1. What Signal to Noise Means
Signal-to-noise ratio (SNR) compares the light from the target with the random variation in the measurement. In astrophotography, faint objects produce few photons, so the random noise in the photon count is a large fraction of the signal. Improving SNR means collecting more signal, reducing the noise sources that add to it, or both.
2. The Basic Relationship
When photon shot noise dominates, which is the usual situation for faint deep-sky targets once a sensible exposure is reached, SNR grows with the square root of total integration time. Doubling the total time improves SNR by a factor of the square root of two, about 1.41, rather than by a factor of two. Quadrupling the total time doubles SNR. This is a consequence of counting statistics, not of any particular camera.
Relative SNR gain with total integration time (illustrative, shot-noise limited)
| Total integration time (relative) | Approximate SNR gain (shot-noise limited) |
|---|---|
| 1x | 1.00 |
| 2x | 1.41 |
| 4x | 2.00 |
| 9x | 3.00 |
| 16x | 4.00 |
The table is an approximation that assumes shot noise dominates. Real images also contain read noise, dark current and sky background, and their contribution changes with exposure length and sensor temperature.
3. Other Noise Sources
Read noise is added each time the sensor is read, so frequent short frames accumulate more read noise than fewer long frames. Dark current grows with sensor temperature, which is why cooled cameras are used for long exposures. Sky background from light pollution adds shot noise of its own, and it can dominate the signal from faint nebulae. Narrowband filters reduce this background, as covered in the companion article.
4. A Practical Integration Plan
Choose a sub-exposure length short enough to avoid saturating bright stars and long enough that read noise is small compared with the sky background. Then collect enough total time to reach the target SNR for the detail you want. Calibrate with dark, flat and bias frames, and reject outlier frames from aircraft, satellite trails and tracking errors. Keep a log of conditions so that results can be compared between sessions.
5. Limits of the Model
The square-root relationship describes how noise averages out, but it does not make a poor image good. Tracking errors, focus drift, optical aberrations and poor calibration all reduce the usable signal, and they must be controlled separately.