Blind vs. trained channel estimation for an SDMA WLAN

Wireless LANs (WLANs) offer an alternative to conventional wired LANs in modern organisations that require a flexible network. However, present-day WLANs yield low capacity compared to conventional wired LANs. In a previous study, we showed that the capacity of a WLAN can be increased by space division multiple access (SDMA). The implementation of a SDMA receiver requires a channel estimation algorithm. Therefore, we compared both a blind and training-based algorithm for channel estimation in an SDMA WLAN. In this paper we present the semi-analytic simulations based on a ray-tracing channel model. For both algorithms we determine (a) the number of symbols needed to obtain a frequency re-use factor of four, (b) compare their robustness and (c) calculate their complexity. As it turns out, the capacity reduction from training can be reduced to a negligible number of symbols (30 in our case study), the training-based algorithm is more robust and the complexity of the blind algorithm exceeds the complexity of the training-based algorithm by one order of magnitude