Elenna, Evan
We installed new damping filters for PR3 and SR3 M1, largely following Jeff's "Level 2" filter designs (G1400151). In general these filters give less sensor noise injection, especially around 10–20 Hz.
For all DOFs, the filters are arranged as follows:
- The new loop design uses FM1, FM2, FM5, FM9, and FM10
- The old loop design uses FM3, FM4, FM5, FM7, FM8, FM9.
Both configurations use an EPICS gain of −1 in all DOFs. If more damping is needed, the new configurations should be tolerant to a 10 dB gain increase at least, and as a last resort the old configurations can be restored as described above.
OLTFs of the old (gold) and new (red) configurations are shown for PR3. The changes were accepted into SDF for PR3 and SR3.
Step responses showed somewhat less damping than the old design, but no mode seemed egreiously underdamped. Since PR3 and SR3 are not controlled by ISC or ASC, this should not result in any loop stability issues. When doing step responses I noticed glitching whenever I sent signal to the SR3 M1 T1 coil, even with the damping loops off (DAC glitch?).
Jeff Kissel loves this. Look at that factor of 40 dB / x100 improvement in yaw at 10 Hz!
I'm also glad to see a bounce / roll filter installed as well; I didn't have that in my model.
Hopefully these are still tuned to the specific HLTS with work we did in ~2019.
I appreciate that all gains are set to -1.0 per convention.
Can you cover the details of where you've stuck the per-DOF gain scaling to match the open loop gain of the model?
Perhaps in the norm_${DOF} bank?
The norm filters in FM5 are unchanged. I tried to match the old and new OLTF gains around 3 Hz, which is roughly where we place the upper UGFs. I did this by hand-tuning the overall gain of the rolloff filters in FM1. I kept the dc/ac gains of the boosts equal to unity as best I could, and the dc gains of the elliptic filters equal to unity.
I've copied over the templates for these measurements from
/ligo/home/evan.hall/Desktop/stuff/
to a standard sharable location, committed to the SusSVN,
/ligo/svncommon/SusSVN/sus/trunk/HLTS/H1/PR3/SAGM1/Data/
2022-07-26_1820_H1SUSPR3_M1_WhiteNoise_L_0p02to50Hz_OpenLoopGainTF.xml
2022-07-26_1820_H1SUSPR3_M1_WhiteNoise_P_0p02to50Hz_OpenLoopGainTF.xml
2022-07-26_1820_H1SUSPR3_M1_WhiteNoise_R_0p02to50Hz_OpenLoopGainTF.xml
2022-07-26_1820_H1SUSPR3_M1_WhiteNoise_T_0p02to50Hz_OpenLoopGainTF.xml
2022-07-26_1820_H1SUSPR3_M1_WhiteNoise_V_0p02to50Hz_OpenLoopGainTF.xml
2022-07-26_1820_H1SUSPR3_M1_WhiteNoise_Y_0p02to50Hz_OpenLoopGainTF.xml
Of course, the amplitude and frequency dependence of excitation for this open loop gain style TF depends on the damping filter you're driving through, so I can't speak to how Evan changed this during the measurement; there doesn't seem to be two excitation options in the template. I would guess that the excitation amplitude / frequency dependence saved in the template works for the current (red trace) data, thus driving through the new "Level 2" filters. So, don't assume that the amplitudes / frequency dependence will work through the old filters; the old filters clearly have much more gain at most frequencies, so I would start by reducing the drive amplitude.