Raw histogram and why does it matter

I'm getting the impression that you think that people are recommending use of a raw histogram to improve OOC jpegs or default conversions. No, the purpose of exposing relative to the raw clipping point is to break one of the vertices of the alleged "exposure triangle" open and separate the two endpoint as "lightness" and "exposure"; you convert an image taken with the ISO 100 setting, for example, as if it could be ISO 47 that you are actually shooting at, or ISO 237 if there are some specular highlights that you want to preserve.

Using the RAW histogram automatically implies using ISO settings and ISO exposure indices independently.

You chose a ready-for-display-with-default-parameters TIFF. That is not the raw image.

A histogram in a raw converter (including those in-camera ones that render the preview), with a few exceptions, is not of a raw data, but of a conversion. That means that, at minimum:
- white balance is applied, thus the red and blue channels are promoted to the right (take a shot of a neutral object, set white balance from that object, see in a converter how the channels are aligned on a histogram in the converter, see how they are not aligned on a raw histogram), which may leave an impression of red / blue channels being clipped
- the conversion conforms to ISO, that is all the channels are usually promoted to the right, by something between 0.5 and 1.5 stops, sometimes more
- the data is converted to some working colour space, which skews the histogram further.
Raw histogram is useful, for example, if one wants to know if the raw data is truly clipped, and what areas of the shot are clipped in raw.
We explain all that in our blogs.

Is RawTherapee's raw histogram one of those few exceptions?
I have used it to see if my photo has clipped parts or not. If not, I should be able to make non-clipped tiffs too.

For OOC JPEGs and default converter conversions with no user input, the in-camera histogram is pretty good. If you want to explore noise vs raw headroom and optimize for both, then the raw histogram is what is most relevant.

You won't visibly see a huge improvement in SNR in some cases, when you expose more to the right, but for some applications, it can make a pretty big difference. Case in point: when the Canon 90D came out, it had the highest pixel density of any APS-C or FF ILC, and a lot of people complained that even in the midtones of low ISOs, you could see noise at the pixel level and it got worse if you needed to sharpen. This was true even if you shot at ISO 100 exposed normally, if you sharpened enough. If your subject tones allowed, however, and you exposed ISO 100 for ISO 25 (imagine a scene with no bright whites in it), then you could sharpen the heck out of it, because you had half the photon noise (and one quarter the read noise) than if you had exposed "properly" for ISO 100.

If that extra SNR is not important to you, and you are happy with the ISO exposure index always being the same as the ISO setting, then you and your in-camera histogram can live happily ever after, together.

I think the answer is here: rawpedia.rawtherapee.com/Editor#Main_Histogram

Perhaps, but the raw data did not necessarily do it. You could have told a converter to darken the output, and then there might not be clipping.

I think I have read that in Rawpedia but I'm interested in if I can trust on it.

I read Your answer I can.

Most raw files have bit depths that allow values up to 4095 (12-bit), 16383 (14-bit), or 65535 (16-bit), and may have offsets for "black" that are 256, 400, 512, 1024, or 2048, etc. In this case, there are clearly values up to about 7000 in the histogram, so it can't be 12-bit, but is probably 14 bit (few cameras are 16-bit; you need really big pixels to benefit from 16 bits). 14-bit raws typically have upper values somewhere around 16000, but could be lower (around 14000) due to firmware clipping, or can be lower because the histogram was already subtracting the value of black. So, there is probably about a stop or a little bit more of unused headroom here.

The x-axis of the raw histogram is the range of counts (DNs, if you please, although I hate that language) that the file can represent. If the number of bits is n, that range is from zero to 2^n-1. For at 14-bit converter, that's around 0 to 16000. It can vary depending on whether the black point is subtracted or not; RD offers both options. In-camera processing can also affect it. But 16000 is a decent estimate for this kind of work if someone else picked the camera and set up RD.

One stop is one binary order of magnitude. So, 8000 to 16000 is a stop. So is 4000 to 8000, So is 2000 to 4000. You get the idea.

I'll have to remember that a brand new thread early in the morning in New York may already be old news across the pond.

If you can expose an extra stop to the right with a red tulip, and the raw does not clip, you could still have that "flat magenta" problem with your converter even if you pull down brightness, however, it is possible to effectively cut the raw values in half, and you still have 1/2 stop less photon noise, and 1 stop less read noise, with the same results. Optimizing for noise and trying to avoid the "flat magenta" effect in a red tulip are theoretically separate issues, which may only be conflated by a particular limited workflow.

To be sure, you can compare it to the histogram in FastRawViewer (trial is free and fully functional for 30 days).

Got it. Thanks. Are the top line values (EV-8 thru EV+3 here) relevant?

Sort of. The assumption is that EV 0 is three stops down from full scale. One EV delta is one stop difference.

Got it. matches this from FastRawViewer

One thing I learned already 20 years ago: never download any free trials of any program. Buy it or leave it!

The main problem for me here is that there is no Linux version of Raw Digger or Fast Raw Viewer. I have left Windows and Adobe behind.

I run both FastRawViewer and RawDigger under wine, very straightforward.
Our trials are safe and non-intrusive, I guarantee it.