What is ISO Invariance and How Can Advanced Photographers Use It to Maximize Dynamic Range in Low-light
I feel like the topic of ISO invariance doesn’t get the attention it deserves, specially considering the very real dynamic range gains it offers to advanced photographers in low-light conditions with high-end cameras. It may have to do with the fact that it’s a complex and confusing topic. At least that’s my sense, based on the discussions that I have encountered in various fora online. I’m going to make an attempt here to clear up some of that confusion by re-framing the way the subject is tackled — instead of assuming ISO as a fact of life, I’m going to take the approach that we really shouldn’t need ISO for digitally photography, the reasons why that’s not the case and finally, why ISO invariance may eliminate those reasons and, therefore, ISO with it.
Digital photographers accept ISO as one of the three pillars of the exposure triangle. But it doesn’t have to be. The concept of ISO is a hold-over from the days of film where one way to control exposure was via the light sensitivity of the film. The latest standard for characterizing this sensitivity came from the International Organization for Standardization, which named it ISO (apparently, it is not an acronym). But there’s no film in digital cameras, the sensitivity of the sensor is fixed and photos can be brightened digitally in post-processing software. So what does ISO mean in the context of digital photography and why is it needed?
To answer that question, we need to understand something called SNR or signal-to-noise ratio (don’t run away, it’s not as scary as it sounds). When a camera sensor captures light, it also captures random noise. This noise is a fact of life and is always there. As the amount of captured light increases, so does the noise but not proportionally and by a much smaller amount. If we were to quantify the quality of the signal captured by looking at the ratio of the signal level to the noise level — SNR — then the more the light captured, the higher the SNR and the higher the SNR, the better the noise performance (less visible noise) of the captured image.
In an ISO-less world then, in a low-light situation, we would capture a signal with a low SNR (low light = low signal). When we boost that image digitally in post, the signal would increase but the noise would increase proportionally. The SNR would stay the same (poor) and we’d get a brightened image that happens to look noisy. And that would be that.
As we know, things are never that simple and there’s another variable that we haven’t considered that changes things quite a bit — noise added by the camera. The process of capturing the incident light and translating it to a digital value (which is then written to a storage device like a memory card) adds a noise component. So when we brighten the image in post, we’re amplifying this noise in addition to the random noise. This will degrade the SNR further.
Now imagine if could avoid brightening the image in post. One way would be to ‘brighten’ the image in camera, by, say, analog amplification. Imagine an analog amplifier between the sensor and the memory card. This amplifier would take the signal and boost it until it reaches the same luminance level we’d have applied in post, but it would deliver a significantly better SNR. Why would the SNR be better? Because only one of the noise components would be amplified (the random noise) and not the camera noise. This process is achieved via ISO on digital cameras. Instead of making the sensor more sensitive to light, we just amplify the signal being received.
There are many noise sources inside the camera. For the sake of simplicity, they are not all considered here. The overall effect presented here is more or less the same.
Sounds like ISO is good thing, then. Yes. Except that dialing up the amplifier gain, reduces dynamic range. It does that because the circuitry inside the camera, due to signal amplification, is operating at higher saturation then it normally would for that amount of light. It, therefore, has less headroom. So yes — ISO is good a thing — it helps us capture less noisy images in low light, but at the cost of dynamic range.
Now imagine a scenario in which there is no camera noise and the only noise we need to consider is the random noise. In this case, having the amplifier in the camera offers no real benefit and we can achieve in post-processing the same result we’d have achieved via in-camera analog amplification. That would allow us to take images without compromising the dynamic range and when we brighten them in post, the result effectively would be the same as if we had use ISO. That behavior is what we call ‘ISO invariance’.
Many high-end cameras on the market today are on a spectrum where some could be considered ISO invariant. There are many resources available online where photographers can learn more about the ISO behavior of their gear.
Another key thing to note is that many cameras have two types of amplification — analog and digital. The digital amplification is no different from what is done in post-processing and therefore offers no value and, on the other hand, significantly reduces dynamic range. The only reason I can see for using it is to generate an image preview in very low-light situations. The analog amplification is sometimes referred to as native ISO. Photographer using ISO invariant cameras should try to use the lowest value in the native ISO range that produces an image that can recovered in post. That might require some experimentation. Lowering the ISO will provide greater dynamic range, as discussed above.
And there you have it: ISO invariance can allow a photographer to preserve a significant number of stops of dynamic range in low-light scenes by using an ISO value that is 4–5 stops below what would be needed for optimal exposure and then making up for that gap by boosting the image in post-processing. As camera sensor noise performance continues to improve, it is possible to imagine a world with ISO-less cameras. And while that’ll certainly be bad news for the exposure ‘triangle’, it’s a development that photographers will certainly welcome.