Chapter 7 Data Rate Estimation

In an autonomous radio operation setting, one of the first parameters that we would like to estimate reliably would be the data rate of the received signal. Knowledge of this parameter is required to carry out maximum-likelihood (ML) detection [1] of other parameters, such as the carrier phase or modulation type. Although ML estimation of the data rate itself is statistically optimal, given that there is little to no a priori knowledge of the incoming signal, this approach is often difficult if not impossible to do in practice. One mitigating factor for the autonomous radio under consideration is the fact that the data rates are assumed to come from a set of known values, such as the data rates used in the Electra radio (see [2] and Chapter 2). In particular, the data rates here are assumed to be related by integer powers of an integer base B. This assumption, as will soon be shown, allows us to estimate the true data rate based on estimates of the signal-to-noise ratio (SNR) computed for various assumed data rates. The method for estimating the SNR here is the split-symbol moments estimator (SSME) discussed in [3] and Chapter 6. This estimator is appealing in that the only parameter required for its operation is the assumed data rate. Hence, estimation of the data rate can be done jointly with that of the SNR. Although this approach provides us with a way to estimate both the data rate and SNR together, it will be shown that it is sensitive to symbol-timing error or jitter. In fact, the presence of symbol-timing error can severely degrade the performance of this estimator, as shown in Chapter 6. To overcome this, a modification is proposed in which the jitter is quantized and estimated alongside the data rate and SNR. This approach, based on a so-called generalized likelihood ratio test (GLRT) [4], is robust in the presence of symbol-timing error and