About that 18%

[Doug Kerr]

In discussions of exposure metering and such we often see mention of the value "18%". But just what that represents, and under what conditions, is often unclear - and some of the statements we see are just wrong.

The situation is complicated for the researcher by the fact that this number emerges from the interaction of several ISO standards, none of them prescribing or assuming it directly, each seemingly assuming that is has been defined or assumed by another standard, often indirectly.

If we follow this trail of shadows, the following seems to be the case:

• The trail of reflective light photographic exposure metering starts with the assumption that the "representative" scene:

•• Has an average reflectance of 18% (aha!).

•• Has a maximum reflectance of 100%

But, as "seen" by the exposure meter, and later the camera, this means:

• The average luminance of the scene is 18% of its maximum luminance.

If we follow the trail through the ISO standard for the sensitivity of digital cameras and the ISO standard for photographic exposure meters, and choose an appropriate value for the reflected light metering calibration constant, K, we find that for a metered exposure, the average photometric exposure on the camera sensor will always be (regardless of the scene particulars) about 12.7% of the saturation exposure (the highest exposure at which the imaging system cam properly determine differences in scene luminance—that is, properly record detail).

Now for our "representative" scene, with an average luminance of 18% of its maximum luminance, that means on the sensor the greatest phtometric expose will be about 70.6% of the saturation exposure - that is, almost exactly 1/2 stop short of saturation. That 1/2 stop "clearance" is viewed as "headroom" to accommodate scenes that do not fit the "representative" model.

Note that this only works out this way if we choose the "appropriate" value of K for the exposure meter. There is no "correct" value of K - that can be chosen by the exposure meter manufacture from a range given in the ISO standard to "tune" the meter to what they feel is the best for users.

Best regards,

Doug

 

[Ted Cousins]

Hello Doug,

In some related research, I sought the theoretical value of "middle gray", commonly said to be 18% reflectance. In particular, I examined the Zone System where the band called Zone V is said to be middle gray and is said to be one step wide. Turned out that the mid-point of that Zone is not 18% reflectance ...

https://www.dpreview.com/forums/threads/mid-tones-what-is-their-range.4818690/page-3#post-68496531

See also: https://kronometric.org/phot/std/ANSI_PH3_49_1971.PDF

 

[Doug Kerr]

Hi, Ted,

Hello Doug,

In some related research, I sought the theoretical value of "middle gray", commonly said to be 18% reflectance.

Yes there are some rationales for that, when we consider the nonlinear scale definitions of various color spaces. In one of them (I forget just now which - maybe the L* of the CIELAB color space), the representation of 18% (of something) is nearly mid-scale.

In particular, I examined the Zone System where the band called Zone V is said to be middle gray and is said to be one step wide. Turned out that the mid-point of that Zone is not 18% reflectance ...

No, we really can't work that out with a logarithmic scale that runs from "black" to "white" (an "infinite" range").

But if for example we hold our noses and take the midpoint of Zone X to be "100% (of something), then the midpoint of Zone V would be about 3.1% of that something.

Best regards,

But like so many things in life this is "art", and we engineers are exhorted to be respectful of that (and not try and work out everything with algebra!).

Best regards,

Doug

 

[Doug Kerr]

Hi, Ted,

Indeed, in the CIE L*a*b* color space ("CIELAB"), where L* is the nonlinear representation of relative luminance, for an L* value of 50 (midway on the 0-100 scale of that value), the indicated relative luminance is 0.1842 (18.42%).

This might be the premise for some saying that the 18% assumed reflectace of a "repreentative" scene was chosen as it corrponds to "mid gray".

Best regards,

Doug

 

[Doug Kerr]

The interaction of the various constants in the ISO standards for (a) determining the "speed" of a digital camera and (b) the behavior of photographic exposure meters leads to this interesting situation:

I assume that the photographic exposure for a certain shot is that recommended by reflected light metering with an exposure meter following the ISO standard, having some value of K (as chosen by the meter manufacturer).

Then the average photometric exposure on the digital sensor will be almost exactly K% of the saturation exposure.

Thus the often spoken of "average photometric exposure for a properly metered shot", 12.8% of the saturation exposure, presumes (this never being said) metered exposure with an exposure meter having K=12.8.

This "ideal" value of the phtometric exposure is based on the notion that, for a scene with an average luminance that is 18% of its maximum luminance, the maximum photometric exposure on the sensor is 1/2 stop below saturation (that is, has 1/2 stop of "headroom").

If in fact the scene had an average luminance of 12.8% of its maximum luminance, then, for a "properly metered" shot, the maximum photometric exposure on the sensor would be right at saturation.

In that case, the average photometric exposure on the sensor would be 18% of saturation. This might be the source of one of the stories about "18%" that we see.

Recall that this is all predicated on an exposure meter with K=12.8.

Best regards,

Doug

 

[Doug Kerr]

In various of my recent essays in this general area, I allude to the ISO specification for determining the "ISO Speed" of a digital camera. (Yes, that is its real name, used in the interest of historical continuity.

That definition, along with the ISO standard for the behavior of free-standing photographic exposure meters, leads to this.

For a shot of a scene whose average luminance is 18% of its maximum luminance, based on reflected light metering with a meter with K=12.8, the maximum photometric exposure on the camera sensor would be 1/2 stop down from saturation. This is said to provide "1/2 stop of headroom" to accommodate situations that did not follow that scene model.

******

But in fairly-modern times, digital camera exposure control systems became mush more sophisticated than the classical exposure meter, and so it came to be thought that this "headroom" was no longer necessary. As a result, the ISO introduced a new measure of digital camera "sensitivity", the ISO Standard Output Sensitivity (ISO SOS).

Simplistically, in most cases this rating would be essentially "1/2 stop" greater than the ISO Speed. In the classical metering context, this would mean that we have "spent the headroom".

At this writing, for many digital cameras, in the manufacturer's specification sheet, it is made clear that the various "ISO" settings that were available were stated in term of the ISO SOS.

That means that if we for some reason wanted to determine the photographic exposure to be used for some shot with our digital camera with an external exposure meter, the resulting exposure would be "1/2 stop hotter" than "in the day".

Best regards,

Doug

 

[Ted Cousins]

<> But in fairly-modern times, digital camera exposure control systems became mush more sophisticated than the classical exposure meter, and so it came to be thought that this "headroom" was no longer necessary. As a result, the ISO introduced a new measure of digital camera "sensitivity", the ISO Standard Output Sensitivity (ISO SOS).

Simplistically, in most cases this rating would be essentially "1/2 stop" greater than the ISO Speed. In the classical metering context, this would mean that we have "spent the headroom".

At this writing, for many digital cameras, in the manufacturer's specification sheet, it is made clear that the various "ISO" settings that were available were stated in term of the ISO SOS.

That means that if we for some reason wanted to determine the photographic exposure to be used for some shot with our digital camera with an external exposure meter, the resulting exposure would be "1/2 stop hotter" than "in the day".

Best regards,

Doug

Hello Doug,

Very popular these days seems to be the REI method which allows manufacturers to state an index based on many tests, in other words: anything they like. It seems rare these days to see ISO speed or SOS in a recent image's exif.

A great pity because, with those methods, one could measure one's own camera "ISO" (been there, done that roughly long ago for Sigma SD9 ... got 95 for setting of 100).

See part 2 of the now-defunct CIPA DC-004 -Translation- 2004 whose methods were absorbed into the ISO Standard around 2006.

 

[Doug Kerr]

Hello Doug,

Very popular these days seems to be the REI method which allows manufacturers to state an index based on many tests, in other words: anything they like. . . .

Yes, I didn't mention REI in my note, mostly for that reason.

It seems rare these days to see ISO speed or SOS in a recent image's exif.

Indeed, often the specs cite the ISO REI.

A great pity because, with those methods, one could measure one's own camera "ISO" (been there, done that roughly long ago for Sigma SD9 ... got 95 for setting of 100).

See part 2 of the now-defunct CIPA DC-004 -Translation- 2004 whose methods were absorbed into the ISO Standard around 2006.

Indeed.

Thanks.

Best regards,

Doug