Thoughts on short term improvements
for Mirror Suspension Control
G.Losurdo - P.Ruggi
Collaboration meeting – Cascina, Feb. 6, 2006 G.Losurdo – INFN Firenze-Urbino 2
Seismic noise vs Duty cycle
C6
C7
C6 duty cycle:89 %
C7 duty cycle:70 %
Collaboration meeting – Cascina, Feb. 6, 2006 G.Losurdo – INFN Firenze-Urbino 3
Seismic noise variability
Seismic noise may vary by ~100 at the microseismic peak
A large soil tilt can be indiced by the action of wind on the buildings
The contribution of tilt is hard to measure (no proper sensing)
Seismometers are not good below 0.1 Hz
Collaboration meeting – Cascina, Feb. 6, 2006 G.Losurdo – INFN Firenze-Urbino 4
Main goal of the recent activity
Many lock losses are associated to angular motion of the mirror
The larger the motion the wider the required control bandwidth
We have worked to reduce the residual angular motions of the mirrors along two paths:
1. Reducing the re-injected seismic noise
2. Reducing the translation-to-angle couplings
Increase the ITF robustness Reduce the control bandwidthReduce the actuation noise
Collaboration meeting – Cascina, Feb. 6, 2006 G.Losurdo – INFN Firenze-Urbino 5
Inertial damping
Inertial sensors:– DC-100 Hz bandwidth– Sensitivity: a few 10-9 m/s2/rt(Hz)
above 1 Hz
Displacement sensors:– Used for DC-0.1 Hz control– Sensitivity: 10-8 m/rt(Hz)– Linear range: few cm
Coil magnet actuators:– Linear range: few cm
Collaboration meeting – Cascina, Feb. 6, 2006 G.Losurdo – INFN Firenze-Urbino 6
Collaboration meeting – Cascina, Feb. 6, 2006 G.Losurdo – INFN Firenze-Urbino 7
Low frequency position control is needed because:– Inertial sensors do not provide DC error signal– Inertial sensors response at f<40 mHz can be spoiled by tilt
Problem: blend the sensors – dominating the tilt effect (…) – minimizing the seismic noise re-injection
Blending the sensorsx
xAccel.
LVDT0x x
dt xHighpass
Lowpass
+x
Highpass + Lowpass = 1
Collaboration meeting – Cascina, Feb. 6, 2006 G.Losurdo – INFN Firenze-Urbino 8
The seismic noise filtering depends on L(s)
The loop design is independent on the L(s) cutoff
02
aH l L x L x
s
Collaboration meeting – Cascina, Feb. 6, 2006 G.Losurdo – INFN Firenze-Urbino 9
Tilt of the IP
Ideal IP: no tilt of the top table.
In presence of tilt, accelerometer response:
Tilt induced by cradle effect is proportional to displacement:
a x g
0 0
0 0
( ) ( )
( ) ( )x
z
a x A x x B z z
a z C x x D z z
0 0
0 0
( ) ( ) ' '
( ) ( ) ' 'x x z
z x z
a x A x x B z z x A l B l
a z C x x D z z z C l D l
Collaboration meeting – Cascina, Feb. 6, 2006 G.Losurdo – INFN Firenze-Urbino 10
Cradle effect subtraction
z
x
z
x
l
l
zs
xs
a
aT
2
2
Before subtraction: (20 mHz tilt-hor crossing) 02.0|| ijt
After subtraction: (5 mHz tilt-hor crossing) 310|| ijt
Drifts on timescale of tens of minutes are of the same order
Collaboration meeting – Cascina, Feb. 6, 2006 G.Losurdo – INFN Firenze-Urbino 11
Effect of tilt and its correction
Before: IP translations are coupled to ACC DC signals
After: the effect is cancelled by proper LVDT subtraction
Before tilt subtractionAfter tilt subtraction
Collaboration meeting – Cascina, Feb. 6, 2006 G.Losurdo – INFN Firenze-Urbino 12
30 mHz crossover
All towers with 30 mHz crossover after subtraction of intrinsic tilt
Factor 10 gain achieved
Reinjected seismic noise
Collaboration meeting – Cascina, Feb. 6, 2006 G.Losurdo – INFN Firenze-Urbino 13
Test on NE/NI
Comparing the performance of different damping configs in same noise conditions:– Cavities locked, standard config on WEST cav.ty, new config. on
NORTH cavity– Compare the zCorr signals to measure the motion of the mirrors
30 mHz crossover
70 mHz crossover
Collaboration meeting – Cascina, Feb. 6, 2006 G.Losurdo – INFN Firenze-Urbino 14
Results
NORTH vs WEST with mid-intensity seism– A factor 10 gained at the microseismic peak!– Noise reinjected below 50 mHz.
Collaboration meeting – Cascina, Feb. 6, 2006 G.Losurdo – INFN Firenze-Urbino 15
Low frequency performance
Stronger wind means larger motion
(obvious, maybe…)
Just Earth shaking orcontrol noise effects?
(i.e. residual cradle effect or seismic tilt?)
Collaboration meeting – Cascina, Feb. 6, 2006 G.Losurdo – INFN Firenze-Urbino 16
IP response:
LVDT response:
ACC response:
Simple IP model
2
02 2 2 20 0
g
l x
22 20
0 02 20
a x g l
2
002 2 2 2
0 0
g
x x
Collaboration meeting – Cascina, Feb. 6, 2006 G.Losurdo – INFN Firenze-Urbino 17
Tilt or translation?
Use the IP as a seismometer (open ID)
Is it possible to understand if thelarger noise is due to tilt or translational seism?
Collaboration meeting – Cascina, Feb. 6, 2006 G.Losurdo – INFN Firenze-Urbino 18
ACC-LVDT TF: - very good coherence with strong wind- TF = 0
2, as expected
The sensors are good also at very low frequencies, (at least in strong wind conditions)
Collaboration meeting – Cascina, Feb. 6, 2006 G.Losurdo – INFN Firenze-Urbino 19
Estimate the input through the model.To reproduce the sensor output one should assume:
~102-103 m/√Hz @10 mHz in case of translation noise
~10-2-10-1 rad/√Hz @10 mHz in case of tilt noise
Strong wind
Weak wind
Collaboration meeting – Cascina, Feb. 6, 2006 G.Losurdo – INFN Firenze-Urbino 20
Model 1: translation only
measurement
Extrapolatedtranslation
Unable to fit the dip
Collaboration meeting – Cascina, Feb. 6, 2006 G.Losurdo – INFN Firenze-Urbino 21
Model 1: translation + tilt
Extrapolatedtranslation
Extrapolatedtilt
Collaboration meeting – Cascina, Feb. 6, 2006 G.Losurdo – INFN Firenze-Urbino 22
Collaboration meeting – Cascina, Feb. 6, 2006 G.Losurdo – INFN Firenze-Urbino 23
If the seism is tilt-dominated at low frequencywe are using the wrong control strategy!
The feedback will push the table in the wrong direction!
Collaboration meeting – Cascina, Feb. 6, 2006 G.Losurdo – INFN Firenze-Urbino 24
Getting rid of tilt
Achieve an angular sensor with sensitivity:
Use it:
– To subtract the tilt from the accelerometer signal
OR
– As error signal for tilt servo (need mechanics modification)
~10-9 rad/√Hz @10 mHz
Collaboration meeting – Cascina, Feb. 6, 2006 G.Losurdo – INFN Firenze-Urbino 25
ACC noise
Equivalent tilt sensitivity:6 10-10 rad/√Hz
If the seism is tilt-dominated at low frequencywe could use our accelerometer to sense it
Collaboration meeting – Cascina, Feb. 6, 2006 G.Losurdo – INFN Firenze-Urbino 26
NI NE
WE
WI
BSPR
3f
I f
I,QCARM
MICHPRCL
f
I DARM
LVDT
ACC+
+
zCorr
What can be done more?
Even with tidal control engagedLVDTs are ON
Collaboration meeting – Cascina, Feb. 6, 2006 G.Losurdo – INFN Firenze-Urbino 27
Removing local signals
LVDT
ACC+
zCorr
NI NE
WE
WI
BSPR
3f
I f
I,QCARM
MICHPRCL
f
I DARM
Use 4 locking signals for theposition control of 4 mirrors
in the beam direction
Collaboration meeting – Cascina, Feb. 6, 2006 G.Losurdo – INFN Firenze-Urbino 28
Summary
The decoupling of the detector from seismic perturbation can be pursued in two directions
1. The removal of the position sensors from the loop, replacing them with the interferometric signals
2. The correction of the seismic tilt, either by subtraction from the ACC signals or by active control of the top stage