Figure 1. Romain has kindly increased the 8MHz modulation depth by 3dB during a few minutes in LN1 today.
Figure 2 shows spectra in purple for the nominal modulation of 9dBm, and in blue/red the modulation with 12dBm of 8MHz. The spectrum of SSFS quadrature increases by about a factor sqrt(2), the B2 6MHz spectrum increases by ~sqrt(2) for both quadratures. Also B4 PD1 56MHz Q (SRCL) increase by ~sqrt(2), this is explained as in the SRCL noise budget frequency noise dominates the error signal. So everything is consistent with the noise at 6MHz demodulations increasing by a factor sqrt(2) for both quadratures due to the increase in 8MHz modulation depth. The B2 8MHz quadrature spectrum doesn't change, which is consistent with the observation that 8MHz noise level depends on the 6MHz modulation depth.
So the issues seems to be that the modulation at 8MHz produce a noise at 6MHz proportional to the 8MHz modulation amplitude. That noise has the same amplitude in both quadrature of the 6MHz. It is very likely a white noise at the EOM level (i.e. constant level over at least a few kHz around the 6MHz frequency), that is the filtered by the IMC cavity pole at ~500Hz.
An open question is what is the origin of this issue.
- It may come from an interaction of the two sidebands, if both the 6MHz and 8MHz are modulated using the same crystal. Noise from the 8MHz modulation electronics, can travel through the crystal, and the become amplified by the resonant circuit of the 6MHz modulation. A way to test it would be to move the 8MHz modulation to another crystal, for example the crystal currently used for the 81MHz sideband.
- It may come from the modulation electronics itself, if the resonant circuit is not selective enough, and we see the tails of the resonance at 8MHz modulating the light at 6MHz.