the different MIMO spatial streams to spatially separated remote antenna units (RAU). This is in contrast to the prevailing paradigm of replicating each MIMO spatial stream at each RAU. Experimental results for two representative DAS layouts show that replicating spatial streams provides an increase of only ~1% in the median channel capacity over merely distributing them. This compares to a 3-4% increase of both strategies over traditional non-DAS MIMO. This result is shown to hold in the multiple user case with 20 users accessing 3 base stations. It is concluded that existing DAS installations offer negligible capacity penalty for MIMO services for small numbers of spatial streams, including in multi-user MIMO scenarios. Finally, the design of DAS to support emerging wireless protocols, such as 802.11ac, that have large numbers of MIMO streams (4-8) is considered. In such cases, capacity is best enhanced by sending multiple MIMO streams to single remote locations. This is achieved using a novel holographic mode division multiplexing (MDM) system, which sends each separate MIMO stream via a different propagation mode in a multimode fibre. Combined channel measurements over 2km of mode-multiplexed MMF and a typical indoor radio environment show in principle a 2x2 MIMO link providing capacities of 10bit/s/Hz over a bandwidth of 6GHz. Using a second experimental set-up it is shown that the system could feasibly support at least up to a 4x4 MIMO system over 2km of MMF with a condition number >15dB over a bandwidth of 3GHz, indicating a high degree of separability of the channels. Finally, it is shown experimentally that when a fibre contains sharp bends (radius between 20mm and 7.2mm) the first 6 mode-groups used for multiplexing exhibit no additional power loss or cross-coupling compared with unbent fibre, although mode-groups 7, 8 and 9 are more severely affected. This indicates that at least 6x6 multiplexing is possible in standard installations with tight fibre bends.For their financial support, I would like to thank the Rutherford Foundation of the Royal Society of New Zealand, the Cambridge Commonwealth Trust and the EPSRC