Hi everyone,
Per my earlier post, I wanted to note that currently there seems to be some cleaned and rotated 20 Hz sonic records that still have “spikes” in the data. These spikes can range in width from 2 samples to around 6 samples, and might throw off some of your results if you are looking at spectral information. I’ve tried a couple detection algorithms and I think I’ve found one that works for all of the strange cases that come up (like sample points halfway up a spike, etc.); after a spike is detected, I linearly interpolate between the beginning and the end of the spike.
There’s an image below of a record with a spike (ignore the second plot).
EDIT: this is for the M4 tower data. Apologies for not making that explicit in the thread title.
- Jenni
