FS7: FS7 II - Does raw really make any difference? (qualitywise)
- Thread starter legrevedotcom
- Start date
Bassman2003
Veteran
Thanks for your reply. Makes sense and too bad the Fs7 does not have a full RAW implementation. Makes me wonder about Canon's Raw Lite. Could be a wonderful opportunity for some segmentation
Anyway, carry on...dmitrizigany
Veteran
There is no raw standard. Every manufacturer handles it differently. Sometimes even different for each camera, or family of cameras in Sony's case.
Also it's "raw" not "RAW", it's not an acronym.
NorBro
Major Contributor
It's officially been "RAW" for a while now. Looks nicer.
Check this thread (post #6) to see almost every company write it as RAW.
http://bmcuser.com/showthread.php?20408-raw-is-not-RAW
Others also brought up how it's grown to be capitalized to differentiate from raw footage, which many understand as unedited.
Check this thread (post #6) to see almost every company write it as RAW.
http://bmcuser.com/showthread.php?20408-raw-is-not-RAW
Others also brought up how it's grown to be capitalized to differentiate from raw footage, which many understand as unedited.
dmitrizigany
Veteran
Ok!
Mark Watson
Well-known member
I disagree about it looking nicer or being "official". Looks illiterate.
How would this work in a verbal discussion, do you say "raw" loudly to indicate straight off the sensor as opposed to unedited "raw"?
How would this work in a verbal discussion, do you say "raw" loudly to indicate straight off the sensor as opposed to unedited "raw"?
RAW has been an acronym for 'Read And Write' since machine language days and it describes a process of collecting and storing data without any manipulation on it.
RAW flavor is an oxymoron since there should be no additional data to the (usually up to 16 bit) value of the AD converter at the specified address of sensor matrix.
After the data has been put through higher level of processing and added metadata (and even compressed) then you get different raw formats (some with baked information).
From AD conversion perspective that isn't any more RAW data, but it's raw from end-user perspective and you have to process it by manufacturers recipe to be able to use it.
So, to be RAW Nazi, RAW can only be a result of AD conversion and allows the data to be put in any context to give you information in the range you want.
Any additional operation on it, where you loose flexibility of manipulation and accuracy, gets you raw.
RAW flavor is an oxymoron since there should be no additional data to the (usually up to 16 bit) value of the AD converter at the specified address of sensor matrix.
After the data has been put through higher level of processing and added metadata (and even compressed) then you get different raw formats (some with baked information).
From AD conversion perspective that isn't any more RAW data, but it's raw from end-user perspective and you have to process it by manufacturers recipe to be able to use it.
So, to be RAW Nazi, RAW can only be a result of AD conversion and allows the data to be put in any context to give you information in the range you want.
Any additional operation on it, where you loose flexibility of manipulation and accuracy, gets you raw.
moldcad
Veteran
Haha - didn't I say I was new to RAW...argh...raw? I bet I saw it capitalized in many places, including camera manufacturers' specs sheets, but your're right - my using it along with proper acronyms (like RAW SDI) wasn't probably fortunate... Anyway, as one of you guys said - just move on
Piotr
PS. Ned, I'd also like to wish you recovering soon! Was the BMW driver reckless, as most of them? Asking, because I'm a BMW guy myself and sometimes cannot believe how much of troublemakers and hooligans my 'BMW mates' are all over the world...
Piotr
PS. Ned, I'd also like to wish you recovering soon! Was the BMW driver reckless, as most of them? Asking, because I'm a BMW guy myself and sometimes cannot believe how much of troublemakers and hooligans my 'BMW mates' are all over the world...
legrevedotcom
Veteran
Well c200 is only 12 bit as Well, so I guess we have to wait and see some more Real World examples.Thanks for your reply. Makes sense and too bad the Fs7 does not have a full RAW implementation. Makes me wonder about Canon's Raw Lite. Could be a wonderful opportunity for some segmentation
Haha - didn't I say I was new to RAW...argh...raw? I bet I saw it capitalized in many places, including camera manufacturers' specs sheets, but your're right - my using it along with proper acronyms (like RAW SDI) wasn't probably fortunate... Anyway, as one of you guys said - just move on
Piotr
PS. Ned, I'd also like to wish you recovering soon! Was the BMW driver reckless, as most of them? Asking, because I'm a BMW guy myself and sometimes cannot believe how much of troublemakers and hooligans my 'BMW mates' are all over the world...
Actually, Piotr, the BMW driver was 84 years old and fortunately driving very slowly. That explains why I just have a broken leg and some shoulder pain.
Harry Lime
Well-known member
ArriRAW is 12bit log, which is not the same as FS7 12bit linear.
F5/F55 is 16bit linear, which is the equivalent of 12bit log.
There is an very long thread somewhere on this site explaining the difference.
SamEdwards
Well-known member
Here's a CDNG file that was recorded with an FS5 to an Odyssey 7Q+.
https://www.dropbox.com/s/jhu8ngin9ib6l1j/CLIP0000139_0000368.dng?dl=0
Note this was exposed at the native 3200iso. So it will have more noise than if it were overexposed 1.5-2.0 stops, as is my current workflow.
It has been widely reported that FS-raw is all pretty similar, but that the FS7 had a 'bug', so YMMV. Go ahead and open it in Photoshop, if you have it. It doesn't appear that the color balance is baked-in for CDNG. Since ProRes and DNX are debayered in the Odyssey, the color balance gets baked in at that stage.
In resolve there are also CDNG controls that can be tweaked, and white balance is adjustable there.
I think for video/web compressed delivery you can get a pretty darn decent image with the compressed recording from an FS7, or going out HDMI/SDI to a recorder. The place where CDNG really shines is in the movie theater.
There is a lot of difference in interpretation/debayering between programs. Photoshop will give one result, Nuke will give another, Resolve is also different.
What I can promise is that for pixel-peepers, and people who want a lot of control over white balance, the raw CDNG is the way to go. I have seen videos where most of the control you have over white balance can be duplicated in Resolve. But if you have a highly complex scene the details will not be as sharp.
Sorry if this seems OT to some. I think an actual FS-raw image to play with may be helpful to the OP and nobody has offered one yet that came from an FS7 II.
HTH
https://www.dropbox.com/s/jhu8ngin9ib6l1j/CLIP0000139_0000368.dng?dl=0
Note this was exposed at the native 3200iso. So it will have more noise than if it were overexposed 1.5-2.0 stops, as is my current workflow.
It has been widely reported that FS-raw is all pretty similar, but that the FS7 had a 'bug', so YMMV. Go ahead and open it in Photoshop, if you have it. It doesn't appear that the color balance is baked-in for CDNG. Since ProRes and DNX are debayered in the Odyssey, the color balance gets baked in at that stage.
In resolve there are also CDNG controls that can be tweaked, and white balance is adjustable there.
I think for video/web compressed delivery you can get a pretty darn decent image with the compressed recording from an FS7, or going out HDMI/SDI to a recorder. The place where CDNG really shines is in the movie theater.
There is a lot of difference in interpretation/debayering between programs. Photoshop will give one result, Nuke will give another, Resolve is also different.
What I can promise is that for pixel-peepers, and people who want a lot of control over white balance, the raw CDNG is the way to go. I have seen videos where most of the control you have over white balance can be duplicated in Resolve. But if you have a highly complex scene the details will not be as sharp.
Sorry if this seems OT to some. I think an actual FS-raw image to play with may be helpful to the OP and nobody has offered one yet that came from an FS7 II.
HTH
Not to search through long thread, could you explain me how RAW data can have log curve?
RAW data is the measured value of voltage produced by photo-sites of the camera sensor, without any processing or manipulation.
To maximize the tonal range of a sensor, log curve can be applied to deliberately remove parts of the image data.
Alexa sensor is 14-bit linear, internal processing is 16-bit and ARRIRAW data is efficiently stored as 12-bit log to retain 14(+) stops of latitude.
To create "16"-bit RGB image, applying reverse curve is a part of performed colour science, followed by final step of debayering.
And all of that with just 1/3 of the required bandwidth. Primary goal of RAW was not safety net for bad DP.
To go back to the FS7 12-bit linear < 12-bit log claim, it's not that simple.
Sony use floating point for storing data in 12-bit linear RAW, otherwise it wouldn't be possible to fit 14 stops.
Limitation is that amount of data available for the shadows is very small.
Data rounding used to retain the whole latitude of the sensor in 12-bit linear resulted in the coarser luma steps in the first two to three stops.
Solution to this is to overexpose up to 2 stops.
Log version has problems above middle grey and it gets worse and worse since each brighter stop has the same limited amount of data.
And without enough light footage will be noisy.
To conclude, 12-bit for "Real World" examples is just fine, for vfx get 16 bit.
RAW data is the measured value of voltage produced by photo-sites of the camera sensor, without any processing or manipulation.
To maximize the tonal range of a sensor, log curve can be applied to deliberately remove parts of the image data.
Alexa sensor is 14-bit linear, internal processing is 16-bit and ARRIRAW data is efficiently stored as 12-bit log to retain 14(+) stops of latitude.
To create "16"-bit RGB image, applying reverse curve is a part of performed colour science, followed by final step of debayering.
And all of that with just 1/3 of the required bandwidth. Primary goal of RAW was not safety net for bad DP.
To go back to the FS7 12-bit linear < 12-bit log claim, it's not that simple.
Sony use floating point for storing data in 12-bit linear RAW, otherwise it wouldn't be possible to fit 14 stops.
Limitation is that amount of data available for the shadows is very small.
Data rounding used to retain the whole latitude of the sensor in 12-bit linear resulted in the coarser luma steps in the first two to three stops.
Solution to this is to overexpose up to 2 stops.
Log version has problems above middle grey and it gets worse and worse since each brighter stop has the same limited amount of data.
And without enough light footage will be noisy.
To conclude, 12-bit for "Real World" examples is just fine, for vfx get 16 bit.
SamEdwards
Well-known member
Here's an unused frame that I shot with the FS5/Odyssey 7Q+ as a VFX background/reflection map for a major studio feature film. Principal photography on the film was F55. The VFX Supervisor thought the plates looked 'pretty darn good' when debayered in Nuke, but '****ing incredible' when processed in Resolve per Convergent Designs recommendations. He was screening uncompressed DPX on a 20 foot wide P3 4k projector. Overexposed about 1.25 stops.
https://www.dropbox.com/s/cvdb9jeo5rslymv/SAMEDWAR060_0000000.dng?dl=0
Crack it open in Photoshop or Resolve. Please don't use for anything but examining technical quality.
https://www.dropbox.com/s/cvdb9jeo5rslymv/SAMEDWAR060_0000000.dng?dl=0
Crack it open in Photoshop or Resolve. Please don't use for anything but examining technical quality.
I have the FS7 as well as an Epic-W and I think compressed codecs and RAW are kind of for different purposes. I think RAW is an amazing codec when it’s well implemented like on Red, but on the FS7 the implementation is quite poor, so it’s not really worth using.
For work, my FS7 tends to go out more since I do a lot of Doc type jobs lately. The compressed XAVC is fantastic and I love the FS7. The Red tends to go on my higher paying jobs and for that, I only use RAW. Pro-Res is an option, but I’ve never used it on a paid job.
They’re two different animals and for different things. Compressed is great for quick turnaround, but the power of R3D / RAW is amazing. Perhaps the image out of camera might not be that different, but I LOVE the flexibility in post. It’s quite incredible. The ability to tweak White Balance alone is amazing for coming up with cool looks for things like music videos. Adjusting the ISO and amazing highlight recovery makes RAW a really good thing. Red R3D are an incredible feat of engineering and the RAW that goes onto the FS7 simply does not compare.
My only experience with shooting Sony RAW is when I had my FS700 and recorded Cinema DNGs to the Odyssey 7Q. Cinema DNGs are very clunky to work with and the fact that it’s uncompressed makes it almost un-usable except for short format type jobs. The files sizes are MASSIVE and require constant card swaps.
I’d say for the FS7 just stick to XAVC as it’s great. If you want to shoot RAW, use a Red and get a super “thick” negative.
For work, my FS7 tends to go out more since I do a lot of Doc type jobs lately. The compressed XAVC is fantastic and I love the FS7. The Red tends to go on my higher paying jobs and for that, I only use RAW. Pro-Res is an option, but I’ve never used it on a paid job.
They’re two different animals and for different things. Compressed is great for quick turnaround, but the power of R3D / RAW is amazing. Perhaps the image out of camera might not be that different, but I LOVE the flexibility in post. It’s quite incredible. The ability to tweak White Balance alone is amazing for coming up with cool looks for things like music videos. Adjusting the ISO and amazing highlight recovery makes RAW a really good thing. Red R3D are an incredible feat of engineering and the RAW that goes onto the FS7 simply does not compare.
My only experience with shooting Sony RAW is when I had my FS700 and recorded Cinema DNGs to the Odyssey 7Q. Cinema DNGs are very clunky to work with and the fact that it’s uncompressed makes it almost un-usable except for short format type jobs. The files sizes are MASSIVE and require constant card swaps.
I’d say for the FS7 just stick to XAVC as it’s great. If you want to shoot RAW, use a Red and get a super “thick” negative.
I can't speak to raw from a Sony. But I can share what it is like to shoot a camera that was designed from the ground up as a maxium quality DCI CDNG raw only camera. The Digital Bolex records a 12 bit linear raw file truncated from a 16 bit AD conversion process. The 4 least significant bits are dropped. When imported into any raw processor, that 12 bit raw file is expanded back into a 16-bit 4:4:4 RGB linear value scale with a black reference level of 8. There are discrete grayscale pixel values in the image that would not be mathematically possible with 12 bit linear capture or a 12 bit linear readout in the processing. This is the only digital camera I ever shot with, still or video, that produced an eerie sensation of actually looking at the physical surface of the sensor when pixel peeping.
The color gamut the D16's Kodak scientific grade CCD can produce is way beyond any current standard video codec. It exceeds DCI-P3 JPEG2000 color gamut by a large margin. The only time I ever got a hint of what I was missing in my monitor setup was running a Cineform 16 bit 4:4:4 RGB clip through a 12 bit DCI-P3 Sony laser cinema projector. Wish I could afford a personal grading theater setup with one of those!
The odd color casts common to early D16 footage before the color matrix was dialed into a more video friendly set of vectors and we learned how to grade the footage were the result of severe chroma clipping in REC709.
Does that kind of color gamut matter for video production in a world where the most common delivery format is 4:2:0 8-bit? Frankly not much. Even exporting graded footage to XAVC-I 10 bit 4:2:2 at 300+ Mbps shows significant degradation, though I think it is an excellent compressed codec that Sony has done a stellar job of engineering.
We are rapidly heading into a world where the current and next generation consumer TV's like LG and Sony and Samsung's top of the line HDR OLED models can actually reproduce DR and color gamuts that make 10 bit 4:2:2 graded log video look a bit thin. Log is just another form of value compression. Useful for sure, to get the incredible DR modern sensors can capture into a much narrower DR video codec with useful shadow and highlight detail that would otherwise be lost.
So the next time you see a scope shot of a DR chart for a "13+" stop camera log file where the bottom four stops are compressed into one stop of a 10 stop codec DR, imagine that curve straightened out extending four full stops below the codec's reference black level with visible texture, color and detail.
At some point in the not too distant future, true high quality linear raw DR and color gamut is going to become very relevant.
The color gamut the D16's Kodak scientific grade CCD can produce is way beyond any current standard video codec. It exceeds DCI-P3 JPEG2000 color gamut by a large margin. The only time I ever got a hint of what I was missing in my monitor setup was running a Cineform 16 bit 4:4:4 RGB clip through a 12 bit DCI-P3 Sony laser cinema projector. Wish I could afford a personal grading theater setup with one of those!
The odd color casts common to early D16 footage before the color matrix was dialed into a more video friendly set of vectors and we learned how to grade the footage were the result of severe chroma clipping in REC709.
Does that kind of color gamut matter for video production in a world where the most common delivery format is 4:2:0 8-bit? Frankly not much. Even exporting graded footage to XAVC-I 10 bit 4:2:2 at 300+ Mbps shows significant degradation, though I think it is an excellent compressed codec that Sony has done a stellar job of engineering.
We are rapidly heading into a world where the current and next generation consumer TV's like LG and Sony and Samsung's top of the line HDR OLED models can actually reproduce DR and color gamuts that make 10 bit 4:2:2 graded log video look a bit thin. Log is just another form of value compression. Useful for sure, to get the incredible DR modern sensors can capture into a much narrower DR video codec with useful shadow and highlight detail that would otherwise be lost.
So the next time you see a scope shot of a DR chart for a "13+" stop camera log file where the bottom four stops are compressed into one stop of a 10 stop codec DR, imagine that curve straightened out extending four full stops below the codec's reference black level with visible texture, color and detail.
At some point in the not too distant future, true high quality linear raw DR and color gamut is going to become very relevant.