Pushed ISO: Let's make it clear

Friday February 04 2011

Sensor Insight
ISO and exposure | RAW ISO values: The fundamental basis for accurate comparisons | RAW ISO measures are inferior to manufacturer ISOs: is this a problem?

Since the launch of the DxOMark website, many debates have arisen about ISO sensitivity: Some manufacturers were accused of cheating, the ISO sensitivity measured by DxOMark was claimed to be meaningless for photographers. The DxOMark team would like to clarify certain points.

ISO and exposure

First of all, why photographers should care about ISO sensitivity? Because ISO sensitivity relates to exposure and photographers certainly do care about exposure. Exposure determines whether shadows still contain some details, if midtones look just right, and if highlights are not burnt.

Now, let’s track back the elements that determine the exposure in the final picture. A first key factor is the sensor exposure: How much light falls on the sensor? Sensor exposure is determined by several factors, at least two of which cannot be adjusted by the photographer :

  • The illumination of the scene (except for studio photography)
  • The transmittance of the lens (see more)

Others are selected by the photographer or the camera auto-exposure system in order to get a correctly exposed photograph:

  • Exposure time and lens aperture (These parameters are also constrained by other conditions as the motion of the scene and the desired depth of field)

There is a last parameter that is almost orthogonal to the ones above and which allows the photographer to obtain a good exposure: ISO sensitivity. Before digital photography, ISO (or even ASA) was directly related to the chemistry of the film. In the digital era, it is now possible to (almost) seamlessly change the sensitivity of the sensor. Therefore, when exposure time is limited by motion blur, and the lens is already wide open and the scene is dark, it is still possible to obtain a well-exposed picture by playing with the ISO setting. The only limit is that with higher ISO comes also more noise, which eventually makes the image too bad to be used.

For a given scene illumination (measured by a luxmeter), there are a number of combinations of exposure times, lens apertures, and ISO settings that lead to the same exposure. Fixing two of them determines the third one.

There are other elements of exposure that happen backstage, unknownst to most users. The image captured by the sensor (the RAW image) is processed by a complex image processing pipeline (the RAW converter) that is embedded on the camera or exported to a computer. The picture exposure can be digitally modified by the RAW converter via amplification gains, tonal curves, etc. Processing the same RAW image using different RAW converters usually leads to photos with different exposures.

DxOMark compares digital camera sensors. In particular, this implies testing independent of any RAW converter. This can be done in two different ways: either actually process RAW images using a single RAW converter, or directly compare RAW images by anticipating the fact that they will be processed by a single RAW converter afterwards.

ISO sensitivity on the RAW image has exactly the same qualitative meaning as explained above. However, the user can usually access the image only after the RAW conversion and a subsequent exposure change. So the RAW ISO value may differ from the ISO on the camera from a user’s point of view. The RAW ISO value is the one that is relevant for sensor comparison, and its measurement is standardized by ISO 12232, and implemented by DxOMark (see more).