Eclipse timing compares measured eclipse epochs with a predicted schedule. Small differences accumulate over many orbits, making this a sensitive probe of binary behavior. The measurement is a timing residual; identifying its physical cause requires additional tests.
Begin with a reproducible eclipse time
Observe the full ingress and egress plus enough baseline to normalize the flux. Fit an appropriate eclipse model or template rather than choosing the faintest individual data point. A flat-bottomed eclipse, finite exposure and correlated noise can make the formal fitted error too optimistic.
State which event was timed: primary minimum, secondary minimum or a specific component’s ingress and egress. In an eccentric orbit the secondary eclipse is not necessarily halfway between primaries. In an accreting system, disk light can move the apparent minimum.
Put every timing on the same convention
Exposure midpoint, clock accuracy, light-travel correction and timescale all matter. HJD and BJD describe different reference corrections; UTC and TDB describe different timescales. Combining published numbers without checking these labels can generate spurious offsets.
For observational practice, consult the AAVSO manuals. Preserve original values and document conversions so another observer can reconstruct the timing record.
A worked residual calculation
Use a linear ephemeris Tcalc(E) = T0 + E P0. Suppose T0 = 2459000.5000 days and P0 = 0.5000 days. At cycle E = 100, the predicted primary eclipse is 2459050.5000. An observed epoch of 2459050.5005 gives O − C = +0.0005 day, or +43.2 seconds.
The positive sign means late relative to this ephemeris. One late eclipse does not establish a changing orbit. A mistaken cycle number changes the interpretation entirely, especially across long gaps.
Fit the simplest adequate timing model
| Residual pattern | Initial interpretation to test |
|---|---|
| Constant offset | Reference epoch correction |
| Straight slope | Reference period correction |
| Persistent curvature | Possible period derivative or part of a longer signal |
| Repeating or irregular structure | Activity, light-travel time, dynamical effects or measurement bias |
Compare models with realistic uncertainties and examine residuals. A more flexible curve almost always fits better; that alone does not establish the extra physical mechanism.
Check both eclipse types and future predictions
The Kepler eclipse-timing study describes timing distorted light curves and identifying candidate multiple systems. Separately comparing primary and secondary events helps distinguish shared timing changes from shape-related biases.
A companion interpretation should survive new timings and dynamical checks. The study of proposed companions in post-common-envelope binaries demonstrates why long timing records and continued tests matter. A wave-shaped residual does not by itself confirm a planet.
For the physical mechanisms, read contact-binary period changes and accreting-binary period changes.
Revised October 26, 2026. The ephemeris and epochs are illustrative and use a single assumed time convention.