Context and Why This Matters
Time-domain astronomy asks not only what an object is, but how it changes with time — and the answer depends entirely on whether the observing cadence matches the timescale of the phenomenon. A real signal can be completely invisible, or badly misrepresented, if the sampling is too coarse.
Learning Goals
- State the Nyquist-style sampling requirement for recovering a periodic or transient signal.
- Evaluate whether a given cadence can resolve a specific event duration.
- Explain the difference between undersampling and aliasing.
Opening Phenomenon
A transiting source produces a dip lasting about 2.1 hours. One observer images the field once per night (24-hour cadence); another images it every 20 minutes during the predicted event window.
Terms Developed in Context
These terms form the measurement chain for Time-Domain Observing: each connects an observable to a model choice, and each has a boundary that must be respected when making a scientific claim.
- cadence
The timing pattern of repeated observations, including the interval and any gaps between samples. Cadence determines whether the light curve resolves the event shape or aliases it into a false pattern.
It must be compared with ingress, event duration, and period rather than only with total observing baseline. Cadence is not synonymous with exposure time; readout and scheduling gaps also determine sample times.
Developed in this lecture: The term helps interpret this evidence: A once-per-night cadence (24 hours) is roughly 11 times coarser than the 2.1-hour event, so the event is essentially invisible in that dataset. It also constrains the working model: To recover the shape of an event of duration T, the sampling interval should be a small fraction of T, not merely shorter than the total observing run.
- sampling interval
The elapsed time between neighboring measurements in a time series. It is the quantitative spacing used to judge whether a changing signal is adequately resolved.
For a regular series, sampling frequency is \(f_s=1/\Delta t\); event morphology often requires many samples across the shortest relevant timescale. A quoted average interval can hide irregular gaps that alter the window function.
Developed in this lecture: The term helps interpret this evidence: A once-per-night cadence (24 hours) is roughly 11 times coarser than the 2.1-hour event, so the event is essentially invisible in that dataset. It also constrains the working model: To recover the shape of an event of duration T, the sampling interval should be a small fraction of T, not merely shorter than the total observing run.
- aliasing
The appearance of an incorrect frequency or timescale when sampling cannot distinguish the true variation from another pattern. It explains how a real periodic signal can appear constant or acquire a spurious period.
Regular sampling mixes the signal spectrum with the sampling window; frequencies separated by sampling frequencies can become degenerate. More total observations do not automatically remove an alias if they preserve the same unfavorable timing pattern.
Developed in this lecture: The term helps interpret this evidence: A once-per-night cadence (24 hours) is roughly 11 times coarser than the 2.1-hour event, so the event is essentially invisible in that dataset. It also constrains the working model: To recover the shape of an event of duration T, the sampling interval should be a small fraction of T, not merely shorter than the total observing run.
- undersampling
Recording too few independent samples across a spatial or temporal feature to recover its shape or frequency. The once-per-night sequence undersamples the 2.1-hour event and therefore cannot characterize its dip.
A practical event criterion is \(\Delta t\ll T_{event}\), stricter than merely obtaining one point during the event. Undersampling is a data-acquisition limitation; interpolation cannot recreate missing information reliably.
Developed in this lecture: The term helps interpret this evidence: A once-per-night cadence (24 hours) is roughly 11 times coarser than the 2.1-hour event, so the event is essentially invisible in that dataset. It also constrains the working model: To recover the shape of an event of duration T, the sampling interval should be a small fraction of T, not merely shorter than the total observing run.
- ingress/egress
The transition intervals during which an occulting body begins to cover or uncover the source. Their duration and shape constrain event geometry and require finer sampling than the full event duration alone suggests.
Cadence and exposure integration convolve with these sharp transitions, smoothing their measured slopes. The flat-bottom duration is not the same as ingress or egress duration.
Developed in this lecture: The term helps interpret this evidence: A once-per-night cadence (24 hours) is roughly 11 times coarser than the 2.1-hour event, so the event is essentially invisible in that dataset. It also constrains the working model: To recover the shape of an event of duration T, the sampling interval should be a small fraction of T, not merely shorter than the total observing run.
- period
The time required for a repeating phenomenon to complete one cycle. It is inferred from repeated timing structure and determines how observations from different nights can be phased.
Frequency and period obey \(f=1/P\), but aliases can create several candidate periods. A single observed dip gives an event time, not a secure period without recurrence or external constraints.
Developed in this lecture: The term helps interpret this evidence: A once-per-night cadence (24 hours) is roughly 11 times coarser than the 2.1-hour event, so the event is essentially invisible in that dataset. It also constrains the working model: To recover the shape of an event of duration T, the sampling interval should be a small fraction of T, not merely shorter than the total observing run.
- duty cycle
The fraction of a relevant time span during which useful observations are actually being obtained. It quantifies coverage gaps that control the chance of capturing short events.
For uniform exposures it is approximately total open-shutter time divided by elapsed campaign time. High duty cycle does not guarantee appropriate cadence if individual samples are too long or poorly timed.
Developed in this lecture: The term helps interpret this evidence: A once-per-night cadence (24 hours) is roughly 11 times coarser than the 2.1-hour event, so the event is essentially invisible in that dataset. It also constrains the working model: To recover the shape of an event of duration T, the sampling interval should be a small fraction of T, not merely shorter than the total observing run.
Evidence
- A once-per-night cadence (24 hours) is roughly 11 times coarser than the 2.1-hour event, so the event is essentially invisible in that dataset.
- A 20-minute cadence during the event window gives roughly 6 samples across the dip, enough to resolve ingress, the flat bottom, and egress.
- A periodic signal sampled at close to its own period can appear to be constant or slowly varying even though it is rapidly changing — classic aliasing.
Model
- To recover the shape of an event of duration T, the sampling interval should be a small fraction of T, not merely shorter than the total observing run.
- Sparse, evenly spaced sampling of a periodic signal can alias to a spuriously long or short apparent period.
- Cadence, exposure time, and total baseline are three separate observing-design choices, each driven by a different aspect of the science goal.
Working Equation
\[ \Delta t_{\rm cadence} \ll T_{\rm event} \]
Worked Example
- Event duration T = 2.1 hours.
- Nightly cadence Δt = 24 hours → Δt/T = 11.4, far too coarse; the event is effectively unsampled.
- Fine cadence Δt = 20 minutes = 0.33 hours → Δt/T = 0.159, giving about 6 samples across the event — adequate to characterize ingress, minimum, and egress.
Common Misconception
Treating a sparse light curve (one point per night) as a complete time series when the phenomenon of interest varies on a timescale of hours, not days.
Active-Learning Guidance
Given a claimed 2.3-hour periodic signal and a proposed 6-hour cadence, decide whether the cadence is adequate and, if not, propose a better one.
Lab / Observing Connection
This unit is assessed through Problem Set 06, which pairs the stacking calculation above with a cadence-adequacy problem for the same kind of dataset.
Synthesis Questions
- What was measured directly in this lecture’s worked example, and what was inferred from the model?
- Which uncertainty or systematic would most change the interpretation?
- How does this technique connect to the technical observing report due at the end of the term?
References
- OpenStax Astronomy 2e, Chapter 19.3 (variable stars and cosmic distances, period-luminosity work) and Chapter 21.4 (transiting exoplanets).
- Local reference copy:
references/openstax-astronomy-2e.pdf.