An O–C diagram plots an observed event time minus the time predicted by a chosen ephemeris. It is a way to inspect whether a repeating star or binary system keeps to an assumed schedule. Positive values mean late arrivals; negative values mean early arrivals, relative to that schedule.
The most useful distinction for a beginner is between using the wrong constant period and measuring a period that actually changes. Both produce a trend, but not the same kind of trend. Work through the example before interpreting a curved line as evidence for a companion.
Build the prediction first
A simple linear ephemeris is C(E) = T0 + P × E, where T0 is a reference event time, P is the assumed constant period, and E is an integer cycle number. For each measured event, calculate O–C = O(E) − C(E).
Use consistent time units and time conventions. Primary and secondary eclipses are different event sequences; do not silently mix them into one sequence. The event being timed, the cycle count, and any conversion between time systems all belong in the record.
Example A: a slope without a changing period
Invented teaching data. Set T0 = 0 day and predict a period of 1.000 day. Suppose the true period stays constant at 1.002 days.
| Cycle E | Calculated C, days | Observed O, days | O–C, days |
|---|---|---|---|
| 0 | 0.000 | 0.000 | 0.000 |
| 10 | 10.000 | 10.020 | +0.020 |
| 20 | 20.000 | 20.040 | +0.040 |
| 30 | 30.000 | 30.060 | +0.060 |
The residual rises by 0.002 day per cycle. That is 2.88 minutes per cycle. Updating the assumed period from 1.000 to 1.002 days removes this idealized slope. The star did not need a changing period to make the original O–C trend rise.
Example B: changing increments
Now consider hypothetical residuals at consecutive cycles of 0.000, 0.001, 0.004, and 0.009 day. The differences between successive residuals are 0.001, 0.003, and 0.005 day. Unlike Example A, the extra delay added each cycle is not constant. This is the intuition behind curvature: the local mismatch with the reference period changes.
Four invented values demonstrate arithmetic, not a statistically established period derivative. Real interpretation requires a longer baseline, uncertainty estimates, and tests of alternatives. Part of a long oscillation can look like a parabola if the observations cover only a short interval.
Read patterns as questions to investigate
| Pattern | First interpretation to test | Important caution |
|---|---|---|
| Flat, with an offset | The reference epoch may be offset | Noise and sparse sampling can hide changes |
| Straight slope | The assumed constant period may be wrong | A slope alone does not require period evolution |
| Curvature | The period mismatch changes over time | A short baseline does not establish the physical cause |
| Wave-like residuals | Test a periodic timing modulation | A companion is one hypothesis, not the automatic answer |
| Sudden jump | Check cycle numbering, timing conventions, and event identification | Do not infer an abrupt physical change before checking the record |
What the diagram leaves out
An ordinary light curve plots brightness against time. An O–C diagram uses event times extracted from such observations. It therefore compresses information. A distorted eclipse profile, starspots, incomplete coverage, or a change in the fitting method can shift an estimated minimum without representing the same physical change in the orbit.
A periodogram asks which frequencies describe the measurements. A phase-folded plot stacks observations using an assumed period. An O–C diagram asks whether particular events arrive on the predicted schedule. These are complementary questions.
A compact checklist for evaluating a published result
- Find the ephemeris: record T0, P, their units, and the stated time system.
- Identify the event: primary eclipse, secondary eclipse, or pulsation maximum; check whether datasets use the same definition.
- Inspect the error bars: a visually attractive trend may be weak relative to timing uncertainty.
- Check cycle counts: long gaps can leave an ambiguity in the integer E.
- Compare simple models: does correcting the epoch or constant period already explain the trend?
- Separate fit from cause: a good mathematical fit is not by itself proof of mass transfer, stellar evolution, or another body.
Sources and the next reading step
The AAVSO guide to analysis of times of minima explains the observed-minus-calculated method. Its student observing projects also discuss the importance of correct cycle counting. These are more useful starting points than a generic organization homepage.
Continue with eclipse timing and orbital changes or contact binary period changes. Browse Variable Stars & White Dwarfs for related stellar questions.
Reworked October 26, 2026: added two explicit teaching datasets, a units calculation, a pattern comparison, and a source-specific reading path while removing repeated general explanations. This is an enthusiast’s educational guide, not an original timing study. About and corrections.