of cause they are thats what makes white. and thats why i just clip the whites so everyone can see a clipping measurement. if i didnt want to clip anything
i would set 1/3 stop less exposure. but its its not part of the experiment.
May 18, 2023
305
-
-
No-no-no...
White in raw is not R=G=B. R=G=B on white is only true after the correct white balance is applied to raw data, and that happens in raw conversion.
BTW I doubt 1/3 would save the day. -
-
from P1016721, clipped in all 4 channels, here is one of the greens:
-
No that is not correct at all.
If you fill the camera's frame with a near black object containing no whites at all you can still get clipping in the raw data if you leave the shutter open for long enough.
It is the demosaicing of the raw data and application of white balance plus whatever other number crunching is applied during the raw conversion into an rgb image that sets what is white.
While the shutter is open, the sensor's pixels are just collecting photons of light from the scene without knowing what the colours in the scene actually are. There could very well be no white in the scene at all and the raw data could still be clipped.
-
Yeh on a tiny part of the shiny peice of ribbon.
-
That.
-
Why do you say that? What, in the processing of the raw data, actually changes the values to induce clipping?
Alan
-
Agreed, Iliah,
Especially with Foveon 'Merrill' sensors ... the raw red channel being about half as sensitive as the other two ...
-
In Lr and ACR, the displayed histogram is after development. A few developing controls that can introduce clipping:
"Exposure"
Contrast
Tone curves
Sharpening
Lighten whites.
Lighten highlights
Detail
Texture -
The brief answer is 'everything'. Processed images have little direct automatic relation to raw ones. If you look at processing as a function, the input and output data are completely different data types. The input (raw numbers) essentially represent per-pixel exposure, which is, at least in theory an open ended scale (though the camera sensor and electronics will clip at some point). The output is numbers in a colour space, which is bounded - and represents not how much light there was but how a human observer should see it. So far as lightness (how light or dark it looks) is concerned there is a definite white point, and within the space there can be nothing lighter than that. The white point in default processing is usually quite a way short of the maximum raw number. The spare raw numbers are called 'raw headroom' sometimes, and are what you have in hand if you set the processing parameters yourself and choose a higher white point.
-
In my hack raw processor I can select any tool in the processing toolchain to display its output (not that useful), and the corresponding histogram (quite interesting). As I incrementally select the tools, the histogram doesn't change appreciably until I encounter either 1) exposure compensation, or 2) a non-linear tone curve. #1 in particular, being a simple multiplication of the data, has the capability to push data past the rendition white point.
Note that it's a different kind of clipping in the software. In the camera, it's about saturating the sensor, trying to record intensities past its capability to resolve. In the post-processing software, it's about pushing data past an arbitrary number called "white". Depending on how the data is internally organized, the over-the-top values may still exist in their relationship to each other (why a lot of software uses floating point numbers internally), but eventually they need to be pulled back somehow to under the white point to be rendered for display or export to an image file. In the camera, the saturation just piles up measurements at the maximum value; in the software, there's the opportunity to bring them back into the fold with a more intelligent algorithm that preserves some notion of the tone gradation.
-
That may well be true in your raw processor, but in Lr and ACR, the "Exposure" control is nonlinear.
blog.kasson.com/?s=lightroom+exposure
In the modern cameras I've tested, the ADC clips before the sensor saturates.
-
It's a matter of a procrustean bed ;)
-
Lions don’t like histograms.
-
Do any monochrome implementations of a sensor use more of the non-linear range of the photosites than the color implementations of the same sensor? If I had a monochrome camera, I certainly wouldn't mind a base or "extended" low ISO setting that was non-linear near raw clipping.
-
At base ISO (100), the Q2M has some nonlinearity. I don't know about the Q2. I consider the base ISO of the Q2M to be 200.
blog.kasson.com/leica-q2-monochrom/leica-q2-monochrom-highlights-at-iso-100-crops/
blog.kasson.com/leica-q2-monochrom/leica-q2-monochrom-highlights-at-iso-100-graphs/
-
We're both wrong:
