(We surely remembered this fact until before the earthquake and planned to replace it but...)
Details of the modification is available in Shimode-san's memo (in Japanese). The 1st, 2nd, and 3rd stages are set as z1:p10, z1.5:p30, and z10:p100, respectively. Because of the modification of the real-time model for reducing thermal noise of HPCD (see also klog#31874), the 3rd one is engaged first.
DAC noise projection with taking into account this fact is shown in the attachment. This plot is made with the assumption of the most pessimistic case in the previous post. Notation of gray, black, magenta, cyan, and lightgreen curves are same as Fig.2 in klog#32123. Darkblue and darkgreen curves represent DAC noise projections with engaging 1 and 2 stages of de-whitening filters taking into account the modification for ETMY HPCD. Though replacing nominal HPCD is not enough for achieving the target sensitivity, DAC noise contribution becomes slightly large by this modification in the case of engaging only 1 stage of de-whitening. In the case of engaging two stages of de-whitening, we cannot large change between nominal HPCD and modified one. This should be because the noise level is limited by the output noise of HPCD, not DAC.
-----
Rough check:
When we use nominal HPCD for all 4 suspensions DAC noise is suppressed by 1 stage of de-whitening as a factor of ~0.14@10Hz (=sqrt(4 * (-17dB)^2) / sqrt(4)) and ~0.1@100Hz. On the other hand, one modified HPCD is mixed DAC noise is suppressed as a factor of ~0.44@10Hz (=sqrt(3 * (-17dB)^2 + (-3dB)^2) / sqrt(4)) and ~0.11@100Hz (=sqrt(3 * (-20dB)^2 + (-17dB)^2) / sqrt(4)). According to these estimations, DAC noise suppression from magenta curve to cyan curve (only nominal HPCD) and darkblue curve (mixing 1 modified HPCD) seems to be reasonable.
