Proposed Work

First we will model and simulate a discrete domain WLAN modem in MATLAB. This will serve as the algorithm design framework. The joint timing recovery and channel estimation algorithm design consists of two phases:

1. The algorithms for channel estimation and timing recovery are studied separately.
   • channel estimation In an IEEE 802.11a packet, long preamble carries BPSK modulated signals on 52 subcarriers, which can be used for estimating the channel frequency response. The estimation is simply obtained by division in frequency domain [2]. The performance is reported to be acceptable.The adaptive algorithm for training 1-tap frequency domain equalizer and channel estimator looks promising to us.It has less computational complexity compared to other non-adaptive methods proposed, which often involve complicated matrix manipulation. LMS should perform better than RLS in terms of MSE, but the slow convergence is the major concern here. We would like to see the performance of some modified LMS based algorithm but with much faster convergence, such as subband NLMS algorithm.
   • timing recovery We will introduce the carrier frequency offset and sampling clock error. We will use the method in [3] to detect the ISI-free section in an OFDM symbol. The method in [a] studies the ensemble correlation across several symbol. [3] demonstrates that this method can detect the whole ISI-free section in an OFDM symbol. But [3] did not provide a way to detect the timing errors and compensate for that. We will use the method in [4] to estimate the carrier frequency offset and sampling clock error to improve the method in [3]. The method in [4] mainly studies the angular changes in certain FFT bins. And it assumes that symbol synchronization has been carried out, which is actually done by the method in [3].

2. The joint approach is developed. We plan to find a way to do the estimation and synchronization jointly. Because the method in [3] will ensemble over several number of symbols, say 200. This means a relatively large delay. We want to shorten the ensemble duration. But definitely this will degrade the performance of the method in [3]. However, the channel estimator can help us. It will provide the channel delay spread information, which can be used to fine the ISI-free section. In turn, the refined ISI-free section can help to get better estimation of the channel. These two procedures can work interactively.
3. performance evaluation: We want to develop an algorithm that will combine channel estimation and timing together. The result can be BER performance graph. We can compare the estimated channel with the real one, to evaluate the performance of the channel estimator.

Depends on how fast we proceed on the algorithm design stage, we would like use ADS to model the data transmission of WLAN in SDF domain if possible.