The dilemma, challenge and paradox of exposure
I started in the late 1960s starting with Sunny 16 from the chart on the Kodak film box and using many different hand-held and in-camera meters including anal retentive spot metering for the Zone System. When switching to Canon I did a comparison test of all 18 of the possible ambient / fill metering combinations on my Canon 20D to discover which was superior. The conclusion: None of them get exposure perfect all the time. Regardless of the metering mode we select with our in-camera Through The Lens (TTL) metering getting the exposure correct is a three step process 1) let the camera metering make its best guess, 2) evaluating that guess, and 3) second-guessing the camera.

There is a reason a camera exposure system gets exposure "wrong" most of the time, but it may not be apparent to someone starting in photography with a digital camera: many scenes exceed the range of the camera sensor. When the sensor range is too short to record everything our eyes see with detail the camera metering must try to guess, based on how the scene is framed in the viewfinder, whether the highlights, middle-tones, or shadows are most important. Since faces fall in the middle range the camera metering will try to get the middle right, even if it means blowing some highlights like the sky and losing detail in the darker shadows.

A significant portion of the learning curve in photography involves developing situational awareness of lighting: where is the sun in the sky and how is it reflecting off the atmosphere and clouds to provide fill. Once a photographer becomes consciously aware that scenes can't be recorded accurately - as the eye sees them - they learn to develop strategies to deal with high contrast lighting. What looks "normal" in a photo is a perceptual judgement made in the brain by comparing way objects are rendered with memories of seeing them in person. The size of the content and the context of the background can change perception as the two photos below illustrate.

The first photo was exposed visually
based on the detail on the highlights
in relatively flat ambient lighting:

DR_FlatLight

I then moved the target and shot from the opposite direction
which put it in strong back-cross lighting, again exposing to
retain detail in the sunny parts of the white towel:

DR_Backlight

Both shots are correctly exposed in the TECHNICAL SENSE of retaining the detail in the brightest highlights. But in the PERCEPTUAL SENSE the first photo appears to be "normal" and well exposed while the second appears badly underexposed. Why? Because our brains expect the towel and card to look like they are rendered in the first shot. Perception of "normal" and "natural" in photos is based on expectations in the mind of the viewer. That simple example above illustrates the dilemma, challenge, and paradox of exposure and metering.

The Dilemma - Scenes exceeding the range of the sensor

When a scene does happen to fit the range of the sensor perfectly the resulting image winds up looking normal automatically because the process is engineered that way. Get the exposure "technically" correct on both ends of the tonal scale and everything in-between falls into place in a way that looks "normal"; as would be seen by eye, meeting the expectation of the viewer. That is why the first shot above seems normal. In flat lighting the range of a camera sensor is just barely able to record the same amount of detail which would be seen by eye.

The difference in the second shot is: 1) the angle of the light, 2) the context of the background relative to the foreground. In the first shot because the card fills the frame the background isn't noticed much an the brain tunes it out almost immediately. But if you look critically at the photo you will notice there are areas in the background and shadows under the towel where detail is lost. But since most of the important familiar content in the photo is rendered "normally" our brains will ignore the loss of detail and consider it normal also. But in the second shot the relationship of the target and background changes, changing the context of the photo from a focus on the target to the overall background. The background is actually rendered very similarly to that in the first shot, and the front of the card is the same tone as the shaded parts in the first shot, but the change in context makes us expect the background to be "normally" exposed and it looks odd because it isn't in photo #2.

The Challenge - Fitting the range of the scene to the sensor

The first part of the challenge is recognizing when the range of the scene exceeds the capabilities of the sensor. The second part of the challenge is finding ways to overcome that technical shortcoming.

My approach for analyzing a scene in ambient light with a digital camera is to get the critical highlights exposed correctly first, using the over-exposure warning in the playback. Using the camera in Av mode I will set the aperture for the desired depth-of-field, set exposure compensation to "0" then take a shot. EC=0 is the camera meter "I think this is correct" baseline. I accept the fact the metering will rarely get the highlights correctly exposed because the metering algorithms aim for getting the middle of the tonal scale where skin tones are correct. In the second test shot in backlight it was necessary to adjust the camera's "zero baseline" by - 2 EC to get the sunny parts of the white towel below clipping.

After adjusting exposure so the highlights are below clipping the left side of the histogram will reveal if the shadows are recorded with detail. The histogram in the second test shot is piled high running off the left side indicating a significant loss of shadow detail in the scene. Because the camera is shooting into the shadow side of the ambient light it is possible to lift the shadow side with a flash placed over the camera:

DR_FlashFill

The histogram in the flash assisted shot is still running off the left side if the histogram because of the background. But when just the target area is selected in Photoshop the Levels histogram for that area reveals the target fits the range of the sensor almost perfectly:
DR_FillHisto

Notice how similar the histogram of the flash lit card is to the first shot taken in the ambient light:

DR_FlatLight

Knowing that outdoor scenes will exceed the range of the sensor I always have a 580ex flash on a bracket...

CGinCO1

...so I can match the scene to the sensor:

CO_BelenFill

CO8551

BelenTree

LillyLake

Exposing for the highlights as illustrated above requires a leap of faith like bungie jumping because before flash is added the ambient-only exposure adjusted to keep the rim-lit areas below clipping make the photo look 2 f/stops under-exposed in the playback. But like the bungie jumper I know the flash on the camera will be able to lift the too-dark foreground up to the point where it looks perceptually normal. "Detail everywhere" is the ethic I learned from the Adams zone system back in 1971. It was easy with B&W film which can be adjusted with development to match the print paper range, but it isn't possible with color film or digital unless flash is used to alter the scene range to match the sensor. See my tutorial "Using Flash Outdoors" for more information on the concepts and techniques.

The Paradox - If it looks normal, its correctly exposed

Experience and professional career on the technical side of photography and reproduction has taught me the huge perceptual dividends which are reaped when exposure is correct technically on both ends of the scale. That's why I carry a flash on a bracket on my camera at all times outdoors and use dual flash indoors, even though its not the most convenient way to take photos. My work in reproduction has also made me realize the weakest link in the process is human perception. The brain is easily fooled into thinking a pattern of contrast in a flat photo is a real 3D object. As long what is most important in the photo is correctly exposed quite a bit of under and over exposure of less important areas will be tolerated.

That's the paradox of exposure. In many cases a photo will have blown out highlights and blocked shadows and not be technically exposed correctly but a viewer will not notice because the face in the photo is correctly exposed. So practically speaking, while detail everywhere is the ideal goal when that is not possible the best strategy is to figure out what in the photo is most important to the message and expose it is rendered as normally seen by eye.

Indoors with flash matching scene to sensor is simple

The discussion above relates to outdoor lighting where the scene exceeds the sensor. But indoors with two flashes in key / fill configuration the photographer has total control over the tonal range. I learned this back in the early 1970s working for Monte Zucker, who was renowned for introducing the dual-flash technique to candid wedding photography. From him I learned a flash over the camera on a bracket to control the shadows and a second off camera on a stand to create the highlights over the foundational fill makes it possible to fit the scene in the flash-lit foreground to the limited range of color film/ prints or a digital sensor.

On a conceptual level Key and Fill works like this:

    1. With a flash over the lens add just enough
      fill light to reveal detail in the darkest shadows

    2. Slowly raise the power of the off-axis
      "key" light until the highlights are just below clipping
    With those two simple steps, overlapping the key light over flat foundational fill the entire scene raised to the level shadow detail is seen in the darkest and the brightest areas in the photo - the definition of technically perfect exposure - with everything in-between looking perceptually normal.

    That's the magic of the photographic process

    when used in the way it is engineered to work.

Holistic Concepts for Lighting
and Digital Photography

This tutorial is copyrighted by © Charles E. Gardner. It may be reproduced for personal use, and referenced by link, but please to not copy and post it to your site.

You can contact me at: Chuck Gardner

For other tutorials see the Tutorial Table of Contents