Time-domain observation filter is erroneous

The calculations of the time-domain optimal observation filters are not correct.

Initial trials have shown that the performance depends on the relation of the filter length to the length of the signals. This is definitely not correct and has to be fixed.

After discussing the problem and going through the code with Jordan Cheer and Aimee Zhang the first candidates for the erroneous calculations identified are cross- and auto-correlation matrices. It seems that the calculations are correct, so the structure of the matrices may be the problem.