In this paper, we analyze optimal (in space and time) adaptive power transmission policies for fading channels when the channel-state information (CSI) at the transmitter (CSIT) and the receiver (CSIR) is available. The transmitter has a long-term (time) average power constraint. There can be multiple antennas at the transmitter and at the receiver. The channel experiences Rayleigh fading. We consider beamforming and space-time coded systems with perfect/imperfect CSIT and CSIR. The performance measure is the bit error rate (BER). We show that in both coded and uncoded systems, our power allocation policy provides exponential diversity order if perfect CSIT is available. We also show that, if the quality of CSIT degrades then the exponential diversity is retained in the low SNR region but we get only polynomial diversity in the high SNR region. Another interesting conclusion is that in case of imperfect CSIT and CSIR, knowledge of CSIT at the receiver is very important. Finally, for the optimal power control policy of the uncoded system we find the error-exponents which provide the rate versus diversity-order tradeoff for this policy. This tradeoff is of an entirely different nature than the well-known Zheng-Tse tradeoff.

This work analyzes the average delay performance of block coding schemes when the arrival stream is stochastic. From classical Shannon Theory, it is known that communication is feasible at all rates strictly below capacity of a channel. However, this reliable scheme of communication is realized with unbounded coding length and hence average delay. This work considers the delay analysis of general block coding schemes over a noisy channel in presence of retransmission requests. Modeling the communication system as a queuing system with bulk service, an expected delay analysis is provided. The expected delay bits experience is then optimized by an appropriate choice of forward error correction scheme.

Data transmission over a discrete memoryless channel is considered when the arrival of data is bursty and is subject to a delay deadline. An exponential decay of the probability of delay violation with respect to a large delay deadline is proved when the block length scales linearly with the deadline. When considered in conjunction with Gallager's error exponents, the first natural consequence of this result is a separation principle: a separated scheme of buffering traffic and blockcoding transmissions achieves arbitrarily high reliability for an asymptotically large delay budget. Furthermore, the exponential decay nature of the result provides some insight as how to budget the delay limit between the coding time and the waiting time in the queue.