Camera Dynamic Range
Dynamic range (DR) is a strong indicator of camera capability and image quality, often tested in camera reviews. Depending on the scene you are recording, good DR can mean the difference between a decent result and failure. It has improved, but not spectacularly, over the digital era.
The term does not refer to the fact that a camera can capture a very dim scene well, something like the Mlky Way. It does not mean a camera can capture extremely brightly lit scenes. These can be handled with long exposures and wide apertures and short exposures with short apertures, respectively. Instead, it's about how well we get both in the same frame.
Dynamic range is indicative of the camera's ability to adequately handle both harsh light and shadow in that same scene. It's the capacity to render both of these aspects without overload or unnatural, blandly-simplified results. If the image range is maintained to the display, results will be more lifelike, a good rendition of the light level variation encountered in real life scenes. There'll be less compression and a more vivid image. The human eye has a better dynamic range than all present electronic-optical systems, so we are still playing catch-up here.
More contrast, but detail and luminance information present across both shadows and highlights, is what we want. HDR needs to be present at all points in the chain from sensor to display. That won't be happening yet unfortunately for most of the material you view, because even if you bought yourself an HDR display, most material won't have been recorded in an HDR format.
Technically, at the camera end, HDR requires sensors and associated electronic systems able to handle a very wide spread of signal levels.
Failure to adequately capture low light levels means they are rendered uniformly black. Good shadow performance renders these low light levels progressively, so subtle shade differences are recorded. In addition, grain or 'noise' may be present, added by the camera sensor and circuitry, also most noticeably in the dimmer parts of the photo. Essentially, the camera is failing to capture nuance of light in the darker areas of the image, and simplifying those areas to solid black plus noise. Any noise present is unrelated to the actual scene being captured, and is simply added over it.
If camera shadow noise is low, and a RAW file with 14 bit depth is available, it is possible to substantially lift the shadow areas of an image without obvious noise appearing. The brightness curve of the original image file is mapped upwards. These pictures are before and after shadow curve adjustment with a RAW file from an EOS 5D III. The camera was set to ISO 1600, and at this setting there's still a lot of latitude for adjustment. The adjusted version is arguably closer to a representation of how the eye perceives the scene.
Conversely, the camera may be overloaded in certain other portions of the image. These areas will likely be rendered solid white. These zones are sometimes described as 'clipped', a term derived from audio, when an amplifier can no longer output a large enough signal. Photographers will most likely talk about 'blown highlights' when significant areas of an image are overloaded in this way.
If a camera has to cope with both shadow and highlight in the same scene, we need to get an optimum exposure, minimising areas of the image with these possible problems. Very often there is no problem and the camera's automatic exposure metering will set an exposure covering the light and dark areas of a shot well and maintaining a progression of shading throughout, without under-exposed shadows or over-exposed highlights.
Dynamic Range restricting factors
We call most modern cameras 'digital'. What does this actually mean? A modern electronic camera uses a pixelated sensor, and so can be considered 'digital' in the spatial sense. In the image plane, directions are quantised, or stepped, giving the 'blocky' pixelated appearance at excessive display magnifications, reminiscent of early computer games. However, the light level actually arriving is limited in resolution only by the fact that light arrives in photons at the sensor. Photons are extremely fine units, but they count up in whole numbers. A single photon would represent a true shade above pure black. If this photon hits a particular pixel during an exposure, ideally that pixel should register a value above pure black in the image output file. Most camera sensors convert only 30-70% of photons into electrons for the camera electronics to process. The conversion ratio is known as the quantum efficiency. So we may need more than one photon to register a light level above dark. Additionally, the analogue amplification circuitry reading the sensor pixel contributes random noise. It may be hard to determine the true brightness from this random added signal. The sensor's own noise and the subsequent electronic 'read noise' are both important; they need to be minimized. At the other extreme, the sensor pixel will eventually reach a light level where it can contain no more electrical 'charge' and will max out. Alternatively the amplification and ADC may limit instead. The pixel or pixels concerned will normally be rendered solid white.*
Particularly when photographing backlit scenes, the dynamic range of a camera may easily be exceeded. Importantly, post processing will not recover these solid black or white regions, because there is no shading information present in the output file in these areas. There's nothing for the software, which re-maps light levels, to work with.
*The real situation is a little more complex because there are 3 colour channels in a digital image, and any or all may overload. Once a colour channel is overloaded, it is impossible to recover brightness or colour balance correctly. Analogue film tends to overload more 'gracefully', maintaining colour progression. Thankfully, even with digital, the normal zone of overload equates to sky sections of the scene, and whited-out areas are less noticeable there.
Dynamic Range measurements
Useful dynamic range is usually measured in 'stops'. A photographic 'stop' refers to a doubling in light intensity. Therefore an analogue film rated at 6 stops dynamic range will cope with a contrast ratio of 2 raised to the power of 6, i.e. 64:1. This is not much, but analogue film 'limits' very progressively, and it is rare to find harsh patches of blown highlight. If you shoot into the sun, and meter your exposure on the face of a person, you'll often get a better result with old -fashioned film, as a lot of fashion photographers know. Especially if you use larger format cameras. The format seems to capture the complexities of scattered sunlight quite naturally with a good film composition. There is very limited ability to alter light curves with film, but it often handles localised over-exposure in high DR scenes rather well. Early DSLRs had a DR of 9-11 stops, and the latest sensors may reach 14 stops or even more. At these figures, and with reasonable care, large clipped areas can normally we avoided. Digital is usually better than scanned film for shadow recovery, though even there large film plates can be very effective.
With Canon, DR moved on a stop or two with the 90D and EOS 5D IV. Sony and Nikon sensors have been slightly ahead of Canon for a few years but Canon have now caught up. Canon's now-replaced EOS 5D III is among cameras with an automated internal three-shot HDR mode which can provide DR exceeding the best available cameras used in single exposure mode. This is a work-round for a camera with non-ideal DR, but the work-round goes further than the problem. A DR of 17-18 stops is achievable. These modes are only suitable for stationary or slow-moving subjects, however, as they rely on frame stacking of successive shots.
Why else does DR matter?
High DR capability allows careless exposures to be corrected in post processing, and allows shadow and highlight areas to be tweaked to taste. We are likely to be tweaking the image for successful display on contemporary monitors and screens based on LCD/LED technologies. These displays often have excellent colour rendition, high resolution and near-zero rectilinear distortion, but actually they don't have brilliant DR. Ironically, while 35mm film has poorer overall DR compared to digital, solid state monitors have poor DR compared to older plasma and CRT ones, although they are getting better again. The maximum light intensity is more limited, and therefore a high DR scene will appear very dim overall unless light levels are compressed somewhat for display. This is probably the one area where older CRT displays were better, and the now obsolete plasma screens generally had better contrast ratios too. I have a large Panasonic plasma screen which works rather well although it is limited to XGA resolution.
All displays require shade to view properly. With modern sensors and screens, we are likely to want to compress real light level into a narrower band of actual displayed brightness. But we still want to maintain shading between light and dark areas. To do this, we need a sensor able to respond in a linear way to all the light levels present in the scene. Only if that succeeds can we adjust them later for optimum display. If displays get better, the overall situation will only improve; we can display more vividly what we captured. Obviously it must be captured first.
In perspective, few users of 12-stop DR cameras like the 5D III complain too much, but 13 or 14 stop DR sensor systems definitely allow a little more room for manoeuvre with high contrast scenes. There is less likelihood of intrusive grain and of blown highlights, and if we boost shadow levels in post processing software we will get cleaner results. At the time of writing, many RAW image data files record at 14-bit resolution, equating to 14 stop DR capability if the sensor and read technology is up to it . Displays on the other hand are likely to be more limited.