For stills it is definitely 14 bit, as it has been with every Canon camera since 2007.
For video work, it is reported to offer 4:2:2 10 bit with Canon Log (a very flat profile for color grading in PP) - which are pro grade specs to shut up all the Tech Toubers.
i would like to agree with all of you. but i have not been able to find an official confirmation that the RAW image bit depth of the EOS R5 is going to be 14 bits.
AFAIK there's not even official confirmation of the sensor resolution, much less bit depth. There are people saying it's 38.5 MP (8192 × 4320) and others say 45MP.
If it's not 14 bit the thing would be DOA, so I'd think 14 bits is a safe bet.
InnomnateViem wrote:
i would like to agree with all of you. but i have not been able to find an official confirmation that the RAW image bit depth of the EOS R5 is going to be 14 bits.
Because there is no need for such a statement. Like we said, every Canon camera, from Rebels to 1 series released since 2007 offers 14 bit RAW files only. Both Sony and Nikon (I believe) offer choices between 12 and 14 bit files and in case of Sony, 12 bit files was the only way how to reach the marketed specs on some models - like 20 fps on A9.
One thing I always liked about my Canon cameras is that 14 bit stayed at 14 bit. My silly Sony a6300 cowers away into its 12 bit bunker every time it thinks it might have to work a little. Auto-bracketing exposures? 12 bit. Burst with speed set to lowest FPS? 12 bit. Sony just blankets 12 bit over everything that isn't a single shot, even when it is completely unnecessary. Sure, drop to 12 bit if it is needed to get max FPS in burst mode, but keep it at 14 bit (well fake 14 bit, but still better than 12 bit) when possible.
dmcphoto wrote:
AFAIK there's not even official confirmation of the sensor resolution, much less bit depth. There are people saying it's 38.5 MP (8192 × 4320) and others say 45MP.
If it's not 14 bit the thing would be DOA, so I'd think 14 bits is a safe bet.
The images of the sensor/shutter show the dimensions to be 3:2.
I think the ~39MP is assuming that the sensor is 7680x5120.
45MP would be 8192x5461 or close to that.
It depends on the definiton of 8K, since this is a primarily video camera.
I'm starting to think that it might be only 39MP, since Canon has been hiding the poor specs of the recent bodies such as low-res 1DX III until launch.
EB-1 wrote:
The images of the sensor/shutter show the dimensions to be 3:2.
I think the ~39MP is assuming that the sensor is 7680x5120.
45MP would be 8192x5461 or close to that.
It depends on the definiton of 8K, since this is a primarily video camera.
I'm starting to think that it might be only 39MP, since Canon has been hiding the poor specs of the recent bodies such as low-res 1DX III until launch.
EBH
It's 45MP sensor since the 8k frame grab is 35MP according to this R5 preview teaser.
Also, Canon released video spec says it shoots 8k DCI.
Chumma wrote:
Pardon my ignorance. Bits come in multiples of 8. Right? Then why do camera manufacturers use 12 bits and 14 bits instead of 16 bits?
Bits can be packed in any way you want in the data file, and be unpacked and zero padded to a bit depth suitable for the processing hardware later.
I would think when the sensor captures light, it captures light within a certain range, from a minimum to a maximum. If you use 8 bits, you can represent that range in 256 increments, if you use 14 bits in 16,384 increments, and if you use 16 bits, then in 65,536 increments. So is it not better to use 16 bits so that you get more granularity? Or, is my understanding wrong?
Chumma wrote:
I would think when the sensor captures light, it captures light within a certain range, from a minimum to a maximum. If you use 8 bits, you can represent that range in 256 increments, if you use 14 bits in 16,384 increments, and if you use 16 bits, then in 65,536 increments. So is it not better to use 16 bits so that you get more granularity? Or, is my understanding wrong?
Yes, more bits will give you finer gradations, but comes at the cost of increased processing power and storage.
Chumma wrote:
I would think when the sensor captures light, it captures light within a certain range, from a minimum to a maximum. If you use 8 bits, you can represent that range in 256 increments, if you use 14 bits in 16,384 increments, and if you use 16 bits, then in 65,536 increments. So is it not better to use 16 bits so that you get more granularity? Or, is my understanding wrong?
Your understanding is correct, but that spec is per color, for red, green and blue. Therefore "8 bit color" is really 24 bits, and so the full spectrum is covered by 16.7 million colors. "14 bit color" is really 42 bits, and so the full spectrum is covered by 4.4 trillion colors, which is far more colors with much finer gradations than human beings can discern. Going to 16 bit color would provide no discernible benefit.
In truth, most experimental data demonstrates that humans can't detect even the fine gradations in 8-bit color (16.7 million). The issue for imaging applications is that when you begin to manipulate the color data (say to adjust gamma, contrast, saturation, and various other transformations), various artifacts can result that are readily visible. Starting with more gradations (10, 12, or 14 bits per channel) allows for a lot more manipulation without the artifacts that become readily visible when starting with 8 bits per channel.
Some of these artifacts can be avoided simply by converting 8 bit/channel images to 16 bits in Photoshop or whatever image editing software you use, making changes, and then converting back with interpolation. But it's better to start with the higher bit/channel images in the first place.
Chumma wrote:
I would think when the sensor captures light, it captures light within a certain range, from a minimum to a maximum. If you use 8 bits, you can represent that range in 256 increments, if you use 14 bits in 16,384 increments, and if you use 16 bits, then in 65,536 increments. So is it not better to use 16 bits so that you get more granularity? Or, is my understanding wrong?
Adding more bits only helps if the sensor itself is capable of capturing light measurements with such a high precision. If you look at say the EOS R sensor, its dynamic range at pixel level is about 12.55 stops at base ISO. If you use 16-bits instead of 14-bits to record those sensor values, you will really only be capturing noise in the lower bits because the measurement is not precise enough to take advantage of the extra bits.
InnomnateViem wrote:
agreed, but what is the RAW image bit depth of the EOS R5? is it 10 bit or 14 bit?
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.
