Tradeoff between channel estimation accuracy and application throughput for in-home MIMO power line communication

In this paper we consider two alternate packet structures for estimating the multi-input multi-output (MIMO) powerline channel and their impact on achievable application throughputs. In the first packet structure, an orthogonal pair of OFDM symbols is used to estimate the channel. When compared to single-input single-output (SISO) transmissions that use one OFDM symbol to estimate the channel, there is an additional overhead of one OFDM symbol in this approach. This additional overhead will reduce the throughput improvements offered by MIMO. In the second packet structure, a single OFDM symbol is used to estimate the MIMO channel and interference between the two transmit streams is avoided by using alternate carriers for each stream. The composite MIMO channel is then obtained by interpolating the estimated response across the missing carriers in each stream. The interpolation may introduce an estimation error in the channel estimate that could degrade the achievable signal-to-noise ratios on each stream and thereby degrade the physical layer (PHY) data rate. However, when compared to the first packet structure the overhead is reduced by one OFDM symbol. This will result in a more efficient transmission that could overcome the loss in PHY data rate. This tradeoff between overhead for channel estimation accuracy and application throughput is studied in this paper using power line channel and noise data that were measured in the field. It is demonstrated that although a higher PHY rate can be achieved when an orthogonal pair of symbols is used to estimate the channel, this approach has a lower TCP and UDP rate when compared to the packet structure that uses only one symbol and interpolation to estimate the channel.