To put it very simply, an ISO Invariant sensor will allow you to increase the 'Exposure' in post-processing with minimal quality loss. An ISO Variant sensor will show severe image degradation if you try to increase the 'Exposure' in post-processing. Therefore, an ISO Invariant sensor gives you a lot more flexibility. You can intentionally underexpose a photo, to preserve details in the highlights, and then brighten the image as needed. You must be shooting in RAW of course.
To understand how this actually works, you need to have a solid understanding of ISO first. It also helps to understand how your camera's sensor actually works. Let's cover that first.
Light enters the lens and travels through to the camera's sensor. The sensor is made up of millions of tiny light wells (pixels). These light wells are designed to capture and store light for the duration of the exposure. As the light wells (pixels) gather more light, they create a stronger electrical charge. Once the exposure has finished, the light wells are covered and a signal is generated. This signal has the measurements from all of the light wells. At this point, the ISO comes into play. Depending on which ISO you selected, the camera will now amplify the signal. If you used a high ISO, the signal will be amplified more. In other words, your camera will brighten up the image. Click here for more information on this initial process.
To be clear, a camera sensor's sensitivity to light cannot be changed, it has a 'quantum efficiency'. This is usually marked as a percentage, for example - 48% QE. The higher the quantum efficiency, the more sensitive a camera will be to light. Click here to see a chart of Nikon cameras and their QE ratings. Unfortunately, most DSLRs are rated around 50% QE, or worse. That means they really aren't that sensitive to light, compared to something like a monochrome CCD-style camera. If your camera has a 50% QE rating, then you are losing one stop of light. That's like using an f/4 lens instead of an f/2.8 lens, or using a 30 second exposure instead of a 60 second exposure. The main culprit of low-sensitivity cameras is the Bayer Array.
The Bayer Array is a grid of colored filters over the sensor's light wells. Think of a checkerboard with Green, Red, and Blue squares. These filters are designed to only allow a specific color through to the light well. For example, a Green filter will only allow green light through, it will block the red and blue light. Most Bayer Arrays have 50% Green, 25% Blue, and 25% Red. This isn't necessarily a big problem during the day. However, for astrophotography this is much more concerning. Since the Bayer Array prevents a lot of light from actually reaching the light wells, it reduces the camera's quantum efficiency.
If you managed to find a DSLR that didn't have a Bayer Array, it would have a much higher quantum efficiency and your images would be cleaner, sharper, and more detailed. (These are almost impossible to find though) The downside of a camera without a Bayer Array is that it can't determine color, so you will only be able to capture monochrome images. However, you could add Red, Green, and Blue filters to the front of your lens and use some post-processing techniques to create a beautiful, full color image!
An IR Cut filter also blocks light from reaching your sensor. Every DSLR has one (unless you mod your camera). This IR Cut filter is designed to block Ultraviolet (UV) and Infrared (IR) light, which our cameras can actually detect. If your camera did not have an IR Cut filter, every image would include Visible Light, UV, and IR. This would create some pretty weird looking photos. Therefore, camera manufacturers always include the IR Cut filter. We want our cameras to capture what our eyes can see, after all. Unfortunately, these IR Cut filters can also block some visible light, usually in the reds. Therefore, an IR Cut filter also reduces the quantum efficiency of a camera sensor. Again, this really isn't a big deal, unless you are doing astrophotography.
New photographers are often told "A higher ISO increases the camera's sensitivity to light". It doesn't work quite that way though. Your camera's sensitivity to light is largely static, and can't be easily changed. This is called the 'quantum efficiency'. This is a rating between 0% - 100% and it measures how sensitive the camera is to light. At 50% QE, the camera is only able to capture and use half the light that reaches the sensor.
As Roger N. Clark states, "ISO is a post-sensor gain". In other words, the ISO has no effect until after the light has been captured by the sensor. Once a signal has been generated by the sensor, at the end of an exposure, the ISO will amplify it. This signal amplification will brighten the image.
As the ISO is increased, the image becomes brighter. However, the actual amount of light does not change. The amount of light captured depends entirely on the shutter speed and aperture. If you take a longer exposure, you will capture more light. If you use a wider aperture, you will capture more light.
To put it very simply, increasing the ISO will brighten the light that was captured during an exposure. It will not change the camera's sensitivity to light.
Have you ever heard white noise or faint static coming from speakers or headphones? Once you start playing music loudly enough, you don't notice the faint hissing noise anymore. In relation to cameras, this is called the Signal to Noise Ratio. Basically, you need to capture enough light to cover up that faint noise generated by the camera's electronics. Older cameras have more 'white noise' than newer ones, meaning you need to capture even more light to cover it up. This is the reason why photos taken in low light are grainy! The faint signal (light) is not strong enough to overpower the camera's ever-present noise. As you increase the ISO, that noise is also amplified, along with the light captured. This makes it even more prevalent. However, if you put the camera on a tripod and use a longer shutter speed, you can capture enough light to create a clean image, without grain.
Therefore, the best way to increase your signal to noise ratio is by using a longer exposure! The comparison below shows example of a 20 second image vs a 4 minute image. As you can see, there's a massive improvement! The key is to capture enough light to overpower the camera's electronic noise, which obscures detail in the shadows.
While we are on the subject of Signal to Noise Ratio, let's expand upon its relation to ISO Invariance. As you can see in the image below, the 20 second exposure looks terrible, even though I used an ISO Invariant Sensor. It's important to understand that an ISO Invariant sensor isn't a magical cure to low-light shooting. If you are shooting in very dark conditions, you will always need to capture as much light as possible. No matter how good your sensor is, extremely low light conditions will always cause problems.
If you continue to scroll down you'll see a photo of an owl, which demonstrates my ISO Invariant camera's abilities. There's a key difference here though. To the naked eye, there was plenty of light when I took the image. I could clearly see the owl, and it was about 20 minutes before sunset. The reason the photo is so dark is because I was at f/6.3 with a fast shutter speed (1/200 roughly). In other words, there was still plenty of light to overcome the Signal to Noise Ratio. However, in the comparison below it was nearly pitch black. With so little light, I could not overcome the Signal to Noise Ratio with a 20 second exposure. However, after 4 minutes, I captured more than enough light.
Some newer camera sensors are ISO Invariant. This means you can shoot at a low ISO and brighten the Exposure in Post-Processing without any quality loss. However, most cameras on the market are ISO Variant. On an ISO Variant Sensor, brightening the exposure in Post-Processing could ruin the image quality. That's why it's generally recommended to use the proper ISO in-camera. Keep in mind, you need to shoot in RAW for this to work properly.
However, an ISO Invariant Sensor isn't a magical fix for low-light scenarios. If you don't capture enough light to overpower the Signal to Noise Ratio, your images will still look terrible. Therefore, it's usually recommended to use a wide-open aperture (when possible), and a long shutter speed to capture as much light as possible. The more light you can capture, the cleaner your images will be.
If you skip down to the Further Reading section you can find different articles showing examples of ISO Variant and ISO Invariant sensors.
This morning I headed down to the local dark sky park, Scenic Vista, to capture my test images. Using a Nikon 14-24mm lens at f/2.8, with a 20 Second Shutter Speed, I took 7 photos. I started at ISO 6400 and decreased the ISO by 1 Stop each time, down to ISO 100.
First, let's take a look at the ISO 100 example. The 'Before' image is ISO 100, clearly way underexposed. The 'After' image is the same photo, taken at ISO 100, but brightened by 6 Stops in Adobe Camera RAW. It's a night and day difference!
If the Nikon D750 is ISO invariant, then the brightened ISO 100 image should look nearly identical to the ISO 6400 photo. Let's do another comparison. The slider below shows the ISO 100 image brightened by 6 Stops on the left, and the ISO 6400 image on the right.
They look nearly the same, but we need to take a closer look. I cropped in heavily into the center of the foreground. You should notice a slight difference now. There is a weird color grain problem affecting the ISO 100 image. After brightening the photo 6 Stops in Post-Processing, it's expected that it would have some quality loss.
Finally, let's look at the stars. I don't see any real increase in noise between ISO 100 and 6400. However, I do see a slight color noise problem. It looks similar to what we saw in the Ground Crop photos above.
After directly comparing the brightened Low-ISO photos and the High ISO photos I am very surprised by the results. There is only a slight loss of quality when brightening the ISO 100 photo by 6 Stops!
Based on these results, it's clear that the Nikon D750 has an ISO Invariant Sensor. This is very important for night photography especially. No matter which ISO you choose, you can always brighten the Exposure in your Camera RAW processor without any major problems. In fact, you can intentionally underexpose the photo to save detail in the highlights, and then brighten the Exposure in post-processing. I will be using this method when I do light painting in my scenes, which tends to get overexposed at ISO 6400.
Another benefit of having an ISO-Invariant sensor is for wildlife photography. For example, the image below was taken at 1/320s, f/6.3, and ISO 200. It was getting dark, and there wasn't much light; plus, I needed a fast shutter speed to freeze any motion on the owl. I deliberately underexposed the photo to see how far I could stretch it in Post-Processing. I increased the Exposure slider in Adobe Camera RAW by 3 Stops to create a beautiful final image! If you accidentally underexpose a wildlife photo, rest assured you can salvage it!
Many new photographers are told that increasing the ISO increases grain. As we've seen, this isn't exactly the case. Your photos are grainy because you didn't capture enough light! There is an important distinction to make here. When shooting at night, the camera settings usually don't change. I use a Wide Open Aperture (f/1.4 - f/2.8), a long Shutter Speed usually around 20 seconds, and ISO 6400 most of the time. If the ISO is the only setting being changed on an ISO Invariant sensor, there should be no noticeable difference in grain between ISO 100 and ISO 6400. That's because the images had the same aperture and shutter speed - I.E. the same amount of light. The examples above proved this.
If I am photographing during the day and I increase my ISO, there will be more grain though! Why is that?
The ISO 100 photo had a Shutter Speed of 1/8 sec. The ISO 12800 photo had a Shutter Speed of 1/160 sec. That's a loss of 4.3 Stops of light (over 16 times less light!) I was also shooting in Aperture Priority Mode. This is a key distinction. When you use an auto mode (Aperture Priority, Shutter Priority, Program), the camera will automatically adjust the Shutter Speed / Aperture when you increase the ISO. Therefore, a higher ISO will ultimately cause less light to reach the sensor; the camera will use a faster shutter speed or a smaller aperture.
This example clearly shows that it is the amount of light reaching the sensor that causes grain, not the ISO. If you want less grain in your photos either use a Longer Shutter Speed or a Wider Aperture.
Of course, this isn't always possible. When I photograph birds I need to have a fast Shutter Speed. My lens can only open up to f/5 - f/6.3. I'm stuck; my Aperture can't get any wider (unless I spend $10,000 on an f/4 lens) and my Shutter Speed can't get much shorter (or I risk motion blur). I have to accept that my images will have more grain in them simply because less light is reaching the sensor.
Gathering more light is critical for higher-quality night photos. The best way to gather more light is a longer shutter speed. The main problem we face is the earth's rotation. The longest shutter speed you can use at night is about 20 seconds. That's nowhere near enough time to capture enough light, especially on a dark night. I recommend taking much longer photos to increase the Signal to Noise Ratio and have noise-free images. You will need a Star Tracker for this method though. Head over to my Star Tracker Tutorial for more information.
With a s star tracker, you can shoot over 4 minute exposures without star trails! Now that you are capturing so much more light, your photos no longer suffer from image grain, thermal noise, desaturated colors, and mottling. This is truly a game-changer for astrophotography! The only downside is that you will always need to take 2 images. One photo will have sharp stars, but a blurry foreground. The other photo will have a sharp foreground and blurry stars. I explain this full process in my Star Tracker Tutorial series.
If you want better photos at night, a star tracker is the best way!
I highly recommend reading this article from Roger N. Clark on ISO. If you scroll down to Figure 2, you will see a graph about the Canon 5D Mark II. As Roger explains, the camera eventually reaches a point where increasing the ISO has no practical benefit. In fact, you begin to lose Dynamic Range (mainly in the highlights) as you continue increasing the ISO. Therefore, you'd be better off leaving the ISO at that point, in this case 1600, and increasing the Exposure in post-processing as needed. This will help to retain detail in the highlights. Instead of being ISO Invariant at ISO 1600, my D750 is ISO-Invariant at roughly ISO 200, allowing me much more dynamic range and flexibility!
This article on Lonely Speck explains how to find the best ISO for Astrophotography. In fact, this is the article that first got me interested in ISO Invariance!
Rishi Sanyal has created a great post on how to effectively use an ISO Invariant Sensor. In this article he covers some real world uses for an ISO Invariant sensor. Read his article over on Digital Photography Review.
Daniel Laan also tested out the Nikon D750 for ISO Invariance in this Fstoppers article. Check out his article for even more comparison photos and information on the Nikon D750 as well as some ISO Invariant tips.
Spencer Cox recently published an extensive look at ISO Invariance for Photography Life. This article is the best resource on ISO Invariance that I've found so far.