Exposure Latitude

In this test, we've shot our low-light test scene at an exposure 5EV below our standard low-light Raw exposure, then push 5EV in Adobe Camera Raw. This illustrates the latitude of each cameras' Raw files, which is also intimately linked to the camera's dynamic range.

In real-world terms, this shows how underexposed tones look when pushed, which you might want to do if you've exposed a very high dynamic range scene for the highlights, or simply made a mistake. Cameras with more dynamic range will show cleaner results when shadows are pushed, while cameras with lower dynamic range will have noisier results, often due to a sensor and/or electronics with a higher noise floor. We're working on a more real-world dynamic range test that is repeatable across cameras and provides visual meaning to photographers; in the meantime, take this test as a proxy for dynamic range assessment.

Here we look at the Nikon D750 next to its closest rival - the Canon EOS 5D Mark III. We will be adding to the list of available cameras in future but at the point of publication it's the 5D Mk III that makes most sense to compare to.

The differences are dramatic, with the Nikon's image showing relatively little noise, even in the relatively deep shadows, which will have received very little light. These images were processed with default ACR settings, as they are set to give a sensible starting point for anyone looking to conduct further processing. With these settings there's a small amount of luminance noise and chroma noise is completely eliminated by the standard +25 chroma noise reduction.

Even in the deepest shadows it's still possible to distinguish between different colors: something that high levels of chroma noise and chroma noise reduction would hamper.

The Canon can't compete with this level of dynamic range - exhibiting not just extensive noise but also significant banding. While the chroma and luminance noise can be reduced with careful use of noise reduction (though usually at the cost of detail), the banding is essentially impossible to remove. This means a D750 user will be able to capture shots with much greater dynamic range or cleaner shots with the same range and will have much more leeway with any mis-exposed images.


It's not just high dynamic range scenes that benefit from a camera with a low noise floor: it can also reduce the need to amplify the sensor's signal to keep it above that noise floor at higher ISOs.

To illustrate this we've done something that may seem counter-intuitive: we've shot multiple images at the same exposure settings but using different ISO settings. The Raw files have then been brightness matched. The reasons for doing this should become apparent.

At first glance, all these images look essentially the same, which suggests it makes no difference whether you achieve the correct brightness by amplifying the signal at the point of capture, or pushing it 'digitally' later. At which point we would say the camera is ISO-invariant (some call this 'ISO-less').

The opposite of this would be the EOS 5D Mark III, which is highly ISO-variant: it really benefits from using hardware amplification to ensure the signal is kept above the noise floor. Once you get to ISO 3200 on the Canon, most of the usable signal has been amplified enough to keep it above the noise floor so again there's no real benefit to be had of increasing the ISO setting (and applying more hardware amplification) above that point.

You can compare the ISO-invariance performance of the two cameras in this state of our widget.

Look closer at the Nikon's results and you can see the ISO 200 image is slightly cleaner in darker tones than the ISO 100 shot, for somewhat esoteric reasons we won't dwell on here (residual downstream read noise, and perhaps a bit of quantization error). Above this point it essentially makes no difference in terms of noise whether you use hardware amplification or post-shot processing.

What does this mean?

The D750's ISO-invariance means that the camera's low noise floor isn't just a benefit when shooting wide dynamic range scenes - it's also useful in circumstances where you might consider using a high ISO setting on the camera.

In terms of noise, there's no difference between increasing the ISO setting in the camera, which applies electronic amplification, and pushing the files later in post processing. However, there is a difference in terms of highlight capture. Amplifying the sensor signal risks pushing the brightest tones so far that some of them 'clip' and can't be recorded. Under-exposing a lower ISO prevents this from happening. As a result, you can get a better image by leaving the camera at ISO 200 and reducing exposure (by up to at least 5EV according to our demonstration above), rather than increasing ISO.

Again, this isn't the way most camera makers expect you to work, so the camera's JPEGs and all its metering and exposure assessment tools that are based around those JPEGs become pretty-much useless. However, this way of shooting confers an image quality advantage if you make the effort to work around these inconveniences.