FS7: FS7 RAW woes & data distribution between: 12-bit RAW, 16-bit RAW, & 10-bit S-Log

starcentral said:
If in the end you have to overexpose by 1 - 2 stops you're effectively using a 12 stop camera. Which is still plenty of course.
Click to expand...
I have purposefully stayed back from contributing to this thread, as I am contractually bound not to share any knowledge about how Sony RAW works. This is not a comment as to the accuracy of any of the statements made on this thread, as there has been a great deal of back & forth and much of it is conjecture not based in factual data.

But I will say that Dennis's point above is paramount. I personally find that Sony (and several other manufacturers) rate their cameras in such a way that there is too much noise in the bottom of the exposure range. So I like to over expose them by a stop or two. And when I do so, I also have the added functionality of pulling away information from the bottom of the exposure range where there is very little data space to hold it. If I'm going to crush those backs down into oblivion anyway then who cares how little room there is for defining the shades of grey in them? Honestly, the camera has more range of exposure then anyone needs and I'd rather toss away the last couple stops of noise and minimal data allocation anyway. This holds true for just about every single one of the RAW cameras out there. They all seem to perform better if over exposed a bit and brought back down in post. Nothing new here.

I will say that the post where someone contended that the camera only had seven stops of usable exposure range was ridiculous. That's what classic straight Rec709 holds, but the camera has way more than that to offer, and it sends way more than that to the Odyssey7Q+ in a usable fashion.
 
There are drawbacks to this, discussed somewhere in this thread.

But basically rate the camera as 1000ISO or even 800ISO and you'll be just fine.
 
I hate Histograms. Never use them. Just not the way my mind works.

I either use a lightmeter (anyone remember those), or trust a waveform. Just make sure the highlights don't blow out, but I consistently expose from shot to shot. I use the Odyssey with exposure compensation LUTs to see how the image is exposing, but did you know that you can do this right in the viewfinder of the FS7 as well? It's quite easy. Nick Shaw did a video tutorial for our website that shows exactly how.

Shoot two stops over and pull it back in post. Yu'll be amazed at how much better the camera looks.

http://www.convergent-design.com/su...ce-of-the-fs7-evf-and-the-odyssey-screen.html
 
^ what a great way to bypass all the confusion a lot of shooters have with what needs to have MLUT turned on or off on the camera when you get plain signal through the RAW output.

Interesting.
 
starcentral said:
^ what a great way to bypass all the confusion a lot of shooters have with what needs to have MLUT turned on or off on the camera when you get plain signal through the RAW output.

Interesting.
Click to expand...
Yes, it's amazing how turned around people can get with all the obfuscation going on. This is areally simple way to think about it, and as the video plainly shows the reduction of noise in the shadows is clear.
 
+1 this.
Nirv said:
Although I don't own an Odyssey this is exactly how I shoot with my FS7.
Although I don't burn in the lut to internal card... (unless I'm going to hand over footage to someone else... and not grade it myself) (and then I might also shoot custom w/ HG in that scenario)
Click to expand...
 
Thats one thing I like about slog1 on the F3: you don't have to worry about noise in the shadows, and can actally lift the shadows about one or two stops without much more noise, go even further and it starts to look like slog3 on the fs7.
 
stokstad said:
Thats one thing I like about slog1 on the F3: you don't have to worry about noise in the shadows, and can actally lift the shadows about one or two stops without much more noise, go even further and it starts to look like slog3 on the fs7.
Click to expand...

The reason there is less noise in the shadows of the F3 is because the pixels are larger, which is one of the reasons why the F35 and the Alexa look so good.
 
Last edited:
Mitch Gross said:
I have purposefully stayed back from contributing to this thread, as I am contractually bound not to share any knowledge about how Sony RAW works. This is not a comment as to the accuracy of any of the statements made on this thread, as there has been a great deal of back & forth and much of it is conjecture not based in factual data.

But I will say that Dennis's point above is paramount. I personally find that Sony (and several other manufacturers) rate their cameras in such a way that there is too much noise in the bottom of the exposure range. So I like to over expose them by a stop or two. And when I do so, I also have the added functionality of pulling away information from the bottom of the exposure range where there is very little data space to hold it. If I'm going to crush those backs down into oblivion anyway then who cares how little room there is for defining the shades of grey in them? Honestly, the camera has more range of exposure then anyone needs and I'd rather toss away the last couple stops of noise and minimal data allocation anyway. This holds true for just about every single one of the RAW cameras out there. They all seem to perform better if over exposed a bit and brought back down in post. Nothing new here.

I will say that the post where someone contended that the camera only had seven stops of usable exposure range was ridiculous. That's what classic straight Rec709 holds, but the camera has way more than that to offer, and it sends way more than that to the Odyssey7Q+ in a usable fashion.
Click to expand...

Which is exactly why the same phenomenon shows up with the URSA vs the Red epic:
https://www.youtube.com/watch?v=gI1ax3oxaBc

You must expose to the right!

Note that even though the URSA has fixed pattern noise in the underexposed shadows, the Red epic exhibits my pet peeve with almost all of Red's cameras - the face lacks color definition, it becomes a wash of whatever color correction you make. It doesn't matter if it's 16 bit. As sam (Morgan Moore) calls this kind of behavior, it is "wooden". Having graded a ton of work for others I can assure you there is very little you can do to change this without serious work in the secondaries and tracking. I will take the URSA or the FS7 any day of the week to this issue with Red. So if you shoot Dragon today, my only piece of major advice - use the Skin Tone OLPF!
 
I think the Alexa betters the Red simply by having fewer bigger pixels that catch more photons.

(im a huge fan of 'low' resolution - as seen in that all the 'top end' stills cameras D4, 1dx have less 'resolution' than their consumer counterparts)

As for Red being 'wooden', I don't think that video really proves anything (apart from that the creators are clueless*) - personally when I tested the Scarlet, that I did not buy, the file looked great to me but the ISO of 320 was not suitable for my applications

Wood is problem when shooting 8bit HG, slog etc where the base file just does not have enough tonal information to bring appropriate contrast to the face.

Yes in this case the Red does look a bit wooden before post but that is just the opening LUT/Look in the NLE.

Any 10-12-14 bit file should have enough graduations in skin to be able to eliminate 'wood' by using an appropriately steep curve in post.

Of course if a deep bit depth camera does not come out of the box with nice 'looks' issued for your NLE one must build ones own nice looks, or waste time on the webble hunting for 'looks' - arri do themselves a huge favour by issuing nice looks from day one of the camera launch; both a time saver and a PR boon as their footage is less often mistreated)

S

*without seeing histos of the base exposure(s) or them revealing any of their post methodology, which clearly did not include paying any concern to the Reds red shadows, the video is useless.
 
Last edited:
Hey Sam,
Nice to hear from you.
Not trying to discount what you have said, but I can assure you that after having graded a ton of RED footage over the years, what you see in that video is pretty much endemic of the problem which I see all the time with skin tones and RED color science. No LUT or look will yield the kind of dimensionality which is necessary for that face. The color information just will not reveal the blush in the cheeks for instance without the rest of the face moving along with any color correction you make. I constantly have to go into secondaries to do "facial reconstruction" as a result. Only the skin tone OLPF on Dragon has made a difference in this regard. So they are aware of the problem. The issue is ironically in the color science around the red layer and their choice of using a smaller pixel array. So it is partially software, and partially hardware. I truly feel that this is why the Alexa has done so well. They got the color science down when it comes to faces as well as using larger pixels for increased color sensitivity, and let's "face" it, 90% of films out there deal with human subjects. For me RED remains a great camera for other subjects, but not so much in this area of content. I even see the same problem in Gone Girl for instance. The faces just lack the kind of nuance and dimensionality that they would have had if they had shot Alexa.
 
Last edited:
I cant really comment further on nuance**, I tested Scarlet for a day at Pinewood some years ago - thought 'yep this is image is fine*.. but the camera too slow (ISO) and too heavy (physically)', moved on. Ive shot other Red jobs but not graded them.

My observation that probably still holds water is that the Scarlet needed to be on 320.. higher values were just 'faux' ISO and many DPs were IMO under exposing.

*'fine' is a complement - I tested it against my personal 'reference' the nikon D3 in raw stills mode.
** im not a nuance guy, for example in (raw) stills I cant tell the difference between canon and nikon - they are both 'fine'
S
 
Last edited:
Maybe you need to work with better Red shooters then? I have never once had this when crafting a scene on MX.

The biggest problem I find with Red, and it is true for the FS7 as well, is that they are used on a ton of low to zero budget works, and often with limited and nasty green spiking cheap LEDs and often by people who don't know their camera or their craft yet. And because t goes online it becomes used as evidence of the shortcomings of a camera.


If you can't get lovely tones out of a Red then you are doing cameranating wrong or there is another external limiting factor.

I will see what I have by way of example R3Ds to post later, from a lowly Scarlet MX.


The skintone olpf thing was definitely a bit of a balls up by Red when changing stuff for the Dragon but just like ketchup being too thick to get out of the bottle, they have managed to spin it into a chief selling point with interchangeable ones including dedicated underwater ones now.
 
Ben Scott said:
Maybe you need to work with better Red shooters then? I have never once had this when crafting a scene on MX.

The biggest problem I find with Red, and it is true for the FS7 as well, is that they are used on a ton of low to zero budget works, and often with limited and nasty green spiking cheap LEDs and often by people who don't know their camera or their craft yet. And because t goes online it becomes used as evidence of the shortcomings of a camera.


If you can't get lovely tones out of a Red then you are doing cameranating wrong or there is another external limiting factor.

I will see what I have by way of example R3Ds to post later, from a lowly Scarlet MX.


The skintone olpf thing was definitely a bit of a balls up by Red when changing stuff for the Dragon but just like ketchup being too thick to get out of the bottle, they have managed to spin it into a chief selling point with interchangeable ones including dedicated underwater ones now.
Click to expand...

Well, I would hardly consider Gone Girl a low budget film and I most certainly did notice the issues I am speaking of in that film. Additionally, I grade a fair amount of advertising stuff and the DP's do know their stuff, but this issue still does come up frequently. They are as aware of it as I am. The faces often lack nuance at about one quarter the frame size. This is especially true in Tungsten situations.
 
Do you have anything you can share image-wise other than that awful Ursa/Epic video to demonstrate what you mean?

I guess we see different things in Gone Girl, because I see a great looking film with a palette chosen to suit a psychological drama.

MX cameras weren't always the best friend of tungsten fixtures if you're underexposing but I've used them plenty. I believe Dragon sensors are improved in this area.
 
Ben Scott said:
Do you have anything you can share image-wise other than that awful Ursa/Epic video to demonstrate what you mean?

I guess we see different things in Gone Girl, because I see a great looking film with a palette chosen to suit a psychological drama.

MX cameras weren't always the best friend of tungsten fixtures if you're underexposing but I've used them plenty. I believe Dragon sensors are improved in this area.
Click to expand...

HeyBen,
I can't really post my client's footage. That would not be professionally ethical of me, so I would have to look through my own footage for an example...
In any case, I do think Dragon is better, even if I still see issues. Tungsten remains a bug-a-boo for Red. Their cameras really function best under HMI conditions. Clearly this is because of the issues in the Red layer which I spoke about.
 
Fair enough, but it is pretty hard to accept the assertions you're making without any evidence to support it, except a really terrible youtube video made by someone who doesn't know what they are doing. What colour spaces and gamuts have you noticed it most in?


Every camera suffers when the sensor is starved of light. Don't under expose under tungsten and you can have a great image. And also, to be honest, almost ALL cameras function best under HMIs :)

Just to add, I'm not having a dig a you personally. I just see a hell of a lot of misinformation out there and the Red workflow from acquisition to delivery is nowhere near as 'plug and play' as most other camera systems if you want to get the best out of it - so I see a lot of people mishandling it at various stages.


DP Patrick O' Sullivan summed this up for me best...

Steps to an Alexa job:
1. Shoot nice images
2. Almost impossible to screw up post pipeline
3. See nice images on TV
4. Collect paycheck

Steps to certain RED jobs:
1. Shoot nice images
2. Someone grades Red Gamma 2/Red Color 1 720p files w/a LUT they bought
3. Disown project



A lot of the blame can be laid at Red's door for not providing a definitive Redlogfilm to REC709 lut like Arri did for their log files.
 
yoclay said:
The reason there is less noise in the shadows of the F3 is because the pixels are larger, which is one of the reasons why the F35 and the Alexa look so good.
Click to expand...

Thats probably one reason, but I would rather say its because of the less flat curve in slog1. If you lift it like slog3 it looks pretty close to the fs7 in my experience.
 
You must log in or register to reply here.
Back
Top