Home / Learn / Science / Astronomy / Lesson 5.2
All lessons

Module 5: Doing real research · Lesson 5.2 · about 25 min

Periodograms: finding periods in messy data

Real observations come on scattered nights with noise. A periodogram tries every possible period and scores how well a repeating wave fits; the tallest honest peak is the period, but beware aliases caused by the daily gaps.

Key points, meanings and formulas on one page: choose "Save as PDF" in the print box.

Module 5 · about 25 minutes · read first: 5.1 Aperture photometry, 3.5 Pulsating stars and the period-luminosity law · maths: sine waves and phase. You’ll be able to fold a light curve on a trial period, read a periodogram, and recognise aliases.

Real observations come on scattered nights with noise; a periodogram tries every possible period and scores how well a repeating wave fits, but daily gaps create false peaks called aliases.

1. Why this matters

Many things in the sky repeat: pulsating stars, eclipsing binaries, spinning asteroids, planets crossing stars, stars wobbling as planets tug them. Finding the period is usually the first scientific result from a set of observations, and it can be done from a backyard or entirely from archive data.

2. Folding and scoring

Folding: pick a trial period P and replace each time with its phase, the fraction of a cycle since some start time. If P is right, all points line up into one clean curve. If it’s wrong, they scatter.

Periodogram: do that for thousands of trial periods and give each a score: how much of the variation a sine wave at that period explains. Plot score against period. The Lomb-Scargle periodogram is the standard version for unevenly spaced data.

Try it: find the hidden period

Simulated data: 60 measurements on random nights over 120 days, from a star with a hidden period and noise of 0.05 magnitudes. Because we observe only at night, false peaks (aliases) appear near periods where 1/P differs from the true value by 1 per day. The periodogram here is a simple least-squares sine fit at each period (like the Lomb-Scargle method).

Things to notice:

  • The highest peak folds the data into a clean wave.
  • Smaller peaks near the main one are aliases: periods that also fit fairly well because of the daily gaps.
  • Halving or doubling the period can look almost as good for some stars (an eclipsing binary with two similar eclipses, for example).

Find the period of a variable star from 40 scattered nights:

Raw data. Plotted against date, the brightness looks like random dots: the nights are too far apart to see the cycle.
Guess a period and fold. Take each time's remainder after dividing by the guessed period (its phase) and plot brightness against phase.
Wrong guess. The folded dots stay a scattered cloud.
Right guess. The dots line up into a clean light curve. The periodogram's score shoots up at this period.
Beware aliases. Because data come once a night, f ± 1 cycle per day fit almost as well. Check the shape and, if possible, observe twice in one night.
Maths and principle

Phase of an observation at time t, for period P and start time t0:

ϕ=t−t0P−⌊t−t0P⌋

Sine model at frequency f=1/P: y(t)=asin⁡(2πft)+bcos⁡(2πft)+c. Fit a,b,c by least squares (Lesson 5.7); the score is the fraction of the variance explained.

Aliases: data taken once a day can’t tell frequency f from f±1 cycle per day, so false peaks appear at |f±1|.

Try it: a true period of 0.5 day (f=2/day) has aliases at f=1 and 3/day: periods of 1 day and 0.333 days.

3. What it means physically

A periodogram is like tuning a radio: you sweep through frequencies and listen for where the signal comes through loud. The gaps in your data (daytime, cloudy nights, the Moon) act like a comb laid over the signal, and the comb’s own rhythm adds ghost stations. The cure is more varied sampling: observers in other countries (Lesson 5.9) or data from space.

φ = frac((t − t₀)/P)
Phase: how far through its cycle the star was at time t, from 0 to 1.
f = 1/P
Frequency: cycles per day.
|f ± 1|
Alias frequencies created by sampling once per day.
Physical meaning
It is like a strobe light on a spinning wheel. Flash at the right rhythm and the spokes stand still; flash at the wrong one and they blur. Folding the data is the strobe, and the periodogram tries every rhythm. A strobe once a day can also make a wheel look as if it turns at a different speed: that is an alias.

“The tallest peak in a periodogram is always the true period.” Daily, monthly and yearly gaps make alias peaks that can be taller than the real one. Check the folded light curve and test with data at different times of night.

4. Worked example

A star’s periodogram peaks at 0.833 days, with a second peak at 5.0 days. Are they related?

f1=10.833=1.20 per day,f2=15.0=0.20 per day

f1−f2=1.00 per day exactly: one is an alias of the other. Fold on both. The true period gives a smoother curve with less scatter; extra nights at different times of night settle it.

5. Practice

More practice (answers below):

  1. Find the phase of an observation at t=10.7 days for P=3 days and t0=0.
  2. A true period is 2 days. Where would the 1-day aliases fall?
  3. Why are space telescopes like TESS less troubled by aliases?
Answers
  1. 10.7/3=3.567, so ϕ=0.567.
  2. f=0.5/day, so the aliases are at 0.5+1=1.5/day (a period of 0.667 days) and |0.5−1|=0.5/day, which is the true peak again.
  3. They observe almost without gaps for weeks, so there is no daily comb.
  • Folding at a trial period lines the data up only when the period is right.
  • A periodogram scores every trial frequency; peaks mark likely periods.
  • Once-a-day sampling makes aliases at f ± 1 cycle per day.
  • Confirm a period by folding and by observing at varied times.

Try it with real numbers: run a periodogram on your own data with the period finder, and fit the solar cycle with the solar cycle fitter.

New words: periodogram, Lomb-Scargle, phase, folding, alias, frequency.

Check yourself

1A periodogram shows:
2Why do night-only observations create false peaks near 1 day?
3Folding data on the right period gives:

Sources

  • Lomb, N. R. (1976), Ap&SS 39, 447; Scargle, J. D. (1982), ApJ 263, 835; VanderPlas, J. T. (2018), ApJS 236, 16

Your ticks stay in this browser. Save them as a file on the course home.

Updated 2026-11-22. Numbers in the examples are made up to show the idea unless a source is given. Simulations run in your browser on made-up data; nothing you type is sent anywhere. Astronomy only: nothing here is astrology.