I check how much loss in the spacer would have to be to have the contribution comparable to the current beat signal. It turns out to be 10^-4 (the current number we are using in the nb calculation is 10^-7. I'll check in the literature to see if the spacer loss angle can go up that much due to the non polished surface and back-action loss or not.

       Since the spacers in our setup is not polished on the side and the supports are on two o-rings which are not optimized on the nodal supports, the mechanical loss in SiO2 substrate might be higher than its usual number. So it is a good idea to check what would be the loss of the spacer to be able to give the thermal noise at the same level as what we measure. By adjusting the loss angle to 10^-4 (in the spacers only, not substrates), the calculated noise is about the same as the measured result.

 fig1: The total noise estimation with loss in spacer = 10^-4 plotted with the latest beat measurement.