A Digital Approach to Controlling Exposure

The aim of this guide is to help understanding exposure in photography in a practical way.

Intentions

One can have lengthy arguments as to the merits or otherwise of various methods of measuring, determining and controlling exposure. We will follow some conventions in this guide – but first, the most important thing to establish is that it is your choice how an image should look.

We assume that you have a camera in hand and are looking at some scenery – in this case, we take a typical semi-urban landscape such as an overview Oban from Pulpit Hill.

We further assume that the standard output method will be a (calibrated) monitor display (as distinct from a projection onto white screen, or a print on some kind of paper).

What is Exposure?

Exposure is the process of taking tones, specifically levels of brightness, in the source scene and mapping them to output levels in a given medium. For example, we identify the darkest shadows regions amongst the poles supporting the ferry terminal, and say that we want them to print as dark as possible subject to retaining some detail; we also identify the brightest white house on the opposite hillside and say we want it to print as bright white as possible without losing detail: so no submerged shadows, no blown highlights.

Why? Notice that we have introduced some convention already. It might be that you want everything so light, the shadow regions come out as bright as we see the buildings; it might be that you want the highlights dark and shadows bright – a negative image. You might want a non-linear mapping, such as a solarisation. The final appearance is up to you, but we are choosing some degree of realism for the purposes of this exercise.

Histogram

A histogram is a graph showing the proportion of an image at each given brightness. Typically we have dark areas on the left progressing to lighter regions on the right; vertically, the proportion of the image’s area at that brightness.

If our scene is a simple single subject, such as an image of grass filling the frame in uniform lighting, then the histogram will have a characteristic shape – it tends to approximate a Gaussian or Normal distribution because of the large number of blades of grass all reflecting light at random angles to the camera.

This immediately gives us two useful parameters: we can choose to place the average brightness of the grass near the middle of the histogram; this corresponds to a midtone, ie the brightness of a piece of card that reflects about 18% of light falling on it – by varying the exposure, we move the red dot on the top of the histogram left and right to reflect how the image comes out darker and lighter respectively. Additionally, we can control the spread of the distribution – when processed for higher contrast, the histogram curve spreads wider (along the red line), or with lower contrast, narrower.

Finally, by checking whether the histogram is shifted so far right or left that a significant part of the curve is “crawling up the axis” rather than tending to zero at either end, we can check for blown highlights (right) or submerged shadows.

Now, in the real world where we want to take a photograph of Oban, the view of the town in no way comprises just one subject. Every building and every tree in the view contributes a little normal-distribution curve of its own, at a particular average brightness and contrast. (For example, the width across the brickwork of one of the hotel fronts contributes a small variety of tones.)

The salient features we can identify from the histogram are:

• points D: all the varying brightnesses of light from the scene can be fitted into the width of the histogram with no clipping of shadows or highlights;
• area C: most of the spread around this area is due to the varying shades from the cloudy sky, with some of the brighter buildings contributing a slight peak;
• point B: a large proportion of the image is a midtone, with not much contrast-separation – this corresponds to the expanse of water;
• point A: a significant proportion of the image is darker than a midtone, with slightly more contrast – this corresponds to the trees spread across the width of the hillside, and darker buildings.

Measuring Exposure

Levels in a scene are typically measured in EV (Exposure Values); this assigns a simple number to the brightness, such as 12 for an outdoor scene on heavily overcast day, rising to 15 and 16 for bright sunny (arctic snow and desert sand) situations. These numbers give an indication how much light we must let onto the camera’s sensor in order to place the midtone correctly around the middle of the histogram.

There are several techniques for measuring the scene’s exposure, but remember, for each of them, some function of the light in the scene is being converted to place the midtone.

• Incident metering: this gives a reading of the amount of light falling onto the meter. It requires you to be in the same light as the subject/scene – so it is frequently used by portrait photographers who ask their subjects to hold the meter pointing back at the camera – and not so often used in landscape. However, its big advantage is that it is independent of the nature of the scene itself, so it’s quick and precise to use.
• Reflective metering: this covers a handful of algorithms, all of which are based on the reflected light coming back from the subject.

• Average: all the light across the scene is averaged and the result converted to a midtone
• Centre-weighted: again, the whole scene is considered, but special weighting is given to a zone in the middle of the composition
• Landscape-centre-weighted: as before, the whole scene is considered, but special weight given to a zone in the middle and lower-middle parts of the scene
• Matrix / multi-segment: the scene is split into several or many regions and some analysis performed to recognise typical scenes
• Spot: a tiny area (1-2 degrees) is measured; this may be fixed at the centre of the scene or bound to one of the auto-focus points

The disadvantage of reflective metering is that the nature of the scene is mostly ignored; a black cat will be normalized up to a midtone, appearing too light, just as much as bright snow will be normalized down to a midtone, appearing dark. (Again, this is convention rearing its head; photographs of snowy landscapes tend to have histograms skewed toward the right, because we see snow as white. It doesn’t have to be so, but this is convention.)

So, to determine what one exposure value to use for your image, you have to juggle the idea of fitting the whole dynamic-range of the scene into your histogram, against the idea of placing tones where you think they should logically be (lighter or darker than a grey-card in the same light, ie a midtone). This is the art of tone-placement: in the case of the Oban scene, one might set the camera to spot-metering and consider:

• a) you know that the brightest highlights should be 2 EV above the midtone in order to preserve detail – if the meter says they have an EV of 14, this indicates you set the camera to an EV of 12 overall
• b) you know that the darkest shadows want to be 3 EV below the midtone to preserve detail – if the meter says they have an EV of about 9 then this corroborates the idea of an EV 12 overall
• c) you know that the tree foliage should appear 1.5 EV below a midtone – if the meter says they have an EV of 10, then this approximately agrees with an EV of 12 overall too
• d) you set your light-meter to read from a wide angle of view and simply point it at the scene in general: this tells you the whole thing averages-out to an EV of 12 as well; you know from the spot-metering that the contrast is not huge (5EV, well within the dynamic range of slide film, let alone a shiny dSLR), so that finalizes it.

Peculiar Cases

In some cases, your chosen composition will have a histogram that looks nothing like a characteristic normal distribution – typically if you have a very bright sky with a person silhouetted in the foreground, for example. This kind of image’s histogram looks like two horns or a spread-out U-shape – a peak at the left (the right half of the normal curve from person, where the left half of the curve is clipped in the shadows) and a peak at the right (the left half of the normal curve from the sky tones, where the full width of sky tones is clipped at the highlights) and not much in between. These kind of scenes do not lend themselves well to calculating one characteristic number to use as the exposure value, because it is hard to say what single number best represents the scene – a simple average of all the values in the composition (forgetting the contrast involved), or a 50-50 average of the brightest and darkest areas, or an average weighted by image-area, or an average that minimizes clipping at either shadow or highlight ends: all are different numbers, with different consequences; these scenes are best handled either with filtration, multi-exposure blending or HDR techniques – or simply avoided.

Exposure Controls

Having gone to the trouble of expressing scenes’ brightnesses as the most appropriate number, we now observe the vast majority of cameras have no means of entering that one number. Rather, they have controls for ISO sensitivity, Aperture and Shutter-Speed; all three of these work together to give the equivalent of an exposure value, and, critically, there are many combinations of these parameters that equate to the same EV, and varying each has its particular consequences.

A Stop

Photographers are to be heard talking all the time about “stops” of exposure. A stop is simply the difference of 1 on the EV scale above (e.g. from 12 to 13), or corresponding to a factor of two in the amount of light being discussed.

ISO Sensitivity

ISO sensitivity is a measure of the number of photons of light required to effect the same change in output brightness. For example, we might say that for a pixel to register a midtone requires 1000 photons at ISO 100; at ISO 200, it requires only 500 and at ISO 800, only 125 photons.

The consequences of varying the ISO sensitivity are manifest as noise: because the sensor itself produces thermal photons randomly, at any point in time a pixel might record as brighter than the incoming light would dictate. When the ISO increases and number of incoming signal photons decreases, the same number of photons coming from the sensor will contribute a greater amount to the random errors – the signal to noise ratio decreases.

Aperture

The aperture is simply the diameter of the shutter iris in the lens. The wider it is, the more light is let through. It is measured relative to the lens’s focal length, so for example f/4 on a 50mm lens is an aperture diameter 12.5mm, the same as f/8 on a 100mm lens.

The consequences of varying the aperture are manifest as depth of field: if you remain focussed at the same distance, but stop-down the aperture (make it narrower) then the depth of field increases, i.e. the difference between nearest and furthest acceptably sharp distances increases. Additionally, lenses suffer from varying intrinsic sharpness due to diffraction at narrower apertures, so any lens has a sweet-spot aperture at which it is sharpest, often a stop or two away from the middle of its aperture range.

Shutter Speed

The shutter speed is measured in seconds, or normally small fractions of a second.

Varying the shutter speed has consequences for motion-blur: fast shutter speeds lead to freezing subject motion and eliminating camera-shake; slow shutter speeds blur subject motion and risk camera-shake.

Sunny-16 and Equivalents

“Sunny-16″ is the name given to a rule of thumb for estimating typical daylight exposures without recourse to a light-meter. The guideline says: when sunny, set the aperture to f/16 and the shutter-speed to the reciprocal of the ISO setting.

Thus, a lot of outdoor photos on bright sunny days can be taken at the combination: ISO 100, 1/100s, f/16.

This gives a certain amount of light as an exposure; we can vary the parameters individually whilst retaining the same brightness overall:

ISO 200, 1/200s, f/16 – (one stop faster) faster film, faster shutter, more likely to freeze motion

ISO 100, 1/200s, f/11 – (one stop wider) faster shutter, wider aperture, less depth of field

ISO 100, 1/400, f/8 – (two stops wider) much faster shutter, wider aperture, much less depth of field (but closer to many lenses’ sweet-spot apertures)

Camera Modes

All serious digital cameras come with a selection of modes for helping set the above parameters. The reasons for choosing between them are purely down to what fits your mode of operation.

Manual Mode (M): it is entirely possible to set manual mode, establish an exposure reading for an overcast day (so the light doesn’t change), and walk around in the landscape in the same light making photographs in varying directions with different subjects, with little or even no change to the settings.
Pros: full control feels good; your choice which of shutter, aperture or ISO to change when change is required; works with manual lenses.
Cons: recalculating in changeable light is not always fast.

Aperture-Priority Mode (A, Av): here you specify the aperture you want to use and the camera calculates the required shutter-speed for you, automatically.
Pros: control over depth of field is frequently important; keeping the lens near its sharpest aperture is also a useful consideration; works with manual lenses.
Cons: reliance on some kind of automated calculation for the shutter speed, with compensation for when it gets it wrong

Shutter Priority mode (S, Tv): here you specify the shutter-speed you want to use and the camera calculates the required aperture instead.
Pros: useful where you know what shutter speed you want, such as either very fast sports photography where action must be frozen, or very slow motion-blur effects (e.g. panning to follow the movement of race-cars)
Cons: reliance on some kind of automated calculation for the aperture, with compensation for when it gets it wrong; could come up with any aperture giving more or less DoF than required.

Programme Mode (P): here the camera automatically calculates a combination of shutter speed and aperture that will give a reasonable exposure; if you want to bias a particular photo towards a wider aperture for DoF or to keep its shutter-speed up to avoid camera-shake, you simply adjust the permutations.
Pros: flexible as to whether it’s aperture or shutter-speed you vary, allows you to think in terms of creative effects and intentions from the outset
Cons: probably won’t work with manual lenses; if you’re starting to settle on either aperture or shutter-speed being more important, you’ll be biassing in favour of that parameter with every shot

Auto: here the camera automatically decides on all the parameters for you, both shutter speed and aperture and ISO as well.
Pros: it might know the lens’s performance better than you do; you’ll get something out of it no matter what, including in a pinch where you don’t have time to think
Cons: probably won’t work with manual lenses; surrendering control can be a heavy, risky, price

Modern cameras also have many other parameters to affect image quality: some have a “creative auto” mode where you specify constraints such as “dark overall, but with narrow DoF” and it works out a combination of parameters to give those effects; most have “scene modes” where you say “it looks like a landscape shot” and it biases towards greater depth of field and boosts the green and blue saturation; some have auto-ISO that can apply in P,A,S,M modes, with guards that the ISO never goes over a given value or the shutter stays above some speed; some have “auto lighting optimisation” that guards against blowing highlights or submerging shadows; some have custom modes where you can store settings to use as start-points for some kinds of photography you use.

Which mode you choose is up to you; there are some forums where a semi-automated mode is frowned upon and people claim manual is a panacea; this is as inaccurate as relying on auto for everything. What matters is that you know how your camera is going to respond in various circumstances and choose modes that work for you.