QR-RLS adaptive digital filter with 18-bit pipeline architecture
A QR-RLS adaptive digital filter provides fast computation without excessive computational resources. 18-bit multipliers enhance speed, and a floating point inverse square root block adjusts dynamic range in 12-dB steps. A memory stores two P-matrix copies, one being delivered with rows shifted according to the clock speed so as to enhance pipeline processing. Embodiments reliably detect modulation schemes, demodulate strong signals by passing feedback bits between multiple stages, remove impulses due to lightening, etc, erase symbol estimates which exceed an error threshold, and add high frequency noise to avoid mathematical divergence caused by excessive S/N. A genetic method is provided for identifying asynchronous spreading codes with minimum correlations, whereby randomly selected candidates compete based on Frobenius norms of their correlations, the weakest being discarded and the process being iterated. A method is provided for selecting optimal filter sampling windows for simultaneously detected symbol streams having relative timing delays.
1. A genetic method for identifying groups of asynchronous spreading codes having minimal mutual correlations, the method comprising: randomly selecting a population of candidate Code Division Multiple Access (CDMA) asynchronous spreading codes; grouping the candidate CDMA asynchronous spreading codes into competition groups; within each competition group, ranking the CDMA asynchronous spreading codes according to their fitness relative to the other members of the competition group; for each competition group, eliminating at least the candidate CDMA asynchronous spreading code having the lowest fitness; replacing the eliminated candidate CDMA asynchronous spreading codes with randomly selected replacement CDMA asynchronous spreading codes; and repeating the steps of grouping, ranking, eliminating, and replacing until a termination criterion has been met; wherein at least one of the preceding steps is performed by a processor.
2. The method of claim 1 , wherein ranking the CDMA asynchronous spreading codes within each competition group includes: for each candidate CDMA asynchronous spreading code in the competition group, calculating its correlations with all other members of the competition group for each I-chip relative time shift therebetween; calculating a Frobenious norm of each calculated correlation; determining a maximum Frobenious norm for each candidate CDMA asynchronous spreading code in the competition group; and ranking each candidate CDMA asynchronous spreading code in the competition group according to its maximum Frobenious norm, the candidate with the highest maximum Frobenious norm having the lowest fitness of the competition group.
3. A method for choosing asynchronous spreading codes having minimal mutual correlations, the method comprising: grouping respective candidate asynchronous spreading codes of a population of candidate asynchronous spreading codes into groups; for each candidate asynchronous spreading code in the group, calculating its correlations with all other members of the group for each I-chip relative time shift therebetween; for each group, calculating a Frobenious norm on a set of calculated correlations of that group; and choosing a group based on comparing the calculated Frobenious norms; wherein at least one of the preceding steps is performed by a processor.
4. The method of claim 3 , wherein each correlation in the set of calculated correlations is the maximum correlation across all shifts between two correlated candidate asynchronous spreading codes.
5. The method of claim 3 , wherein choosing a group based on comparing the calculated Frobenious norms includes choosing the group with the lowest Frobenious norm.