The X-arm interferometer test of HEPI at LIGO Livingston

Transcription

1 The X-arm interferometer test of HEPI at LIGO Livingston J. Giaime, Louisiana State University & LIGO Livingston. 1 G D, LSC meeting, LIGO Hanford, 18 August 2004.

2 Development history Decades of R&D on quiet hydraulics with Dan DeBra at Stanford, focussing on use of laminar flow oil to actuate machine tool assemblies. Recent development & prototyping of zero-stiction balanced bellows quiet hydraulic actuators, by DeBra, Hardham, Lantz et al, intended for use in Advanced LIGO pre-isolation stage. 2-DOF test stand experiment. Study by Hua et al of effective control filter techniques for sensor correction active seismic isolation at sub-hertz frequencies. Design of third-generation actuator, payload suspension springs, and external housing for HEPI by Hardham, Hammond, Mason, Kern, Lacour, etc. Tests at LASTI (ongoing) by Mason, Hardham, Coyne, Lantz, Mittleman, Ottaway, Sarin, Macinnis, etc. New safe fluid in use, tested at CIT. Re-implementation of control system and electronics for LIGO/VME environment and GDS by Bork, Sarin, Abbott(s), etc. Mass production and installation at LLO, by Kern, Abbott, Spjeld, Lacour, Traylor, Overmier, Mailand, Hanson, Carter, and many more. Hardware/software commissioning at LLO by Abbott, Traylor, Overmeir, Hanson, Fyffe, Wooley, Sellars, Parameswariah, etc. Controls commissioning/ testing at LLO by Mittleman, O Reilly, Coyne, Lantz, Giaime, Frolov, etc. 2

3 Feedback y =(I + GK) 1 GK r Feedforward command tracking + (I + GK) 1 G d d disturbance suppression (I + GK) 1 GK n. noise K ff G ff G = G d noise cancellation Active noise reduction environment sensor noise n (ω) d + + environmental disturbance d(ω) Sensor Correction M corrects error signal within servo loop command input r(ω) + measured environment feedback controller K(s) K (s) ff G (s) ff - + feed-forward controller system dynamics G(s) G (s) + + d "real" output y(ω) y (ω) m M(s) sensor correction other measurements measured output + + n(ω) sensor noise 3

4 Low-frequency pre-isolation At each tank corner pier, there is a sensor/actuator set, vertical and horizontal. Each DOF controlled with respect to HEPI displacement sensors and geophones. Displacement sensor corrected for floor motion as measured by Streckeisen STS-2., in x, y, z DOF s. payload Geo disp STS-2 geo K offload spring supports payload STS-2 displacement sensor blend and correct (Optimal FIR highpass) ground 4

5 Hydraulic bridge actuation 1. Pressure-stabilized pump. S 2 2. four-valve flow-resistance bridge. 1 pump C1 C2 3. pipes connect bridge to actuator. 3 4 R Stiction-free bellows on each side of actuated plate. 5. Actuated plate connected to payload through 1-DOF linkage. 5 5

6 Valve issues Electrically-controlled valve bridge is central to the design. Three valve-related failure modes have been observed. Gross imbalance in actuation with zero drive; may be due to particles in the fluid path or blocking the armature. In some cases this has shown to be leakage in the non-valve parts of the actuator. abnormally low gain. Not understood, but may be due to crud or particles. oscillation (due to too-high fluid pressure.) Flow in the DYP-2S Parker DYP-2S valve The new nozzle Ps nozzle C1 C2 flapper torquer motor Pr DYP-2S valve 6 original new

7 Installation and Commissioning 7

8 Pier actuation system 8

9 LVEA pump stations 9

10 Commissioning procedure 1. Manual sensor & actuator check-out, platform alignment. 2. Automated system identification of 8 input, 16 output, plant. 3. Feedback servo design and implementation for x, y, z, rx, ry, rz and two overconstrained DOFs. 4. Sensor correction sys-id, using portable witness geophones. 5. Sens. correction filter design and implementation for x, y, z. 10

11 Polyphase highpass FIR for sensor correction (W. Hua) Seismometers cannot easily distinguish between horizontal acceleration and ground tilt & thermal artifacts. Below 0.1 Hz, there is very little coherence between STS-2 seismometer signals and the LIGO detector DOFs. Challenge: low-frequency cut-off of seismometer-based sensor correction signal, to avoid tilt and thermal pickup from seismometer. This filter should roll up as steeply as possible, while allowing magnitude and phase accuracy above 0.1 Hz Hua s design implemented by R. Bork for LIGO/vx-works front end code Notch filter 10-1 In Decimation filter P-FIR Interpolation filter + Out filter gain Original FIR Interpolation Decimation Notch Compensation New FIR Compensation filter freq (Hz)

12 Vertical and transverse performance ITMX: Vertical TF from floor to crossbeam (out-of-loop) Y TF from floor to y modal L-4C signal (in-loop) 20 Z GS-13 between Piers 3 & 4 / Z STS-2, HEPI on Z GS-13 between Piers 1 & 2 / Z STS-2, HEPI on Z GS-13 between Piers 3 & 4 / Z STS-2, HEPI off Z GS-13 between Piers 1 & 2 / Z STS-2, HEPI off 10 0 Y modal L-4C signal / Y STS-2, HEPI on Y modal L-4C signal / Y STS-2, HEPI off (db) 0 (db) -10 Magnitude Magnitude Coherence Coherence Coherence Coherence

13 X, yaw and pos performance ITMX: X TF from floor to crossbeam X noise on crossbeams Magnitude (db) GS-13 on pier 1-2 crossbeam X /.STS-2 X GS-13 on pier 3-4 crossbeam X /.STS-2 X GS-13 on pier 1-2 crossbeam X /.STS-2 X, HEPI off GS-13 on pier 3-4 crossbeam X /.STS-2 X, HEPI off ) 1/2 (m/hz Magnitude GS-13 on pier 1-2 crossbeam X GS-13 on pier 3-4 crossbeam X GS-13 on pier 1-2 crossbeam X, HEPI off GS-13 on pier 3-4 crossbeam X, HEPI off ) 1/2 (radian/hz Magnitude Yaw optical table noise, from SUSYAW X optical table noise, from SUSPOS HEPI on HEPI off HEPI on HEPI off ) 1/2 Magnitude (m/hz rms, HEPI on rms, HEPI off

14 10-5 Detector disturbance levels 10-6 ) 1/2 (m/hz Displacement L- L+ l- l+ L- RMS L+ RMS l- RMS l+ RMS Data from R. Adhikari s MIT Ph.D. thesis (2004) of the LLO detector. Bulk of RMS disturbance comes from Hz band. 1 μm rms is consistent with detector operation. Also, 1 μm/s rms velocity is the practical limit for reliable lock acquisition. 14

15 X-arm length disturbance, quiet evening X-arm displacement noise with HEPI ) 1/2 ASD (m/hz HEPI on, sensor correction on ", rms HEPI on, sensor correction off ", rms HEPI off, sensor correction off ", rms

16 X-arm length disturbance, quiet evening X-arm velocity noise with HEPI ) 1/2 ASD (m/s/hz HEPI on, sensor correction on ", rms HEPI on, sensor correction off ", rms HEPI off, sensor correction off ", rms

17 X-arm length disturbance, noisy afternoon 10-5 ) 1/2 ASD (m/s/hz HEPI on, sensor correction on ", rms HEPI off ", rms Noisy afternoon of Aug 10, 2004 had a BLRMS ground velocity 1 3 Hz monitor value between the 90th and 95th percentiles. With HEPI in use, we expect the LLO detector to work on such a day, with factor of 2 headroom. 16

18 Band-limited rms velocity monitor statistics Analysis of 600+ days of BLRMS data from LIGO PEM seismometers: E. Daw et al, Class. Quantum Grav. 21, (2004) 1 3 Hz: 4 7 x higher at LLO Hz: 5 7 x higher at LLO Hz: 3 x higher at LLO. example: 1 3 Hz 90th percentile values site chan 90%, µm/s llo/lho lvea x LLO LHO lvea y ex x ey y lvea x lvea y mx x my y

19 Torture-test data Data taken during very noisy episodes during S2, when we could not reliably lock the LLO detector. RMS acceleration, velocity and displacement calculated between 20 mhz and 16 Hz tabulated, for EY Y - EX X + LVEA X - LVEA Y. Worst day that we observed, if suppressed by HEPI as currently performing, would probably permit interferometer lock. data file Displacement Velocity Acceleration Enormous µseism Day Train Borderline day 63 µm p-p 35 µm/s p-p 180 µm/s 2 p-p 11 µm rms 4.8 µm/s rms 17 µm/s 2 rms 13 µm p-p 13 µm/s p-p 150 µm/s 2 p-p 1.7 µm rms 1.6 µm/s rms 17 µm/s 2 rms 30 µm p-p 18 µm/s p-p 150 µm/s 2 p-p 4.6 µm rms 2.5 µm/s rms 17 µm/s 2 rms 18

20 What about the train? 11 AM CST, without train EY - Y Data from day with high microseism, with and without train, looking at EY Y seismometer, which bears the brunt of the train. Also, these data show a set of nightmare microseism graphs. Train vibration energy falls mainly in the 1 2 Hz band, which is reduced well by HEPI. Some falls just above the HEPI band. ) 1/2 Magnitude (µm/s/hz *T0=26/02/ :00:00 Avg=6 *BW= PM CST, with train ) 1/2 Magnitude (µm/s/hz EY - Y *T0=26/02/ :02:06 Avg=6 *BW=

21 Remaining tasks Complete basic functionality on 6 more payloads Optimized sensor gains and whitening to make saturation less likely during extreme storms. Lock/unlock scripts, interfaced with watchdog function, to automate HEPI operation. 3-stage watchdog, switches among servo & sensor correction, servo only, offset only, or HEPI off. Simplified operator s EPICS screen. 20

22 Methods for improvement Resonant gain in the geophone-based inertial-feedback controller to lower the stack mode excitation, and/or the test mass bounce mode. This is a challenge, as it makes sensor correction filter performance more sensitive to small plant changes, perhaps involving non-minimum phase zeros; we will try it of course. Control reallocation from test mass suspension OSEMS to HEPI. This will certainly be done at tidal frequencies, where the blend effects will have only a small effect on sensor correction. Adaptive sensor correction, to adjust the correction filter as conditions change. This is under study at LASTI. 21