Exposure Length vs Integration Time

Astrophotographers often use “exposure time” to mean two different things:

• Sub-exposure length is how long the camera records one frame, or sub.
• Integration time is the total exposure time in the accepted frames that are combined into the final image.

Those decisions are related, but they solve different problems. Sub length determines how much signal, sky background, and risk enter each file. Integration determines how confidently faint signal can be separated from random noise.

What one subframe contains

Each pixel records charge produced by photons, plus unwanted contributions from read noise, thermal effects, sky brightness, and other sources. A longer sub usually records more target signal, but it also records more background and gives tracking errors, wind, cloud, aircraft, and saturation more time to spoil the frame.

The goal is not the longest sub the camera permits. It is a sub long enough that the target and sky signal are usefully recorded, while short enough that bright structures and stars do not saturate unnecessarily and a reasonable percentage of frames remain usable. Open the original sub-length explanation

The histogram helps, but there is no universal magic position. Its useful interpretation depends on sensor read noise, gain, sky brightness, filters, and the target.¹

How integration time changes signal-to-noise ratio

Signal-to-noise ratio, or SNR, compares the repeatable information from the target with the random variation that makes that information harder to distinguish. A higher SNR does not automatically make an image more artistic, but it gives faint structure, subtle color, and smooth gradients a more reliable foundation.

When compatible exposures are registered and stacked, target signal reinforces from frame to frame while independent random noise only accumulates by its square root. Under simplified conditions, the relationship is:

SNR multiplier ≈ √(new usable integration ÷ original usable integration)

This gives a practical set of reference points:

• two hours compared with one hour provides about 1.4× the SNR;
• four hours provides about 2× the SNR;
• nine hours provides about 3× the SNR;
• and sixteen hours provides about 4× the SNR.²

Open the original square-root explanation and long-integration example

The important consequence is that SNR improves predictably, but the time required grows quickly. Doubling the SNR requires about four times the usable integration. Halving the random noise relative to the target signal also requires about four times the integration. Adding one hour to a one-hour project makes a much larger difference than adding one hour to a fifteen-hour project.

That is a diminishing rate of improvement, not a hard point where more data suddenly become worthless. A longer project can also give you a deeper pool from which to reject weak subs and retain the clearest conditions. The square-root comparison only applies cleanly when the added data are compatible; ten excellent hours and fifteen badly degraded hours are not automatically a twenty-five-hour equivalent.

These numbers are a planning model, not a guarantee. They assume that the added frames are compatible and that random noise is the main limit. Changing transparency, Moonlight, focus, framing, filter, gain, or sensor temperature can make one hour of data unlike another. Gradients, walking noise, poor calibration, clipped highlights, and tracking errors are structured problems; they do not disappear according to the square-root rule.

This is why more clean integration—and, when possible, a darker sky—is the most reliable capture-side response to random noise. Noise-reduction software can make recorded data easier to view, but it cannot replace target photons that were never captured. Open the original integration-and-noise note

Why many shorter subs can work

The square-root relationship is driven mainly by total accepted exposure, not by frame count alone. Once each sub is long enough that read noise is a modest part of the total noise, 120 accepted 30-second subs and 60 accepted 60-second subs both provide one hour of integration and can produce broadly similar final SNR. Longer subs can retain an advantage when very short exposures are still read-noise limited, but not simply because each individual file looks brighter.

Shorter subs have practical strengths:

• one ruined frame costs less time;
• bright stars and cores are less likely to saturate;
• modest tracking systems can retain more frames;
• and frame selection can respond to brief wind or cloud.

The risk is easy to picture: if a gust or tracking error ruins a 60-second exposure at second 59, the entire minute may be rejected. With shorter subs, an equally brief disturbance may cost only one small part of that minute.

They also create more files, more read-noise contributions, and more registration and stacking work. Extremely short exposures can be inefficient when the background has not risen enough above read noise.

What longer subs change

Longer subs reduce the number of readouts for the same total integration and can be helpful behind strong filters or under dark skies. But the trade is concentrated risk:

• tracking drift or wind can spoil more time at once;
• bright stars and target cores can clip;
• skyglow can consume dynamic range;
• field rotation can become visible in Alt-Az capture;
• and fewer files give rejection algorithms fewer samples.

Equatorial tracking can make longer subs possible, but it does not make them automatically better. ZWO’s own Seestar EQ guidance warns that longer exposure choices can increase frame loss if alignment or conditions are not good enough.³

Usable integration is the real total

If a two-hour session produces 90 minutes of accepted frames, the usable integration is 90 minutes. The rejected half hour may explain what happened, but it does not contribute to the final signal.

This distinction matters when comparing modes or nights. A 30-second setting that rejects half its frames may produce less usable signal than 10-second subs that survive reliably. See the original accepted-frame comparison

Track both:

• wall-clock capture time;
• accepted frame count and duration;
• rejected frame count and likely reason;
• and the conditions during each block.

Factors that influence sub length

Sky brightness and altitude

Bright sky raises the background more quickly. A target low in the sky is seen through greater airmass, which increases extinction and often amplifies skyglow.⁴ Under these conditions, a shorter sub may protect dynamic range even though the final project needs more total integration.

The Moon and the target’s spectrum

Moonlight affects broadband targets more strongly because it overlaps their useful wavelengths. A narrowband emission target behind an appropriate filter may tolerate a brighter Moon, but the filter does not eliminate gradients or scattered light.

Filters

A narrow filter reduces the photon rate reaching the sensor. That can justify longer subs, higher gain, more integration, or some combination. It does not justify changing settings by habit without checking the actual frames.

Gain and sensor behavior

Gain changes the conversion between collected charge and recorded values; it does not create photons. Higher gain can reduce usable dynamic range and make saturation easier. Camera-specific read-noise and full-well behavior matter more than a generic “high” or “low” label.

Focal ratio, sampling, and target brightness

A faster optical system concentrates more light per unit sensor area for an extended target. A bright core reaches saturation sooner than faint outer dust. Dense star fields can constrain exposure before the nebula does.

Mount, wind, focus, and dithering

Tracking accuracy sets an upper practical limit. Wind, cable drag, field rotation, and focus drift can lower it on a particular night. Dithering improves rejection of fixed-pattern noise, but the settle time between frames also affects efficiency.

A practical test

1. Choose a representative target and the filter you intend to use.
2. Capture a short run at two or three plausible sub lengths.
3. Inspect star shape, saturated pixels, background level, and accepted-frame rate.
4. Compare equal amounts of usable integration, not equal frame counts.
5. Choose the shortest setting that records the background and faint signal efficiently without creating an unnecessary saturation or rejection penalty.

Repeat when the filter, gain, sky brightness, mount mode, or target type changes. The right answer is a working range, not a sacred number.

How much integration is enough?

Stop when the data support the image you are trying to make. A bright object intended for a small web image may need much less time than faint dust intended for a large print. Continue while additional data are improving background smoothness, faint structure, color, or processing flexibility enough to justify the capture cost.

More integration cannot repair poor focus, severe tracking errors, clipped highlights, or a bad field-of-view choice. Quality and quantity multiply each other; neither substitutes for the other.

Notes and external sources

Footnotes

1. SharpCap, Smart Histogram and Sensor Analysis. Connects sub-exposure choice to sky background, sensor read noise, gain, and diminishing returns. ↩

2. STScI, Sensitivity, Count Rate, and Signal-to-Noise, and Siril, Image Stacking. STScI derives the square-root exposure-time relationship for photon-counting observations without read noise; the Learn rule is an approximation for compatible, background-limited stacked data. Siril documents the practical aligned-combination, normalization, and rejection workflow. ↩

3. ZWO Seestar, How to Switch to EQ Mode. Documents available exposure choices and the risk of longer-exposure frame loss. ↩

4. AAVSO, Guide to CCD/CMOS Photometry. ↩