Essays: Photographic Exposure
Abstract
Over the years, I've encountered several methods for calculating exposure. Most have left me cold at one time or another, so I present a short essay on my favoured approach and the backgrounds behind exposure choices for a given photo. Disclaimer: I'm mostly a landscape photographer by inclination, so the following does not yet cater for flash at all.
Technicalities and Consequences
What is Exposure?
An image's exposure is its overall brightness, the amount of light reaching the film or sensor. Too little light, and a photograph loses all detail in the shadows; too bright, and highlights wash or blow-out.
There are 3 factors affecting exposure: film-speed (ISO rating), shutter speed and aperture. Keeping ISO setting constant, if you open the aperture twice as wide (by `1 stop'), then to achieve the same exposure / brightness, you must either halve the duration of the shot or halve the ISO rating. Likewise, if you shoot twice as fast, you must either double the aperture or the ISO rating to compensate.
These ideas apply equally well both to film and digital cameras; shutter-speed translates directly, but depth-of-field requires wider apertures with smaller camera-formats, and ISO sensitivity is an emulation (the number of photons required to hit a photo-site to be recorded as the same brightness change in the final image - where film gives grain with high ISO, digital sensors give extra noise).
Varying each of these quantities has consequences.
If you choose a low ISO rating, photographs will generally need longer shutter-speeds, meaning you will be unable to freeze motion on darker days. On the plus side, they will tend to have lower grain, greater sharpness and greater colour saturation. If you use high ISO settings, you get more obvious grain, and can more easily freeze motion with a fast shutter-speed, but on the flip-side, you cannot choose to take a long exposure to blur water without needing filters (since you won't be able to stop-down the aperture far enough to get exposures of the order 2-8 seconds).
If you choose a fast shutter-speed, you can freeze motion. A slow speed will introduce motion-blur in the subject, or risk hand-induced camera-shake.
Aperture controls the depth of field present in a photo. For a given zoom-length, focus distance and camera format, the depth of field is the area both slightly in front of and mostly behind the focussed distance at which points are acceptably sharp. With a narrow aperture, the zone of acceptable sharpness can range from very close (a handful of meters) to infinity. With a wide aperture, you can focus selectively such that only some parts of the image will be sharp - typically throwing the background out to emphasize the subject detail.
On-Camera Controls
Most cameras these days have controls for selecting a metering algorithm as well. One thing to remember: whatever the algorithm, you're telling the camera "this function of the scene approximates a midtone 18% grey in brightness". The camera then tells you, typically for a given ISO, what the valid aperture and shutter-speed combinations might be to achieve this brightness.
With "average" metering, the entire image is used with even weight to approximate the midtone.
With centre-weighted metering, you're telling the camera that everything in the scene, but particularly with importance weighted towards the centre, approximates a midtone grey. (It might be that the centre portions count for 60% or so of the final calculation.)
With matrix or multi-segment metering, all the area in the viewfinder is used, but some intelligence is applied to identify common scenes - for example, bright top and dark lower-half could be a landscape, or bright sides and dark middle could be a silhouette.
Finally, with spot-metering, a tiny fraction of the image (maybe as small as 1 or 2 degrees' field-of-view) is used to calculate the midtone.
Correct exposure
The question most people face in the field is "how do I expose for this scene?". There are also several methods you can apply, categorized into reflective (using light coming from the scene) or incident metering (using light arriving on the scene).
Reflective Metering
Initially, it's tempting to choose a matrix or multi-segment mode and trust the camera's "intelligence" to choose settings that will work. The downside of this is that it will often fail, if your scene does not comply with one of its presets; it assumes that "correct" is the same as "realistic" is the same as "approximates a midtone overall".
Centre-weighted averaging is little better. It is frequently, but unfortunately, true that important subjects in a photo will be placed in the centre, and frequently assumed they should approximate a midtone. Certainly you can set your exposure off this and then use an AEL feature to lock it whilst recomposing the photo. Obviously, this is less likely to work the more you deviate from the norm, both in choice of composition or of subject (black cats, snowy landscapes).
Spot-metering used naiively can also be a problem: if you just put it on centre-spot and hope, the chances are the brightness of a small point will vary wildly across any one scene, so it becomes a complete lottery as to the suggested exposure.
Compensation, et al
The trouble with these reflective techniques is the tendency to make crude theories for various scenes, such as "expose for sunlit green grass", or "for snow, open up 1-2 stops", or "for Caucasian skin-tones, expose for the skin and open up 1 stop". Alternatively, if you have a wide-angle reflective meter, you might find the instruction manual recommending you point it straight at a subject, or if it's a landscape, tilt up or down a small amount to compensate for whether the land or sky occupies more of the frame. These approaches and sentences are unhelpful and inaccurate. The phrase "expose for" omits the fact that the quantity in question will come out as a midtone. The business of opening-up 1, 1.5 or 2 stops only works for certain canned scenes or subjects. In a fast-changing scene such as a race-track or the jungle, you'll find yourself setting an exposure and then compensating back down to more or less the same value, over and over again.
Incident Metering
Incident metering is one answer to choosing a good overall exposure for a scene, independent of the brightness of the subject. By holding an incident light-meter at the subject facing towards the camera, you measure the incoming light hitting the subject, and by using its recommended settings, you know you'll get a conventional exposure of the scene - the same as if you'd replaced the subject by a mid-grey reference card. (Thus, you should not expect a reflective and incident meter-reading to give the same exposure unless the scene approximates a mid-tone grey.) You can also often use "poor man's incident metering" for landscapes: instead of running to a hill to meter back at the camera, you assume the angle of inclination of the sun's light hitting the ground is the same both at the distant scene, and where you're standing - the result being that you hold the incident meter back over your shoulder. This only works if you're in the same light as the subject, ie the subject is not back-lit, or you don't have a large object immediately behind the camera. It also only works if you assume you want a conventional exposure with objects about as bright as you perceive them to be.
Sunny-16 Metering
Included here as an odd-ball method, Sunny-16 is where you set the speed equal to the reciprocal of the ISO setting (so for ISO 100, you set speed=1/125th as the closest approximation, etc), and then if it's a bright sunny day, you set an aperture of f/16, opening-up to f/8 for grey days, f/4 for clouds, etc. This makes huge assumptions, both that you can identify the kind of day and direction you're shooting in relative to the sun, and that you want to start from a given specific shutter-speed, and again that you want a shot to come out conventionally exposed. It probably helps to be shooting on colour-negative film where you have the latitude for huge exposure-errors, and/or to bracket a stop either way to be safe (the blunderbuss approach).
Putting it all together
The one question I've not answered yet is what a correct exposure actually is. I've taken care to emphasize the idea of a conventional exposure as one that approximates what your eyes might see in terms of brightness of various objects in a scene, but that doesn't allow for artistic choice.
Rather than conventional exposure, as above, it might be that you intend a photograph to be "low-key" (ie mostly much darker than a midtone) or "high-key" (mostly much lighter than a midtone). None of the automatic methods will achieve this for you, as they all assume you want some approximation to a mid-grey. Compensation is an unhelpful mentality with which to approach a shot, too; how do you know that the desire for "low-key" means any given number of stops' "under-exposure" in a given photo? What if your snow is lit by low glancing gold sunlight, or by high-power flood-lights? How can you convey thoughtful meaning rather than a simple "it was there, it looked like this" if you rely on the camera's metering to do half the job for you?
Hence, I propose the following simple definition: a photograph is correctly exposed when the tones (shadows, midtones, highlights) are placed where you intended them to be in your pre-visualization of the final photo. You take some care to place objects spatially in the frame; you should take some care to place light-tones where you want them to come out in the finished medium.
To this end, I recommend choosing between the various possible metering algorithms according to the situation at hand. If you're in the same light as your subject and want a conventional representation, incident-metering will get you there immediately (and one reading will be good for as long as the lighting stays much the same, up to several hours). If you want to see how all the tones in the scene will come out, watch the exposure-guide whilst pointing the spot-meter at various parts of the scene; don't think of making a given point be a midtone and then compensate, rather, say "I want this point to be a highlight 2 stops bright" and then line-up the dial with "2 stops over-exposed" when spot-metering off that small part of the scene. If you have to respond fast to a scene, such as at a race-track, why bother including cars, with all their variety of brightnesses and colours in your metering considerations, if the lighting is near-constant and you know you want the tarmac to come out half a stop lower than a mid-tone? And if your meter's bust, use Sunny-16.