Angular control of Advanced Virgo suspended benches Michał Was for the DET and SBE team LAPP/IN2P3 - Annecy Michał Was (LAPP/IN2P3 - Annecy) GWADW, Elba, 2016 May 25 1 / 12
Suspended benches in Advanced Virgo Support: photo-diodes quadrants wavefront sensors cameras,... read-out electronics Michał Was (LAPP/IN2P3 - Annecy) GWADW, Elba, 2016 May 25 2 / 12
Angular control requirements - end of arm benches QPD2 beam shift/angle bench shift/angle QPD1 DC quadrant signals used to control the arm soft mode Beam angle/position bench angle/shift Losses in arm < 10 3 < 110 nrad of beam angle < 33 nrad of bench angle (3 times better than test mass local controls achievement) RMS @ 10 Hz angle 3 10 8 rad 3 10 15 rad/ Hz shift 2 10 5 m 2 10 12 m/ Hz shift requirements 3 order of magnitude easier focus on angle Mantovani 2012, VIR-0101A-12 Michał Was (LAPP/IN2P3 - Annecy) GWADW, Elba, 2016 May 25 3 / 12
Bench suspension concept inverted pendulum double pendulum LVDT local controls Pre-isolated double suspension (MultiSAS - NIKHEF) 1 inverted pendulum 2 pendulums + vertical isolation GAS blade stages local controls: 2 LVDTs in 4 corners Differential position sensor, sensitivity 1 nm/ Hz Maxwell pair of coils Magnet actuator 2 redundant signals: 1 horizontal, 1 vertical null combinations (stretching, twisting the bench) Michał Was (LAPP/IN2P3 - Annecy) GWADW, Elba, 2016 May 25 4 / 12
Bench control scheme Translation control Translation control from the top stage Angular control on the bench angular actuator & sensor between ground and bench Angular control Z Yaw - ty Y Pitch tx credit: Joris van Heijningen Michał Was (LAPP/IN2P3 - Annecy) GWADW, Elba, 2016 May 25 5 / 12
2 redundant sensors sensor gain cross calibration Apr 29 2016 20:00:00 UTC bench angle (rad/sqrt(hz)) 10 2 10 4 10 6 10 8 Pitch TX Null Vertical Yaw TY Null Horizontal Angular control Z Yaw - ty Y Pitch tx 10 3 10 2 10 1 10 0 10 1 Sensors absolutely calibrated at 10% level Assume perfect geometry (location, orientation of sensors) Cross-calibrated vertical sensors, cross-calibrated horizontal sensors Degree of freedom coupling at resonances is 10 3 horizontal-vertical coupling 10 2 geometry is not perfect Michał Was (LAPP/IN2P3 - Annecy) GWADW, Elba, 2016 May 25 6 / 12
Is an imperfect sensor geometry a problem? May 10 2016 16:00:00 UTC bench angle (rad/sqrt(hz)) 10 6 10 7 10 8 10 9 Pitch TX Yaw TY Null Horizontal Null Vertical Z/100 Y/100 Angular control Z Yaw - ty Y Pitch tx 10 10 10 2 10 1 10 0 10 1 Shifts: ground motion wrt static bench 1% coupling of ground motion to angle sensing well below the locked spectrum not a problem Null combination is close to locked spectrum electronic noise at large offsets? (> 10 6 above noise floor) Michał Was (LAPP/IN2P3 - Annecy) GWADW, Elba, 2016 May 25 7 / 12
Closed loop performance - low frequency bench angle (rad/sqrt(hz)) 10-6 Pitch - RMS 121 nrad Yaw - RMS 101 nrad specification 33 nrad 10-7 10-8 10-9 Phase (deg) Magnitude (abs) Open Loop Transfer Function 10 2 Bode Diagram Yaw - TY 10 1 Pitch - TX 10 0 10-1 10-2 180 90 0-90 10-2 10-1 10 0 10 1 A few days of work -180 10-2 10-1 10 0 10 1 Traditional approach of moving poles and zeros around until it works RMS is a factor 3-4 above specification (33 nrad) moderate microseismic Michał Was (LAPP/IN2P3 - Annecy) GWADW, Elba, 2016 May 25 8 / 12
Closed loop performance - in band bench angle (rad/sqrt(hz)) 10-12 10-13 10-14 10-15 Pitch - TX Yaw -TY specification 10-16 5 10 15 20 25 Magnitude (abs) Phase (deg) 10-2 Bode Diagram Yaw - TY Pitch - TX 10-4 10-6 180 90 0-90 -180 5 10 15 20 25 Traditonal approach: adding low-passing once the low-frequency part looks ok Good for Yaw, needs improvement for Pitch Measurement done in air with airflow shaking the tower, should be better in vacuum Specification is 3 10 15 m/ Hz Michał Was (LAPP/IN2P3 - Annecy) GWADW, Elba, 2016 May 25 9 / 12
How to get the remaining factor 3: global cost function https://git.ligo.org/rana-adhikari/moderncontrols/tree/master/optimalfeedback/globalcost Bode Diagram Magnitude (abs) 10 0 10-5 Global optimization Pitch - TX Phase (deg) 270 180 90 0-90 -180-270 -360-450 10-1 Frequency 10 0 (Hz) 10 1 Nice framework, highlights different aspects of a good loop Blindly optimize the loop using a cost function Starting from random or current filter Stuck at local minima solutions no control, just let the system free no stability, just put 3 poles at 0 Hz (phase looks good at unity gain) Cost function require lots of tuning not better than by hand filter tuning? Not a magic solution... yet? Michał Was (LAPP/IN2P3 - Annecy) GWADW, Elba, 2016 May 25 10 / 12
How to get the remaining factor 3: better sensing 7.5 m QPD2 QPD1 Optical lever with long arm Should have a 10 better sensing noise at 10 10 m/ Hz Measure directly the angle between bench and end mirror No issues with tower shaking Bench follows the mirror, signal usable only well above 1 Hz? blend the signals and increase the gain figure out the coupled 3 body alignment of 2 cavity mirrors + bench more complicated, no longer local control Michał Was (LAPP/IN2P3 - Annecy) GWADW, Elba, 2016 May 25 11 / 12
Summary Angular control Z Translation control Yaw - ty Y Pitch tx A simple angular sensing & control from the ground works Redundant sensors are useful cross-calibration understanding couplings measuring sensing noise Brute force optimization might be useful Better sensing is possible but more complicated credit: Joris van Heijningen Michał Was (LAPP/IN2P3 - Annecy) GWADW, Elba, 2016 May 25 12 / 12
( xqpd1 w QPD1 x QPD2 w QPD2 ( xqpd1 w QPD1 x QPD2 w QPD2 ) ( ) ( ) 85 1.05 10 5 x = 121 1.30 10 5 θ ) ( ) ( ) 54 2.4 10 4 x = 88 2.9 10 4 θ bench beam θ + = 0.6470θ IM 0.7625θ EM θ = 0.7625θ IM + 0.6533θ EM RMS(θ + ) 1 nrad if bench locked to end mirror θ bench = θ EM = 0.7594θ + + 0.6444θ 0.6444θ ( xqpd1 w QPD1 x QPD2 w QPD2 ) ( ) ( ) 54 9.2 10 4 x = 88 5.5 10 4 θ beam Michał Was (LAPP/IN2P3 - Annecy) GWADW, Elba, 2016 May 25 i
Locking the bench to the mirror might actually work, just amplifies the error signal for soft mode alignment Michał Was (LAPP/IN2P3 - Annecy) GWADW, Elba, 2016 May 25 ii
Free SNEB horizontal DoF - msas controlled TY (yaw) torsion pendulum resonance at 6.3 mhz resonance cross-coupling 10 3 sensing noise 10 9 m level anti-alias at 10 Hz Apr 29 2016 20:00:00 UTC bench angle (rad/sqrt(hz)) 10 2 10 4 10 6 10 8 TY X Z null 10 10 10 3 10 2 10 1 10 0 10 1 Michał Was (LAPP/IN2P3 - Annecy) GWADW, Elba, 2016 May 25 iii
Free SNEB vertical DoF - msas controlled TX (pitch) resonance at 189 mhz TZ (roll) resonance at 163 mhz TX/TZ resonance cross-coupling 10 4 TY cross-coupling is 10 2, don t know why Apr 29 2016 20:00:00 UTC bench angle (rad/sqrt(hz)) 10 4 10 6 10 8 Y TX TZ null 10 10 10 3 10 2 10 1 10 0 10 1 Michał Was (LAPP/IN2P3 - Annecy) GWADW, Elba, 2016 May 25 iv
SNEB TY angular control Angular requirements for end benches (VIR-0101A-12) 3.3 10 8 rad rms 3.3 10 15 rad/ Hz above 10 Hz Current performance on TY (yaw) - after a few hours of work filter with 10 8 gain at 10 Hz sensing resonances 10 7 rad/ Hz loop reintroduce noise at 10 15 rad/ Hz level lock RMS at 1.9 10 7 rad, factor 6 above specification excess gain below 10 mhz Magnitude (abs) Phase (deg) 10 5 10 0 10-5 10-10 180 135 90 45 0-45 -90 Bode Diagram System: "flt_ty" : 9.97 Magnitude (abs): 2.28e-08 "flt\_ty" bench angle (rad/sqrt(hz)) 10 5 10 6 10 7 10 8 10 9 May 02 2016 17:00:00 UTC TY null -135-180 10-3 10-2 10-1 10 0 10 1 10 2 10 10 10 3 10 2 10 1 10 0 10 1 Michał Was (LAPP/IN2P3 - Annecy) GWADW, Elba, 2016 May 25 v
SNEB TX/TZ angular control Angular requirements for end benches (VIR-0101A-12) 3.3 10 8 rad rms 3.3 10 15 rad/ Hz above 10 Hz Current performance on TX/TZ (pitch/roll) lock RMS at 7 10 7 rad, factor 20 above specification haven t worked on that loop yet No translation (X, Y and Z) control Ground shakes and the bench is still May 02 2016 17:00:00 UTC May 02 2016 17:00:00 UTC bench position (m/sqrt(hz)) 10 5 10 6 10 7 10 8 10 9 X Y Z bench angle (rad/sqrt(hz)) 10 5 10 6 10 7 10 8 10 9 TX TY TZ 10 10 10 10 10 3 10 2 10 1 10 0 10 1 10 3 10 2 10 1 10 0 10 1 Michał Was (LAPP/IN2P3 - Annecy) GWADW, Elba, 2016 May 25 vi