Posted is a comparison of the frequency noise coupling to DARM from Oct 2016 as opposed to Nov 2016, both calibrated into m/Hz (Plot 3).

We calculate the coupling slightly differently, with Ballmer-Hall calibrating the REFL9 PD into m/W, then calculating m/Hz using a CARM calibration, while we used REFL_SERVO_ERR cts, the error point on the Common Mode Board, to measure m/cts.  There's some phase differences, but the magnitude results are consistent.

My calibration of REFL_SERVO_ERR into Hz is as follows:
CARM DC Optical Gain = 31.3 mW/Hz
CARM pole = 0.6 Hz
Transimpedance = 2900 V/W
SumNode Gain = 2 dB
REFL_SERVO_ERR Whitening Gain = 200 V/V
ADC = 2^16/40 cts/V

The DARM calibration is the Goetz recalibration.

The frequency noise coupling is around 2.5 times higher now than in 2016.  Our beliefs about our CARM optical gain is is also 2.5 times higher now than in 2016 (Ballmer-Hall used 13 mW/Hz).  The frequency dependence of the coupling has not changed.

Is this real?  Is the CARM optical gain really higher?  And if so, is the CARM coupling to DARM higher by the same factor?

It's reasonable to believe that the CARM optical gain is really higher, with our  ~20% improvements in the power-recycling gain.  An improvement to the PRC gain is primarily indicative of improvements to the arm reflectivity, and since CARM optical gain depends on the derivative of the arm reflection, it's possible that 20% PRC gain improvement could produce 2.5 times more CARM gain.  Deep thought (#modeling) is required to verify this statement.

So, is the CARM to DARM coupling really higher?  This is far less clear, since this coupling also depends on the arm reflectivity, but also DARM offset, contrast defect, radiation pressure, SRC detuning, practically all the dirty aspects of interferometry that are hard to know.  Very deep thought, and a bit of wrangling, is required.

My impression is that CARM to DARM coupling is probably worse now because of the improved PRC gain, as that is the clearest indicator of a changed interferometer from 2016 and 2018.  The radiation pressure coupling (low frequency rise) is also 2.5 times worse, same as the high frequency increase, so rad pressure coupling probably is not very different now than then.  

Team TCS is sitting next to me and tell me contrast defect is about 0.1% now for a cold, dark simple michelson.  We're not sure what this number was in 2016 (Dan Hoak quotes 70 ppm for a 22 W IFO in 2015), and we don't know what the 2018 contrast defect is for high power, more wrangling required.