www.scantips.com

18% Gray Cards - What's the Idea?

Some significant ideas first:


We might use 18% gray cards in a couple of ways.


The idea for the 18% gray card became popular in the early days of B/W negative film because it was perceived to represent 'middle gray' for a film. And it has caused us so much confusion since. :) The plan on this page is to help with "what does middle gray mean?" Or not mean?

Actually, it was earlier than film. There is a history.

Early Days

It was believed that 18% was the gray tone level that the human eye recognizes as "middle gray", meaning our brain thinks this tone looks like it should be about middle of the range between black and white. It reflects only 18%, not 50%, but our brain tends to change things, in some cases to be what we think we should see. Look up optical illusions. (Here's two amazing ones: Here and There). Fun, and about what eyes and brains see and don't see, and sometimes just makes up. But the reflectivity of our gray card ink is 18%, which is not 50%. The response of the eye is a different subject, whereas our histograms are instead actual data... sort of.

Specifically, history in older days (before electronics or sensors or anything), the 18% gray card was used by ink press printers to judge their halftone ink flow (ink level adjusted so that a 18% halftone was judged to create this middle gray tone we think we see). Halftone patterns could easily design the number of black dots to be 18% of the total dots, but then dot gain will affect its printed result. Dot gain is the spread of the ink dots (to be larger), as the ink soaks into uncoated paper, not so much in coated papers. And too much ink was an issue. The ink flow to an 18% halftone was adjusted until it looked about the same darkness as the 18% card. So 18% is an old concept.

Then in the 1930s, Ansel Adams developed his popular Zone System about controlling grayscale tonal values. The 18% card already existed, and was called middle gray, and Ansel declared the 18% card was Zone V at the middle of his system, eleven tone levels, 0 (zero, white) to X (10, black), even if the Romans had no number for 0. The system and the card were widely used then, and Ansel described details of the use of the card in his book "The Negative", 1948, and several later improved editions. One goal was that middle Zone V will look about middle gray to our eye. The actual plan of the Zone System was his belief that photos should end up containing some actual black and some actual white, and Ansel's prints certainly do that (contrast is extremely beneficial to grayscale work, but may be a little detrimental for average color work).

Ansel achieved his work with extreme darkroom manipulation, but in digital, we can achieve this baseline with something like shown in Levels at right (or maybe a bit more and tighter clipping than this, to the main edge of the data). Then the center slider adjusts overall brightness level. This treatment is speaking only of grayscale. Individual photos may vary. Setting the Black and White Points specifies the new 0 point for black and the new 255 point for white. Setting them to where the major data actually begins is clipping, a minor thing here in grayscale, because part of the goal is larger areas of extreme black and extreme white (Clipped areas lose their detail. In Adobe, hold the ALT key down while moving the sliders, and you will see the pixels that are being clipped. There is no color for clipping to change.) The point is that blacker blacks and whiter whites is greater contrast, which dresses up grayscale to make it sparkle. Try it, and look at it.

Then color film came along, and really complicated the Zone System. :) Clipping grayscale for contrast only slightly changes the gray tone, lighter or darker, but still gray. But it won't turn blue or pink like color might. Correct color requires no clipping. Color differences are many, luminance values, different subject reflectivities, different contrast goals, a different ball game. Color gives contrast in color images, more so that just tonal variations.

Then decades later, digital came along. I suspect Ansel never had opportunity to see a digital histogram, it's a modern digital thing. Gray cards were a film thing, about "looking middle gray" to the eye. Digital histograms instead analyze the specific binary tonal data, the real data. We tend to get confused when we try to numerically combine the two concepts (below).

Metering on the 18% Gray Card

Metering on the scene and subject, containing random colors and such, can sometimes give skewed results. I do use the camera meter, and do compensate it as necessary. We eventually learn a few things, to do and to avoid. :) But instead, metering on the 18% gray card in the same light is a plan to substitute a stable situation, independent of the scenes colors. Which is what an incident meter does too of course, metering the light directly, at the subjects position, but aimed toward the camera.

Why? How Camera Light Meters Work shows that if we meter on a white card, it comes out middle gray. And if we meter on a black card, it comes out middle gray. And if we meter on a middle gray card, it comes out middle gray. The (reflective) meter is always seeking a middle gray result (meaning the subject colors will be averaged to a middle tone). And in this last case, when middle gray comes out middle gray, then white will come out as white should, and black will come out as black should. That's the idea of metering on an 18% card that is in the same light as the desired subject. We hope to place the middle correctly, and then the rest will be correct too.

Kodak always said if metering on their 18% card, we should increase exposure 1/2 stop to come out proper (which converts to 12%). Ansel himself (same "The Negative" book) said 1/3 stop increase (meters of his day). They are speaking of the presumed average reflectance of typical scenes (but all scenes of course do vary).

When using reflective meters (like camera meters) on the exceptional scenes (aren't they all?),
Kodak has always said (and photographers have always had to learn):

Note that Kodak sold all of their printing business in 1995 (owned by Tiffen today), and while third party 18% cards bearing the Kodak name are still available, this "Open 1/2 stop" info is lost now. Still true though, still supposedly 18%.

Light and dark (as used by Kodak) means the reflected color, not lighting intensity. A white wedding dress is a light colored subject, which reflects a lot of light, and so the meter will read high, and the exposure is adjusted back to middle, and the dress will be gray. A black tuxedo is a dark colored subject, which does not reflect well, and so a meter will read low, and the exposure will be adjusted up to middle, and the tux will be gray. However, both standing together may be an averaged gray subject seen together, and exposure probably comes out correct. We do have to pay attention to what we are metering. This has always been true, from the day of the first reflected light meter. We do need to learn how the meter works.

For Incident meters, Kodak reverses the words increase and decrease. That reversed concept would also include the gray card, in addition to the first 1/2 stop. Both of these methods meter the light more directly, independent of the subjects colors.

Reflective meters measure the reflections from subject, from the camera position, affected by the subjects colors.
Incident meters measure the light source directly, from the subjects location, but independent of the scenes colors, and is usually about right for most scenes.

Reflective meters will read light colored scenes too high, and then drops that value back to middle (causing underexposure of light colored scenes, or maybe a white wall background, etc).
Incident meters (and gray cards) read the light directly, usually about right, but light color detail may show better if less bright.

Reflective meters will read dark colored scenes too low, and then raises that value back to middle (causing overexposure of dark colored scenes, maybe dark colored backgrounds, etc).
Incident meters (and gray cards) read the light directly, usually about right, but dark color detail may show better if brighter.

Our reflective meters today (Sekonic, Nikon, Canon) do 12.5% (it is not a debate, it is ANSI standard, and K=12.5 is printed in specifications in all Sekonic manuals, and also matches Kodak's "open 1/2 stop" for those scenes other than gray cards). Ansel claimed the first meters in the 1930s were 18%, but it is lower today.

Note that this value for meters is a judgment about some average mixed tone of typical scenes (if possible exists), and is NOT related to any value of middle gray (nor to any one specific image). Nothing says the average scene is or should be middle gray. The shadows in a green forest or a bird in a blue sky will vary. It depends of course on the specific scene colors, however the mix in many usual scenes is in fact close enough to the meters, so it often sort of works out (which is the basic idea, but also often not as well). We have compensation buttons on our cameras.

The meter is a dumb device. It can measure the blob of light it sees, but it cannot recognize the scene to know what it means, or ought to be. Our brain is our best tool. We can recognize things, and have experience to know what it means. The meter's only goal is to place the average of whatever metered scene at a middle tone, hoping for greatest range either direction. This does work pretty well for many average scenes, but of course, some unusual scenes will require manual compensation (when reflective metering is affected by too much dark or light colors, meaning low or high reflectivity of colors like black or white, Not meaning in dim or bright lighting). Our human brain can recognize the unusual scenes, then we can help the meter.

Here's the real deal about metering.

We know that mostly white or light-colored content in a scene (snow, or brides dress, or light-colored walls) will reflect light very well, so a reflected meter will read too high, and will reduce exposure to show it as middle gray. Also a dark-colored scene (grooms tuxedo, or for flash, a distant dark non-reflecting background) will read too low, and will be adjusted up to show as middle gray. So the reflected meter does not place these high or low as we would hope, it places everything in the middle. A high reading might mean the light was bright, or it could mean the scene colors reflected unusually well (and should be bright), like white or yellow. The meter is a dumb chip that does not know the difference. It only sees a high reading, and it can only assume all scenes are an average scene (it cannot contemplate things). So according to a reflected light meter, all scenes should and will go to the middle (the scenes average color will). We must learn that the cameras reflected meter is fooled by scene colors.

Preachy here maybe, but intended as hopefully helpful. This fact about scene colors is Photography 101, perhaps not obvious, but basic and clearly evident, and one of the first things we should learn. Our best tool is a human brain that can see and actually recognize the scene. Brains and photographers have experience to know and recognize the difference. We should think about how we work. If we can see there is a white background, or a white brides dress, we know to expect underexposure, so we would compensate to boost metered exposure a bit. Or maybe even a stop or two if the scene is mostly all white. For example, most pictures in the snow probably need +1 EV, and if the scene is entirely snow in bright sun, maybe consider dialing in +2 EV exposure compensation. (Always do what is seen needed, because you will be disappointed if you imagine the camera should always get it right.) Experience lets us "already just know" when we first walk up to the scene. It does require we look, and think a little about we're doing. Do Not turn off the brain while the camera is engaged. :)

Or easier, an incident meter directly meters the light itself (independent of the subjects colors), and in that light level, light and dark scenes will seek their proper high and low levels then, same as we see them.

And metering on a gray card is about the same deal as the incident meter (standard reflection from the gray card, representing the light, and independent of the subject colors). If we did not have an incident meter, this would be the reason we might meter on a gray card (so metering will be independent of the subjects colors).


General guide lines for metering exposure on the 18% gray card (or White Balance would be very similar):


There is a similar 18% gray card discussion at the bottom of the gamma page


The Middle of Histograms

We are aware of a few specific known and certain facts, which don't always match our simplest assumptions about "middle". Two facts are:

The definition of a "stop" is a 2x change, so we do know that one stop underexposure is 50% of the previous proper level.

This part is a bit tedious perhaps, but we also know our image is linear analog light in the lens and at the camera sensor. Then it is digitized to ones and zeros binary data, but raw data is still linear for example. But when the image is converted to RGB data for use, then the exponent gamma modifies all of the RGB data we ever see. The histogram shows this gamma data (for raw images too). "Linear" in math means that functions graph on a straight line (a 2x change has a 2x result, i.e., not exponential), but in computer systems, the word linear is also used to say "not gamma encoded" (and is also my use here). Gamma is sort of a no op for us - our eyes will hopefully see the linear data again later, as the reproduction of the original linear scene. Later when the image light is in the air, sent traveling to our eye, the reproduction will necessarily be linear analog light again, suitable for our eye (same as when we saw the original linear scene).

This Test Case to find the middle point: Suppose we intentionally expose an image so that the brightest white tone reaches histogram 255 level (specific value for this example). Then we intentionally underexpose by one stop to drop this data to 50%, which would be from 255 to be at 128 level, if we assume linear data at the digital sensor. However, all of our RGB picture data we see is always gamma encoded data, which boosts the data numeric values. And histograms only have the gamma encoded data to show, so the histogram data we see is gamma encoded. This means that in our histogram, we see that one stop lowers this 255 end to only about 3/4 scale in gamma data. You'd think that ought to be called "middle". :) Nevertheless, it is handy to know when viewing the histogram and contemplating a one stop compensation, this is the degree of change you can expect (relative to the 255 end).

You can and should easily check this test yourself: (seeing is believing)

One detail first to clarify the "why": Gamma is an exponent, so the [0..255] image data is first normalized to be fractional values [0..1], because 0 or 1 to any exponent is still 0 or 1 (so these end points don't move... same range and no clipping). So 255 is still 1 in gamma. Then one stop down is 0.5 midpoint if in linear. But if gamma 2.2, then it is 186 in the histogram data we see.
  -1 stop is 255/2 = 128 linear, or 0.5.   But gamma is 0.5 (1/2.2) = 0.73,   and 73% of 255 full scale is 186.

So, to check what we see when we create 50% by intentionally underexposing by one stop, then we see 255 fall to about 186, around 3/4 scale. It won't be exactly 186, because the digital camera is also making other tonal changes, white balance and color profile and contrast, etc. But certainly we do NOT see one stop go down to 50%, because gamma data boosts the histogram value. One stop down may be 50% in linear, but we don't see linear data, our RGB images only have gamma data. So point is, we are mistaken if we imagine 128 is the middle of any histogram data that we see. The histogram is semi-log data, so to speak. The elementary articles we see discussing histogram 128 either oversimplify or don't know, but virtually none mention gamma. That's not much help, because gamma in RGB histograms is all that we can see.

Anyway, 18% is 18% and is not 50%, and is not half of anything in digital. And 18% of 255 is 46 in linear data, however gamma raises it to 117 at 46% in our photo data, which is due only to gamma, it is NOT because it is half of anything. Yet then we see that 46% is nearly 50%, so we imagine 18% must be middle of something. We heard it was middle, but we're not sure what middle means (could be it's what the eye might imagine it sees later, but 18% is the actual data now, before gamma changes it to 46%). Likewise we know 128 is half of 255, so 128 must be middle of something too (however remember middle 50% at 128 went to 186 after gamma). So therefore (?), we assume middle must mean middle, so 18% is 50%. :) I'm just having fun teasing, but some accounts are a bit messy. :)

The facts might confuse placing the histogram middle numbers somewhere, but otherwise should not interfere with our normal uses of the gray card (top of this page). Except my peeve with the "middle" thing is that we even hear people telling us to calibrate our meters so a 18% card is at histogram 128, which turns out by coincidence of gamma to be not far wrong (about 0.3 stop, and gamma 2.5 could hit it). But they clearly have no clue about what any of this means. They say "yeah, yeah, but it works for me". But I'd rather leave it alone, because it doesn't make sense, since gamma has nothing to do with determining exposure. And worse, it was NOT even about the 18% card, because a white card or a black card would give exactly the same results (examples here, maybe start at the top of that page), because reflective meters simply put stuff in the middle. It's hard to defend the unique meaning of 18% when a sheet of white copy paper will give the same result (different numeric exposures, yes certainly, but which produce the same photo result). We often don't really know what we mean by "middle", and we do make crazy assumptions. The 18% card might seem different in an incident meter, but here is Sekonics incident calibration procedure. No gray card is used, and they do know a thing or two about meters, and according to Sekonic (and the ISO standards organization), the right idea to calibrate your meter is that IF it repeatedly and continually gives a consistent error on a wide range of scenes, THEN adjust the meter to compensate. Any one single reading is not fail-safe anyway, too many variables.


Metering a precise exposure is a hard problem. Incident meters work well, easiest (if not always convenient). Reflected camera meter accuracy can sometimes be a close ballpark, but reflective meters are really just a guide, not an absolute (depends on scene colors). So after we avoid clipping, then what looks right on the camera rear LCD is really about the best plan. Meaning, if we need to compensate, then we need to compensate.

But when checking for clipping, we always have to look at each of the three RGB histograms. The single gray histogram is NOT valid color data, it is a mathematical manipulation (called luminance, about how grayscale can reproduce brightness of color values). It is NOT true data, and it CANNOT show RGB clipping, usually not at all. The colors register differently, and then White Balance shifts red and blue oppositely. To see clipping, we must look at each of the three RGB histograms. More about the three RGB histograms.

White Balance

18% cards can work for this, but they vary, and they are dark. Kodak sold that printing business and has not manufactured a gray card or a photo book or a filter for 20 years (other companies bought rights to use their products and their name, currently Tiffen, but they sublease it too). Most 18% cards are somewhat neutral (neutral is any gray tone that has equal RGB components, so no color tint). General gray colors out in the world often have pink or blue color casts, but if they look very gray, most 18% cards are pretty close to neutral. But they do vary. They are only intended to reflect 18%. They claim no specs about neutral color. And while they can work, they are pretty dark for white balance purposes. Actual white balance cards are either white, or very light gray.

Actual white balance cards are available, with the goal to be neutral white or light gray. A color tint can show better on a light card. It is not rocket science, it is just white without a color cast. I like the Porta Brace White Balance Card, $5 at B&H. The WhiBal brand card is good too, using light gray pigments which then has to be tested, but they claim to test each card, and it costs more. Not a thing wrong with it, but I have two of each, and I use the Porta Brace.

Use of a good WB card is the plan, and a good plan, but actually, another skill we should learn is this: Many of our photo scenes already contain white objects (white paper or envelopes, signs, porcelain dishes, T-shirts, white collars, church steeples, picket fences, even the white polka dots in the kids pajamas). Many of these are intended to look white (not off-white), and they mostly work rather well too (maybe a few exceptions when we must Undo), but normally much better than if no other planning was done. If no other choice is available, then is this choice better than it was? Often it is real good. However better planning is always a good thing.


But White Balance can actually be quite easy. Simply clicking a good WB tool on a good WB card (in the same light with subject) is trivial to do. The philosophy is that we know the card is neutral (equal RGB, so no color cast). Then after the computer makes that spot actually be neutral color, therefore no color cast in our image. Couldn't be any better, or easier. Lots more about White Balance.


Menu of the other Photo and Flash pages here


Copyright © 2014-2016 by Wayne Fulton - All rights are reserved.

Previous Menu