Back in the good old days when we were shooting film, ISO was a pretty simple concept. Each film type had a given speed and that’s what you’d use. Lower ISO films gave better results, with finer grain and detail and more saturated and realistic colour rendition (or with black & white, better tonality). High ISO films were, to tell the truth, pretty grainy: I remember using colour print and slide films around the ISO 800-1000 mark and being hugely disappointed by the results. I sometimes used Ilford Delta 3200 black & white film to shoot indoors without flash, but only to make small prints.
The ISO setting on a digital camera does something rather different than changing to a different speed film In essence, the film speed rating was a fundamental property of the emulsion – high ISO films were physically and chemically different from their low ISO cousins. If you processed your own black & white film you could adjust the ISO by manipulating the processing, but only by a stop or two. However colour film used standardised processes so there was little scope for manoeuvre.
Because of this, changing the ISO sensitivity during shooting meant using a different film. But if you just had one camera body this wasn’t really practical on a shot-by-shot basis. You’d have to rewind your film, making a note of the next frame number, and load another of the speed you wanted to use next. Then to change back you’d have to reload the first film and fire off a series of blank shots to advance to the next available frame. So serious photographers would often be caught carrying several
cameras loaded with different films. Those lucky enough to use medium format had the luxury of interchangeable backs. Then of course came digital and the ability to change the ISO on a whim. But what does this mean, in terms of how the camera and sensor actually work?
Digital and Variable ISO
 |
| Digital cameras make it easy to change the ISO for each shot |
When I got my first digital camera, the ability to change ISO on a shot-by-shot basis brought a wonderful new freedom to shoot in a wider range of light conditions. In 2017 we’re now pretty disappointed if a camera doesn’t give good results at ISO 1600 at the very least. Of course higher ISOs suffer from familiar penalties, with ever-increasing noise bringing decreased detail and poorer colour rendition. But even so we’ve seen some extraordinary cameras recently that give reasonable results right up to ISO 51,200 or so, such as the Nikon D500.
Despite our familiarity with ISOs that can be changed at will, and cameras that work across a huge range of settings, it’s fair to say that many photographers probably haven’t given much thought to how the ISO control on their camera actually works. But understanding this can help with using your camera to get the best results.
How Does Digital ISO Work?
It’s tempting, when changing your camera’s ISO setting, to think that it’s like changing the film, with the control manipulating how the sensor reacts to light. But in fact it does nothing of the sort – changing the ISO has no influence whatsoever on the sensor itself. Instead, alongside changing the light metering, it changes the way the camera processes the data that’s read from the sensor after the exposure, with the aim being to give the best possible results with the weaker signals that are generated when the sensor is exposed to less light.
To understand what this means, we need to consider how a digital camera actually makes an image file – either raw or JPEG. When the sensor is exposed to light, each of the pixels builds up an electrical charge, dependent upon how much light it receives. This is read off the sensor as an analogue signal, and needs to be converted to a digital value so the camera’s processor can convert it to an image file. This is done using one or more analogue-to-digital (A/D) converters. The digital image data can then be recorded as a raw file, or further processed to make a JPEG.
However, each stage of the process can add undesirable noise to the image. Crucially, if the signal fed into the A/D converter is too weak, then it can be overwhelmed by the noise added at this stage, making the image fi le unusable. For this reason, an amplifier is placed between the sensor and A/D converter to boost the signal. In the simplest terms, increasing the ISO setting on your camera introduces more amplifi cation at this stage, so that weak signals from the sensor can be successfully converted to digital output. The amplifi er may itself add noise, but as long as this is much lower than the signal it outputs, the process is beneficial.
 |
| Raising the ISO setting increases the amplification of the analogue signal read from the sensor, prior to its conversion to a digital output |
The various forms of electronic noise added by the camera are collectively referred to as ‘read noise’. In practice, they tend to limit the highest usable ISO, and the maximum dynamic range at low ISOs. On older cameras this could be signifi cant, but more recent models have done a remarkable job in reducing it to very low levels indeed. This means that the fi les they produce are very accurate representations of the data originally recorded by the sensor, regardless of how much light any individual pixel received. This brings some key benefi ts – not only are very high ISOs more usable, but photographers can also recover more detail deep into shadow areas of low ISO files.
Where Does Noise Come From?
Of course even the best cameras still show ever-increasing noise in their image fi les as the ISO is increased, so where does this come from, if not from the camera’s electronics? The answer is from the nature of light itself.
Light is made up of discrete particles called photons, and most of the noise in our images is due to the fact that adjacent pixels don’t capture exactly the same number of photons, and therefore record exactly the same signal, when they’re exposed to an even intensity of light. If the light levels are high, the differences are barely visible, so noise is minimal. But as the illumination falls, the variation
between the pixels increases, so noise becomes increasingly more obvious. Put simply, the lower the light, the higher the noise, due to the particulate nature of light.
Where noise from the camera’s electronics does matter is mainly in the dark tones of the image file. Here, read noise from any step of the signal pathway can swamp details in the image itself. In practice this tends to limit both the highest usable ISO setting, and the deepest shadow detail that can be seen in raw files.
Pixel Count and Noise
In the early days of digital, it became received wisdom that the more pixels a sensor had, the higher the noise levels would be. With older cameras this looked self-evident: higher pixel count sensors could certainly give more resolution, but at the expense of poorer results at high ISOs. For example, in late 2008 the 12-million-pixel Nikon D700 full-frame DSLR gave clearly better results at ISO 3200 or higher than the 24-million-pixel Sony Alpha 900, even when the latter’s fi
les were downsampled to match the D700’s. But sensors have progressed signifi cantly since then, and this ‘rule’ no longer necessarily holds. In fact, if we look at the evolution of cameras since then, megapixel counts and high ISO capabilities have gone up hand-in-hand.
Today’s 24-million-pixel APS-C cameras give vastly better high ISO capability compared to the 12-million-pixel cameras we were using 10 years ago. It’s still true that if we examine images at the pixel level, and particularly if we refuse to apply any noise reduction, sensors with higher pixel counts give noisier images. But if we compare them at the same output size, the differences now all-but disappear.
 |
| Comparing files from the A7 II , A7R II and A7S II at ISO 12,800 shows that higher pixel counts don’t necessarily always mean increased image noise |
The best example is Sony’s Alpha 7 II triplets: the 12-million-pixel A7S II, 24-million-pixel A7 II, and 42.4-million-pixel A7R II. Of these, the A7 II actually performs worst at higher sensitivities of ISO 6400 or more, with the A7R II’s incredibly impressive sensor only starting to fall behind the A7S II at ISO 51,200 or more. So it’s the quality of the sensor that matters, not the number of pixels.
Sensor Size and Noise
The fact that most of the noise in our photos comes from the light itself also explains very well why larger sensors perform better at any given ISO setting. There are several ways of framing the explanation, but the simplest is that for a given shutter speed, aperture setting and light level, a larger sensor simply captures more light overall, essentially in direct proportion to its area. So there’s roughly a stop’s worth of improvement at each step in the progression of the most common sensor sizes: 1-inch, Four Thirds, APS-C, and full frame.
This means that in theory, images will look similarly clean at ISO 100 on a 1-in sensor camera, ISO 200 on Four Thirds, ISO 400 on APS-C, and ISO 800 on full frame – and in practice these do indeed all tend to look clean and practically noise-free. Likewise you can expect similar amounts of high ISO noise at ISO 1600 on 1-in, ISO 3200 on Four Thirds, ISO 6400 on APS-C, and ISO 12,800 on full frame. Again, in our experience this refl ects the highest settings that usually give good results in practice, before noise becomes overly intrusive. However some cameras do give usable results at higher ISOs.
 |
| This comparison illustrates how high ISO noise depends on sensor format. Each step-up in size equates to approximately double the sensor area and a corresponding 1-stop improvement in high ISO performance |
ISOLESS CAMERA
 |
| On the Nikon D810, underexposing by 5 stops at ISO 100 and adjusting in raw processing gives the same noise as shooting at ISO 3200 |
Now you might be wondering if modern cameras can record low-level signals very accurately, why do they still add an amplification step prior to A/D conversion. As it happens, on the best modern cameras it’s not really necessary, and in terms of image quality you’ll get similarly good results to using a high ISO by shooting with a low ISO setting (and therefore underexposing), then pulling-up the image brightness in post-processing.
Indeed this is the basis of the highlight dynamic range expansion modes found on Canon, Fujifilm and Pentax cameras. The advantage of shooting this way is that you can avoid clipping highlight detail, and retain a very broad tonal range in your raw files. Naturally some cameras work better than others in this respect, and you’ll find some people refer to the best performers using the somewhat confusing term ‘ISOless’.
However working in this fashion can have operational disadvantages: for example you’ll often end up with files that are too dark to be useful for viewing images in playback.
0 comments:
Post a Comment