Another signal that sees ~1/sqrt(f) noise is B1s.

Figure 1 show the spectrum at 3 different time, red is nominal, purple is with the 6MHz modulation depth increased, and blue is when the RAMS servo was put in place.

The increase in noise with the modulation depth shows that the spectrum is dominated by the 6MHz noise.

Figure 2 shows the projection of this 6MHz noise. It is simply the spectrum of B1s multiplied by 20 (as B1s sees only 5% of the power reflected by the OMC), then multiplied by 4.4e-3 (the theoretical transmission of the OMC at 6MHz), and added as noise onto B1 and calibrated into h(t). That projection is about a factor 3 below the sensitivity curve.

It would be interesting to check in LN2 the impact on the sensitivity of increasing the 6MHz modulation depth. The result are likely to be unclear in the present situation, as increasing the 6MHz modulation depth will increase this noise, but reduce the laser frequency noise, so the two effect might compensate each other. The best would be to achieve a good CMRF first and a working feed-forward to remove the spurious noise affecting the SSFS sensing.

Another interesting test would be to inject directly amplitude and/or phase noise at 6MHz.