Use a Lightmeter for Exposure
That is not meant to come across as cheeky or snobby at all, that is not my intent — but the histogram does not (necessarily) tell you if you are "filling the buckets" and really does not tell you your "exposure" relative to the RAW data.
When you change "picture settings" the RAW data is not at all affected (assuming real RAW). What IS affected is the jpeg preview, and that is what the histogram is calculated from. Thus that only shows you the exposure relative to the preview settings, not relative to the RAW data.
On the sensor, think of each RAW pixel as a bucket that can hold a certain number of photons of light1. Once the bucket is full of electrons, it can not take any more photons (clipping) and when the bucket is empty or close to it, it has a random number of electrons which is never exactly zero (noise). Your useable image data is between these two extremes of clipping and noise.
The way the buckets 'fill up" is linear relative to the light hitting it. Twice as much light means twice as many photons hitting the sensor and thus twice as many electrons in the bucket are stored.
Film is decidedly NOT linear in response to light, and neither is human vision, nor jpeg image data.
- Film is essentially logarithmic over much of its range, but at the high and low ends tapers so there is a very soft clip at which and black.
- In contrast, RAW sensor data is linear and when it gets to the clip
point, it rather abruptly stops.
- JPEG image data is usually stored with a transfer curve such as sRGB, which is approximately a power curve with a 0.4545 exponent.
- Human vision is way outside the scope of this post, but for the sake of discussion let's say photopic vision is approx. a power curve with a 0.43 exponent.
There are plenty of theories on exposure best practices. Ansel Adam's zone theory suggested Zone 5 (middle, 18% grey card) be exposed to 50% density, i.e. right in the middle of the exposure/density curve. This allows for the highlights to gracefully blow out with film's extended high end/soft clip, and also retain enough density in the negative for the dark areas of the image.
But zone 5 should definitely not be at the 50% point of "buckets full" on a CCD, as the "data in the bucket" is linear to light. A 1:1 relationship would mean an 18% luminance would be a bucket that is 18% full (or 82% empty for the pessimists... 😆)
But setting the 18% grey card so that it is 18% in RAW data may be correct, as that can still result in poorly clipped highlights depending on the subject matter, or the corollary of too much noise for the darker areas of the image it that is more important than highlights for the given subject.
The various preset picture settings are just ways the manufacturer to debayer the RAW data into a useable image, weighting the highlights and shadows differently. None of these affect the RAW data, but the settings do affect the histogram, and how highlights are clipped/rolled off and black is handled.
The distance between where the white level soft clipping starts, and the hard clip of the sensor (full bucket) is a headroom the manufacturer chose based on their ideas of how an image should look and feel (i.e. Nikon, Canon, Sony all have different ideas here — Nikon seems to under-expose the sensor to preserve more highlight data, and as Nikon sensors are excellent in low noise they have plenty of room to do so).
The implication here is that the histogram is not necessarily telling you exactly how full the buckets are. Thus, their utility as an exposure reference may not be ideal.
So with RAW it comes down to:
- At what point can you tolerate the hard clip of highlights.
- Consider how much headroom you want above where you'd like to start rolling off highlights, to the "brick wall" of "full".
- How much noise can you accept in the darker areas.
If you can make a custom "picture setting" so the histogram shows you this information in a way that's valuable to you in your workflow, that might work. But I'd be cautious of "standard" or other presets.
Personally, I much prefer to use a Sekonic and meter to find the exposure the old fashioned way. If you are going to use a meter, I do recommend finding how the meter correlates to the camera's brick wall highlight clipping point — meter and shoot a white card increasing the camera's exposure in 1/3rd stops till you find the point the RAW data clips at least one of the channels.
Then make a note as to how the camera exposure settings at that clipping point relate to the reading on the meter. A useful target setting for shooting then is a nominal exposure with an 18% grey card at least five stops lower than that clipping point. Seven stops is probably better if you have highlight details you want to preserve.
I find I shoot about a stop darker per the camera's standard preview image so I boost exposure in post in Lightroom — I prefer this as it gives better control for how highlights are handled, and I hate losing highlight detail if I can avoid it.
1) The "bucket" does not actually hold photons, but when the sensor's pixel is hit with a photon, it creates a charge by knocking free one or more electrons. The "bucket" can hold a certain number of electrons.