jeffrey.kissel@LIGO.ORG - posted 13:04, Tuesday 15 October 2024 - last comment - 16:07, Monday 28 October 2024(80683)
Measurements of HAM2ISI to/from SUS PR3 Sus Point and M1 Stages Successful, But Incomplete
J. Kissel WP 12140 I've completed 6 SUS + 4 ISI = 10 of 12 total DOF excitations that I wanted to drive before I ran out of time this morning. Each drive was "successful" in that I was able to get plenty of coherence between the 4 DOFs of ISI drive and SUS response, and some coherence between 6 SUS drive DOFs and ISI response. As expected, the bulk of the time was spent tuning the ISI excitations. I might have time to "finish" the data set and get the last two missing DOFs, but I was at least able to get both directions of LPY to LPY transfer functions, which are definitely juicy enough to get the analysis team started. Measurement environmental/configuration differences of the HAM2 ISI from how they are nominally in observing: - PR3 M1 DAMP local damping loop gains are at -0.2, where they are nominally at -1.0. (The point of the test.) - CPS DIFF is OFF. (needed to do so for maintenance day) - Coil Driver z:p = 1:10 Hz analog low-pass (and digital compensation for it) is OFF. (need to do so to get good SNR on SUS M1 drive without saturating the SUS DACs) Interesting things to call out that are the same as observing: - The PR3 alignment sliders were ON. P = -122 [urad]; Y = 100 [urad]. (Don't *expect* dynamics to change with ON vs. OFF, but we have seen diagonal response change if close an EQ stop. Haven't ever looked, but I wouldn't be surprised of off-diagonal responses change. Also DAC range gets consumed by DC alignment request, which is important for driving transfer functions.) - Corner station sensor correction, informed by the Bier Garten "ITMY" T240 on the ground. (the h1oaf0 computer got booted this morning, so we had to re-request the SEI_CS configuration guardian to be in WINDY. The SEI_ENV guardian had been set to LIGHT_MAINTENANCE.) - PR3 is NOT under any type of ISC global control; neither L, P, or Y. (global ISC feedback for the PRC's LPY DOFs goes to PRM and PR2.) There are too many interesting transfer functions to attach, or even to export in the limited amount of time I have. So -- I leave it to the LSC team that inspired this test to look at the data, and use as needed. The data have been committed to the SVN here: /ligo/svncommon/SusSVN/sus/trunk/HLTS/H1/PR3/SAGM1/Data/ 2024-10-15_1627_H1SUSPR3_M1_WhiteNoise_L_0p02to50Hz.xml 2024-10-15_1627_H1SUSPR3_M1_WhiteNoise_T_0p02to50Hz.xml 2024-10-15_1627_H1SUSPR3_M1_WhiteNoise_V_0p02to50Hz.xml 2024-10-15_1627_H1SUSPR3_M1_WhiteNoise_R_0p02to50Hz.xml 2024-10-15_1627_H1SUSPR3_M1_WhiteNoise_P_0p02to50Hz.xml 2024-10-15_1627_H1SUSPR3_M1_WhiteNoise_Y_0p02to50Hz.xml /ligo/svncommon/SusSVN/sus/trunk/HLTS/H1/PR3/Common/Data 2024-10-15_1627_H1ISIHAM2_ST1_WhiteNoise_PR3SusPoint_L_0p02to50Hz.xml 2024-10-15_1627_H1ISIHAM2_ST1_WhiteNoise_PR3SusPoint_T_0p02to50Hz.xml [ran out of time for V] [ran out of time for R] 2024-10-15_1627_H1ISIHAM2_ST1_WhiteNoise_PR3SusPoint_P_0p02to50Hz.xml 2024-10-15_1627_H1ISIHAM2_ST1_WhiteNoise_PR3SusPoint_Y_0p02to50Hz.xml For the SUS drives templates, I gathered: Typical: - The top mass, M1, OSEM sensors, in the LTVRPY Euler Basis, calibrated into microns or microradians, [um] or [urad]. H1:SUS-PR3_M1_DAMP_?_IN1_DQ [Filtered with the 64x filter, then downsampled to to fs = 256 Hz] - The top mass, M1, OSEM sensors, in the T1T2T3LFRTSDD OSEM Sensor/Coil Basis, calibrated into microns, [um]. H1:SUS-PR3_M1_OSEMINF_??_OUT_DQ [Filtered with the 64x filter, then downsampled to to fs = 256 Hz] - The top mass, M1, OSEM coils' requested drive, in the T1T2T3LFRTSD OSEM Sensor/Coil Basis, in raw (18 bit) DAC counts, [ct_M1SUS18bitDAC]. H1:SUS-PR3_M1_MASTER_OUT_??_DQ [Filtered with the 32x filter, then downsampled to to fs = 512 Hz] For this set of templates: - The bottom mass i.e. optic, M3, OSEM sensors, in the LPY Euler Basis, calibrated into microns or microradians, [um] or [urad]. H1:SUS-PR3_M3_WIT_?_DQ [Filtered with the 64x filter, then downsampled to to fs = 256 Hz] - The bottom mass i.e. optic, M3, optical lever, in PIT YAW Euler Basis, calibrated into mircoradians, [urad]. H1:SUS-PR3_M3_OPLEV_???_OUT_DQ [Filtered with the 64x filter, then downsampled to to fs = 256 Hz] - The ISI's Stage 1 GS13 inertial sensors, projected to the PR3 suspension point LTVRPY Euler basis, calibrated into nanometers or nanoradians, [nm] or [nrad] H1:ISI-HAM2_SUSPOINT_PR3_EUL_?_DQ [Filtered with the 4x filter, then downsampled to to fs = 1024 Hz] - The ISI's Stage 1 super sensors, in the ISI's Cartesian XYZRXRYRZ basis, calibrated into nanometers or nanoradians, [nm] or [nrad] H1:ISI-HAM2_ISO_*_IN1_DQ [Filtered with the 2x filter, then downsampled to to fs = 2048 Hz] Note: The six M1 OSEM sensors in the Euler Basis are set to be the "A" channels, such that you can reconstruct the transfer function between the M1 Euler Basis to all the other response channels in the physical units stated above. As usual the excitation channel for the given drive DOF (in each template, that's H1:SUS-MC3_M1_TEST_?_EXC) is automatically stored, but these channels are in goofy "Euler Basis (18-bit) DAC counts," so tough to turn into physical units. For the brand new ISI drive templates, I gathered: - The ISI's Stage 1 super sensors, in the ISI's Cartesian XYZRXRYRZ basis, calibrated into nanometers or nanoradians, [nm] or [nrad] H1:ISI-HAM2_ISO_*_IN1_DQ [Filtered with the 2x filter, then downsampled to to fs = 2048 Hz] - The ISI's Stage 1 GS13 inertial sensors, projected to the PR3 suspension point LTVRPY Euler basis, calibrated into nanometers or nanoradians, [nm] or [nrad] H1:ISI-HAM2_SUSPOINT_PR3_EUL_?_DQ [Filtered with the 4x filter, then downsampled to to fs = 1024 Hz] - The top mass, M1, OSEM sensors, in the LTVRPY Euler Basis, calibrated into microns or microradians, [um] or [urad]. H1:SUS-PR3_M1_DAMP_?_IN1_DQ [Filtered with the 64x filter, then downsampled to to fs = 256 Hz] - The bottom mass i.e. optic, M3, OSEM sensors, in the LPY Euler Basis, calibrated into microns or microradians, [um] or [urad]. H1:SUS-PR3_M3_WIT_?_DQ [Filtered with the 64x filter, then downsampled to to fs = 256 Hz] - The bottom mass i.e. optic, M3, optical lever, in PIT YAW Euler Basis, calibrated into mircoradians, [urad]. H1:SUS-PR3_M3_OPLEV_???_OUT_DQ [Filtered with the 64x filter, then downsampled to to fs = 256 Hz] - The ISI's Stage 1 actuators' requested drive, in the H1H2H3V1V2V3 ISI actuator basis, in raw (16-bit) DAC counts, [ct_ISIST116bitDAC]. H1:ISI-HAM2_OUTF_??_OUT [Didn't realize in time that there are DQ channels H1:ISI-HAM2_MASTER_??_DRIVE_DQ stored at fs = 2048 Hz, or I would have used those.] Note: Here, I set the number of "A" channels to twelve, such that both the ISI's Cartesian basis and the PR3 Suspoint basis versions of the GS13s can be used as the transfer function reference channel.
Thank Jeff!
You were right - this looks much more interesting than I had hoped. We'll run the scripts for the SUS to SUS TFs and put them up here, too.
Transverse to Pitch at 50 rad/m on resonance. Maybe "only" 10 when you turn up the damping to nominal? Ug.
I've also taken a look at how much the ISI moves when Jeff drives the BOSEMs on the top stage of PR3. The answer is "not very much". I've attached two plots, one for the top mass Yaw drive and the other for the top mass length drive. note - The ISI reponses need to be divided by 1000 - they are showing nm or nrad/drive, while the SUS is showing microns or microradians/drive.
So - the back reaction of the osem drives can be safely ignored for PR3, and probably all the triples, as expected. (maybe not for the TMs, not that it matters right now).
It raises 2 questions
1. How do I divide a line by 1000 in a dtt plot? (I feel so old)
2. Why does the green line (SUSPoint) look so much noiser that the cart-basis blend signals? I would expect these to look nearly identical above about 1/2 Hz, because the blend signal is mostly GS-13. The calibrations look right, so why does the TF to the GS-13 signal look so much worse than the TF to the blend output?
These plots are at {SUS_SVN}/HLTS/H1/PR3/SAGM1/Results/
2024-10-15_length_to_length_plot.pdf
2024-10-15_yaw_to_yaw_plot.pdf
I grabbed the remaining ISI drive degrees of freedom this morning, V and R. The color and strength of the excitation was the same as it was on Oct 15th, where I used the L drive excitation params for V, and the P drive excitation params for R. PR3 damping loops gains were at -0.2 again, Sensor correction is ON, CPS DIFF is OFF. PR3 alignment offsets are ON. For these two data sets, the PR3 top mass coil driver low pass was still ON (unlike the Oct 15th data), but with the damping loop gains at -0.2, there's no danger of saturation at all, and the low pass filter's response is well compensated, so it has no impact on any of the ISI excitation transfer functions to SUS-PR3_M1_DAMP_?_IN1_DQ response channels. It's only really important to have the LP filter OFF when driving the SUS. There was the remnants of an earthquake happening, but the excitations were loud enough that we still got coherence above at least 0.05 Hz. Just for consistency's sake of having a complete data set, I saved the files with virtually the same file name: /ligo/svncommon/SusSVN/sus/trunk/HLTS/H1/PR3/Common/Data/ 2024-10-15_1627_H1ISIHAM2_ST1_WhiteNoise_PR3SusPoint_L_0p02to50Hz.xml 2024-10-15_1627_H1ISIHAM2_ST1_WhiteNoise_PR3SusPoint_T_0p02to50Hz.xml 2024-10-15_1627_H1ISIHAM2_ST1_WhiteNoise_PR3SusPoint_V_0p02to50Hz.xml # New as of Oct 23 2024-10-15_1627_H1ISIHAM2_ST1_WhiteNoise_PR3SusPoint_R_0p02to50Hz.xml # New as of Oct 23 2024-10-15_1627_H1ISIHAM2_ST1_WhiteNoise_PR3SusPoint_P_0p02to50Hz.xml 2024-10-15_1627_H1ISIHAM2_ST1_WhiteNoise_PR3SusPoint_Y_0p02to50Hz.xmlToday I also gathered another round of all six DOFs of ISI excitation, but this time changing the color of the excitation to get more coherence between 1 to 20 Hz -- since this is where the OSEM noise matters the most for the IFO. In the end, the future fitter may have to end up combining the two data sets to get the best estimate of the plant. In the same folder, you'll find /ligo/svncommon/SusSVN/sus/trunk/HLTS/H1/PR3/Common/Data 2024-10-23_1739_H1ISIHAM2_ST1_WhiteNoise_PR3SusPoint_L_0p02to50Hz.xml 2024-10-23_1739_H1ISIHAM2_ST1_WhiteNoise_PR3SusPoint_P_0p02to50Hz.xml 2024-10-23_1739_H1ISIHAM2_ST1_WhiteNoise_PR3SusPoint_R_0p02to50Hz.xml 2024-10-23_1739_H1ISIHAM2_ST1_WhiteNoise_PR3SusPoint_T_0p02to50Hz.xml 2024-10-23_1739_H1ISIHAM2_ST1_WhiteNoise_PR3SusPoint_V_0p02to50Hz.xml 2024-10-23_1739_H1ISIHAM2_ST1_WhiteNoise_PR3SusPoint_Y_0p02to50Hz.xml Happy fitting!Here is the set of plots generated by {SUSsvn}/Common/MatlabTools/plotHLTS_dtttfs_M1 for the data Jeff collected on Oct 15.
(see above, the data set is in 6 text file with names like 2024-10-15_1627_H1SUSPR3_M1_WhiteNoise_L_0p02to50Hz_tf.txt (L, P, Y, etc)
These are funny looking because the damping loops are only running at 1/5 of the normal gain. This gives higher-Q peaks and less OSEM noise coupling. This is done as part of an exercise to run the detector with a combination of real OSEM signals (ie the ones here) PLUS model-based OSEM estimators. I've set the script to show all the cross terms, and these are clearly present. It remains to be seen how much the various cross terms will matter. This is the data we will use to help answer that question.
I've also attached a slimmed-down version of the cross-coupling plots which just shows the coupling to yaw. These are the same plots as above with some of the lines removed so that I can see what is happening to yaw more easily. In each plot the red is the measured cross-coupling from dof-drive to Yaw-response. For reference, these also include the light-blue yaw-to-yaw and the grey dof-to-dof measurements.
These plots and the .mat file are in the SUS SVN at {SUS_SVN}/HLTS/H1/PR3/SAGM1/Results/
2024-10-15_1627_H1SUSPR3_M1.mat
2024-10-15_1627_H1SUSPR3_TFs_lightdamping_yawonly.pdf
2024-10-15_1627_H1SUSPR4_M1_ALL_TFs_lightdamping.pdf
On a side note, the ISI to ISI TFs are not unity between 0.1 and 1 Hz. I think they should be. This is a drive from the blended input of the control loop (well, several, because it's in the EUL basis) to the signal seen on the GS-13, in the same EUL basis, converted to displacement (so it will roll off below 30 mHz, because the the realtime calibration of the GS-13s in displacement rolls off, and it has a bump at 30 Hz because this is really the complementary sensitivity, and that has a bump because of the servo bump)
But it should be really close to 1 from 0.1 to 3 Hz. The rotational DOFs (right side, red line) look pretty good, but the translations (L, V, T) all show a similar non-unity response. Jim and Brian should discuss. They look similar to each other, so maybe it's a blend which isn't quite complementary?
I've plotted the TFs from the SUSpoint drive to the M1 EUL basis TFs. Note that in the plots, I've adjusted the on-diagonal model plots to be -1 + model. The model is the INERTIAL motion of the top stage, the measured TFs all show the RELATIVE motion between the ISI and top stage. So you want to model Top/ISI - ISI/ISI or -1 + model. This is only true for the on-diagonal TFs.
The code to do this lives in {SUSsvn}/HLTS/Common/MatlabTools/plotHLTS_ISI_dtttfs_M1.m
I've attached a big set of pdfs. The cross couplings look not-so-great. See the last 5 plots for the cross-couplings of dof->Yaw. in particular, L->Y is about the same as Y->Y. (pg 22)
The pdfs and the .mat file have been committed to the SVN at
{SUSsvn}/HLTS/H1/PR3/SAGM1/Results/
2024-10-15_1627_H1SUSPR3_M1_SUSpointDrive.mat
2024-10-15_1627_H1SUSPR3_M1_ALL_TFs_lightdamping_SUSpointDrive.pdf
(Also, see in the previous comments, there was a file which I named ...PR4... this is now corrected to ...PR3... )