stefan hild 1ilias wg1 meeting, sep 2005, perugia title geo 600 commissioning progress...

Stefan Hild 1 Ilias WG1 meeting, Sep 2005, Perugia

Title

GEO 600Commissioning progress

Max-Planck-Institut für Gravitationsphysik(Albert-Einstein-Institut)

Stefan Hild Ilias WG1 meeting, Sep 2005


GEO 600 layout

1.5-2 kW

0.09


Tuning signal recycling to 300 Hz

Better knowledge of IFO parameters:

more accurate prediction from simulations

downtuning to 200 Hz realized

Fixed instability of MI auto-alignment by setting up a new telescope for DWS.

We will stay at 330 Hz !


GEO 600 design sensitivity


Calibration and demod phase @ 300 Hz

Using signal recycling

GW signal is split in P and Q quadrature

P-response independant of demod phase

Q-response only for larger than 20 to 30 deg demod phase compatible with current calibration(zero + complex pole)

Decided to use 40 deg for time being.

Drawback: ‚P‘ used for MI servo is not really only ‚P‘


High RF on HPD / Q-compensation

Observations:

• Found huge RF level after MI diode (1 Volt)

• Large DC level in Q after mixing.

Q-compensation:

Injected RF directly into resonance circuit of the HPD to make Q DC level zero.

Improved sensitivity above 700 Hz

Q-comp OFFQ-comp ON


RF setup

OLD setup

New setup:

• Using a less noisy splitter(0/90°)

• Using a cable as phaseshifter

NEW setup

Old setup:

• Dominated by phase noise in the LO path (phaseshifter and splitter(0/90°)

• Due to this noise from Q was mixed into P quadrature.

• That is why Q-compensation worked.


Shot noise

Green line is the shot noise level calculated from the DC current of the photodiode.

We are within a factor sqrt(2) shot noise limited above 1kHz !


Future RF setup

Proposed by Rana and others

GOALS: • Using only high quality hardware • Guaranteeing high RF levels in all components


Noise projections 21st of july


Noise in the SR-long loop

Noise in the detection band is entirely limited by front-end-noise (shot noise from the detector).Limits sensitivity from 100 to 500 Hz

PROBLEM

• Large locking range• OLG of 10^6 at 0.1 Hz• No feedback noise above100 Hz

Requirements

1. Servo has a locking mode and a running mode2. Lowering UGF to get less gain in detection band3. Using lowpass filter with very steep roll off above UGF

(implemeted using a dSPACE system)

Solutions


Analog SR electronics

Red = acq amp, green = acq phaDark blue = run amp, brown = run phaseLight blue = run2 amp, pink = run2 pha


Signal Recycling digital

Phase @18 Hz = -147 deg


SR feedback: analog vs. digital


SR loop filter future design

Bode Diagram

Frequency (Hz)

Phase (

deg)

Magnitude (

dB

)

10-3

10-2

10-1

100

101

102

103

-360

-270

-180

-90

0

90

180

270

To:

long s

tage 3

-150

-100

-50

0

50

100

150From: Input Point

To:

long s

tage 3

Filter not used at the moment. Can be implemented when more gain is needed around pendulum resonances

Four complex integrators:2.25 (2x),1.3, 0.56 Hz


Sensitivity improvement July to August

High frequency improvements: reduction of RF phase noiseLow frequency improvements: reduction of Signal recycling feedback noise


PR-bench

Historically grown layout of PR-bench not optimal !

• beam clipping• too many

transmissive optical components

• too many polarizing components

• acoustic coupling

Goal:Shorten and simplify the HPD path


Sensitivity after PR bench work

Removed resonance structures between 100 and 200 Hz.


Noise projections 21st of july


Michelson length control

< 0.1Hz


< 10 Hz

< 0.1Hz


Reaction Pendulum:

3 coil-magnet actuators at intermediate mass, range ~ 100µm


< 10 Hz

> 10 Hz

< 0.1Hz


Reaction Pendulum:

3 coil-magnet actuators at intermediate mass, range ~ 100µm

Electrostatic actuation on test mass bias 630V, range 0-900V= 3.5µm


Using ESD as actuator

• Force is proportional to square of applied voltage.

• high voltages are needed

• for bipolar acting a bias voltage needs to be applied

Noise is introduced:

• loop electronics

• sqrt circuits

• intrinsic HV amplifier noise


Sqrt circuits in MI loop

ESD: F U^2

Sqrt circuits are necessary to give full linear force range for acquisition.

Drawback: sqrt circuits are noisy 1µV/sqrt(Hz)(=100µV/sqrt(Hz) @ ESD)


ESD: F U^2

Sqrt circuits are necessary to give full linear force range for acquisition.

Drawback: sqrt circuits are noisy 1µV/sqrt(Hz)(=100µV/sqrt(Hz) @ ESD)

Bypassing sqrt circuits after lock is acquired.

Sqrt circuits in MI loop


Noise in MI loop

HVA noise = 100nV/sqrt(Hz) (=10µV/sqrt(Hz) @ ESD)

HV-amplifier noise can be reduced by decreasing bias voltage or active noise suppression.

Suppressing noise introduced by loop electronics needs whitening


MI loop whitening / dewhitening

dewhiten

dewhiten

dewhiten

Whiten

Whitening right after mixer:zero 3.5 Hzpole 35 Hz

Dewhitening for both split passes

Passive dewhit-ening done in HV path (0-1kV)


Sensitivity after fixing noise from MI loop


Michelson servo design


MI loop gain problem

Strange observations:

• We are not able to increase the low frequency loop gain, even though it should work from loop model (oscillation around 8 Hz)

• OLG measurements show that there is nearly no gain around 8 to 10 Hz.

• We needed to turn down the MI crossover gain continiously for the last few months

Findings:

• Influence of MI AA gain on maximum cross over frequency (AA-tilt , cross over possible.

• phase margin of crossover, can crossover frequency, but adding integrator still causes 8Hz oscillation

• IM gain ( crossover frequency), less noise in servo 4 to 8 Hz.

• IM gain, still low gain around 8 to 10 Hz (low gain can‘t be caused by IM)


Injecting LF noise in MI loop

Injected noise from 5 to 11 HzBetween 6 and 7 Hz the noise is not suppressed !


Measurement of open loop gain (PRMI)

Resonance structureis clearly visible

Very low gain around 9 Hz

Design gain: 7 @ 15 Hz 20 @ 10 Hz 80 @ 6 Hz


Mi servo design detail


TF: Alignment to longitudinal

Tilt couples a factor of 10 stronger than rot @ 10 Hz.

(IM-FF tilt2long)

Why are the two rot-TF so different?


How to go on with the MI loop gain problem

Measure MI loop gain for various conditions to decouple crossover and AA-tilt.

Implementing digital AA control to be more flexible (nearly done).

Measure tilt2long for single suspensions, with the goal of setting up a FF system.

Investigate possible advantages of using the ESDs for angular alignment.

....


Segmented ESD for alignment


Sensitivity improvement of GEO

102

103

10-21

10-20

10-19

10-18

10-17

Frequency [Hz]

h(t

) [H

z-1/2

]

Jan 02

Aug 02 (S1)

Jan 04 (S3)

Sept 05

Feb 05 (S4)

Aug 04


Latest noise projections


Current sensitivity vs. Design sensitivity

102

103

10-24

10-23

10-22

10-21

10-20

10-19

10-18

10-17

Freq. [Hz]

ASD

[h/ H

z]GEO600 Theoretical Noise Budget

SeismicSuspension TNSubstrate TNCoating TNThermorefractiveShot 350HzTotalShot Sept 1 h(t) Sept 1


Discussion

Michelson loop gain problem:

• What experience exists in VIRGO with angular to longitudinal and longitudinal to angular couplings ?

• Is somewhere gain lost ?

• ...

Digital filtering:

• What kind of filters are used within VIRGO for steep roll off?


E n d

stefan hild 1ilias wg1 meeting, sep 2005, perugia title geo 600 commissioning progress...

Documents