made some DCPD cross correlation plots to compare with LLO.

Time was from gps=1252872096 with duration 2954 seconds, when there was no squeezing and no glitches.  With a 1 Hz binwidth, this corresponds to 2954 effective averages, or a factor of 54 of uncorrelated noise suppression.

Seems like our correlated noise is close to the thermal noise floor at 200 Hz, similar to Livingston.

Something correlated is going on at 1.5 kHz.
We were suspicious of the photodetector dark noise curve, as it cuts through at nearly the right amplitude.  It is uncorrelated, though, and should integrate away.
I checked the dark noise cross spectrum of the DCPDs now, when no correlation should be present (third attachment).  
Some correlated dark noise seems to exist at HF, but it's not enough to explain the correlated spectrum we see at Hanford.

Probably the intensity noise is lower at the time of this spectrum than at the time we took the trace for the noisebudget.  We will continue to observe this when we commission the last week of October.

Code lives in /ligo/home/craig.cahillane/Git/o3-commissioning-paper/code/lho_xcor.py