I have some of the M4 20Hz data from software version 1.21. The NREL website lists the “launch date” and fixes made for the version 1.21 release.
The software version now has been upgraded to version 1.23. My own software seems to work fine with the data from either version. Has anyone noticed any data issues between these two versions?
Also, I was wondering if it would be possible to post the specific changes made for the version 1.23 release?
Thanks for posting this question. This was partly an update to the code that was used to write out the data files (I fixed some bugs that were shifting columns around), and partly an update to the code that processed the sonic anemometer data. I apologise for not posting more useful information!
From my notes, this is what I was up to:
- If there is less than 5% missing sonic data, the measured sonic data are mapped to the nearest 0.05-second timestamp and then gaps are filled in the ideal 20-Hz time series using linear
interpolation. There is no second interpolation before the rotation, which could happen in version 1.21. For better or worse, this is the closest I can find to a standard approach in the eddy flux community.
- Extrapolated data at the end of the time series are replaced with the mean.
- The raw, cleaned and rotated sonic time stamps are now explicitly identified. You should find Sonic_cleaned_timestamp, which should be used with Sonic_<x,y,z,T>clean*m and Sonic_rotated_timestamp, which you should as the timestamp for Sonic_<u,v,w>_*m.
If you want to have a look at the data side by side to see what the different timestamps are, check the vertical wind speed on the 15-m sonic. The following is an example using MATLAB code:
plot(24*60*60*(time_UTC.val-time_UTC.val(1)),Sonic_z_15.val,'ko') % raw data
plot(Sonic_cleaned_timestamp.val,Sonic_z_clean_15m.val,'r+') % cleaned data
plot(Sonic_rotated_timestamp.val,Sonic_w_15m.val,'bx') % rotated data