Transcript MSC - ego
Virgo suspension control progress E. Majorana INFN Mirror Suspension Control workgroup ILIAS Geneva 29 March 2007 The usual “standard-super-attenuator” suspension … Soft isolator concept: 1. very efficient passive attenuation 2. active controls for normal mode damping EM-ILIAS-290307 2 The mission of Mirror Suspension Control workgroup: commissioning-oriented activity Virgo sensitivity (at LF !) Virgo duty-cycle In fall 2006 the majority of main ITF control issues had been addressed: - lock acquisition strategy - automatic alignment (~) -suspensions and local controls allowed all above To start noise hunting, stable operation was needed: => MSC performance started to be integrated in ITF issues EM-ILIAS-290307 3 Our plan was clear thanks to the previous efforts of MSC Workgroup, during the first part of the commissioning, and to some ideas matured meanwhile. Unavoidable acknowledgments to my friends and colleagues Giovanni Losurdo and Paolo Ruggi EM-ILIAS-290307 4 Interface with ITF Noise budget Angular control strategies (Local/automatic) Suspension control chain Beam centering (read-out,numerical,digital,actuator) NOW Disturbance rejection Sensor blending optimization Global Inverted Pendulum Control EM-ILIAS-290307 All-the-possible-handles (completion needed!) Lock force re-allocation (BS) SA-chain damping from ground Vertical damping Safe-and-soft operation interfaces 5 Focus on: - improvement of disturbance rejection - reduction of control noise through the suspension - reduction of ITF coupling with the control noise EM-ILIAS-290307 6 > disturbance rejection: vertical damping needed (1/9) V-damp ON/OFF vs alignment EM-ILIAS-290307 7 > disturbance rejection: vertical damping needed (2/9) One-By-One O-B-O OFF ALL OFF ALL ON O-B-O OFF ALL OFF ALL ON No crucial Improvement during OBO OFF Test. Clear Improvement All OFF/ON Relationship with Angular fluctuations EM-ILIAS-290307 8 > disturbance rejection: vertical damping needed (3/9) One-By-One (detailed) TOP stage Last stage ITF EM-ILIAS-290307 9 > disturbance rejection: top stage sensor blending (4/9) Winter issues: STD scheme and blending used until Dec 2006 ACC HP + cross IP LP LVDT fcrossover = 50 mHz Trade-off between: 30 mHz (wind disturbance through ACC) and 70 mHz (mseism disturbance through LVDT) EM-ILIAS-290307 10 cavities locked independently unexpected ! wind-noise 30 mHz crossover N-arm 70 mHz crossover W-arm pulling back the crossover: benefit expected mseism (sea) produces angular excitation of the payload. wind (reasonably through the tilt) fake acceleration re-injected. Key solutions : - smart experiments to simulate the disturbance ; - tools to optimize the blending on-the-fly ; - other smart ideas …. EM-ILIAS-290307 11 Sensor blending: TWO-SIDE OPTIMIZATION characterizing Suspension-control_vs_disturbance EM-ILIAS-290307 12 > disturbance rejection: top stage sensors, first attempts (5/9) old old new new hybrid filters (on-the-fly tuning) ACC HPw LPw HPms + mix cross IP + LPms LVDT mix =0.5 : “medium” attenuation of mseism noise mix =0 (wind): not worse than old filters against mseism + tilt noise attenuation below 50 mHz mix =1 (mseism) : “strong” attenuation of mseism noise with slightly EM-ILIAS-290307 worsened tilt noise attenuation. OUR TARGET How much is it enough ? The answer is worked out through a deeper analysis 13 > disturbance rejection: top stage sensors (6/9) hybrid filters in use: noise percolation paths Pitch excitation (payload mode) yaw bump (due to large zM correction) mix=1 Ineffective @ 0.45 Hz, (IP notch) Large zM due to tilt disturbance through accelerometers Calm period 3 mechanisms (at least): F0_z EM-ILIAS-290307 zCorr, F0_z qx, F0_z qy 14 > disturbance rejection: top stage sensors (7/9) WSR7, heuristic threshold for mix adjustment in-line wind (mix 0) EM-ILIAS-290307 hic sunt leones sea (mix 1) 15 > disturbance rejection: top stage sensors (8/9) LP/HP optimization, given the standard actual corrector (cross) LP/HP ratio versus CL TF mix=0.70 mix=0.90 mix=0.95 mix=1 CL TF ACC HPLP mix cross IP seism LVDT Hz V/sqrt(Hz) LVDT noise projection into zCorr (NE@step 1) Two regions can be distinguished: A) LP/HP ratio plays a role attenuating LVDT sensor disturbance (mseism) B) No effect of LP/HP attenuation Hz EM-ILIAS-290307 16 WSR7 => WSR8 (optimization example) patches added to LP to reduce mirror tilt excitation due to SA mseismic noise disturbance HP: anti-wind (tilt) previous WSR7 Hz Hz WSR8 LP: accel. comparison mms-2 Hz Hz note: cleaner (LP+HP=1) blending allows to re-tune HP to reduce tilt re-injection (=> heavier computation…) EM-ILIAS-290307 17 WSR7 => WSR8 optimization LP: anti-mseism previous WSR7 Hz Hz WSR8 accel. comparison mms-2 Hz EM-ILIAS-290307 Hz 18 WSR7 => WSR8 overall tuning (mix 0-1) LP/HP mix 1 => 0 sea => wind mix 0 => 1 wind => sea A) assessing how much it is enough to enhance LP/HP to beat tilt re-injection is not easy. B) more data needed in windy-only conditions… C) optimization not significantly efficient around the crossing frequency. EM-ILIAS-290307 19 > disturbance rejection and noise budget: 4-marionette (9/9) Marionette locking force can be distributed to both input and End mirrors Since WSR8 the lock correction is applied to four marionettes (NI,NE,WI,WE) instead of two (NE,WE): A) reduction of direct locking force budget and B) non-linear torque negligible in case of wind EM-ILIAS-290307 20 (=> reduction of large zM by a factor 2 lightens gain request to AA). Novel strategy (…one smart idea): GLOBAL OPTIMIZATION anti-mseism plus coherent wind compensation EM-ILIAS-290307 21 GIPC (Global Inverted Pendulum Control) benefits: wind disturbance rejection, lock acquisition, duty-cycle. Once the ITF is locked, the mirror position, provided by the Global Control, can be used instead of LVDTs, referred to the ground. GC (reconstructed z) Features: - automatic engagement (soft!) - anti-wind blending in NE-WE-BS-PR - anti-mseism in NI-WI Splitted anti-mseism /anti-mseism EM-ILIAS-290307 LVDT LVDT 22 Example1: NE_GIPC,WE_GIPC only step9 NE “follows” NI EM-ILIAS-290307 WE “follows” WI 23 Example 2: NE_GIPC, WE_GIPC only step9 EM-ILIAS-290307 24 Example 3: NE_GIPC, WE_GIPC, BS_GIPC, PR_GIPC step9 EM-ILIAS-290307 25 Example3: NE_GIPC,WE_GIPC only step9 Thanks to GIPC GIPC benefit to Automatic Alignment Short suspension (i.e.beam) improvement necessary ? EM-ILIAS-290307 26 A curiosity: in November GIPC had already been tested but we Nextour improvements andtostudies needed to improve knowledge and optimize the blending A major effort necessary at LF during stormy weather! EM-MSC-041206 EM-ILIAS-290307 27 Example 4:environmental (central Bld. shock absorption) EM-ILIAS-290307 28 > noise budget: actuation noise & reallocation -Suitable resistors are used to reduce coil-drivers noise (ln=lownoise/HP=highPowe to accomplish sensitivity/locking) -a trade-off with DAC noise using Emphasis/deemphasis was found. Still some work: Too large correction directly on the mirror Low frequency component that can be moved upwards (marionette) BS requires marionette reallocation. WE,NE have a mseism peak that should be reduced by further force reallocation to the marionette. EM-ILIAS-290307 29 > synthesis: net improvement of recent activity In this example: - red data (now) show up much larger mseism Much more mseism and similar peak at 130 mHz In spite of this: => the force reallocated to the marionette is smaller EM-ILIAS-290307 30 > conclusions It was mandatory to accomplish this task by considering the suspension system as a single component of the interferometer The development was a bit delayed by the requirement of checking the performance under actual environmental disturbance. Tools to emulate crucial situation have been developed. A large effort was spent on sensing optimization and on “soft operation” Now it is possible to reduce by a factor 3 the rms disturbance due to the wind, while mseism (sea) does not seem to be the main source of problems. To be done before (possibly before the MegaRun): -further technical noise reduction -improvement of short suspension performance (InjB,MC,OutB) EM-ILIAS-290307 31