The attachment shows REFL WFS signals on 28 May and 8 August. The relevant interventions between these two dates are:

• Locking down RM blade springs (63552, 63567, 63940)
• Swapping the WFS A diode head (63544)
• Lowering the fraction of reflected power reaching WFS A and B, and the LSC REFL diodes (63544, 63625)
• Improved damping loops on the IMs (64007, 64017), RMs (63656), PR3/SR3 (64070, 64152), MC2 (64195), and main interferometer HSTSs (64093)
• Improved analog filtering of IM and RM suspension drives (64150)

Ground motion around 0.8 – 2 Hz was about 5× higher in the May data than the August data, so the improvement seen in that frequency band is not entirely due to the above activities.

The changed noise floor on the 9 MHz signals is consistent with shot noise. I did not observe a similar change on the 45 MHz signals, but anyway they appear to be limited by optic motion below a few tens of hertz. In 9 MHz at least the noise at 10–20 Hz is still limited by HAM1 motion (RF9 yaw is highly coherent with HAM1 RX, for instance), but obviously at a much reduced level from previously.