Way back when I shot film I was a serious devotee of The Zone System. I lived and died by my spot meter. I have a much harder time doing this in HD, however, so I'm rediscovering the joys of using an incident light meter. I've picked up a few new tips and tricks, and now I'm going to share them with you.
A DP once told me, "I put all the important stuff within a five stop range, +/- 2.5 stops from middle gray. Everything else can do what it wants." He'd light by eye, make sure that the important stuff (flesh tone, important set pieces, etc.) fell within that five stop range, check the bright areas and dark areas with a spot meter to see if they would retain detail, and shoot.
I became very good at using The Zone System, which is a great way to pre-visualize what you want and then determine an exposure that will allow you to capture it just that way. Better yet, I'd light a set by eye, place the values that I saw by eye on The Zone System, and then calculate an exposure to preserve that. For example, if I lit a set in a way that looked great by eye I'd look for something in the set that appeared to be middle (18%) gray, measure it with my spot meter, and then put that stop on the lens. I'd then measure a few other things, like important highlights and shadows, and see what zones they landed on. If they fell within the range of the film stock (I discovered +/- 3 stops was usually safe) and then judged the rest by eye I generally got great results.
When I say I "let the other stuff go" I mean that I didn't worry too much about the extreme highlight and shadow values unless they took up a huge part of the frame. A small area of highlight was no big deal, but a row of windows that were +7 stops might be a concern.
One trick that helps in judging shadow contrast is to squint. Shadows become darker and you'll get an idea as to how film will see them. Over time you can become very good at determining fill levels just by squinting.
HD doesn't really work the same way. Film's great advantage is that it is predictable: once you learned the limits of a film stock they would never change unless you processed the film differently. The gamma curve and contrast were set chemically in the factory. HD cameras vary in contrast and gamma not only by manufacturer but also by setting, so unless you're willing to memorize ten cameras with five gamma curves each it becomes very difficult to judge exposure with a spot meter.
The five or six stop rule still applies to some extent. If you're shooting for broadcast you're still working with a six stop container in Rec 709. While most cameras will capture much more dynamic range than that (typically at least ten stops from noise floor to white clip) you're really only going to see the middle three or four stops clearly. This is where the S-curve of color correction comes in: by stretching out the mid-tones so that contrast is high in the mid-tones, and compressing the extra stops in the highlights and shadows to make them fit in the Rec 709 container, we're reproducing how our brains see the world, which is to see the middle tones most clearly while compressing tonal ranges in extreme highlights and shadows.
This graphic hopefully illustrates what I'm trying to describe:
"A" shows the Rec 709 "bit bucket."
"B" shows that the bit bucket was designed to encompass roughly six stops without any sort of contrast compression at either end: a difference of one stop in brightness in front of the lens resulted in a difference of one stop on the screen.
"C" is the dynamic range of an average modern camera.
"D" shows what happens when we try to cram those 12 stops into a six stop "bit bucket" without any contrast modification. The steps (or "stops") are much, much closer together, which results in a flat looking image. If your container is designed to show six stops of dynamic range with appropriate contrast, and you cram twice as many steps into that container, your contrast is going to drop by half. This results in a flat-looking image, because there's not enough contrast between each step to be pleasing to the eye. (In a very simplistic way this describes why log curves can look so flat.)
"E" shows what happens if we apply an "S" curve to increase contrast. The middle stops are stretched back out so that they match more closely the spacing between the original steps in the Rec 709 example. The higher and lower stops are pushed closer together, or flattened out, because that's what our eyes do anyway: they compress shadows and highlights and give the mid-tones the most contrast.
"F" shows that those middle six stops have more contrast than those above or below, and are the "sweet spot" of exposure for any HD camera.
Every camera made can see the Rec 709 six-stop dynamic range easily, so it's fairly simple to pre-light if we know that the really important stuff should appear within that range: flesh tones, important shadow detail, and important highlight detail. We'll typically light for a greater dynamic range because bright highlights and dark shadows make images much more interesting, but for ballpark pre-lighting of a set before the camera arrives it's safe to put everything important within that range.
This brings me back to incident meters. What I discovered recently (and maybe I'm a bit dense for not figuring this out earlier) is that Rec 709's roughly six stop limit encompasses typical highlight and shadow detail under normal circumstances. The easiest way to see this is to look at a color chart, either a DSC Labs Chroma du Monde or a Macbeth Color Checker.
In both cases, white is 2-2.4 stops brighter than middle (18%) gray because that's how bright something is when it reflects 100% of the light hitting it. Here's the math:
18% * 2 = 36% reflectance (average caucasian flesh tone, one stop brighter than 18% gray)
36% * 2 = 72% reflectance (bright caucasian flesh tone, two stops brighter than 18% gray)
72% + (.4 * 72%) = 100% (white, or 2.4 stops brighter than 18% gray)
The standard for 100% reflectance is barium sulfate, which is a very white powder indeed. (The only time you'll see an object lit brighter than 100% reflectance is if there's another light hitting it, or if the surface of the object is reflective.) The white chip on a Chroma Du Monde chart, due to its glossy nature, is whiter than can be printed on a matte surface and reflects about 85% of the light hitting it. Matte white and matte black generally end up being about two stops brighter and darker than 18% gray, both of which are well within range of even the cheapest and most crippled camera.
Barium Sulfate. A diffuse surface will never reflect more light than this.
What this means is that if you're outside with an incident meter and the sun is at your back, an incident meter will tell you an exposure that will allow any HD camera to capture the dynamic range of everything you can see, excluding specular highlights and deep shadows. Everything the sun touches that has a matte surface can be captured with the exposure information on your meter. The brightest highlight that you will see off a matte surface will have a reflected value of no more than 2.4 stops brighter than your incident meter reading.
It's important to note that highlights and deep shadows are the exception to this rule. Excluding fabrics like deep black velvet, there are very few materials dark enough to go completely black on a modern camera; similarly, any recent camera is easily able to hold detail in highlights that are 2.4 stops brighter than 18% gray. Any shiny surface, though, will reflect the light source and will be MUCH brighter than a matte surface; similarly, a shadow is likely to be much darker than -2 stops below 18% gray simply because the amount of light in that area is dramatically reduced.
This is important to know because the incident meter measures how much light is falling on it, so it really has no idea how light affects the objects within the frame. All it knows is how to tell you what the exposure is to make matte surfaces appear roughly the same to the camera as you see them by eye as long as they are bathed in the same light as the meter. As long as you know how bright the things in the frame are, an incident meter can tell you what stop to put on the lens to reproduce the scene so that it looks as close to what your eye sees as possible.
An incident meter is most effective within that six stop dynamic range that every camera, no matter how crippled, can see--assuming that everything in the frame is lit by the same light, at the same intensity. This rarely occurs.
In the past I never trusted incident meters because they didn't tell me what was actually happening in the scene. They predicted how "normal" things should look when lit "normally," but that kind of lighting tends to be very boring. An incident meter didn't tell me how bright the window in the background was, or how dark a distant wall was. Reflected meters tell me, more or less, exactly what's happening in the shot, and as long as I intimately understand the gamma curve used by either my camera or film stock I can calculate an exposure that will capture the scene exactly the way I want it to appear.
As cameras increase in complexity, however, reflected meters become less predictable: one built-in gamma curve may hold highlight detail in a distant window, for example, but another gamma curve may not. The number of possible settings is so great that unless one works with the same camera with the same settings constantly it can be very difficult to use a reflected meter to predict what the camera will see.
HD cameras are a lot less predictable than film.
What I do know is that if I hold an incident meter at a point in space and measure the light falling on it, it will tell me an exposure that will capture anything that falls within the Rec 709 six-stop range… which is most everything, barring specular highlights and deep shadows.
Of course it's a bit boring to light a scene that way as it'll turn out to look very flat and two dimensional, so the meter alone doesn't give us great results. We have to know how to interpret what it tells us. Turn the page...