I tried to evaluate the level of extra noise for different DARM offsets, using the data taken in the experiment performed by Annalisa and Antonino. In order to do that, I made some hypothesis about the residual noise below the extra noise, then I fitted the floor of the sensitivity curve with a model of total noise which includes two parametric model, one for the mid frequency and one for the high frequency. As free parameters, I used only the amplitude. For the high frequency, the expected shape has been taken (flat * ZERO@56Hz). For the mid frequency, the hypothesis of a perfectly flat noise has been discarded (in the past I reported some evaluation based on the 'flatness', but I made a macroscopic error in the modeling of the residual noise). A model which seems in a good agreement with a large set of data is ~f^0.25. In particular for the best available data (fig 1, DARM SET 0.0006, 47 Mpc), that slope seems the best to have a good agreement at 60 Hz and at 300 Hz at the same time.

In fig 2, fig 3 and fig 4 is reported the evaluation for DARM SET 0.0008; 0.0015; 0.002, which are the values taken one close to the other.

Fig 5: 'FLAT NOISE' level vs DARM SET (B1 power).

Fig 6: 'FLAT NOISE' level vs  SQRT(DARM SET) (longitudinal offset)

Both sets of data lie on a straight line, but the second fits better (the accuracy of the evaluation is probably too low to give a good level of confidence to this statement).

In any case, the level of noise is not going towards a saturation: it seems that we can discard the hypothesis that the good effect of DARM SET lowering is vanishing, due to some constant noise, not included in the model, which is replacing the flat noise.

Unfortunately we have a lot of data with DARM SET at 0.00045 (fig 8), but we cannot find there the expected improvement.

The data has been used also to evaluate the level of electronic noise (fig 7, red dots), which begins to be well visible for low offsets. That level has the expected trend vs the optical gain (fig 7, blue dots) if the level of the shot noise is fixed at 2.58e-24.