The Ups and Downs of ISO


Adjustments to aperture, ISO and speed all have different advantages and disadvantages, so understanding what happens with each element of the triad is important in understanding which one you want to adjust when setting your exposure. In this post, I am focusing on ISO.

The native ISO on most cameras is 100 and this setting means you will get the most noiseless image possible. It is possible to set the ISO slightly lower which will have a minor impact on the noise in the image. If the ISO is set higher then the noise will increase as the ISO increases, and this is the downside. Noise, of course, can be reduced in post-processing but this will reduce the resolution or soften the image. The upside of ISO is you can create increase the sensitivity of the sensor to light, allowing you to get great exposure in darker lighting conditions.

What is High ISO Noise

Eventually, as you raise your ISO you will get noise showing up in your image, and when this happens will vary depending on your camera. Increasing your ISO or the gain on the sensor making it more sensitive to light, is much milk turning a speaker volume up at some point you get a distorted noise. In the case of a camera, the distortion to the image comes called chrominance noise or colour noise and luminance noise, the latter usually being referred to as just noise. If you see small red and green blotches on your image that is Chromance noise. If you see a kind of black-and-white fine grain in your image this is called luminance noise.

The first part of the video below starts by describing how noise levels as you raise ISO vary depending on your camera, then what the noise looks like, and how to reduce noise in your images.

Other Ways to Deal with Low Light beside ISO

In the video below Tony Northop uses a provocative title “ISO does not cause noise” to get across the idea that noise is caused by a reduction in light going to the sensor, and increasing light in other ways should be considered first. He suggests you should make sure you have considered lowering your speed or lowering your ISO first before resorting to the ISO setting. The video below suggests an exercise to understand this relationship to light but letting your ISO float while keeping your Aperature and Speed manual. Then adjusting these two settings until you get the lowest possible ISO and still make the image you intended.

ISO as Speedometer

As I always shoot manually so I can control my camera more accurately, as a camera is an average instrument, I don’t let my ISO float. Letting it float in order to understand the relationship is a good exercise to understand how to set your f-stop and speed to an optimal setting, but my suggestion, especially if you are learning, is to have all three exposure settings in manual mode, otherwise, you may never get out of the automatic setting.

ISO Invariance

In the video below Tony Northop illustrates how some cameras are somewhat ISO invariant. ISO invariant means if take an image at a high ISO and then another one with the same aperture and speed with the ISO set to the base ISO, you will get a perfectly exposed image and a black image. If you take the dark image and correct the exposure in “post” and you find only minor differences in the noise level then you know the camera is somewhat ISO invariant. Why might this be important, in a situation where you have a camera like this and you have a situation where the light might vary a lot during the exposure such as photographing the Arora Borealis or fireworks, you may wish to dramatically underexpose to protect your highlights. One limitation you should be aware of is Lightroom does not, according to Tony, allow you to brighten the exposure more than 5 stops. In the video, he used a preset to do this and there is other software that you could use to overcome this limitation.

So are all cameras ISO invariant? The answer, of course, is all cameras are different so vary throughout the ISO range. Photos for Photos is one possible resource for researching your camera, but you could as Tony has done simply test your camera through your ISO range. This is likely not a bad idea as it will also tell you at what point your personal do not like the noise level. Through this process, you may discover your camera has dual gain architecture. In the next section, I will outline the advantage of these kinds of cameras and the potential drawbacks.

Dual Native ISO

The chart above, from 2024, shows most new sensors have a point where the noise increase reverses itself partly, this is called Dual Native ISO technology. If you examine this chart carefully it would appear it occurs as follows on the following cameras: GFX: 650 = 300, Nikon: 700 = 450, Canon: 800 = 350, Sony: 600 = 250, and Hasselblad: no dual ISO.

What does this mean practically? The GFX for example at 650 ISO has the same noise signature as 300 ISO. So if you find yourself at 300 you may as well go to 650 as there would be no difference in image quality. The following goes into more depth.

Two Optimal ISO settings (Dual Native ISO)

If you have a camera that was designed a year after 2015 it is likely you may have a camera that has two optimal ISO settings rather than one, this is found in cameras that have dual gain architecture. Sony in addition to making cameras also make camera sensors, and so often new innovations in sensors appear in their cameras usually a year before they appear in other cameras. If you look at the chart below you can see the evolution of the A7R series. In 2015 they came out with the A7R Mark II represented in black, which has two optimal ISO points, whereas the earlier model the A7R represent in black does not. The rising lines indicate the increase in noise levels on these cameras.

https://photonstophotos.net/Charts/RN_ADU.htm

In the chart below we are looking at a popular camera on the market the A7R III which came out in 2017, in the diagram below which graphs the noise level of the camera you can see at an ISO level of 640 an image will have a roughly similar noise level as the ISO 200 setting. This also means between above 200 ISO all the way to 500 ISO, photographs taken at these ISO will be noisier than ISO 640. This is the result of dual gain architecture. The link below the image will take you to a site where you can check your own camera.

https://photonstophotos.net/Charts/RN_ADU.htm

In the chart below we can see an APSC camera that was designed in 2020 by Fujifilm the X-T4 has a similar dip.

https://photonstophotos.net/Charts/RN_ADU.htm

Out of curiosity let’s go back to a camera that was designed in 2014 the Nikon D3300, you can see from the chart below that it does not have this dip in the ISO performance. The noise level increases steadily as the ISO goes up.

https://photonstophotos.net/Charts/RN_ADU.htm

You can see the age of the design has a relationship to whether or not there are two optimal ISO settings or just one. However, not all new cameras have new sensors if you look at the Fujifilm GFX50s II which was designed in 2021 but used a sensor that was much older you can see it does not have two optimal ISO settings.

https://photonstophotos.net/Charts/RN_ADU.htm

The newer GFX 100s has a dual gain sensor, so you can see in the chart once you reach ISO 200 you are better off moving directly to ISO 500, as it will have less noise than those settings between 200 and 500.

ISO Invariant Cameras are Not Perfect

We should not assume cameras that are somewhat ISO invariant produce images that are Identical whether you raise the ISO in the camera or in your post-processing software. In the following video Fstoppers Video, the following RAW processing software was tested: Capture NX, Luminar, Capture One, Darktable, Alienskin, Photoshop and Lightroom. Using his Nikon he underexposes an image four stops and took the same image at the correct exposure. He then compares these images in all the different software. He finds that his camera is not perfectly ISO invariant, as the underexposed files no matter which software he uses have more noise than the correct exposure. He suggests that this is the case with all cameras, in other words, there is no camera that is perfectly ISO invariant, just some closer than others. But he does say that with some skillful processing the overexposed images can come close to replicating the correct exposure. So if it is necessary you could use an underexposing strategy, if the situation calls for it, and do a good job of correcting it in your post-processing software but in most cases, the correct exposure is best.

References: