Closed pasha-ponomarenko closed 7 years ago
Instead of a magic number can we replace 3 with SIGMA_LEVEL and then define it in the header to replace ACF_SNR_CUTOFF?
Yep, that's a good idea! Let's call it NOISE_SIGMA_THRESHOLD. It seems to be loong but would not require much explanation.
We also can use the threshold later for re-filtering the fitted data because some of the fitted power will go below the assigned threshold due to statistical fultutations. Howeve, I should think about this carefully because at that stage we will have two "powers", fit.p_l and fit.p_s, and they are not the same...
Removing low-power ACFs from analysis by using an upper three-sigma level of the model noise distribution. The statistical variance (sigma) depends on the number of averages as 1/sqrt(prm.nave) so using this parameter makes more sense than utilising an arbitrary cut-off level. This also makes ACF_SNR_CUTOFF parameter redundant, so I just removed it from this code. This constant should also be removed from preprocessing.h file in order to avoid further confusion.