There is a danger to use DTT together with NDS2 as described here: https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=22128

When you look at the data of a channel from the past, if the sampling rate of that channel was changed in  the past, DTT could be confused and could output a totally bogus calculation. If you're only looking at the spectrum it's just a matter of wrong frequency axis, but if you look at the coherence between a channel with this problem and a channel without, your coherence is totally gone. Recently I was hit by this behavior again and spent a day to figure it out.

In the first attachment, I was looking at the coherence between PEM magnetometers and DARM using DTT. On the bottom row left half,  you can see that DARM and H1:PEM-CS_MAG_LVEA_OUTPUTOPTICS_QUAD_SUM_DQ are sometimes coherent with each other (BTW this cannot be magnetic coupling into HAM6, because the level of magnetic noise is too small according to Robert) but indivisual X, Y and Z channles don't show any coherence (top row right half, middle row). Strange thing is, QUAD_SUM is made inside the frontend, adding square of X, Y, and Z.

In the second attachment, I did the same measurement using matlab, and  X coherence is actually larger than QUAD_SUM. This is not as mysterious as I thought from the DTT plot.

The difference, it turns out, is that the sampling rate of X, Y and Z channels were increased from 4096Hz to 8192Hz (but that's not the case with SUM). DTT cannot handle this and assumes that they were always 4096Hz. If you look at the spectrum of, say, X channel in DTT, you'll see that the first mains line peak is at 30Hz, not 60.

This is really annoying.