This has come up before⊠2^14 = 16384 so I doubt the sensor would be that good and that linear basically up to the max possible value⊠I suspect the other values for the Sony cameras ie 15300 - 15700 are more likely right
Empirically, 15360: if you use 16383 (darktable default) with an image with overexposed areas (sky), you get the effects of clipping (magenta) without the raw clipping indicator showing any over-exposed areas.
Set the raw white point to 15360, and the areas indicated by the clipping indicator correspond to the visually over-exposed areas (more or less).
@priort : the sensor doesnât fill up to a specific digital value, thatâs the job of the amplifier/ADC circuitry. I guess itâs much easier to make sure the valid values are somewhere high enough, and then indicate the maximum reliable value, than to tweak the circuitry to exactly fill up the available range of valuesâŠ
What you say makes intuitive sense to me, but there are also lots of entries in cameras.xml where black = zero and white = 2^N exactly.
I am new to dt and trying to make sense of the RAW & EXIF data so am still unclear as what is completely raw or perhaps massaged-raw data to make a product perform/look good.
I would trust the exif dataâŠ. Problem for some cameras is they donât offer a white value at allâŠthe sony ones have it, its just in the maker notesâŠ
You can of course do the simple thing and try both but I suspect the one in the maker notes lower that the max possible sensor data is correctâŠ
Some of the Canon ones can be even more confusing as they provide no white level or it changes with iso and they give a value around 10000 as max linear ie I assume something like it is linear up to that limitâŠ. But I really donât know about how this linear value gets used if at all by any raw editorsâŠ
Well I suspect you could apply offsets or gains or whatever across that 2^14 range and there would be the question of how linear it is across that rangeâŠin the end you are going to have in theory 2^14 available for resolution from the sensor but I wonder if most sensors vary enough or batch to batch there is some variability and so this gets pulled back a bit for safety??
Apparently this varies based on design - MOST cameras have the ADC clip point set before the sensor saturates, but apparently not all. There was a recent post somewhere on DPR regarding this and âextended ISOâ from one particular Panasonic camera - for that camera, the âextended ISOâ actually exposed a region of sensor performance where the response was nonlinear.
No just that you were in the parallel post for the G90 and that was talking black pointâŠthis was Sony but you might have been crossing over with your comments⊠I wasnât sureâŠ
Use the command line flag that lists the groups.
Youâll see that there are multiple sources for metadata inside a raw file , and these white level tags are inside different sources of the same file .
In the cheatsheet is an example to âlist all tagsâ, because otherwise exiftool will only list one of these kind of duplicate tags.
Anyway, for Sony there is one in the exif data. Thatâs the 16384 one . And that is somewhat meant to tell the theoretical max white value you could encounter , based on the bitdepth and file format.
The later one is inside âmaker notesâ, a manufacturer set of tags , that differ per camera maker. For instance it can describe settings of your Sony camera at the time of the shot .
The white level there is IN THIS CASE the better one, because it tells the maximum THIS MODEL can do with these settings.