We been commissioning the CRS this week, so far we've done a tilt to tilt measurement:

• We excited RY on HAM3 for about 40 minutes starting around 1468604538 (2026/07/20/17:42:00). The excitation template is located at /ligo/svncommon/SeiSVN/seismic/HAM-ISI/H1/HAM3/CRS/Templates/dtt/ryex2.txt, and we ended up finding that setting the gain to ~2000 works well. 
• The dtt template used is CRS_tilt_TF_072026.xml and is located in /ligo/svncommon/SeiSVN/seismic/HAM-ISI/H1/HAM3/CRS/Templates/dtt/CRS_tilt_TF_072026.xml 
• Because the excitation window was so short (~40 minutes), we can only get 2 averages, and I've attached a picture of the result of the measurement (picture 1). We used these measurements to calculate parameters of the response function using a Matlab script which fits the data gotten from the excitation.
• The script used to get these parameters are in /ligo/svncommon/SeiSVN/seismic/HAM-ISI/H1/HAM3/CRS/Scripts/CRS_TransferFunction.m
• To run the script, we got a Q measurement for the CRS in vacuum using /ligo/svncommon/SeiSVN/seismic/HAM-ISI/H1/HAM3/CRS/Scripts/ring_down.m and looked at a 12 hour period starting at 1468630128, which gave a Q measurement of 4365.
• Using the parameters from CRS_TransferFunction.m, we got the zeros and poles of the CRS response and used those to make a new SensInv filter (SensInv_Meas) for the CRS (H1ISI_HAM3_CRSRY_THETA, located in the THETA block of the CRS_OVERVIEW). We used a Q of 50 for the zpk to determine ring down times. 

Summary:

• In vacuum Q of 4365 (picture 2)
• Parameters: delta = 1.04e-5m, w0 = 0.0222Hz, wR/(2pi) = 0.0253Hz (which is what we expect), wg=0.0764
• Found response from both fitting the data and from zpk parameters (picture 3)
• Created new inverted response filter called SensInv_Meas (ring down plot in picture 4)
• Looking at the in air (1467529134) vs in vacuum (1468744818) performance (picture 5, with in air being green), the noise has decreased significantly
  • On a slightly related note, it seems like there is an issue with dtt displaying the spectrum at higher (>1Hz) frequencies. When plotting the same channels at the same time, and applying the response inversion filter offline in Matlab, the post inversion signal (THETA_OUT) doesn't flatten out and follows the input (THETA_IN1) (picture 6)
  • Arnaud also found that adjusting the windowing settings in dtt gives a better measurement (see SEI alog 2582) I've applied the change to the most recent CRS template.