A data-dependent algorithm was developed for estimating symbol signal-to-noise ratio (SNR) with a non-integral number of samples per symbol. The classical split symbol SNR estimator algorithm was adapted to incorporate the whole symbol by removing the data polarity.
The Whole Symbol Moments SNR Estimator (WSME) was developed for adaptive data rate (ADR) control, and is required to operate at all symbol rates between 1 and 4,096 ksps. This is particularly challenging at the highest symbol rates, especially when there is not an integer number of samples per symbol. For example, a nominal 19.18-MHz sampling rate and a 4,096-ksps symbol rate correspond to approximately 4.68 samples per symbol, in which case each symbol output from the symbol tracking loop (DTTL — data-transition tracking loop) will comprise either four samples (32% of the time) or five samples (68% of the time). Originally, it was envisioned that a split symbol SNR estimator could be used for ADR. However, given that each half-symbol at 4,096 ksps comprises only two or three samples, the resulting SNR estimates exhibit an extremely large variability that render this technique impractical for relay operations with the Mars Reconnaissance Orbiter (MRO). Consequently, a modification to the split-symbol SNR estimator has been developed that allows the utilization of whole symbols. This is accomplished by removing the data polarity of the BPSK (binary phase-shift keying) data symbols, thereby resulting in more samples used in generating the sample SNR estimates.
This algorithm provides a novel extension to the classical split-symbol SNR estimator by incorporating data over each entire symbol. This improves the stability of the algorithm, especially at high data rates.
This work was done by Igor Kuperman and Edgar H. Satorius of Caltech for NASA’s Jet Propulsion Laboratory.