If you insist

On the previous page, I strongly suggested using an L* workflow in which you employ working color spaces with L* tone curves, calibrate your monitor to an L* curve, use the LAB color space to the extent possible for post processing, and so on. But most of you won’t. I get it. You’re used to AdobeRGB or ProPhotoRGB or sRGB or whatever. Your print service has a hard enough time dealing with anything but AdobeRGB or sRGB, and if they saw another color space tagged to your files, they’d toss it. Chances of them accepting an LAB file are the usual slim/none function.

So, you want to know what Zones mean in standard RGB language. This is fairly straightforward to get at. You take theoretical or practical L-channel Values for Zone exposures, and convert those into RGB Values using the tone curve for the color space that you’re interested in. I distinguish between theoretical and practical L-channel Values because we’ve already seen that a real DSLR has a film-like S-curve. [I’m not introducing some new variable, S; I’m just saying the shape is like an “S”.]

For theoretical L-Values, we might consider starting with Value V at 0.18 in terms of Y/Yn and going up by 1-stop steps to 0.36, 0.72, and then the fatal 1.44, for Zones VI, VII, and VIII. We could translate these Y/Y_n numbers into L-Values with the equation for L. By Zone VIII, our theoretical goose is cooked. With a real DSLR though, we found that the L-Values for Zones VIII and higher didn’t blow up off the scale: they just got compressed into the available range of Values at the upper end of the scale as they must. The same occurs at the low end of the scale as well: tonal compression. There is one big difference between compression at the upper and the lower end of these tone curve scales. At the upper end, a Value greater than maximum is impossible. As we go up by each stop, we are doubling luminance, and there is a limit to how many times we can double luminance, theoretically and practically. Going down from Value V though, we are dividing by 2 at each stop; and there is no theoretical limit to how many times we divide. The practical limits to this division process are the resolution of the analog-to-digital converter in the camera and noise at the ISO setting that we are using.

Current professional-grade DSLRs can deliver 14-bit RAW images. It is natural to process these as 16-bit files in applications like Photoshop. At base ISO, the combination of quantization noise and sensor noise is usually equivalent to a few bits. That implies that of the 14-bit RAW range, 13 bits are useful data before any noise reduction. With noise reduction, at base ISO, for most images, the full 14-bit range may well be accessible. This implies that we may be able to do a better job of pulling information out of shadows than out of highlights, at least for images at or near base ISO. Expanding this 14-bit RAW file into a 16-bit file when applying some color space to the data in initial post processing and subsequent editing means that digital artifacts will be minimized and the proper texture and detail can be maintained.

To summarize, don’t shoot JPG.

OK. Here is the table that you’re looking for. It shows the various tone Values as converted from the Zone exposures I got from the previous exercise with my D700.

Zone

LAB

LStar-RGB

sRGB

AdobeRGB

ProPhotoRGB

0

1

3

4

10

6

I

2

6

9

17

10

II

6

15

19

26

16

III

17

44

42

46

32

IV

33

84

78

79

61

V

55

141

132

131

113

VI

77

196

190

188

176

VII

92

234

231

231

225

VIII

97

247

246

246

244

IX

99

254

254

254

253

X

100

255

255

255

255

 

I want to emphasize this one point. This table is based on actual values for my D700. Note first off that it’s reading about 1/4-stop high for Zone V. This translates into similar offsets for the other Zones. Other cameras may well give different results than what I am showing here. The theoretical value for Zone V should be L=50, not 55. I could have entered a 1/3-stop exposure compensation to get the result closer to 50; but I thought it was more instructive to show the slight variation in metering that my D700 has.

It is straightforward to reproduce a table like this for your own camera, if somewhat tedious. All you need is your camera, your computer, and Photoshop. In fact, I strongly encourage you to do just that.

The first thing to notice is that the values in the LStar-RGB column are precisely what you get by multiplying 255 by the number in the LAB column, taken as a percent. This is because the two color spaces have the same tone curve. It’s just that LAB’s L-channel reads out of a maximum of 100 while the LStar-RGB column (expressed as 8-bit values) reads out of a maximum of 255. For the other color spaces, I performed a relative colorimetric conversion (with black point compensation) in Photoshop to get the values in the table. I did not notice any differences in the conversion intent, presumably because the images were all neutral. I would not guarantee the general truth of this equality for images with color information.

Now, many other digital Zone System models are based on the idea of equal steps along the tone curve within a color space. The idea would be as simple as starting in the L-channel with 50 in the middle, and working up and down by steps of 10 to define Zones. You can see here that the step size per Zone in the middle of the range of light is closer to 20, and it is compressed at the high and low ends of the scale. This is exactly like the case of film.

In theory, the exact L-Value that correspond to 1-stop steps up and down from Value V would look like what is in the table below:

 

Value

Y/Yn

L

0

0.00575

5

I

0.0115

10

II

0.023

17

III

0.046

26

IV

0.092

36

V

0.184

50

VI

0.368

67

VII

0.736

89

VIII

1.472

116

IX

2.944

150

X

5.888

193

Plainly, this table runs off the edge of the world for Values north of VII. As I have shown earlier in this set of pages, it is quite possible in the real world to get specular highlights or direct lighting values that are several stops higher than Value V. In one way, it makes perfect sense to talk about these Values as so many stops more than Value V. In another sense, given that our aim is to present images as prints or rendered on screens, we want to compress these extreme Values down to Zones VIII or IX and yet do so without inflicting damage on our mid-tones.

You’ll see from the table above that the theoretical 1-stop changes in Value up and down from Value V are about 15. The changes given by my camera are somewhat greater; namely, 20. As well, the theoretical table shows no compression. This too is an effect of processing in a real camera. Well, we don’t take pictures with equations, we use cameras. Hence, it makes sense to me to consider the equipment that we are actually using, as opposed to theories about how they might work in some idealized reality.

As well, this is the behavior of my D700 with RAW conversion through Adobe Camera Raw and into Photoshop. If I use Phase One’s Capture One Pro v7 instead, I find the following set of L-Values for my Zone exposure sequence. [Conversion was done using the Nikon D700 Generic V2 ICC profile and the Film Standard V2 Curve. The working proof profile was LStar-RGB.]

 

Value

L*

0

0.16

I

1.22

II

4.98

III

15.0

IV

28.9

V

51.8

VI

75.1

VII

91.6

VIII

98.5

IX

100

X

100

 

But in Capture One, the Film Standard V2 is just one of 6 curves and the Generic V2 ICC profile is one of 2 profiles for the D700, meaning that there are 12 different ways to do RAW conversion. Each one shifts the tone curve to a greater or lesser extent. In effect, even with a single camera, these different combinations of profile and curve produce different conversions that may suit any given image and your personal taste more or less. These variations allow for more or less mid-tone contrast, shadow or highlight compression, and so on. Note that the Zone V exposure is much closer to 50 in this conversion than it was using ACR and Photoshop. The central 1-stop change in L-Value is here about 23, yielding strong mid-tone contrast. With Capture One’s Portrait Curve and the Generic V2 ICC profile, the 1-stop change in the mid-tones comes down to less than 21, yielding softer contrast for portraiture.

This is to emphasize, once more, that there is no absolute relationship between RGB values on a tone curve and Zones. Zones will depend upon the conversion method and the camera, at a minimum. Rather than this being a weakness, it is, IMHO, a strength, since it gives you as the photographer more control over your image than you would otherwise have had. In the days of film, you would have had to commit to the tone curve of the film in your camera, and your most significant option once an exposure had been made would be in development; that is, normal, push or pull. Now, with digital, you can defer the choice of a tone curve to RAW conversion, and you can non-destructively select any number of options depending on the subject matter. And all of this before we have barely touched on the power of curves adjustment. More of this to come.

Previous Page … ••• Next Page …