No abstract

No abstract

No abstract

No abstract

Statistics obtained during demodulation are used in a direct-sequence spread-spectrum packet radio network to govern the adjustment of transmitter power within the first few packets of a session. The statistics are also employed by an adaptive transmission protocol to select the modulation parameters and the rate of the error-control code for each packet. The power-adjustment protocol uses the statistics to ensure that the transmitter power level is high enough to satisfy the signal strength requirements at the receiver and low enough to prevent unnecessary interference to nearby radios. We describe a protocol for which modulation and coding parameters are adapted to achieve the most efficient com bination for the given channel conditions. We evaluate the performance of each protocol for several channels. The protocols are not given any information about the type of channel or the channel parameters, yet they are able to use the demodulator statistics to efficiently set the initial power level and adapt transmissions throughout the session.

Next generation packet radio networks will have far greater processing capabilities than current radio systems. We propose and evaluate decoding techniques that make use of such capabilities to increase the probability of successful decoding. We propose a metric derived from statistics collected during demodulation in a binary CDMA receiver. We investigate several methods to apply the proposed metric to the demodulator's softdecision outputs prior to decoding. Our soft-decision decoding techniques are designed to mitigate the effects of interference from other signals in the frequency band. We compare the performance of our proposed metric to the log-likelihood ratio (LLR) metric, which requires that the mean signal level and noise variance are known for each bit position. Rather than attempt to estimate these parameters directly, our metric uses demodulator statistics and thus does not require pilot symbols or training sequences typically required by an LLR-based metric.

No abstract

No abstract