One of the main basic advantages of using a digital still camera is the ability to chose which ISO setting to use. In film camera days the type of film you used in the camera controlled this, indeed film was classified according to it's ISO rating as well as whether it was colour or B&W, print or slide. So what is the ISO rating, what does it mean, and how does it affect how a camera takes a shot? In particular how does the ISO rating for digital cameras differ from that for film? And what are the image quality implications involved?

ISO

The acronym ISO stands for the International Standards Organization which is the body that sets and oversees the control and maintenance of international standards used worldwide in many differing spheres. Usually the ordinary person is unaware of these but sometimes they are quoted or used in conjunction with products and services that routinely affect everyday life. Often a product or service is quoted or marked as confirming to a particular ISO standard to convey the message that it meets certain standards required of it, and consumers are advised/warned not to purchase goods which do not meet these standards. 

ISO film standards

In photography ISO has become synonymous with film rating. All films are given an ISO value, generally known as the speed rating, although in actual fact the ISO rating signifies the Exposure Value (Ev), the aperture and shutter speed combinations needed to correctly expose the film. So it has long been understood that if you wanted a high quality image with fine detail you used a 'slow' film, one with a low ISO rating, and if the light levels were poor, one with a 'fast' rating, a high ISO. When using a high ISO film it was accepted that the trade off was a reduction in detail and overall image quality.

The reason why films are so rated is easy to explain, and is related to the size of the chemical grains used on each particular type. The smaller the grains used, the finer the possible detail reproduction, but the longer they need to be exposed to light. By contrast the larger the grains the less exposure they need, but the coarser the image quality, the less detail visible. This has translated for the majority of camera users into the concept that 'slow' films mean long exposures/slow shutter speeds, whilst 'fast' films mean just the opposite, short exposures/fast shutter speeds.

As a result the type of film you choose is important since it determines the overall image quality that results, low ISO films providing fine detail and richly saturated tones and colours, high ISO films having coarser detail with muted tones and colours. 

Digital Camera ISO sensitivity - and noise

Although digital camera ISO is perceived as being exactly the same as that of film, because virtually all digital cameras are designed so that their sensor sensitivity ratings align with those of the familiar ISO film ratings, this is not actually so, although it appears to the camera user to work in much the same way, raising the ISO resulting in higher shutter speeds and smaller apertures, and lowering it longer one's and larger apertures. The basic image quality also follows the same general pattern, in that quality falls as ISO speeds rise, and rises as they lower, but all is not quite as it seems.

Just like film, digital camera sensors are sensitive to light. When light falls on them they produce (analogue) electrical signals which are used to construct the image information. And as they are of fixed construction and the pixels they contain are of a fixed size the amount of light they require remains constant and so does the signal produced in any given situation. But you don't change the sensor and thus the size of the pixels every time you change the ISO setting as you do with film when choosing which speed rating to use.

So the first notable difference is that changing the ISO has no bearing on the detail captured by the sensor, the pixels on the sensor staying exactly the same size, and the level of detail produced remaining the same, being dependent solely on the number of pixels present on the sensor. The second is that whilst resolution is constant, image quality does reduce as ISO speeds rise due to increased noise levels, giving a worsening effect to image quality.

As the sensitivity of a sensor can't change, enabling a range of ISO settings to be provided is achieved by amplifying the signal from the sensor. Like nearly all electrical circuits digital camera sensors are far from perfect, and produce random errors in the signal, commonly known as interference or noise. Noise in digital images appearing as 'graininess' or specks of false colour. Although it increases with longer exposures and at higher temperatures, for normal photography the amount of noise produced by a sensor is fairly consistent. At low ISO settings the level of noise is usually very low in relation to the signal so it doesn't have a huge impact on the picture quality.  As the ISO settings rise the camera's processing has to amplify a weaker signal, and unable to distinguish what is image information and what is noise, everything gets amplified, with the result being a signal with a greater proportion of noise in it. The relationship between the relative strength of the signal containing image information and noise is known as the Signal to Noise ratio - S/N - and the lower this is the more noisy images appear.

Most DSLR's produce essentially noise-free images at their lowest ISO settings, usually ISO 100-200, and because of their relatively sensitive (large pixel) CCD or CMOS sensors don't suffer too badly from noise at higher settings either. There are a few DSLR's that can produce perfectly usable results at ISO 1600 or even ISO 3200. Small sensor compacts by comparison tend only to be able to produce low noise images at their very lowest ISO setting, usually ISO 50-100. At higher settings noise becomes very intrusive, and once you get to ISO 400 or higher the results are so noisy that they are only suitable for very small prints. This is particularly the case with the high megapixel digicams now available.

Normal and extended ISO ranges, noise and pixel over-saturation

As a general rule a sensor produces the best image quality at the lowest ISO rating available, the lowest ISO rating provided needing the least signal amplification and thus having the lowest noise levels and producing the best image quality. Some DSLR's, and mainly the Sony sensored 6mp CCD APS-C types until the arrival of the Sony sensored 12mp CMOS types, have sensors with a very high sensitivity because of the design and size of the pixels and ISO200 is the lowest sensitivity setting available. As pixel counts have risen lower base ISO sensitivities have resulted due mainly, but not entirely, to the smaller pixels which are used. ISO100 has been the normal base ISO for most DSLR's and digicams.

In the main the ISO range provided for use with a digital camera reflects the quality and design of the sensor, the size of the pixels, and the processing used in it. Until recently most have had a range covering 3-5 stops, digicams normally ISO100-400 and DSLR's ISO100-1600. The newer digicams now have ranges matching those of DSLR's, anywhere from ISO50-3200. How useful most of these higher ISO sensitivities actually are is the subject of much debate. 

The larger the pixels on a sensor are the more sensitive they are to light and the better signal they collect. However, there has long been demands from many camera users for higher/lower ISO sensitivity settings than those usually provided, one reason being the misguided perception that lower sensitivity means higher image quality as in film use, whilst another is the requirement for both longer shutter speeds when 'fast lenses' are used in bright light, and shorter one's in low light, so this is changing, especially with digicams. Unfortunately doing so actually results in much reduced image quality in many cases as a result of higher noise and the over-saturation of the pixels which takes place. To understand why it occurs you first have to look at the concept of how digital camera ISO sensitivity is provided.

If a sensor produces a certain signal level at it's base or optimum sensitivity setting, say ISO100, then setting ISO200 halves the exposure and thus the signal generated is weaker. So the signal then has to be amplified to twice it's original level to get the correct value. If ISO400 is set it's quadrupled since at this ISO the exposure is a quarter of ISO100. This carries on up the ISO scale so that at ISO3200 the signal is being amplified by x32, and at ISO 6400 by x64. Not surprising then that images taken at these settings aren't thus as good as those of lower ISO's. 

If you the work the other way around, down the ISO scale instead of up, and past the optimal signal level whatever that is, a different scenario occurs. The signal needs reducing, halving with each stop, because with lower ISO's the exposure is longer, it keeps doubling. But there is another problem here. Because the pixels are a constant size, whilst reducing the light means a weaker signal, increasing it beyond the pixels maximum collection capacity means over-saturation of the pixels occurs, they can't cope with amount of light, it's just too much information. This ends up leading to reduced image quality via a lower dynamic range. A good analogy to use is a liquid container. Pouring in less is not a problem, but pour too much in and the surplus overflows and is lost. The same thing happens when lower ISO's are used than are optimal for the sensitivity of the sensor. Information is lost and when the subsequent signal is reduced, the wrong information results.

It is for these reasons that camera makers restrict the ISO ranges provided to those they feel give acceptable image quality for the setting used. Some give two alternative ranges, standard and extended. The extended range being offered on the basis that the image quality level might not be all that the user expects, and it is their choice as to whether they use it or not. This is a known problem with Canon's pro DSLR's provided with an ISO50 option. And Sigma's SD14 which has also been given one via a firmware update - along with a warning about reduced DR and image quality. It also seems to exist with the new Sony 12mp CMOS sensors now being used. They appear to have optimum ISO ratings of ISO200 but are being provided with as standard - not part of an extended ISO range -  a base ISO of 100, which has a reduced DR range compared to the ISO200 setting. This is certainly the case with the new Sony A700 DSLR. Nikon's D300, which uses the same sensor, appears to have a standard range of ISO200-1600. Nikon's D3 also starts at ISO200 but goes up to ISO6400 reflecting the lower noise levels a full frame sensor with large pixel sites produce. In it's extended range it goes to ISO12,800 and ISO25,600, quite unbelievable settings.

Noise Reduction

Most digital cameras have processing algorithms that are used to reduce excessive noise. The amount of reduction, the strength of the effect produced, varies according to what the camera maker concludes is the best balance between the reduction achieved and the loss of sharpness that occurs, because reducing noise results in blurring of the image details, which many now refer to as 'smearing'. To offset the smearing that occurs extra sharpening is often applied. Together these can have undesirable consequences for an image. It may be less noisy, but with less detail and of higher contrast since sharpening is achieved by increasing contrast between adjacent pixel values.

Until recently these levels of sharpening and noise reduction could not be altered. Occasionally the user had the option of whether to apply noise reduction or switch it off, but nothing else. Now the latest cameras, mostly DSLR's, are increasingly offering choices as to the levels of both noise reduction and sharpening applied individually. This is one of the better and more useful options to arrive because often the noise reduction and sharpening has a worse effect on the image than the original level of noise. It gives the user the opportunity to 'tune' the level of noise reduction and subsequent sharpening applied to suit their own particular preferences. 

At present the level of noise reduction applied with some of the newer sensor designs to emerge that apparently have low noise levels cannot be altered. This is because this process now occurs on the sensor as part of the basic design, and before the information leaves the sensor.

Noise levels in images

The noise levels found in digital camera images vary not only from camera to camera, the size of the sensor and it's pixels, the processing used, but also according to the lighting conditions under which they are taken. A basic tenant of digital camera sensor capture is that as exposure times increase so do noise levels. This is irrespective of the ISO sensitivity used. So in bright lighting conditions, where tonal values are high and shadows are few there will be low amounts of noise. In these circumstances the difference in noise levels between low and high ISO's may be small and hard to spot. But in lower lighting levels, where darker tonal values predominate, noise will be present to a much larger degree, with a far wider gulf between low ISO and high ISO use.

So beware when setting ISO levels with a digital camera. Wonderful thought the concept of changeable ISO speeds undoubtedly is, a measure of practical use to establish the best settings to use is always advisable. Unlike with film use, a higher/faster ISO setting may in some circumstances actually turn out to produce a superior image quality to a lower/slower one. The best ISO varying with each and every camera and the particular conditions under which it is used.

For the conundrum is that whilst DSLR users wish for lower ISO's and the sensors optimally use higher one's, the reverse is true for digicams with their slow lenses. They need high ISO speeds but decent image quality only results when low ISO's are employed, in some cases as low as ISO50.

ISO image comparisons

To finish with here are a few examples to show the difference in image quality at various ISO speeds from both a digicam and a DSLR. It should be understood that making comparisons is not quite as clear cut as some would have you believe. Valid comparisons are only really possible when the images used are all taken at the same time under the same lighting conditions, and any processing variables are taken into account. Getting all the images to the same tonal values for each tonal step. Because noise levels within any image appear worse as tonal values are darkened, and appear better as they lighten.

The digicam used for this is an Epson 3mp digicam from 2000, we don't have a higher resolution digicam than this because we just haven't found the need to acquire one. So the noise levels will not be as high as those from the newer high resolution digicams nor does it have a wide ISO range, just 100-400. The DSLR is our 8mp Canon 350D from 2006. The noise levels this produces are about the same as our 6mp Pentax *ist-D DSLR from 2003 but the 350D has been used as the base ISO matches that of the digicam. The 350 having a range 100-1600. The *ist-D 200-1600 (200-3200 extended)

This is the test chart used. Six base colour squares, RGB blue, green, red, yellow, magenta, and cyan, along with a greyscale wedge on a mid-grey (50%) background. The accurate colour reproduction of the squares is not of particular importance in this case where noise levels are being assessed, but the level of noise arising in the three primary squares blue, green and red, and mainly the noise present in the greyscale wedge. It is crops of the wedge we will use as indicators of the noise levels produced at the various ISO sensitivities for each camera, making comparison easier at the sizes used for web images. As a rule the highest noise levels emanate from the blue channel, less from the red, with green being the 'cleanest'. 

Test chart

Epson 3mp digicam. Iso 100, 200, 400, top to bottom

Canon 8mp DSLR. Iso100, 200, 400, 800, 1600, top to bottom.

These examples clearly show how 'cleaner' the DSLR images are than the digicams. At ISO1600 the Canon DSLR's images aren't really any worse than the Epson digicams at ISO400, and probably as good as those at ISO200. So a 2-3 stop advantage. At both cameras base ISO's the Canon 350D is far better. Indeed at ISO200 and 400 it's images look cleaner that the digicams ISO100 setting. Considering the large difference in pixel site size between the two cameras, 3.5µm for the Epson against 6.4µm for the Canon - nearly four times the area coverage - that a clear visible difference exists is not to be unexpected. And remember, this is a low resolution 3mp digicam with relatively large pixel sites by comparison to those of today's models. A 9mp digicams pixels are only 2.0µm in size.

So overall this proves two things. That image quality reduces as ISO sensitivity is increased because of higher noise levels. And that the larger the pixel sites are, the lower the level of noise that results. This is why much time and effort is spent trying to keep noise levels down, and why larger sensors are looked upon as being better, of delivering higher image quality. And this is the main reason so much fuss is made over the desire to have affordable full frame DSLR's.

But this last aspect needs to be put into perspective. It takes a large difference in pixel area coverage for differences in noise levels to be seen to have a viewable effect. This readily occurs between digicams and DSLR's because there are substantial sensor size differentials. This does not occur to anywhere the same extent between APS-C DSLR's and the full frame types. The advantage full frame DSLR's have is of the order of just 1 stop at higher ISO settings, and much less as base levels of sensitivity. This is simply because when the state is reached that little or no noise is visible, it doesn't matter how much larger pixel sites become because you can't improve on an absence of noise.