Simulation Study for

Cross-Talk Noise
between Two Detectors of LCGT
on Detection of GW
Nobuyuki Kanda
Osaka City University
for the LCGT collaboration

The 9th Gravitational Wave Data Analysis Workshop

12/16/2004, Annecy, France
 2

Plan of Talk

Brief Introduction to LCGT

Motivation
Coincidence and Cross Talk Noises
Simulation
Model of Cross Talk Noises
Generation of Noise
Process Noise by Matched Filter for GW event search
Evaluate the fraction of Accidental Coincidence
 3

LCGT
Large-scale Cryogenic Gravitational Telescope
(near!) Future Plan in Japan

Cryogenic interferometer
Mirror temperature: 20K
• Reduce thermal noise
Underground site
Kamioka mine,
1000m underground
• Reduce seismic noise
stable operation
Large-scale interferometer
Two interferometers
Baseline length: 3km
High-power laser
• Better sensitivity
Fake-reduction
with coincidence
 4

LCGT design sensitivity

LCGT design
-14 TAMA dt9
10 binary inspiral: 1.4-1.4 Msolar, 200Mpc
-15
BH ringdown: fc=100, Q=20 (=100Msolar)
10
-16
10
-17
10
-18
10
-19
10
-20
10
-21
10
-22
10
-23
10
-24
10
-25
10
1 10 100 1000
frequency [Hz]

Single detector of LCGT will reach to 200Mpc for binary inspiral.

 5

LCGT Board
Institute for Cosmic Ray Research, University of Tokyo (ICRR)
High Energy Accelerator Research Organization (KEK)
National Astronomical Observatory JAPAN (NAOJ)
Department of Physics, University of TOKYO
Department of Advanced material Science, University of TOKYO
Earthquake Research Institute, University of TOKYO
National Institute for Advanced Industrial Science and Technique
Kyoto University
Osaka University
Osaka City University
Communication Research Laboratory
Electrical Telecommunication University
Waseda University
Niigata University
Ochanomizu University
 !400. The 400 time delays are chosen from "12000 sec to 6
li
Reason of two detectors : 12000 sec in increments of 60 sec. The distribution of acci-
ci
dentals is shown in Fig. 12. In Table IV, we also list the
Coincidence for Fake Event Reduction expectation values of the number of accidental coincidence
th
and the standard deviation after each selection procedure. As
th
can be seen from this, the number of coincident events after
Fake events due to noise from complete independent detectors
each selection procedures will bewith the expected !1
is consistent gr
accidental coincidence only. number of accidental coincidences within the statistical fluc-
ev
tuations. Thus, we conclude that no statistically significant co
signals of real coincident events are observed in our search.
#
Requirement of the coincidence between GW candidates ta
triggers
---> Reject Fakes

example:
TAMA -LISM(Kamioka20mIFO)
Reduction Rate ~ 10-4

FIG. 10. ( # TAMA / !$ TAMA

2
, # LISM / !$ LISM
2
) scatter plots. The
crosses !&" are the events survived after the time selection, and the
circled crosses ( ! ) are the events survived after the time, mass and tim
amplitude selections. an
H.Takahashi et. al., PRD70, 042003 042003-1
 7
Possible Problem:
Cross Talk between two Detectors
However, closed detectors have
common source of noise,
cross talk of electric signal
mechanical coupling
--> cross talk (common component) of noise be contains in the data.

Schematic detector 1
laser optics servo signal transfer, DAQ

vacuum anti-vibration

external noise source

(seismic, acoustic, etc.) “cross talk”

“common” noise Noise due to instruments

of one detector
detector 2 (electric spike, etc.)
laser optics servo signal transfer, DAQ

vacuum anti-vibration
 8
Spike like noise cross talk

--> generate fake which amplitude be proportional to the coupling

simulation example
!"

Matched Filter
Output (SNR)
Matched Filter 4,5+.6/78+98+/:/%&';+<
/=.+.2+76/$

detector 1 signal (example: !# /=.+.2+76/!/;$#>/26711/+?5@/7A/19,@./B7,1.<

TRIVIAL!
ringdown search)
$"

%&'
detector 2 signal Matched Filter $#

(example:
ringdown search) "

#
*(" )(# )(" $#(# $#("
+,-./01.23
time [sec]
• However, it is possible to reject with the requirement of GW amplitude consistency.

• Even same spike in electric raw data, it might be not equal in h(f) or h(t).

Stationary bulk components of cross talk /common source of noise

--> How appear after event selection ?

inspiral GW search, Black-Hole ringdown GW search : easy to understand,

but the quantity of fake coincidence is not trivial after Matched filtering.

Aim of this study: to evaluate the influence of cross talk noise in GW

event search (= check the statistical characteristics of cross talk.)
 9

Simulation of Cross Talk Noise

Noise model:
two detectors, each signal s1 and s2
bulk noise cross talk (stationary noise) s21 = s21,independ + s2common
all frequency band
common only seismic noise s22 = s22,independ + s2common
2
scommon
cross talk :R =
s2
spike noise
s1 = s1,independ + sspike
coupling factor : R
s2 = s2,independ + R sspike

Monte-Carlo
use LCGT design spectrum
generate noise in Fourie domain
amplitude <-- design average, PDF
phase <-- random
 10

LCGT expected power spectrum : Sh(f )

Randomize (amplitude<- spectrum, phase<-uniform)

S̃1,independ (f ) S̃common S̃2,independ (f )

√ √
S̃1 = 1 − R S̃1,independ + R S̃common

√ √
S̃2 = 1 − R S̃2,independ + R S̃common
 11

example of simulated noise (Bulk)

-15 -15
10 10
LCGT design LCGT design
-16 simulated noise -16 simulated noise
10 10
common component (30 %) common component (30 %)
-17 -17
10 10

-18 -18
10 10

-19 -19
10 10

-20 -20
10 10

30% smic
sei
-21 -21
10 10

frac noise
-22 -22
10 10

-23 -23

tion
10 10

-24 -24
10 10

of
-25 -25
10 10

1 10 100 1000 1 10 100 1000

frequency [Hz] frequency [Hz]

All frequency region Seismic component only

 12

Matched filter processing for simulated noise

detector 1 signal : s1(f) detector 2 signal : s2(f)

Matched Filter !
inspiral GW h̃∗ (f ) · s̃(f ) i2πt Matched Filter
(h, s) = e df
ringdown GW
Sh(f )
! √
SNR(t) = (h+ , s)2 + (h× , s)2 / 2

SNR1(t) SNR2(t)

Exceed YES YES Exceed

Threshold ? Threshold ?
AND

NO
NO

Accidental Coincidence
Check the fraction of accidental coincidence for the variance of cross talk amount !
 13

Example of simulation: Accidental coincidence of Fake Event

complete independent two detectors

5 5

4 4

3 3

SNR2
2 2

1 1

0 0
0 4 8 12 0 1 2 3 4 5
3
x10

12

0
0 1 2 3 4 5
! √
SNR1 = (h+ , s1 )2 + (h× , s1 )2 / 2
 14

Example of simulation: Accidental coincidence with Huge cross talk

cross talk with 80% (so much!)

5 5

4 4
Contamination
by cross talk
3 3

SNR2
2 2

1 1

0 0
0 4 8 12 0 1 2 3 4 5
3
x10

12
Take note:
8 Each PDF of SNR looks same as the
4 case of no cross talk !
0
0
!
1 2 3 4
√ 5

SNR1 = (h+ , s1 )2 + (h× , s1 )2 / 2

 15

Note:

fraction for the case of independent detectors (analytic expectation)

3.5

-3 -6
3.0

1.93 x 10 3.72 x 10
2.5

2.0

1.5

1.0 -3
0.5
1.93 x 10
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
 16

Results 1: Cross talk VS contamination

-3
10 8
threshold: 2.5 sigma for each detectors
6
inspiral 10-10 Msolar
4 inspiral 1.4-1.4 Msolar
BH ringdown fc=20Hz, Q=20
2

-4
10 8
6

contaminate
4

-5
10 8
6
4
independent two det.(analytic)
2

-6
10

0 10 20 30 40 50 60 70
cross talk (common component) [%]

30% cross talk complicate 10 times accidental coincidence.

 17
Results 2:
Seismic component cross talk VS contamination
Seismic component is biggest concern.
inspiral search : assume larger mass source -> integrate from flow < 10Hz
BH ringdown : source near seismic cutoff
-3 threshold: 2.5 sigma for each detectors
10 8
6 All frequency region:
4 inspiral 10-10 Msolar
inspiral 1.4-1.4 Msolar
BH ringdown fc=20Hz, Q=20
2

Seismic component only:

-4
10 8 inspiral 10-10 Msolar
BH ringdown fc=20Hz, Q=20
6
inspiral 1.4-1.4 Msolar
4

-5
10 8
6
4

-6
10
0 10 20 30 40 50 60 70
cross talk (common component) [%]

If the cross talk is seismic only, no significant contamination !

 18

Summary and Future

We tried practical study of cross talk between two LCGT detectors.
• Even stationary bulk noise, 30% cross talk contaminate 10 times for
accidental coincidence.
• Seismic component looks as no problem. GW source

inspiral 1.4-1.4 inspiral 10-10 BH ringdown

Msol Msol (20Hz,Q2=0)

all frequency
30% -> x 10 30% -> x 10 30% -> x 10
band
seismic
cross talk model no effect no effect no effect
component only
proportional to proportional to proportional to
spike
cross talk cross talk cross talk

Future and More...

• Consider instrumental mechanism more (e.g. up-conversion of seismic
trough the servo system, scattering light, etc.)
• Study for stochastic GW