Should accept an option to calculate SNR from coherence, possibly from multiple coherence estimates.
Could be combined with an SNR or coherence cutoff level too, to reduce datasets with a lot of points.
