The summing of the first antennas can bias the k's for later antennas.
Since the set of k's is limited to -3,-2,..,3 this influences the amount of 'correct' k's to try.
It is better to use the strongest signal only to keep the same reference point
This makes sure to actually use interpolation to find the best period offset instead of rolling the samples.
The latter won't work when the period is not dividable by the sampling period.
