We dis some investigation of the DARM sideband noise around the mains lines. In the 35Mpc lock of June 25 the bumps around 100, 200, and 300Hz  are particularly large and simmetric (Figure 1). The Noise budget projection of demod noise from B4_56MHz fits well the 100Hz and 200Hz sidebands (not so much the 300Hz which might have a different origin): Figures 2,3,4. Therefore a possible explanation of the sidebands (Michal) is that clock signals are polluted by (too) large phase noise at the mains harmonics, and this is added to demod PD signals in the demodulation process. The recorded phase noise of the clocks indeed shows large noise at the mains harmonics, particularly the even harmonics: Figure 5, Figure 6 and Figure 7.

A naive check is shown in Figure 8 (Barbara): the low frequency part of B4_56MHz_I signal is modulated using the phase noise of the SPRB 10MHz clock, and the resulting signal is used to compute coherence with B4_56MHz itself. See that there is good coherence at the sidebands around all 50Hz harmonics. The coherence of the same modulated signal and DARM is large mainly at sidebands of 100 and 200Hz. It is possible that DARM noise around other mains harmonics is dominated by other kind of non-linear paths. 

To be noted:

- The phase noise of the clocks (in CEB) overall reduced with the swap to the modified receiver (elog41991) except that at the mains frequencies, where it increased by a factor 3 or so (Figure 5, Figure 6 and Figure 7). This increase looks consistent in all distributed clocks. The phase noise of NEB and WEB clocks got worse with the swap, but (Alain info) this is expected because of cross-talk between as the distributed 10MHz and IRIG-B signals which is on the same wire for end of arm benches, but separated for central building benches.

- The online hrec subtraction (elog 41668) looks efficient for the sidebads of 200Hz, but less efficient for those at 100Hz (see Figure 1). This subtraction of the 100Hz sidebands can be tuned to be made more efficient (Michal): Figure 9.

Conclusions:

- the 100Hz and 200Hz sidebands in DARM seem associated to large 100Hz and 200Hz phase noise of the clocks.
- a reduction of at least a factor 10 of this noise in the clocks is needed.
- most of the noise is subtracted in the h(t) reconstruction.
- nothing yet we can say about sidebands around other mains harmonics (50, 300, ...) that are large in both DARM and hrec.