Identification and removal of glitches in GEO 600

1
Identification and removal of glitches in GEO 600

• Joshua Smith and Stefan Hild for the
  GEO600 team GEO Meeting
  Glasgow, 11.10.06

2
Overview
After start of 24/7 S5 emphasis shifted toward
data quality Main glitch identification strategy
automatic characterization, mining this,
following up to find sources

• Main Tools
• GEO Summary pages
• HACR
• Coincident analysis
• Noise projections
• Now, KleineWelle

• Glitch Families
• The Owl
• Sine-Gaussians
• Dust
• Laser frequency
• MID slow FB
• Magnetic field/mains

3
The Owl

• Narrow line(s) with (sometimes) rapidly moving
  frequency
• Is partially controllable, but not in
  well-understood way
• Can cause poor data, glitches

4
Nevermore?

• Not seen after work on the signal-recycling
  longitudinal error signal module around May
• Broken switch from one error signal to another
• This cant be the ultimate cause of the owl owl
  existed before this module

5
Owl and analysis

• No good veto found yet
• Could remove the areas of strongest owl presence
  by eye (cyan squares)

6
Sine-Gaussians

• Few per day this summer. Source unknown
• First look indicates some vetoed by null-stream
  (a combination of two GEO calibrated detector
  outputs that yields close to no GW signal)

7
Power glitches vs. dust
Particle count
Dark port DC power glitch count
Dark port DC power glitch count
Dark port DC power glitch count
Input port DC power glitch count
0 Time
m 180
8
Air Conditioning particle filters
140 (90)?20 factor 5-7
45 (25)?5 factor 5-9
(Particles/cu.ft)/1000
3 (2)?0.2 factor 10-15
A a
9
Dust made loudest glitches
Det. output BLRMS _at_ 170Hz
Time h
Time h
Before filter
After filter
10
Dealing with dust

• Statistical vetoes look promising

11
Freq. Glitch Characteristics
0s
60s

• Show up strongly in all frequency stabilization
  signals
• Couple to H via frequency noise (MIC EP)

12
Removing MC2 glitches
2kHz

• Traced origin to mode-cleaner 2 loop. Suspected
  saturation.
• June 22nd Glitches drastically improved with 2x
  reduction of RF modulation depth (compensated by
  digital gain)
• Surprise! Glitches return in July.
• Found to be a switching comparator in
  modecleaner module, fixed glitches for good by
  changing threshold for this device

June
Freq. Hz
30d
0d
13
Vetoes for frequency noise glitches

• Noise Projection vetoes (see Ajiths talk)
• Null-stream veto (Martins talk)
• Apply both, or in series?

14
Magnetic field glitches
H glitches
B glitches
Linear coupling, coherence
15
Magnetic field projection

• B-field nearly limiting noise around 100Hz?
• Coupling not yet understood
• Checks magnetometer measures B-field, IFO
  back-coupling ruled out

16
Hourly mains glitches

• Hourly series start
• 24 secs after GPS hour
• 10 secs after UTC hour

17
Mag./mains coinc. with H
Magnetometer central building 50coinc./8hrs
Mains monitor central building 30coinc./8hrs
Time-freq glitch maps
Frequency Hz
Frequency Hz
0
0
8
8
Time h
Time h
50
50
Time-shifted coincidence
0
0
1
1
-1
-1
Time shift s
Time shift s
18
Dealing with mains glitches

• Could take out seconds 20 to 100 of every GPS
  hour.
• Statistical veto
• Filtering of Mains signal proposed for future

19
MID slow path glitches

• Looks like a switch!

20
Coincident with mains glitches

• Identify point of coupling already narrowed
• Filtering mains may help elsewhere

21
Glitch rate at Start/end of lock
Glitches about equally distributed over the lock
stretches. No excess at the beginning or end (for
May)
22
HACR History Plots
May
June
July
August
23
Summary

• Owl not seen after May 28
• Sine-Gaussians seen mainly in May-June
• Dust reduced drastically after May 22 by AC
  filtering
• Laser frequency removed after Sept 7
• MID slow FB traced to after digital feedback
  last week
• Magnetic field/mains mains filtering a possible
  commissioning task
• Current HACR glitch rate is about .035Hz
• Often 5 times that in May

24
New KleineWelle G1 Triggers

• Triggers online for S5 up till September
• http//lancelot.mit.edu/cadonati/S5/online/Kleine
  Welle/
• Thanks to
• Laura Candonati, Lindy Blackburn for trigger
  generation, summary pages
• Chris messenger for segment list generation

25
END
26
Frequency stabilization