Histograms and Raw

[bjanes]

I still don't get your "BaselineExposure offset" argument, since you are essentially saying that you are applying negative EC to ACR's default -- this strikes me as being a semantical disagreement.  Also (as I already said), there are other RAW converters, and the use of negative EC is close to universal to all of them for RAW conversions of ETTR files (again, assuming a lower contrast scene).  Of course for high key photographs you will probably skip negative EC in RAW conversion.

I would venture that most scenes are not low contrast. Also, negative exposure is also not need for images where the dynamic range of the scene is equal to that of the camera. Thus, negative EC is hardly close to universal.

BaselineExposure is defined in the Adobe DNG Specification on page 32, which is quoted below:

BaselineExposure

Camera models vary in the trade-off they make between highlight headroom and shadow noise. Some leave a significant amount of highlight headroom during a normal exposure. This allows significant negative exposure compensation to be applied during raw conversion, but also means normal exposures will contain more shadow noise. Other models leave less headroom during normal exposures. This allows for less negative exposure compensation, but results in lower shadow noise for normal exposures.

Because of these differences, a raw converter needs to vary the zero point of its exposure compensation control from model to model. BaselineExposure specifies by how much (in EV units) to move the zero point. Positive values result in brighter default results, while negative values result in darker default results.

When you take a picture, you probably want mid gray (L* = 50) in the scene to be mid gray in the image. In a 2.2 gamma space, L* of 50 has a pixel value of 118. If you take a picture of a gray card (or any other uniformly reflecting surface) you would want the raw file to have a saturation of 18% (mid gray), which would represent a pixel value of 2949 in the 14 bit raw file or 739 in a 12 bit file.  When you take such a picture according to the light meter reading, the resulting pixel value will reflect the calibration of the system, which includes the meter calibration and the tone curve. Nikon (and the ISO standard) specify that light meters are calibrated for 12% reflectance (see Thom Hogan). This allows 0.5 EV for highlight headroom. However, for ETTR exposure, one does not want much if any headroom.

To test my own D3, I took 3 shots of an 18% gray card and looked at the raw pixel values with Rawnalize.
 
[attachment=19043:Rawnalize.png]

The 12 bit green pixel value was 497; the saturation is 497/4095 = 12.1%, corresponding to a pixel value of 99.8 in a gamma 2.2 space, as shown by Rawnalize. The system is at spec for exposure. The sRGB pixel value of the file rendered with Capture NX and the Standard Picture Control was 149. Now that is a hot tone curve and mid gray is lighter than it should be. This is why many photographers complain that the camera "overexposes". With ACR and the Camera Standard calibration, the pixel value was 149. Using an ACR exposure of -0.5 EV gave a pixel value of 126, which is close to the desired value of 118.

[attachment=19045:ACR_scrCap.png]

Using the Adobe Standard calibration and a linear tone curve and compensating for the baseline offset with -0.5 EV, the rendered value in Adobe RGB was 99, just where it should be with a gamma 2.2 space. With a full ETTR exposure with the raw value just short of clipping, the rendered value in sRGB would also be just short of clipping with the exposure offset of -0.5 EV, but the highlights would be clipped in the rendered file without the compensation. If you didn't use the offset, the highlights would falsely appear to be blown and you would think that you overexposed and might cut back exposure in the future. If you judge exposure with ACR (or any other raw converter, including the camera JPEG) you have to know the BaselineExposure offset. Many inexperienced photographers naively assume that the raw converter histogram at default setting represents the raw histogram.

According to tests done by Bill Claff, the D300 behaves in the same way as the D3. If you want 18% saturation for the metered value, you have to use an ISO of 138 rather than the nominal 200. In other words, you have to increase the exposure by 0.5 EV just as Thom described. If you use the real ISO for your Nikon camera (recent digital models), you wouldn't need the -0.5 EV compensation. Of course, I am applying a negative exposure value to the ACR default to bring the rendered values down to where they should be. The raw values were correct, so this is not correction.

You may consider the difference to be one of semantics, but a failure to understand a positive BaselineExposure may result in underexposure. Other cameras may have a negative BaselineExposure. For example, the ISO of the D70 is near the nominal value; however, I don't know what value Adobe uses for that camera. The important thing is to know the behavior of ones camera and raw converter and not be hung up on semantics.

[Tony Beach]

I would venture that most scenes are not low contrast. Also, negative exposure is also not need for images where the dynamic range of the scene is equal to that of the camera. Thus, negative EC is hardly close to universal.

A.) Most of the scenes I shoot are, the primary exception being midday under harsh light.
B.) Redefining what I said does not change what I said.  I said, "Generally with moderate contrast scenes, if you are not applying negative EC in the RAW converter and/or reducing the Brightness slider, you are working with an underexposed NEF file."

Many inexperienced photographers naively assume that the raw converter histogram at default setting represents the raw histogram.

I do not, which is why I recommend that most of the time you should set negative EC in the RAW converter to get middle gray to the middle, and I would add (with NX for example) that is in addition to lowered brightness or (with Capture One for example) no added brightness.

If you want 18% saturation for the metered value, you have to use an ISO of 138 rather than the nominal 200. In other words, you have to increase the exposure by 0.5 EV just as Thom described. If you use the real ISO for your Nikon camera (recent digital models), you wouldn't need the -0.5 EV compensation.

I use positive EC in the camera, followed by negative EC in the RAW converter; usually more of the latter than the former -- YMMV.

Of course, I am applying a negative exposure value to the ACR default to bring the rendered values down to where they should be. The raw values were correct, so this is not correction.

Compensation is not "correction" according to you whereas I see them as interchangeable in this context (you have to compensate the ACR default, which is the same as "correcting" it; the same as what happens with nearly all the other RAW converters).  Spare me from this semantical debate and stop obfuscating this.

You may consider the difference to be one of semantics, but a failure to understand a positive BaselineExposure may result in underexposure.

At some point it would be useful to get back to why it is important to adjust the settings in your camera so that its histogram gives you a reasonable representation of the exposure based on the RAW data.  For this, we can go way back in this thread and disregard everything that has been written since then.

Other cameras may have a negative BaselineExposure. For example, the ISO of the D70 is near the nominal value;

Right, less headroom in a D70 file, and in a D200 file too.

however, I don't know what value Adobe uses for that camera. The important thing is to know the behavior of ones camera and raw converter and not be hung up on semantics.

Ignore the RAW converter for the moment, and (as I said above) get back to the camera's settings.  Once you have a well exposed file, then configure the RAW converter to take advantage of that.  Something I found interesting that Iliah Borg has stated more than once is to expose for the RAW converter.  For example, ACR does better than others at actual highlight recovery (which is where highlights are actually overexposed), so in some high contrast scenes where highlights need to be sacrificed know what ACR can or can't do and expose with that in mind.

[Guillermo Luijk]

For example, ACR does better than others at actual highlight recovery (which is where highlights are actually overexposed), so in some high contrast scenes where highlights need to be sacrificed know what ACR can or can't do and expose with that in mind.

Unless things have changed in the last versions, the only thing ACR does for highlight recovery is to produce monochrome detail in the areas where some RAW channel got clipped, by copying the information from the non-clipped channels onto the clipped ones in a neutral (R=G=B ) fashion. That is the standard behaviour expected from any decent RAW developer. ACR does not even have an option to try to recover colour, which DCRAW (a simple command line RAW developer written by a single guy) has:

Left ACR, right DCRAW: