Canon R5 new 1.3 firmware ads CLOG3, Raw light and IPB light

[UncleBobsPhotography]

I really don't understand how RAW CLOG works, especially when the press release includes this line:

Quote

Canon Log and Log 3 always transform EOS R5 video files to 10-bit, 4:2:2 recording

How is it possible to record RAW in 4:2:2? As far as I knew 4:2:2 was only used for debayered RGB. I guess it's possible to use it for bayered pixels as well?

At first I thought it would revert to h265 when choosing RAW and clog at the same time, but then I saw the video linked below where they are talking extensively about RAW CLOG.

So what is the difference between RAW and RAW CLOG?

I assume RAW is 12 bit "4:4:4". Is RAW CLOG 10 bit 4:2:2? In that case I would think the bit rate would be reduced since the bit depth and chrome subsampling is smaller, but the camera reports the same runtime with CLOG enabled and disabled. I am confused. 

 

 

 

[gt3rs]

3 hours ago, UncleBobsPhotography said:

I really don't understand how RAW CLOG works, especially when the press release includes this line:

How is it possible to record RAW in 4:2:2? As far as I knew 4:2:2 was only used for debayered RGB. I guess it's possible to use it for bayered pixels as well?

At first I thought it would revert to h265 when choosing RAW and clog at the same time, but then I saw the video linked below where they are talking extensively about RAW CLOG.

So what is the difference between RAW and RAW CLOG?

I assume RAW is 12 bit "4:4:4". Is RAW CLOG 10 bit 4:2:2? In that case I would think the bit rate would be reduced since the bit depth and chrome subsampling is smaller, but the camera reports the same runtime with CLOG enabled and disabled. I am confused. 

 

RAW is RAW 12bit 4:4:4, if you have CLOG on or off it changes the base iso (clog 400 or clog3 800) to give more headroom to the highlight, If I get it right it is just an exposure trick.

From the manual:

 

[independent]

Right, "gradation in highlights" = better highlight rolloff. 

That's the purpose of the clog3 gamma curve, to allocate data across another stop (or so). Clog has a steeper curve in the highlights, so you don't get as smooth of a transition. But clog3 has a penalty in the shadows though.

[gt3rs]

4 minutes ago, independent said:

Right, "gradation in highlights" = better highlight rolloff. 

That's the purpose of the clog3 gamma curve, to allocate data across another stop (or so). Clog has a steeper curve in the highlights, so you don't get as smooth of a transition. But clog3 has a penalty in the shadows though.

There is no curve in RAW as it is impossible as is not debeyered

[Video Hummus]

4 minutes ago, independent said:

Right, "gradation in highlights" = better highlight rolloff. 

That's the purpose of the clog3 gamma curve, to allocate data across another stop (or so). Clog has a steeper curve in the highlights, so you don't get as smooth of a transition. But clog3 has a penalty in the shadows though.

 

8 minutes ago, gt3rs said:

There is no curve in RAW as it is impossible as is not debeyered

That's why they find the wording in the manual from Canon confusing.

[Video Hummus]

Lots of questions answered in this video about R5 RAW and CLOG3.

 

[gt3rs]

15 hours ago, Video Hummus said:

Lots of questions answered in this video about R5 RAW and CLOG3.

 

Partially as they also don't have explanation how it works. What we know that RAW, RAW with Clog (more headroom for HL but more shadow noise) and RAW with Clog3 (even more headroom for HL but even more shadow noise) are different but stilly RAW as you can apply WB,  any log curve, etc.. They process RAW with Clog3 on by applying a Clog2 curve in Resolve.

Now I'm even more confused.... would be nice if Canon would publish a white paper that explain what they are doing....

[Video Hummus]

10 hours ago, gt3rs said:

Partially as they also don't have explanation how it works. What we know that RAW, RAW with Clog (more headroom for HL but more shadow noise) and RAW with Clog3 (even more headroom for HL but even more shadow noise) are different but stilly RAW as you can apply WB,  any log curve, etc.. They process RAW with Clog3 on by applying a Clog2 curve in Resolve.

Now I'm even more confused.... would be nice if Canon would publish a white paper that explain what they are doing....

They also mention that this is strange and question if it is true RAW data as something is being processed in a certain chosen gamma. 

[UncleBobsPhotography]

On 4/9/2021 at 11:04 PM, gt3rs said:

There is no curve in RAW as it is impossible as is not debeyered

Technically, I guess it is possible to apply a curve even though they are not debayered. A log-curve is just changing the representation of the values (usually compressing the range from what has been read out from ADC to reduce the bit-depth requirement), which can be performed on non-RGB values. However, as long as RAW uses 12-bits, I don't think there is any reason to apply a log-curve.

On 4/10/2021 at 1:13 AM, Video Hummus said:

Lots of questions answered in this video about R5 RAW and CLOG3.

That's the video that confused me in the first place. The only explanation I can think of for how RAW CLOG works would be for the CLOG to simply add metadata to the RAW file (which is still recorded in linear 12 bit), and that the different base ISO's is just to make the user protect the highlights by adjusting exposure (which is basically what gt3rs said in a post a bit earlier). I guess that doesn't match with the findings in the video, but as far as I could see they adjusted the exposure settings between the clog1 and clog3 recordings, which would make the results different. To check if they are truly just metadata you would have to record with the same settings but different base ISO, and seemingly overexposing clog3.

 

[Video Hummus]

9 hours ago, UncleBobsPhotography said:

That's the video that confused me in the first place.

Yeah, I posted it and then realized you posted just a few questions up. I had clicked on it in a new tab, forgot about it, then watched it later. Sorry!

 

[gt3rs]

9 hours ago, UncleBobsPhotography said:

Technically, I guess it is possible to apply a curve even though they are not debayered. A log-curve is just changing the representation of the values (usually compressing the range from what has been read out from ADC to reduce the bit-depth requirement), which can be performed on non-RGB values. However, as long as RAW uses 12-bits, I don't think there is any reason to apply a log-curve.

That's the video that confused me in the first place. The only explanation I can think of for how RAW CLOG works would be for the CLOG to simply add metadata to the RAW file (which is still recorded in linear 12 bit), and that the different base ISO's is just to make the user protect the highlights by adjusting exposure (which is basically what gt3rs said in a post a bit earlier). I guess that doesn't match with the findings in the video, but as far as I could see they adjusted the exposure settings between the clog1 and clog3 recordings, which would make the results different. To check if they are truly just metadata you would have to record with the same settings but different base ISO, and seemingly overexposing clog3.

 

Ok now I get it, it is just metadata and exposure trickery. 

I did a test aperture and shutter speed fixed, exposed to almost clip highlight in RAW Clog3 iso800, then re-shoot the same scene in RAW with nolog iso100 and changed the iso to 800 in Resolve Camera Raw and you get basically the same image and parade. So what the camera is doing is capturing at iso 100 and fake the metadata to iso 800 so when Resolve process the raw the image is correctly exposed by bosting the exposure by 3 stop. Zebra, Meter and Histogram are tricked to think that the camera is at iso 800 so you end up  underexposing the image by 3 stop.

Also shooting RAW Clog3 CinemaGamut or RAW Clog3 Rec709 in my quick test the image looks the same. In fact nowhere in the doc is mentioned that the Log color space as an effect on RAW.

 

[UncleBobsPhotography]

11 hours ago, gt3rs said:

Ok now I get it, it is just metadata and exposure trickery. 

I did a test aperture and shutter speed fixed, exposed to almost clip highlight in RAW Clog3 iso800, then re-shoot the same scene in RAW with nolog iso100 and changed the iso to 800 in Resolve Camera Raw and you get basically the same image and parade. So what the camera is doing is capturing at iso 100 and fake the metadata to iso 800 so when Resolve process the raw the image is correctly exposed by bosting the exposure by 3 stop. Zebra, Meter and Histogram are tricked to think that the camera is at iso 800 so you end up  underexposing the image by 3 stop.

Also shooting RAW Clog3 CinemaGamut or RAW Clog3 Rec709 in my quick test the image looks the same. In fact nowhere in the doc is mentioned that the Log color space as an effect on RAW.

 

In that case, the only real reason to use RAW CLOG would be to match it with other CLOG clips. I would rather protect my own highlights in ISO 100, and as long as nothing is clipping, ETTR with normal RAW should provide optimal results.

I still haven't looked much into the different colour spaces. With RAW, I would guess it has no impact. Even without RAW (at least for rec709 delivery), I am not sure it would make much of a difference since the bit depth stays the same, but I would be like to see if someone could show some objective advantages of choosing one colour space over the other. I have found some YouTube-videos recommending shooting in Cinema Gamut for maximum latitude, but their arguments were not very convincing.

[gt3rs]

FW 1.3.1 is out at fixes the view assist in clog3.