Usable Stops of Dynamic Range

[Zac Halberd]

I don't think it's bad form to clip some areas of the image. I shoot with my Scarlet-X the same way I shoot with film. I over expose by up to 2 stops and bring my ISO back down in REDCINE. Im cramming as much information into my image as I can within a certain contrast ratio and I get zero noise, very seldom clip beyond repair and my images glow in post. Im capable of anything now. I love this camera. It's effectively a digital frame of celluloid. Im blown away every day.

 

[Uli Plank]

Of course you can clip local highlights, like shiny spots on reflective surfaces or light sources, or you would underexpose massively. But with large areas it looks too video-ish.

One recent experience: we were set to shoot in a small hotel room, it was a sunny day outside, the actors were there, the Scarlet was there. BUT the truck with the lighting equipment got stuck in traffic. Only one small Kino-Flo was in the other car and available. For one scene we had to shoot an actor against the window in MCU. We didn't want to blow the window, which only had a light curtain. I said to the crew "OK, we have a Scarlet, not a video camera. Let's try." We used the single Kino-Flo as fill and started shooting and stayed just below clipping for the window, but couldn't use HDRx because we needed 25p in good rez.

The result was absolutely useable with some grading, skin tones were fine and the window with the curtain was there, not just an empty white area. I love this cam!

 

[Lee Saxon]

The "purple" is conservative. It says - there may be noise down here, be careful before you place majorly important image detail in this region. It doesn't mean "image in purple region is un-usable so avoid."

And that's just one example of these 'objective' IQ test being fairly useless in absolute terms.

It's only in relative terms (comparing multiple products tested under similar conditions) that such charts are interesting.

 

[Will Keir]

fix it in post?

The result was absolutely useable with some grading, skin tones were fine and the window with the curtain was there. I love cookies

 

[Timur Civan]

Where do you place mid grey indeed.... There's lots of room for user-choice here, and nothing fixed in stone about where. Always happy to work through stuff with you Timur.

Graeme

It seems to change per project. Just today i was shooting an interview for a Documentary, and i set the camera to ISO320 RLF, and placed the skin tones at 40ish IRE. Expand that to 800ISO, and boom. Perfect. That said, it did have a tiny bit of grain. Its about what you prioritize and how you like to work. I'm trying something different every time.

Like the DVX100, HVX200, 5DmkII, EX1, F3 and now Scarlet, each camera i have owned has had a way it likes to be treated with regard to light. When you find that sweet spot, just keep it there.

 

[Kevin Marshall]

Hm... That was a strange question, really...

You need to start off by having the full latitude avaiable. best purely technical startingpoint for that would be redlogfilm or linear light at 320 (or less if your wb is off senter).

But you know what? For most images that is really a bad advice these days, especially if you are not used to or don't have the tools for linear or log grading.

Best advice, methinks, is to start off with RG2 or 3 and RC3, and for the images where you think the highlighta are clipped or the lowlights are crushed, you could try to start from the log or lin space at 320. Your lin/log grade should be at least as good as your RedGamma3 grade for it to make sense...

Just out of curiosity, do you find anything fundamentally wrong with processing RLF at ISO 800 instead of 320 (assuming footage shot at ISO 800 RC3/RG3)? Or is it a case-by-case consideration?

 

[Gunleik Groven]

Not at all!

If it is shot within the goalposts of RG3 800, there is nothing wrong with keeping it that way

There is basically never anything wrong with an image you are happy with.

If you need to recover highlights in post, you may have to lower the ISO or change gamma.

 

[Kevin Marshall]

Cool, thanks! About to meet on Sunday with my editor and director to prep for color correction. Still ironing out a couple kinks in our workflow...

 

[Gunleik Groven]

For longer form stuff, we usually prep the images before grade, and in that case, redlogfilm can be a very nice place to start.

 

[Kevin Marshall]

This project is a short - maybe 15 minutes long. Colorist is working from the R3Ds, though we're baking in some speed changes and stuff ahead of time. Either way, I've noticed (in Color, at least) that even off the RAW, using RG3 and trying to bring up darker shadows results in more noise than using RLF + a LUT and bringing them up - so that's our game plan at the moment.

 

[Gunleik Groven]

This project is a short - maybe 15 minutes long. Colorist is working from the R3Ds, though we're baking in some speed changes and stuff ahead of time. Either way, I've noticed (in Color, at least) that even off the RAW, using RG3 and trying to bring up darker shadows results in more noise than using RLF + a LUT and bringing them up - so that's our game plan at the moment.

Smart plan!

If you are a bit more concious, thinking of the signalchain is a good thing. What distorts the signal where, from aqusition to delivery. If you want to have "all" available RLF or Linear are good startingpoints. I just don't argue too much for that, as those are very confusing startingpoints for a lot of users, even though they technically are the best.

If you are in resolve (or most/all graders that have picked up on this these days), the signal out of the Rocket is an RGB signal and you don't then grade "on the RAW" - which is only good. You grade on the RGB signal that is developed from the RAW. So all rules as to how you would like your DPX/EXR look before grading applies. The ETTR thingy is a bit of a fun thing to discuss, but the basic theory is good. But you need to protect all your signal, and anything which is basically black&dark out of the RAW file (that's when it's translated into 444 RGB) will basically be noisy if you try to lift it in grade. Thus RLF is good for both the top and the bottom.

WB affects noise as well, and going too far off the sensornative WB will also introduce some noise, and risk clipping in one of the channels, as the wb'ing happens through limiting one channel and expanding another. Thus, make sure you have headroom in all channels in the developed file before grade.

BUT
STill
If it looks good to you at RG3/RC3 and you cannot get "that" look out of RLF/Linear - by all means, do what make your images look good, and not what is in a theoretical way technically "right".
CC is a very non-religious practice. The technicalities are only important tools to solve issues when you see them, and not the focus of the process...

 

[Kevin Marshall]

Thanks for the insight. I also like to look at it as problem solving rather than dogma - though with a few best practices to fall back on. RLF has worked out pretty well for me so far, with a log tolin LUT/transform. Now linear light...that's definitely beyond my abilities/know-how.