High ISO, friend or foe?
People have the habit of being afraid of what they don’t understand. They may even dislike that thing, or god forbid, hate the thought of something foreign to them. That’s what high ISO shooting is like to many a photographer, some even experienced with years, or decades, of experience. Is it perhaps that they were so used to shooting film like Velvia or Kodachrome that they never wanted to accept that modern technology allows us to utilise our equipment at a level never thought possible?
I learned not to be afraid of using high ISO a long, long time ago. Why?
Firstly, high ISO can save you at a moment when your trained mind tells you it’s impossible to get an image, because there is just not enough light. If you consider wildlife photographers before the digital age when film was king, and for them ISO800 film was considered way too fast and grainy. They had to shoot with ISO100 or ISO200 slide film and may have even pushed the film one stop during taking the shots to be able to get one extra stop of exposure during development. These were not necessarily the times for good action photos in low light, where lower ISO films meant slower shutter speeds, blurred images and nothing in comparison to what we can do today with the latest digital camera technology available to us. I now regularly shoot at ISO values of 3200 and 6400 and even up to 25600 if I really, really have to.
Secondly, I learned the ETTR theory. What is that you ask? Expose To The Right, that’s what ETTR stands for. It means you push your digital exposure as far right (on the histogram) as you possibly can, making the image look overexposed, then pull that exposure back during post processing. This method of RAW file capture really reduces the amount of digital noise generated during the image taking process. It won’t necessarily remove it all, but it will make further processing more effective and less visually overpowering.
Consider this: you want to use high ISO when the light is low, so you can get a fast enough shutter speed, even if it’s just to be able to hand hold your super telephoto lens and take a beautiful low-light portrait of your subject. How does your sensor record the light? It is covered in millions of photons (pixels) that capture light, which the camera then converts into a colour image. That is truly putting it in very simple terms. Noise will be more visible in dark areas of the image where there is less light and more amplification of the signals, especially as you push ISO higher and higher.
More about high ISO use
Take a look at this chapter of an e-book I wrote for one of my photography workshops. I’ll repost it here for clarity and to help you become a better digital photographer.
You often read that a good digital histogram will contain an even spread from dark to light tones. However, to get the best possible spread of data (this is digital remember) you need to shoot with the histogram pushed as far right as possible. While this just about throws the idea of metering and exposing out the window in the traditional sense of photography. Understanding how the camera meters will get you arriving at a better base exposure that you can then push to maximize the quality of captured data. Here is why.
Digital sensors record data in a number of stops, usually seven in 12-bit mode. Although, you now have cameras that can capture 14-bits of data, which is enormous! However, from what I have read having 14-bits over 12-bits does not give a huge advantage. In fact, it slows down your image making process. That is because the camera has to process more data than in 12-bit mode, 4 times as much actually.
A 12-bit sensor gives you 4,096 tonal values (2 to the power of 12 = 4,096). This allows for excellent tonal transition. The ability to record such a large number of tones should guarantee smooth transitions between the tones within the resulting image you may think. I am afraid it is not quite that simple.
Whilst you might think that each of the seven stops in the range of the sensor record an even number of tones throughout the dynamic range, you would be mistaken. F-stops are logarithmic in nature meaning that each stop records half of the light of the previous one. Practically, this means that the brightest stop records half of the possible number of tones, i.e. 2048, the second stop records half again (1,024), the next half again (512) and on and on until you reach the seventh stop that records a measly 32 tonal levels. Here is the news! If you underexpose an image and correct the exposure during in post processing, the tonal transitions in the darker areas will not be as smooth, and the risk of degrading your image quality is much higher. If you overexpose your image, by pushing the histogram to the right, you will capture much more tonal information that results in much better image quality when correcting the exposure in post processing.
I will now demonstrate with some images I took of one of my trusty dogs – Biscuit – on a dreary winter afternoon when it was dark and cloudy. I deliberately chose to shoot all images at ISO1600 with my Canon EOS 1DmkIIn camera body. Sure it’s a 1D series, but the principle remains the same for any digital SLR camera and should yield similar results overall. I took shots at five exposure correction settings in Av mode, using evaluative metering, which is my standard daytime operating method (EC = exposure compensation).
Two things to consider: I now only shoot in FULL MANUAL mode for complete control. I have no need to use any other mode than Bulb (B) when shooting very long exposures during star trails or auroras. And if I get a few blinking highlights I will accept those, because the review image on the LCD is based on a jpeg file that is generated from the RAW file in the camera for review. In reality, the RAW file will be far more effectively managing the dynamic range of each image and I have found more often that those images with few blinkies will have no blinkies once I open them in the RAW conversion software.
That image above shot at -1 EC actually means that there is loss of detail in the black tones. Also, looking at the image overall, it appears underexposed, that is dark. This is a time when I have seen and heard or read of many photographers who resort to Adobe Photoshop to improve their exposure. I also believe that many may be very surprised at the sheer amount of digital noise that is in the dark areas of the image as they brighten it. On the next page you will see what happens to the image quality when you use high ISO, underexpose it then believe that Adobe Photoshop will be the panacea to your fundamental exposure error.
You can see a 100% crop of the image here showing clearly that it is way underexposed. 100% crop basically means that you increase the viewing magnification to 100%, which will give you the detail that you need. There is clearly digital noise that is visible already. So many people would now increase the exposure during the RAW conversion process, assuming they use RAW files, which most serious photographers should really be using.
Here is the image after an exposure correction of +1 stop was applied in Digital Photo Professional. I feel that the image quality is really poor here as this increase in exposure also exacerbated the amount of noise. No wonder many folks complain that their cameras are noisy at high ISOs.
Below image is with no exposure compensation applied.
The histogram looks far better, but there is still an apparent loss of some of the blacks due to the histogram bunching up on the left hand side of the histogram. The right side is still empty so that indicates that the exposure can be pushed some more. However, most folks when seeing this image on the LCD will accept this image as well exposed. I am afraid that is not the case. It is not a transparency we are looking at, but a digital file. Let’s see the 100% crop from this image. There is still some noise, but not as severe as the image underexposed by 1 stop (-1EC). This means we could push it even further!
I am getting more data on the right side, with far less on the left. This is almost how I want it. I feel it could be pushed even more. Let’s look at what noise we can see when we view the image at 100%.
The last example below was pushed as far right as I could go without blinking highlights, that is: overexposed bright tones. It looks so bright, completely nothing like the atmospheric conditions were on the day, but I am trying to push exposure to the right to obtain the best digital file.
Now to get the image looking a bit more natural as to represent the real scene, I reduce the exposure by 1 stop in the post process.
Below, I will compare the first image (taken with a -1EC) with the last (taken at +1 2/3EC). In both instances the exposures were adjusted in Canon’s Digital Photo Professional software to provide the visually correct exposures.
You can see for yourself which is the best image to work with!
Those images of Biscuit were taken with a Canon EOS 1DMkII camera in 2012. That camera by then was already 8 years old technology, having been released in April 2004. Fast forward to 2023 when things like the Olympus OM-1 lead the way in the M43 scene with this amazing camera body.
Below are two examples of images captured at ultra high ISO values then processed carefully using Topaz Photo AI to get very, very useable images.
I saw this Grey Butcherbird at dusk when light was very low. The image below shows actual amount of light.
Below is an image of a White-bellied Sea Eagle captured using ISO20000. The result of the carefully processed image is quite remarkable.
We hope the above information, lots as it may be, will become useful in your understanding of managing the dreaded digital noise during your RAW file captures.