Instrument Configurations and Observing States

An interferometer is not described completely by its arm length and site. Optical hardware, control filters, calibration models, operating mode, and data-quality state determine what digital output means at a particular time. Those facts change on different timescales. A scientifically usable record therefore separates stable design principles from versioned configuration and time-dependent observing state. This page gives that record structure and the checks needed before comparing data across detectors or epochs.

Required background. The Wave Equation and Characteristics supplies phase propagation and frequency-domain transfer functions.

Helpful background. Gravitational Radiation and Waveforms fixes the strain and polarization conventions. Probability and Inference for Gravity supplies nuisance-parameter propagation and conditional inference.

Five layers of detector description

The word “configuration” is often used for several different objects. Keeping the following layers separate prevents a design paper, a control-room state, and a calibrated data product from being treated as interchangeable.

Layer	Scientific question	Typical evidence	Timescale
Reference design	What optical and mechanical topology was intended?	Design paper, as-built technical description	Years
Installed configuration	Which optics, actuators, sensors, and filters were actually in place?	Versioned configuration records and commissioning notes	Days to months
Control and observing state	Was the interferometer locked, stable, and designated for astrophysical observation?	Timestamped state vectors and segment lists	Seconds to hours
Calibration product	How were digital channels mapped to differential arm strain?	Response-function version, uncertainty model, validity interval	Minutes to epochs
Analysis selection	Which times and frequencies are usable for this analysis?	Data-release mask, veto policy, event-specific choices	Analysis dependent

The Advanced LIGO instrument paper is authoritative for its documented design and early instrument realization, not for every later observing epoch. Conversely, a current data-quality flag does not redefine the optical topology. Each source answers a different layer of the description.

From free motion to a released strain channel

At the physical level, a gravitational wave drives an effective differential arm displacement $\Delta L_{\rm free}$. Feedback holds the interferometer near its operating point, leaving a residual displacement $\Delta L_{\rm res}$ and applying a controlled displacement $\Delta L_{\rm ctrl}$:

$$\Delta L_{\rm free}(f) =\Delta L_{\rm res}(f)+\Delta L_{\rm ctrl}(f).$$

Sensing, digital filtering, and actuation convert these quantities into recorded channels. A calibration response then produces

$$\widetilde h(f;t) =\frac{\widetilde{\Delta L}_{\rm free}(f;t)}{L} =R(f;t)\widetilde d_{\rm err}(f;t),$$

for a declared response $R$, epoch $t$, arm-length normalization $L$, and error signal $d_{\rm err}$. This compact equation hides a chain of time-dependent model parameters. A data stream labeled “strain” is therefore a derived physical product, not a raw photodiode voltage.

Optical configuration shapes $R(f;t)$. Arm cavities, power recycling, signal extraction, readout quadrature, laser power, and thermal compensation affect bandwidth and quantum noise. Mechanical and control configuration add suspension resonances, feedback features, and actuation paths. The functional roles of these systems are described in LIGO Scientific Collaboration 2015, §§2–4, PDF; values used for an observation must instead come from the applicable epoch record.

Observing state is a conjunction, not one switch

“Locked” means that control systems hold the optical degrees of freedom near their operating points. It does not by itself imply astrophysical readiness. A useful observing-state predicate has the logical form

$$\mathcal O(t) =\mathcal L(t)\cap\mathcal I(t)\cap\mathcal C(t)\cap\mathcal A(t),$$

where $\mathcal L$ denotes the required lock state, $\mathcal I$ the intended interferometer configuration, $\mathcal C$ valid calibration, and $\mathcal A$ authorization for astrophysical observing. Data-quality conditions then refine $\mathcal O$ for a particular search or measurement.

This separation matters during transitions. A detector can remain optically locked while calibration is temporarily invalid, can be calibrated while a hardware injection is active, or can produce usable data outside an official observing segment for a specialized instrumental study. “Usable” must always be completed by “for which analysis?”

The public O4 documentation describes released channels, state vectors, calibration information, and caveats for that release. The statements on this page are frozen to the GWOSC O4 technical details available on 2026-08-18. They do not assert a detector’s live state after that cutoff.

A minimum reproducible configuration record

For each analyzed detector interval, record at least:

• detector identifier, geocentric geometry, and timestamp convention;
• dataset and release version, GPS start and stop, sample rate, and file checksum;
• observing/state-vector requirement and any excluded subsegments;
• calibrated strain channel, response or calibration version, validity interval, and uncertainty representation;
• optical or control configuration only to the granularity needed by the claim;
• frequency band, line list, conditioning operations, and data-quality/veto policy;
• software environment and the code or query that reconstructs the selection.

The record should be proportional to the inference. A sky-localization calculation needs timing and calibration phase; a narrow spectral search needs line and frequency-dependent state information; a broad duty-cycle statement needs a precise segment definition but not every servo coefficient.

A compact machine-readable representation can be thought of as

$$\mathscr C_I =\bigl(D,V,[t_0,t_1),f_s,H,R,Q,P,E\bigr)_I,$$

where $D$ is dataset, $V$ release version, $f_s$ sample rate, $H$ strain channel, $R$ calibration product, $Q$ quality selection, $P$ processing recipe, and $E$ software environment. The tuple is an analysis input, not a substitute for the source metadata from which each field was obtained.

Comparing configurations without conflating causes

Suppose two epochs have amplitude spectral densities $A_1(f)$ and $A_2(f)$. The ratio

$$\mathcal R_A(f)=\frac{A_2(f)}{A_1(f)}$$

is a useful observational summary only after matching estimator, units, window, segment length, frequency resolution, and data-quality selection. A change in $\mathcal R_A$ is not automatically attributable to an advertised hardware upgrade. Environmental conditions, commissioning state, calibration, nonstationarity, and selection effects can also change the spectrum.

A reproducible comparison should therefore include three levels of evidence:

1. an official configuration record establishing what changed;
2. a like-for-like spectral or transfer-function comparison with uncertainty and matched processing;
3. an independent physical check, such as a measured injection transfer, known line frequency, or limiting response.

For public data, pin the exact files and metadata through the GWOSC API documentation, version available 2026-08-18. Repeat the comparison under reasonable changes of segment length, spectral averaging, and quality mask. If the conclusion changes, report that dependence rather than selecting one processing choice.

Common pitfalls

Using a design sensitivity as measured performance. A design curve is a target under stated assumptions. Measured spectra require dated data, estimator details, and quality selection.

Equating lock with science readiness. Lock is necessary for normal operation but not sufficient for valid calibrated astrophysical data.

Assigning a spectral change to one subsystem from chronology alone. Temporal association suggests a hypothesis; a coupling measurement or controlled configuration comparison is needed for attribution.

Exercises

1. Diagnose an incomplete observing claim. A report states, “Detector X was locked from GPS $t_0$ to $t_1$, so the entire interval is analysis ready.” Give three independent ways the conclusion can fail and the evidence needed to repair it.

Solution

First, the lock may not be the designated astrophysical observing configuration; check the configuration and observing-intent state. Second, calibration may be invalid or outside its stated epoch; check the response version, validity interval, and calibration state. Third, acquisition, environmental, or hardware-injection flags may exclude subsegments; reconstruct the analysis-specific segment intersection. The corrected claim names the data release and defines the precise conjunction of states rather than using lock alone.

2. Make a fair spectral comparison. Two public strain files have sample rates $4096\,\mathrm{Hz}$ and $16384\,\mathrm{Hz}$. State a procedure for comparing their noise between $30$ and $1000\,\mathrm{Hz}$.

Solution

Choose equal-duration, analysis-ready intervals and verify compatible calibration metadata. Either estimate both spectra at their native rates with the same physical segment duration, window, overlap, averaging rule, and final frequency grid, or anti-alias filter and downsample the higher-rate file before applying one common estimator. Use identical line handling and report amplitude spectral density in $1/\sqrt{\mathrm{Hz}}$. Vary segment duration and quality cuts as robustness checks. Directly decimating without an anti-alias filter or comparing unmatched estimators would confound configuration with processing.

References

• Gravitational Wave Open Science Center. “API Documentation.” Version available 2026-08-18. Official documentation.
• Gravitational Wave Open Science Center. “O4 Technical Details.” Version available 2026-08-18. Official documentation.
• LIGO Scientific Collaboration. “Advanced LIGO.” Classical and Quantum Gravity 32 (2015): 074001. DOI. Open PDF.