Control strategy in the vibrtion
isolation system for KAGRA
Ryutaro Takahashi
(Institute for Cosmic Ray Research, Univ. of Tokyo
/ National Astronomical Observatory of Japan)
The 3rd Korea-Japan workshop on KAGRA
Sogang University, Seoul
21-22 December, 2012
Contents
1.
2.
3.
4.
5.
6.
Vibration isolation system for KAGRA
Control system
Low frequency disturbance
1. Seismic noise
2. Ground strain
3. Thermal drift
4. Newtonian noise
Control strategy
1. Hierarchical contol
2. Common mode rejection
3. Feed-forward control
Consideration
Summary
1. Vibration isolation system
for KAGRA
Schematic
view of
Seismic
attenuation
system
(SAS)
Top filter
[Filter0]
Pre-isolator
Inverted Pendulum (IP)
Filter chain
Geometric Anti-Spring (GAS)
Filter1 (Filter1~3 in Type-A)
Type-A/B
Bottom Filter (BF)
Payload
Intermediate Mass (IM)
Intermediate Recoil Mass (IRM)
Test Mass (TM)
Recoil Mass (RM)
Disposition of vibration isolation system
Type-A
IP + GAS Filters (5 stages)
+ Payload (23kg, cryogenic)
Type-B
IP + GAS Filters (3 stages)
+ Payload (10kg/20kg)
Type-C
Stack + Single/Double-pendulum (~1kg)
Type-A (2-layer structure)

Upper tunnel containing
pre-isolator (short IP
and top filter)
2F
7m
5m

1.2m diameter 5m tall
borehole containing
standard filter chain
1F
8m

Lower tunnel containing
cryostat and payload
Type-B

IP base is supported
by the outer frame.

Pre-isolator is the
same as Type-A’s.
Type-B payload
on rigid table
Type-C
Type-B Payload
Type-C Payload
Rigid table
Stack
2. Contol system
Pre-isolator
Type-B Payload
Linear variable differencial transformer (LVDT)
and Voice coil actuator
Secondaly coil
Primaly coil
VC
Yoke
Coil
Magnet
LVDT
Embeded LVDTactuator unit on
the inverted
pendlum.
Evaluation of LVDT. The noise level was
less then 0.1mm at 0.01-0.1Hz. The noise
level is proportional to DC offsets.
Inertial sensor
(Geophone)
Vertical
Ground 1
Ground 2
Differential
Velocity responce of L4-C geophone.
Evaluation of L4-C geophones in Kamioka.
The noise level was 5x10-11 m Hz1/2 at 1Hz.
Test of geophones and preamplifirs
in Kashiwa.
Motor slider and Hydraulic leveler
Position of the IP is
tuned by the motor
sliders to compensate
the DC component of
the feedback signal to
the actuator.
Level of the IP base is tuned by the
tripodal hydraulic piston.
Optical sensor and electro-magnetic actuator
(OSEM)
The linear range is ~ 1mm.
The sensitivity is ~ 2.5x10-10 m/Hz1/2 at 1
Hz, and ~7x10-10 m/Hz1/2 at 0.1 Hz.
Optical lever
Type-B
by K. Agatsuma
ACC, LVDT
Control of IP
(example of
TAMA)
q
ACC
Y
Actuator
PS
q
LVDT
X
Length
X
 Global control of
cavity Length
after cavity lock
 Damping of
excited torsion
mode using
Position Sensor
3. Low frequency disturbance
3-1. Seismic noise
Peterson noise model
Global high (NHNM) and
low (NLNM) noise models
represent upper and lower
bounds of a cumulative
compilation of
representative ground
acceleration power spectral
densities.
Acceleration spectrum
Dashed lines: Peterson high and
low noise models
Solid lines: noise spectral level for J. Havskov and G. Alguacil, “Instrumentation in Earthquake
IRIS station (3 components)
Seismology”, Springer, 2009
3-1. Seismic noise
Origin of seismic noise
Man made noise (Cultural noise)
•Originates from traffic and machinery with
high frequencies (>2-4Hz).
•Propagates mainly as high-frequency
surface waves which attenuate fast with
distance and decrease strongly in amplitude
with depth.
•Has a large difference between day and
night.
Wind noise
•Makes any object move.
•Usually high frequency, however large
swinging objects can generate lower
frequency signals.
Horizontal noise attenuation (dB, spectral
acceleration density) as a function of depth
and period
Ocean generated noise (microseisms, microseismic noise)
•Seen globally.
•Long period (10~16s): generated only in shallow waters in coastal region.
•Shorter period (peak~5s): generated by the superposition of ocean waves
of equal period traveling in opposite directions.
3-2. Ground strain
Tidal variation
JGW-G1000063 by A. Araya
3-3. Thermal drift
Stack isolation in Type-C system
R. Takahashi, et al : Rev. Sci. Instrum. 73 (2002) 2428
3-4. Newtonian noise
Ambient seismic waves induce density perturbations, which
produce fluctuating gravitational forces.
Naive
x»
estimation
4p G r
b ¢W
2
(2p f )
æ 1Hz ö
-9
W = 1´10 m / Hz ç
÷
f
è
ø
2
Generally
this noise is
smaller in underground.
Large amount of moving
water due to melted snow
may make effective noise
in spring around Kamioka.
Estimated Newtonian noise in LIGO (Hughes
and Thorne, PRD 58 122002)
4. Control strategy
4-1. Hierarchical control
Displacement
horizontal
vertical
Actuation
Sensing
Control
Band
Actuation
Sensing
Control
Band
Moter Slider
on IP
Offset
of VC
1/day
Moter Slider
on Filter0
Offset
of VC
1/day
Voice Coil
on IP
LVDT
<0.1Hz
Voice Coil
on Filter0
LVDT
<1Hz
Voice Coil
on Filter1-3
LVDT
0.1-1Hz
Geophone 0.1-1Hz
Global
<0.1Hz
Intermediate
Mass
OSEM
<1Hz
Global
0.1-1Hz
Test Mass
Global
1-1kHz
Intermediate OSEM
Mass
<1Hz
4-1. Hierarchical control
Angle
pitch
yaw
Actuation
Sensing
Control
Band
Actuation
Sensing
Control
Band
Hydraulic
leveler on IP
Offset
of VC
1/day
Moter Slider
on Filter0
Offset
of TM
1/day
Moter Slider
on IM
Offser
of TM
1/day
Voice Coil
on IP
LVDT
<0.1Hz
Intermediate
Mass
OSEM
<1Hz
Intermediate OSEM
Mass
<1Hz
Test Mass
Optical
Lever
<1Hz
Test Mass
Optical
Lever
<1Hz
Global
<0.1Hz
Global
<0.1Hz
Geophone 0.1-1Hz
4-2. Common mode rejection (CMR)
The
interferometer senses
not local displacement but
length between mirrors.
The mirrors move in
common mode at low
frequencies.
Microseismic noise was
reduced by the common
mode rejection in TAMA or
CLIO.
Such a reduction is not
expected in the 3-km cavity
of KAGRA.
4-3. Feed-forward control
Comparision with feed-back control
Witness
sensor is effected by the feed-back control.
Witness sensor is NOT effected by the feed-forward control.
Target
signal
Target
signal
feed-back
Filter
Witness
Sensor
feed-forward
Witness
Sensor
Filter
Target
signal
Filter
Witness
Sensor
feed-forward (offline)
5. Consideration

There are many kinds of servo loops. Control bands
are limited by the sensing noises.

A large disturbance at low frequencies should be
fedback to the upper reaches considering phase delay.

Sensing signals must be diagonalized well for
independent controls.

CMR at the low frequencies is not effective in the
length control of the 3-km cavity, it is expected only in
the center area.

Feed-forward control is useful in the cace of having
independent monitor like seismometer (seismic noise),
strain meter (tidal variation), or gravity gradiometer
(Newtonian noise).
6. Summary

KAGRA employed the SAS which consists of an
inverted pendulum and geometric anti-spring filters.

There are some kinds of displacement noise sources at
frequencies lower than 1Hz as well as observation
band.

The vibration isolation system is controled by multiple
servo loops using many kinds of sensors and
actuators.

Hierarchical control is required considering sensing
noises, CMR and feed-forward loops.