moondigger wrote:
As others have already pointed out, Canon has not produced a camera with less than 14 bit capture in over a decade. Why on earth would they revert to 10 bits now?
ETA: And even before 2007, Canon's 5D (first generation) had 12 bit color. So you'd have to go back even further to find a camera with a bit depth of 10 bits/channel.
Even 10D (2003) had 12-bit RAW, so you have to go back even further
it will be 14-bit when using the physical shutter (EFCS and full) and 12-bit for all electronic modes.
'if' is the biggest 2 letter word in the human language.
for confirmation purposes, can you supply us with your source for this information relevant to the R5?
without reference sources this is speculation not fact. although i tend to agree, the bit depth should be at least '14 bits'.
however, Seriously, it is not impossible that 10 bits is a compromise made in order to read out 47 MP x 2 for dual pixel continuous AF at 20 fps. or even 12 bits is a very real possibility.
the underlying premises for the OP question. how will a lower bit depth effect large format printing? you gain larger print sizes from the 45 mp sensor but you loose fidelity and colour gradient transition at a bit depth less than 14 bits. so what will be the effect on the resolution and quality of very large format and archival prints?
depending on the answers to the question in the above paragraph. it may be better to consider the 30 mp 14 bit EOS R or EOS Ra with a RAW file that is capable of producing a higher print quality for my applications. astrophotography and extreme macro work that will be printed large format. the logical choice the EOS 5DrS is a good choice for large format printing but is not good for DSO astrophotgraphy.
InnomnateViem wrote:
'if' is the biggest 2 letter word in the human language.
for confirmation purposes, can you supply us with your source for this information relevant to the R5?
Go to the Canon Website and grab the "EOS-1DX Mark III Advanced User Guide" (eos1dx-mk3-ug_en.pdf). On page 949 you'll find a table with the 14/12-bit info:
"JPEG (8-bit), HEIF (10-bit), RAW (14-bit Canon original)
RAW+JPEG simultaneous recording possible
RAW+HEIF simultaneous recording possible
* 12 bit A/D conversion processing is applied for RAW images shot with the electronic shutter."
10 bits RAW files will normally not affect large print quality at all.
The required bit depth for preserving tonal gradations above noise varies with the sensor resolution. The higher the resolution, the lower the number of required bits. The R5 at 10-11 bits will give the same normalized tonality as the 1DX III does at 12 bits.
It is even likely that the the R5 at 12 bits gives better tonality than the R at 14 bits, because the R5 sensor may give cleaner signal to the bits you have.
The R5 will guaranteed be possible to shoot at 14 bits in some modes even if it goes to 12 or 10 in others.
A much more important question for high detail printing and astro is; will the camera support electronic first curtain shutter?
InnomnateViem wrote:
'if' is the biggest 2 letter word in the human language.
for confirmation purposes, can you supply us with your source for this information relevant to the R5?
without reference sources this is speculation not fact. although i tend to agree, the bit depth should be at least '14 bits'.
however, or even 12 bits is a very real possibility.
the underlying premises for the OP question. how will a lower bit depth effect large format printing? you gain larger print sizes from the 45 mp sensor but you loose fidelity and colour gradient transition at a bit depth less than 14 bits. so what will be the effect on the resolution and quality of very large format and archival prints?
depending on the answers to the question in the above paragraph. it may be better to consider the 30 mp 14 bit EOS R or EOS Ra with a RAW file that is capable of producing a higher print quality for my applications. astrophotography and extreme macro work that will be printed large format. the logical choice the EOS 5DrS is a good choice for large format printing but is not good for DSO astrophotgraphy....Show more →
I don't understand why you keep shutting speculation down and asking the posters for sources. There are no sources. You asked a question that can only be answered with speculation. As has been stated numerous times in this thread, Canon hasn't released this information. People are doing their best to answer your question given all the information at hand.
InnomnateViem wrote:
how will a lower bit depth effect large format printing?
It wouldn’t. Anybody who claims they can see a difference between a print made from an image that was captured at the camera sensor with 14 bits and a print made from an image that was captured at the camera sensor with 10 or 12 bits is fooling themselves. Many printers don’t even accept more than 8 bits/color through the print driver.
This particular spec gets outsized attention compared with the value it provides. I'm not saying capturing 12 or 14 bits/channel isn't better than capturing 10 bits. It's better because of the flexibility and resistance to visual artifacts you get in editing. But it doesn't matter once editing is done and you're making a print.
you gain larger print sizes from the 45 mp sensor but you loose fidelity and colour gradient transition at a bit depth less than 14 bits.
No, you don’t. Color gradient banding can happen when working with 8-bit/channel images (i.e., starting with a JPEG), if you aren't careful when adjusting gamma or do some other transformation that stresses smooth gradients. If you start with 10 bits, you have a lot more resistance to these artifacts. 12 bits is even better -- I have never once seen these kinds of artifacts when working with my old 20D and 5D (classic) files. But the value of 14-bit capture is more theoretical than practical. I've never seen a legitimate demonstration of benefit there. In any case, you're concerned with output.
Wavelengths of light can be captured in extremely subtle gradations by a camera sensor. But printers don't output wavelengths. They spray ink on paper, using an extremely limited number of fixed-color inks. My printer has 11 different ink colors, 9 if you don't count the black inks and 7 if you don't count the black and grey inks. Other colors in the spectrum have to be simulated by spraying those inks in tiny halftone (or similar) patterns. It is not possible to spray 7 or 9 or 11 different colors in such a way as to accurately reproduce those kinds of imperceptibly subtle color gradations. And it simply isn't necessary, given the limits of human visual perception.
If you want to print huge, or if you want to have the freedom to crop a lot and still print large, capture as many pixels as you can. Either use a high megapixel sensor, stitch together multiple images, or both. Don't worry about whether the sensor is capturing 10, 12, or 14 bits/channel. Even if it only captures 10, import the RAW file and make sure your editing software operates in 16 bit color mode (so that the captured 10 bits/channel aren't truncated to 8 bits).
I don't understand why you keep shutting speculation down and asking the posters for sources.
1. speculation the forming of a theory or conjecture without firm evidence.
2. assume suppose to be the case, without proof.
speculation ~ assume where assume = ass-of-u-&-me
in science and engineering i was taught we don't assume and speculate. scientific knowledge is based on fact not speculation or altered fact and assumptions.
i can see in print the difference between 8 bit and 16 bit input to the printer in the color gradients in large prints. i can not speak to 10 bits because i don't have any 10 bit files to compare.
if i'm going to stretch data which we do in astro image processing. i would rather stretch 14 bits RAW to 16 bits RAW vs stretching 10 or 12 bits to 16 bits. it stands to reason that one will get the least amount of artifacts produced with the 14 bit to 16 bit data stretch.
ideally we would like to see the R5 with a RAW bit depth of 16 bits. but that's a stretch, no pun intended, lol.
When increasing the number of bits, the first bits you gain are most important. Going from 8 to 10 bits represents most if not all the difference you see between 8 bits files and 16 bits files. This is mathematically evident and not an assumption. The remaining step between tones is only 25% of what it was. This might even be below the threshold of the difference you are able to see.
Going from 10 to 12 bits is almost invisible compared to going from 8 to 10 bits. Going from 12 to 14 bits is barely visible at all in tests designed specifically to show the difference. It is not visible at all in large prints that are even extremely tweaked.
InnomnateViem wrote:
ideally we would like to see the R5 with a RAW bit depth of 16 bits. but that's a stretch, no pun intended, lol.
No, that would be a waste of processing power, battery life, conversion speed and storage space with no benefit to the result, as the two last bits would only be filled with noise. 14 bits is plenty and 12 bits often a better balance.
InnomnateViem wrote:
speculation ~ assume where assume = ass-of-u-&-me Speculation most certainly does not have the same meaning as assumption. For example, I might speculate, based partly on your location, and partly on your comment "speculation ~ assume," that English is not your first language*, but I would not assume it
assume: To take as true or as a fact without actual proof.
speculate: To form an opinion from little or no evidence.
*That was for illustration purposes Your English seems pretty good to me.
alundeb, are we talking apples to apples or apples to oranges?
Radiometric Resolution is refering to RAW RGB data. where each individual color channel per pixel R G or B has a defined number of bits per colour ie 14 bits per channel representing 16,385 colors per RGB channel or over 4 trillion shades. When you combine the RGB channels you have: 16,385 x 16,385 x 16,385 = 4,398,851,866,625 colors.
When increasing the number of bits, the first bits you gain are most important. Going from 8 to 10 bits represents most if not all the difference you see between 8 bits files and 16 bits files. This is mathematically evident and not an assumption.
this can not be true when referring to the RAW colour depth and the colour gradient. all the bits in a give channel have an assigned colour value and therefore can not be noise. 0 would be a shade of a particular colour not noise. just as 16,385 would be a shade of a particular colour not noise.
me thinks you are talking buss width or data word length not Radiometric Resolution or RAW colour depth. can you please qualify my assumption.
InnomnateViem wrote:
are we talking apple to apples or apples to oranges?
Radiometric Resolution is refering to RAW RGB data. where each individual color channel per pixel R G or B has a defined number of bits per colour ie 14 bits per channel representing 16,385 colors per RGB channel or over 4 trillion shades. When you combine the RGB channels you have: 16,385 x 16,385 x 16,385 = 4,398,851,866,625 colors.
this can not be true when referring to the RAW colour depth and the colour gradient. all the bits in a give channel have an assigned colour value and therefore can not be noise. 0 would be a shade of a particular colour not noise. just as 16,385 would be a shade of a particular colour not noise.
me thinks you are talking buss width or word length not Radiometric Resolution or RAW colour depth. can you please qualify my assumption. ...Show more →
Dear friend. This place is full of people who know what we are talking about, from inside out.
I am a Master of Science in electrical engineering, and work with signal processing for industrial infrared inspection camera systems. I have designed electronic hardware for analog signal interfaces to image sensors, and write software for signal processing.
I could write a long technical explanation, covering everything in detail, but I don't think it will help anybody.
maybe that's the problem. i'm an electronic engineer and SIL. but that isn't relevant.
you have qualified my assumption. I have designed electronic hardware for analog signal interfaces to image sensors, and write software for
we are talking apples and oranges. analog signal interfaces to image sensors, is not the same topic as the sensor's Radiometric Resolution.
thank you.....
I could write a long technical explanation, covering everything in detail, but I don't think it will help anybody.
try me i'm all ears. and you can write it from the assumed skill set, engineer to engineer. no need for you to break it down to the level of joe six pack.
InnomnateViem wrote:
~ ≠ =. assumption ≠ speculation. however, any way you slice it there is a commonality. no evidence/without proof. end of line.....
The only commonality is that speaker does not know X for a fact. All of the people above who have been speculating do not need the proof/facts you insist on. If they had proof it would not be speculation. That's the end of the story If they assume it will be X bit, that is different than having an opinion that it will be X bit. It's an important distinction in English.
Here's a good answer from an English forum on the difference When you assume something, you decide that something is right, or is the probable answer, usually on a definite basis.
I saw him walking up his driveway in shorts and running shoes, looking tired, so I assumed he was returning from a long run.
"Speculate" refers to wondering about something and what could be the answer.
People don't know how the disease is transmitted and they keep speculating abot possible ways.
If you prefer some quantitative terms. then "assume" represents that the speaker is >90% sure that X is the case while speculate is a much lower % confidence. Either way, when people discuss expectations/possibilities/speculationsd of things like X that are not yet "known as fact" they do not need to prove or provide a source, because they are not stating things as "facts".
Your English seems pretty good to me.