Whatever camera or phone you have, it’ll report the ISO value it used to take its photos. Despite its ubiquity, ‘ISO’ probably doesn’t mean what you think it does. Worse still, it may be holding your camera back, both in terms of the images it takes and in the tools it provides you. This means it's potentially holding your photography back, too. Part of the problem stems from the fact that ISO sounds like something you were already familiar with.

At first glance, ISO settings look just like the sensitivity ratings used for film (to the extent that there are some people who still refer to ASA: the US standard incorporated into the ISO standard for film). But ISO in digital isn’t the same as film. it’s essentially a metaphor for the way film sensitivity worked, if you got it processed in a minilab machine. This is a problem.

It causes confusion

The apparent familiarity and simplicity of ISO setting leads to a number of common misunderstandings. Despite what you may have heard or read, changing the ISO of your camera does not change its sensitivity.

ISO changes the lightness of the final image but it doesn’t change the fundamental sensitivity of your sensor. Nor is it an indicator of amplification being applied: although many cameras do increase their amplification as you increase the ISO setting, this isn’t always the case.

"Why can't I use ISO 100 in Log mode?" The answer is that a log gamma curve is so flat that it requires very little light to achieve middle grey, which means it's considered a high ISO. Strictly speaking, though, you can't really calculate an ISO value for log at all, since the standard is based on a different colorspace and gamma. It's a similar story for Raw.

This may sound like semantic nit-picking, but it causes a lot of misunderstandings. It’s widely thought that the additional noise in high ISO image comes from the ‘background hiss and hum’ of the sensor’s amplifiers. This feels right: we’ve all heard more noise if we turn up the volume on an audio amplifier. Unfortunately it’s simply not true: most noise actually comes from the light you’re capturing, so it primarily depends on your shutter speed and aperture.

The ISO standard doesn't specify that amplification needs to be used, nor does it specify what happens in the Raw file

The ISO standard doesn't specify that amplification needs to be used, nor does it specify what happens in the Raw file. All it does is relate initial exposure to output JPEG lightness, however that is achieved. The only thing you can be sure of, at the Raw level, from an increase in ISO is that if it prompts a reduction in exposure, you'll collect less light and therefore see more noise for each tone from the scene.

There's an ISO standard that's slightly more pertinent to Raw files, which looks at when the sensor becomes completely saturated, but this doesn’t correspond to the standard used by your camera. So next time you see a graph comparing ‘Manufacturer’ and ‘Measured’ ISO, what you’re actually looking at is the ‘JPEG ISO’ vs ‘Saturation ISO.’ Any differences between the two mainly tell you how many stops above middle grey the manufacturer's JPEG tone curve is designed to deliver.

It encourages poor exposure

As well as giving a false sense of simplicity, ISO’s increasingly tenuous attempt to mimic film ratings can mean making poor use of sensor response.

Film (particularly negative film) has a very distinctive response curve that gives lots of latitude for recovering highlights. Digital is very different: it offers a much more linear response but with a hard, unrecoverable clipping point in the highlights. And no, your favorite software doesn't really recover completely clipped highlights from your Raw file*.

This graph shows the signal-to-noise ratio (essentially the noisiness) at different brightness levels of film and digital. The film response peaks and then gradually declines, with plenty of scope for recovering highlights from the right-hand side of the curve. The digital response rises to much higher levels than the film, then cuts-off abruptly. So why would you expose these two media in the same way?

Illustration based on DxO's analysis

And yet, despite these differences, the digital ISO standard is based around ‘correctly’ exposing JPEG midtones**. A 2006 update to the standard gave manufacturers some flexibility in terms of how many stops of highlights they wanted in their JPEGs above middle grey***, but it still encourages exposure based on midtones, with a pre-set number of stops above this for highlights.

That’s not the best way to expose digital. The best results are achieved by giving as much exposure as possible without clipping the brightest tones you care about: a process called ‘exposing to the right.’ This maximizes the amount of light, and hence signal which, in turn, optimizes the signal-to-noise ratio (essentially ‘noisiness’).

And yet, by worrying about the JPEG middle grey, cameras end up giving every image the same number of stops for highlights, even though this is wasted in low DR scenes (that highlight space isn’t used and exposure is lower than optimal) or insufficient in high DR situations: the lovely colors of the sunset you’re shooting are lost, unrecoverably, to clipping.

Fujifilm’s DR modes essentially give you a choice of amplification and tone curve combinations that include different amounts of highlight information. These end up being rated as different ISO settings.

The ISO 200 / DR 100 example on the left has the least noise. The ISO 400 / DR 200 image has a shorter exposure, bringing more noise, despite having the same amount of amplification as the ISO 200 image. This low level of amplification means it has retained more highlight information than the ISO 400 / DR 100 image on the right, which used the same exposure but more amplification.

This problem isn’t easily solved: there are times that exposing-to-the-right will result in noisier midtones than you want. In these situations, you have to let the highlights go. However, fixating on JPEG midtones isn't helpful.

It warps camera development

This brings us to the biggest problem with using a clumsy metaphor for film sensitivity as the way of setting image brightness in digital: it means we aren’t given the tools to optimally expose our sensors.

ISO ends up conflating the effects of amplification and of tone curve, meaning you have to do your own research to find out what your camera’s doing behind the scenes, and what the best way to expose it is.

We aren’t given the most basic tools: Raw histograms or Raw clipping warnings that would help optimize exposure

The preview image your camera gives, the histograms it draws and the exposure meters and guides it offers are all based on JPEG output and their midtones, because ISO says that’s what matters. This means we aren’t given the most basic tools we need: Raw histograms or Raw clipping warnings that would help optimize exposure. It means no development has been done to create more sophisticated tools that would help you judge the quality implications of exposing to the right, and when to let the highlights go.

In short, ISO is an increasingly shaky metaphor that promotes misunderstanding, obscures what your camera is doing and robs us of the tools we need to get the most out of our cameras. Isn’t it time for something better?

Thanks to Bobn2 for feedback and fact-checking

* Highlight recovery sliders usually rely on only one of the color channels having truly clipped, and try to guess the value of the clipped channels, based on the remaining, unclipped one, so tend to be limited in their effectiveness. [Return to text]

** We put the word "correctly" in inverted commas because the more you think about it, the harder it becomes to pin down what 'correct' exposure might be. If you're certain that you know what 'correct' exposure means, then you should probably check through the assumptions that underpin it. [Return to text]

*** This change is why the JPEG ISO ratings used by manufacturers don’t need to coincide with clipping-based Raw ISO numbers. We’ve previously written an article about how it works. [Return to text]