ISO Noise Calculator

ISO invariance describes sensors where the read noise at base ISO is so low that amplifying the image digitally in post-processing produces similar or better results than using analog gain (high ISO) in-camera. For these sensors, shooting at ISO 100 in a dark scene and brightening in Lightroom can yield equivalent or better noise performance than shooting at ISO 3200. Not all cameras are ISO-invariant — Canon DSLRs historically have higher read noise at base ISO, making actual in-camera ISO adjustment more important. Check your specific camera's SNR curves at sites like Photons to Photos.

Base ISO is the sensor's native sensitivity — the ISO at which it achieves maximum dynamic range and signal-to-noise ratio. For most cameras, base ISO is 100 or 200. Some cameras have a dual native ISO with two base points (e.g., Sony FX3: base ISO 800 and ISO 12800 for video), where a second amplification circuit engages at the higher value. Shooting at base ISO gives the cleanest files; any ISO increase sacrifices some dynamic range in exchange for exposure sensitivity.

Larger pixels collect more photons per unit of exposure, resulting in more shot noise photons (absolute) but much better signal-to-noise ratio because signal increases faster than noise. A pixel with twice the area captures twice the photons, giving √2 more shot noise but 2× the signal — resulting in √2 better SNR (approximately 3 dB). This is why full-frame cameras with large pixels generally outperform smaller sensors with tiny pixels in low-light situations, even if the smaller sensor has more megapixels.

Luminance noise appears as random brightness variations (grain-like texture), while color noise appears as random colored speckles — red, green, and blue pixels scattered randomly. Color noise is more visually distracting than equivalent luminance noise. Most RAW converters (Lightroom, Capture One, DxO) have separate luminance and color noise sliders. Color noise reduction is typically more aggressive than luminance (which also reduces apparent detail). Modern cameras and AI denoise tools (Lightroom Denoise, Topaz DeNoise AI) handle both types effectively at higher ISO settings.

Generally: expose to the right (ETTR — Expose to The Right of the histogram without clipping) at any ISO, then adjust in post. An underexposed image at ISO 100 lifted 4 stops in post will have more noise than a correctly exposed ISO 1600 image, because the shadow regions capture fewer photons. The optimal approach is to use the ISO needed to correctly expose the scene (not underexpose at low ISO), then apply noise reduction in post. The exception is ISO-invariant sensors where the low-ISO read noise is exceptionally low.

AI-based noise reduction tools (Adobe Lightroom's Denoise, Topaz DeNoise AI, DxO PureRAW, ON1 NoNoise AI) apply trained neural networks to identify and preserve real detail while removing sensor noise. Results are dramatically better than traditional luminance/color noise sliders — typically recovering 2–3 usable stops of ISO range. A noisier ISO 12800 image can often be cleaned to match a non-AI processed ISO 3200 image. These tools work best on RAW files rather than JPEGs.

Industry benchmarks: SNR ≥ 40 dB is excellent (clean, commercial print quality). SNR 35–40 dB is good (acceptable for most uses, slight grain visible at 100%). SNR 30–35 dB is fair (noise visible, suitable for web and social media). SNR below 30 dB shows obvious grain, often suitable only for artistic use or heavily noise-reduced output. Camera sensors are measured by DxOMark's SNR 18% metric, which tests at a standard 18% gray target.