3.4 LIGO and Virgo detectors’ noise
3.4.1 Gaussian versus non-Gaussian noise
The noise curves presented in the previous section are the result of time-averaging the noise and do not represent the non-stationarity of the noise at either long or short timescales. The long-timescale non- stationarity is evident in the binary neutron star range as a function of time, Figure 3.4. The main sources of short-timescale non-stationarity are laser intensity fluctuations and alignment issues, which generally mani- fest as low-frequency non-Gaussian noise. On the other hand, the quantum noise at higher frequencies tends
to be more Gaussian. Here, we distinguish between two types of non-Gaussianity. The first are excursions
101 102 103
10−20 10−19 10−18 10−17 10−16 10−15
Frequency [Hz]
Displacement [m/!Hz]
H1 (DC) at 2010−07−21 11:50:00, (963748215)
UGF = 186Hz, Req. PWR = 20 W, range: 25.8 MPc pred, 19.2 MPc meas
created by makeNoisePlot on 22−Jul−2010
Shot Dark Intensity MICH PRC BS ETM ITM ASC OpLev OSEM Seismic RadPress IntTherm SusTherm totalnoise SRD DARM
Figure 3.17: The noise budget for H1 during S6. The shot noise is due to the Poisson fluctuations in the laser light hitting the anti-symmetric port’s sensing photodiode. The dark noise is the noise that is measured on the photodetector when the laser is turned off; it is due only to the electronics themselves. The intensity noise is due the fluctuations in the laser intensity, whose power emitted is nominally 20 W. The MICH noise is from the control signal that keeps the anti-symmetric port dark. Similarly, the PRC noise is from the control signal that keeps the laser resonant in the power-recycling cavity. The BS (beam splitter), ETM (end test masses), ITM (input test masses) and ASC (angular-sensing and control) noise is residual noise from control systems that monitor and control the mirrors’ positions and orientations. The OpLev noise is from the optical lever servo, which senses and controls the mirror angular positions (pitch and yaw). The OSEM noise is from the optical shadow sensor and magnetic actuator, which locally damp the pendulum motion of the mirrors. The seismic noise is due to a variety of sources that produce displacement noise at the mirrors (ITMs and ETMs).
The IntTherm noise is the thermal noise internal to the test masses themselves. The SusTherm is the thermal noise in the suspension wires at the violin mode frequencies of ˜340 Hz and harmonics; it also includes the pendulum mode at 0.74 Hz (off the scale of this plot) and1/f2falloff. The totalnoise curve is the sum of all the listed noise sources (which were already transformed into displacement noise), added in quadrature.
The DARM curve is the total noise measured at the anti-symmetric port; the gap between the DARM curve and the total noise curve, especially noticeable below 60 Hz, is not quantitatively understood. The SRD is the strain sensitivity goal listed in the science requirements document [14], presented to the National Science Foundation in 1995.
of 3 to 5σfrom the mean. These alone are not problematic, because they are rarely found in coincidence between detectors. The second are the extremely non-Gaussian excursions of many moreσfrom the mean;
these are the glitches that really limit our sensitivity because they need only be found in coincidence with a 3 σexcursion in another detector. The next chapter goes into the causes of several types of glitches, as well as methods used to veto them.
Chapter 4
Glitches, their effect on data quality, and the need for vetoes
As discussed briefly in Section 1.3, glitches are a problem for the high-mass search because they cause events with a large signal-to-noise-ratio to be found by the matched-filter algorithm. These events not only obscure potential astrophysical GW events, but also would lower our detection confidence in a true event.
Of course, the effect is not limited to the high-mass search; glitches present problems for every search done by LIGO and Virgo. This chapter discusses glitches and glitch-finding algorithms in general, gives specific examples of glitches in LIGO S6 data, and explains the traditional methods of mitigating the effect of glitches on a search like the high-mass search. There are two main titles given to research in this realm —data quality anddetector characterization. Although much of their work overlaps, they can be distinguished by the direction the information learned travels — data-quality information tends to go downstream to astrophysical search pipelines, while detector characterization information tends to go upstream to detector commissioners.
4.1 Glitches and glitch-finding algorithms
Glitches are short duration events recorded by the GW channel that can be attributed to an environmental or instrumental disturbance and, as such, we are confident they are not GWs. For the high-mass search, for example, glitches are spurious events that are picked up by the matched-filter algorithm, which compares the data in the GW channel to short-duration templates that model waveforms from high mass binary black hole coalescence. Therefore, glitches with duration and frequency content comparable to the the waveforms listed in Table 2.3 cause the most difficulty for the high-mass search, because the high-mass templates are so short that aχ2test does not work well (see Section 7.3.6 for an in-depth discussion of this test).
Glitches can be identified with various algorithms. Two such algorithms were used in this thesis; they are described in the following subsections. The shared goal of these algorithms is to do a fast transform in a wavelet basis that is essentially performing a matched-filter for shapes that look like glitches. These glitches can be found not only in the GW channel, but also in the auxiliary channels; Figure 4.1 shows a glitch in
one of the auxiliary channels. Like many glitches it can be characterized as a ringdown — an abrupt change followed by a decay described by normal modes.
Figure 4.1: An extremely loud glitch seen in an auxiliary channel recording the sum of the photode- tectors in the output mode cleaner. Image courtesy of the Detector Characterization group Wiki page https://wiki.ligo.org/DetChar/OMCSpikeGlitches. Note the characteristic ringdown shape.