Two different stellar periods can predict the same brightness at every time you observed. In that situation, another precise measurement taken on the same schedule may repeat the ambiguity. The useful question is when the competing predictions separate.
Daily aliases arise when an observing pattern makes different frequencies difficult to distinguish. This guide builds one exact, invented example, then uses it to plan a discriminating follow-up. It concerns periodic brightness measurements, rather than deciding whether a transit-like dip comes from a planet. You need only a calculator with cosine in radians to check the numbers.
THE GOLD STANDARD OF ASTRONOMY
Choose the question you need to answer
Understand the competing clocks · Chapters 1–2. Build and calculate an exact nightly alias.
Choose useful evidence · Chapters 3–5. Plan a discriminating visit and keep the claim in scope.
Practise and continue · Chapters 6–8. Change the schedule, check your reasoning and read further.
Open the chapter contents · Try the transfer checks · The Gold Standard of Astronomy
Chapter contents
Understand the competing clocks
Choose useful evidence
3. Choose a new time from the competing predictions
4. Read a periodogram as evidence with a sampling history
5. Separate alias selection from detection and physical interpretation
Practise and continue
6. Test the schedule before trusting another decimal place
7. Questions about repeated observations
8. Continue from a candidate period to an observing decision
A variable source has a brightness history. Your observations sample that history at selected times. The sampling schedule is another structure in the problem: daylight, weather, visibility and telescope access determine which parts of the history reach the dataset.
Frequency counts cycles per unit time; period measures the time for one cycle. A frequency of 0.2 cycles per day corresponds to a period of 1 ÷ 0.2 = 5 days. A frequency of 1.2 cycles per day corresponds to about 0.833 day, or 20 hours. Those are physically different clocks.
An alias is a competing frequency that the sampling pattern makes resemble another frequency in the recorded data. This is different from ordinary random scatter around one model. An exact alias can remain even when the measurements have no noise at all.
For a real astronomical dataset, preserve the time standard, exposure duration and any corrections specified by its archive. Our example uses elapsed days from an arbitrary origin and instantaneous readings. It deliberately removes those complications so we can see the sampling issue clearly.
Consider two proposed models for normalised flux. Model A is F(t) = 1 + 0.1 cos(2π × 0.2t). Model B is F(t) = 1 + 0.1 cos(2π × 1.2t). Time t is in days. Both have a baseline of 1, an amplitude of 0.1 and the same chosen starting phase.
Suppose we observe at t = 0, 1, 2, 3 and 4: exactly one day apart. For any integer n, Model B has an extra angle of 2πn compared with Model A. Cosine is unchanged by a whole number of complete turns. The predictions therefore agree at every visit on this grid.
At day 0 both predict 1.1000. At day 1 both predict approximately 1.0309. At day 2 both predict 0.9191. The two curves differ between observations, but that difference is absent from the nightly measurements.
Extending this exact grid to day 100 does not resolve this particular pair. More observations would help estimate the shared sampled pattern, but no integer-day observation can select A over B under these assumptions. This is a concrete limit of the available information, not a failure of arithmetic.
| Elapsed days | Model A flux | Model B flux |
|---|---|---|
| 0 | 1.1000 | 1.1000 |
| 1 | 1.0309 | 1.0309 |
| 2 | 0.9191 | 0.9191 |
| 3 | 0.9191 | 0.9191 |
| 4 | 1.0309 | 1.0309 |
Try t = 0.5 day. Model A uses cos(0.2π), approximately 0.8090, and predicts 1.0809. Model B uses cos(1.2π), approximately −0.8090, and predicts 0.9191. Their predicted fluxes differ by about 0.1618.
That time is informative because the models disagree there. If measurement uncertainty and calibration errors are small relative to that separation, the new observation can challenge one model. A reading near 1.0809 would favour A over this particular fixed B model. It would not establish that every other possible model has been eliminated.
Half a day later may be daylight at the original observatory. The calculation identifies an information need; the observing plan must also be feasible. Another longitude, an existing archive with different timing, or several accessible times within a night might help. Recalculate the predictions at the actual available times before requesting observations.
An effective request states the target, candidate periods, predicted difference, timing tolerance and required measurement quality. “Obtain more data” is less useful than explaining which unresolved alternative the data should test.
A periodogram compares candidate frequencies with a time series. Lomb–Scargle methods are widely used with uneven observations. An irregular schedule, however, does not guarantee that aliases disappear. Nightly and seasonal structure can remain inside an uneven dataset.
VanderPlas explains how the observing window affects period searches and why Lomb–Scargle does not automatically remove sampling aliases. Use the actual observation times when evaluating candidate frequencies. A smooth, densely drawn frequency plot does not supply missing measurements between those times.
Compare phased light curves at plausible alternative peaks. Phase folding places observations according to where they fall within a proposed cycle. A tidy folded curve is useful evidence, but two aliases may both look tidy when the schedule has not sampled their disagreement.
Keep the original chronological plot too. It can reveal trends, changing amplitude or instrument offsets that a folded view conceals. Record the model family, fitted parameters, excluded observations and search range so a reviewer can reproduce the comparison.
CHAPTER 5 OF 8
5. Separate alias selection from detection and physical interpretation
Three questions require separate answers: is there structured variation, which period best describes it, and what physical process causes it? Evidence for the first does not automatically settle the second or third.
A small false-alarm probability is conditional on a specified null model and search procedure. It does not supply the probability that a particular alias is the true stellar period. Likewise, a narrow peak around one candidate describes local resolution without necessarily accounting for a distant competing peak.
Real signals may not be sinusoidal. Two unequal eclipses can make a half-period interpretation tempting; evolving starspots can change the apparent pattern over time. A more detailed model must be justified by the source and evidence rather than selected only because it makes the preferred peak look convincing.
The exact equality in our worked example relies on integer-day times and the chosen sinusoidal models. Actual nightly times drift, observations have finite durations and noise matters. These differences may weaken the equality, but whether they resolve the ambiguity depends on their size and the data quality. Preserve that distinction when moving from the teaching case to a survey result.
Transfer check: skip nights
Keep the same models but observe only at days 0, 2, 5 and 9. Do the gaps resolve the ambiguity? No. Every time is still an integer, so the added whole-cycle term remains invisible. Missing nights reduce the dataset without creating the required phase information.
Transfer check: improve precision
Suppose every integer-day measurement becomes ten times more precise. Which problem improves? Random uncertainty in those fluxes. Which remains? The exact model equality at the sampled times. Explain the difference before calculating a more precise period.
Transfer check: change the sampling interval
If samples occur at t = nΔ, two frequencies separated by k/Δ, for integer k, have the same additional whole-cycle phase on that grid when the remaining model terms match. For Δ = 2 days, a frequency difference of 0.5 cycles per day can create the corresponding ambiguity. This statement concerns the explicit sinusoidal sampling construction; it is not a universal frequency limit for every irregular dataset.
Practical evidence check
Make a table for A and B at days 0, 0.25, 0.5, 0.75 and 1. Circle a feasible time where their predictions differ clearly. Then increase the assumed measurement error and decide whether your proposed observation still discriminates. State what result would leave the choice unresolved.
Can more nights ever help resolve an alias?
Yes. Real schedules may add different phases, extend the baseline or improve precision enough to distinguish imperfect aliases. Our exact integer-day example identifies a case where repeating the same timing structure cannot help.
Does the tallest periodogram peak have to be correct?
No. Compare alternative peaks with the observing window, model fit, uncertainty and independent evidence. Peak height alone does not complete the argument.
Should every programme observe at half-day intervals?
No. That time was selected for two particular invented models. Useful timing depends on the candidate periods, available observing windows and scientific question.
Is a daily alias a daily physical process in the star?
Not necessarily. The daily pattern can enter through observation timing on Earth. Identify which clock belongs to the source and which belongs to the measurement system.
A useful analysis ends with a reviewable decision: retain these candidate periods, identify where their predictions differ, and request measurements that can test that difference. If no suitable observations are available, report the unresolved alternatives.
The related astronomy guides below address different ambiguities: identifying the source of a transit-like signal, separating parallax from motion, and maintaining responsible observation records. The sampling and audio articles help transfer the mathematics, while the astronomy case keeps the observing constraints visible.
Subject and learning map
The Gold Standard of Astronomy
Education: learning that transfers into life
How X Works: connected subject hub
Other astronomical inference problems
Vetting an exoplanet transit candidate: identify the source of the dip
Parallax and proper motion: choose epochs that separate contributions
Astronomers: reliable observations and honest discovery
Transfer the sampling idea
Digital audio sampling, Nyquist rate and aliasing
Palaeoclimatology: sampling intervals and false climate periods
Authoritative sources
Sources checked on 21 October 2026. Numerical cases and practice tasks are original teaching examples, not observations from a published experiment or official statistics.
VanderPlas: Understanding the Lomb–Scargle Periodogram — primary methodological account; see the observing-window, failure-mode and uncertainty discussions. The numerical models here were constructed separately.