The message:

Even if the excess low frequency noise on AS45Q when we are locked on DC readout is due to real DARM offset fluctuations, these are not large enough to cause any significant noise in band.

More details:

One of the ideas which has been suggested is that there may be some kind of lock point error in our DARM loop, which is modulating the DARM offset and causing noise through upconversion.  We see some excess noise in AS45Q compared to the DCPDs at frequencies below 1Hz, this could be a because of some lock point errors on either AS45 or the DCPDs.  

The first attached plot shows the two signals with and without a DARM boost and resonant gain that Evan tried out a few weeks ago.  You can see that boosting the DARM loop reduced both signals at frequencies where they agree, from about 7-2 Hz.  At lower frequencies boosting the DARM loop reduced the DCPD spectrum, but has no impact on AS45, the out of loop sensor.  From this we can say that one of the sensors has some kind of lock point error at these frquencies, but we don't know which it is.  

If we assume that the AS45Q sees the true DARM motion and the DCPDs have lock point errors, this could cause upconversion in DARM by modulating the DARM offset.  One idea would be to blend the two sensors, but we can also get an idea if this comes close to the DARM noise anywhere by looking at our data.  

I low passed the AS45 data at 5Hz, and added it to the DCPD residual to get an estimate of what the DARM fluctuations are if the AS45 noise is real (blue and green lines in second attachment). Then I repeated the estimate up upconversion due to DC readout from 25053

The projected noise (pink curve) doesn't come close to DARM except at a narrow line at the calibration line.  So it seems like we wouldn't expect to learn much by blending AS45 and DCPDs for DARM control.