ISO, noise and the exposure value

Aperture and shutter speed control how much light reaches the sensor. The third exposure setting, ISO, does something different: it controls how strongly the sensor's signal is amplified before it becomes an image. Raising the ISO lets you use a faster shutter speed or a smaller aperture in dim light, at the cost of more visible noise, the grainy speckle that degrades detail and color. This lesson explains what ISO really does, where noise comes from and why the amount of light, not the ISO number itself, mostly decides how noisy a picture looks, how ISO affects dynamic range, how to use auto ISO well, how noise reduction works, and how the exposure value (EV) system ties aperture, shutter and ISO into one number that describes the light in a scene.

What ISO does

In film photography, ISO (formerly ASA) measured the sensitivity of the film itself: ISO 400 film needed a quarter of the light that ISO 100 film did. A digital sensor's basic sensitivity is fixed by its design. The ISO setting tells the camera how bright to make the image for a given amount of light, mostly by amplifying the sensor's analog signal before it is converted to numbers, and partly by digital scaling. Doubling the ISO doubles the brightness of the final image, one stop, just as doubling the shutter time or opening the aperture one stop would.

The crucial difference is that ISO does not change how much light the sensor actually captures. Aperture and shutter speed do. Brightening a dim exposure with ISO is like turning up the volume on a quiet recording: the music gets louder, and so does the hiss.

Every camera has a base ISO, usually 100 (sometimes 64 or 200), at which it delivers its best dynamic range and lowest noise. Cameras also offer "extended" ISOs below base (such as 50) that are produced by deliberately overexposing and pulling back, at some cost in highlight range, and above the normal maximum, which are mostly digital multiplication of a very noisy image.

Where noise comes from

Noise has several sources, of which two dominate:

• Photon shot noise. Light arrives as discrete photons, randomly, like raindrops. If a photosite expects to collect N photons on average, the actual count fluctuates by about √N. The signal-to-noise ratio (SNR) is therefore about N/√N = √N. A bright area that collects 10,000 photons has an SNR of about 100, and looks smooth; a shadow area collecting 100 photons has an SNR of about 10, and looks grainy. Shot noise is a property of light itself and cannot be engineered away; the only cure is more light.
• Read noise. The electronics that measure and amplify each photosite add a small, fairly constant amount of noise. It matters mainly in the deepest shadows and in very dim scenes, where the signal is tiny.

Other sources include thermal noise in long exposures and hot sensors (seen as colored speckles and "hot pixels"), and fixed patterns from the sensor's structure. Cameras subtract some of these with long-exposure noise reduction, which takes a second, dark frame of the same length.

This explains why larger sensors are less noisy (Lesson 1): at the same f-number and shutter speed, a full-frame sensor collects about 2.3 times more total light than APS-C and 4 times more than Micro Four Thirds, so its whole-image SNR is about 1.5 and 2 times higher, roughly one and two stops of noise advantage at the same output size.

ISO and dynamic range

Dynamic range is the range of brightness, from the deepest shadow that still shows detail to the brightest highlight before it clips to pure white, that a camera can record in one exposure, measured in stops. Good modern sensors capture about 12 to 14 stops at base ISO. Each doubling of ISO amplifies the signal so that highlights clip one stop sooner, while the shadow noise floor improves less, so dynamic range falls by nearly a stop per stop of ISO at higher settings. Bright, contrasty scenes are therefore best shot at base ISO, with exposure set to protect the highlights (Lesson 6).

Many current sensors are largely "ISO-invariant" above a certain point, often with a second, higher-gain mode switching in around ISO 400 to 800 (dual gain). For such cameras, an image shot at ISO 800 and brightened two stops in raw processing looks almost the same as one shot at ISO 3200, which lets photographers protect highlights in high-contrast night scenes. The general principle remains: get as much light as the picture allows, then use ISO to set the brightness.

Choosing ISO in practice

In practice, choose the aperture the picture needs for depth of field and the shutter speed it needs for motion, then raise the ISO only as far as necessary to reach a correct exposure. A sharp, noisy picture is almost always better than a clean, blurred one: noise can be reduced in editing, motion blur cannot be removed.

Auto ISO automates this. The photographer sets the aperture and shutter speed (in manual) or the aperture and a minimum shutter speed (in aperture priority), and a maximum acceptable ISO; the camera adjusts ISO to keep exposure correct as the light changes. It is ideal for events, weddings, sport and street photography where light changes quickly. Exposure compensation (Lesson 6) still works with auto ISO in manual mode on most cameras.

Noise reduction

Noise appears in two forms: luminance noise, variation in brightness that looks like film grain, and chroma noise, colored blotches, which is usually more objectionable. Cameras apply noise reduction to JPEG images, and raw processing software (Lesson 10) offers adjustable noise reduction, which trades noise for loss of fine detail: too much makes skin and foliage look plastic. Recent software uses machine-learning models trained on pairs of noisy and clean images and achieves remarkable results, often recovering one or two stops of usable ISO. Other techniques include averaging several frames of a still scene, which reduces random noise by √n for n frames, and exposing generously ("exposing to the right") so shadows collect more light (Lesson 6).

Moderate luminance noise is not always a problem. At normal viewing sizes, it is less visible than on a magnified screen, and a fine, even grain can look pleasing, particularly in black-and-white. Judge noise at the size the picture will be seen.

The exposure value

Aperture and shutter speed can be combined into one number, the exposure value (EV), which identifies all settings that give the same exposure. At ISO 100:

where N is the f-number and t the shutter time in seconds. f/8 at 1/125 s gives EV log₂(64 × 125) ≈ 13, and so do f/5.6 at 1/250 s and f/11 at 1/60 s. Each higher EV is one stop less light admitted. Light meters often report the brightness of a scene as the EV needed at ISO 100: a bright, sunny scene is about EV 15, and a dim interior about EV 6 or 7. At a higher ISO, the required EV rises by one for each doubling of ISO.

The sunny 16 rule

Before built-in meters, photographers relied on the sunny 16 rule: in bright sun, set the aperture to f/16 and the shutter speed to the reciprocal of the ISO, so 1/100 s (or 1/125) at ISO 100. From there, open up for dimmer conditions: f/11 for slight overcast, f/8 for overcast, f/5.6 for heavy overcast, f/4 for open shade at sunset. It still works, and it is a quick check that a meter reading is sensible, especially with bright or dark subjects that fool meters (Lesson 6).