We locked the OMC in a few different configurations tonight, to try to understand the noise features in DARM between 100Hz-1kHz.

With the ITF locked and DARM on RF, we locked the two OMCs on one of the 56MHz sidebands.  See Figure 1; purple is sideband, blue is the usual carrier lock (with DARM on DC).  The structures around 100Hz and 150Hz are well-matched between the two locks, these structures also appear in MICH and in the 112MHz signal.  When locked on the 56MHz sideband the B1_DC power is about 3mW (on the carrier, with the DARM offset, it's 25mW).  The sideband doesn't appear to have the dense features above 200Hz.

While the OMCs were locked on the sideband we noticed the PRCL line injection at 67.1Hz was quite strong in the sideband power.  This implied the PRC is not perfectly on resonance, so we scanned the PRCL offset by +/-1e-4 counts.  The line height did not change, but the arm power dropped by 30% at the maximum offset (data between 19:13 and 19:18 UTC).  The effect of the offset on the carrier buildup appeared to be symmetric around zero offset.

Then, with a single bounce off NI and all other mirrors misaligned, we locked the first OMC on the sideband and the carrier, see Figure 2.  Again, purple is sideband (about 0.13mW of light on B1s2), blue is carrier (7.5mW).  In this state we think the lock on the sideband is a little strange, and probably includes a contribution of p-polarized carrier light (Michal says this resonates with the same optical frequency in the OMC as the 56MHz sideband).

It seems like we're missing some power incident on the OMCs, in single bounce there should be 14W * 0.04835 * 0.5 * 0.5 ~= 160mW arriving at the dark port, but with the first OMC locked we only get 130mW in the sum of B1s1 and B1s2, after correcting for the fraction of light in each beam that's dumped.  (A large fraction of light is rejected by the first OMC in single bounce, so including the B1s1 beam is important.)  The B1p photodiodes report more light (between 140mW and 160mW), so maybe there are some losses between the B1p cutoff and the OMCs.  Or, we don't know our pickoffs as well as we should.