Measuring the COMM noise with the refl DC bias path engaged gives repeatable results, while we did not repeatable results without it.  However, something seems to be wrong with our calibration.  Today, when the refl bias path is not engaged we wander over the peak in about 10 seconds, and all th way off resonance, so we know that the COMM noise is at least a fwhm of the IR resonance in the arm, 164Hz. 

When we lock the refl bias path, measure a spectrum and correct for the gain of the refl DC bias path, the rms should be the noise of the refl bias path.  However, we are consistently measuring a noise around 35 Hz (with cavity pole corrected for) or around 9 Hz without taking out the cavity pole.  However, we know this is wrong. 

At least with a repeatable measurement we can evaluate what make the noise worse and better, even if we don't yet trust the calibration. Attached is a plot of the COMM noise measured with refl DC bias engage, the gain of that path corrected for but not the cavity pole.  In the blue trace the OpLev damping was on for both pitch and Yaw, in the green trace the pitch damping was off on both the ETM and ITM.  The coherence are shown in the bottom panel, the solid lines were measured with no damping, the dashed lines with damping.

With the damping on, the noise at 0.45Hz is reduced by a factor of 5.5, although it doesn't seem to change the rms much in this plot. 

The green trace was measured with lower bandwidth because the lock was not lasting more than a few minutes today, especially with the damping off.