Truth or .... False Color

[paulherrin]

stop lights are sensor clip (raw)

 

[R. Gonzales]

I for on agree! We shot with two Dragons last weekend and ending up having this very same conversation. Thanks for bringing up Mark.

Here's what I want RED to change to help everyone with exposure:

#1: Under TOOLS - when you select EXPOSURE - and get the "false color" display - RED, GREEN, PURPLE - but it SHOULD NOT be downstream from IRE processing (it should NOT change when you move the ISO) - it should, perform exactly like it does when you are in the RAW view mode now. This way, your monitor path is set how you want it - but the EXPOSURE tool is showing you want is happening at the "raw/sensor" level - like the clipping stoplights.

#2: Under TOOLS - eliminate RAW view. You can't view RAW - this RGB representation only confuses people and really serves no meaningful purpose IMO. It does much more harm than good - with the ONE critical exception - which is if you switch to RAW view mode and select EXPOSURE tool - you will see a VERY accurate representation of what's about to clip/clipping and what's hitting the noise floor. The problem is - to use this tool - you have to switch to RAW view mode - which sends this horrific, ugly "RAW" view out the monitor path and the clients freak out - or the ACs forget to switch back, etc. -

Make the EXPOSURE tool work like it should - (see #1) - and eliminate RAW view mode (see #2)

If RED would change these two things - we would be able to pull up the EXPOSURE "false color" tool anytime - and not disrupt the monitor path - and REGARDLESS of ISO - we would know what's happening at the top and bottom of our image at our exposure choice - in a much more meaningful, detailed way than the stoplights. And there would be less confusion.

 

[David Battistella]

stop lights are sensor clip (raw)

Who cares if you don't know what pixels or which part and how much of the captured pixels are clipping.

The best tool for the is EXPOSURE, where you see what is in the red and you can decide to let it go or not.



Battistella

 

[Jacobo Martinez]

Does anyone know what is dragons base exposure in the EXPOSURE TOOL in RAW mode? MX is 320.

 

[Mark L. Pederson]

If RAW EXPOSURE MODE in MX is based on 320, without metadata, when you see RAW EXPOSURE MODE on dragon is it also based on 320?? Digging deeper, are the stop lights based on 320 on the dragon?

Guys - the please stop saying the RAW EXPOSURE MODE in MX is based on 320 - the RGB representation is (close to that but not exactly that) but the tools in that mode are responding to the amount of light or lack thereof hitting the sensor. NOT to any ISO.

Please DO NOT think of stoplights as being based on any ISO.

 

[Jacobo Martinez]

Mark, thanks for clarifying this. I went to reduction hollywood in 2011 and I had the impression they said that RAW was based on the sensor base ISO which is 320. I'll go see my notes.

 

[Mark L. Pederson]

Mark, thanks for clarifying this. I went to reduction hollywood in 2011 and I had the impression they said that RAW was based on the sensor base ISO which is 320. I'll go see my notes.

The RBG representation (image) is - but the tools (meters) are not.

 

[Jacobo Martinez]

Mark, is the RGB representation ( image ) in dragon based, I know no exactly, at 320?

 

[Mark L. Pederson]

Mark, is the RGB representation ( image ) in dragon based, I know no exactly, at 320?

I think it's around 500-640ish - would need to look at a scope. But ... who cares? It's a useless RGB representation that should die a swift, quick delete of code

 

[ericyoung]

+1 to NOT limiting high ISO settings - it's useful for checking black calibration, and to setup focussing in dark conditions.

 

[Hans von Sonntag]

+1 to NOT limiting high ISO settings - it's useful for checking black calibration, and to setup focussing in dark conditions.

This is a valid point. Oversaw that.

Thanks,

Hans

 

[Adrian Jebef]

David, do ISO changes have something to do with where 18% gray is located within the dynamic range?

Yes. Exactly.

Not that I enjoy disagreeing with you, but...

No. Not during scene referenced exposure based on a camera's Base ISO. During development ISO via RCX your scene referenced exposure continues to remain the same (ie: fixed) but relative middle-grey will shift within the dynamic range scale initially captured during exposure.

This chart will help. You can read it one way by seeing how 18% Grey is always a constant value no matter what ISO you change the camera to. The static range of DR (about 14 Stops) shifts up and down relative to ISO set in-camera. In Post (RCX) via development ISO this chart can be read an entirely different way. When you adjust development ISO you are now re-scaling 18% Grey to the mid-point of your camera's overall DR (about 14 Stops). Yet it is important to realize that just because you have re-scaled 18% Grey in Post doesn't mean you have recovered any additional highlight or shadow detail that was lost during initial scene referenced exposure. There is only 14 Stops of total DR. It is a fixed range that will not change no matter what ISO you set your camera to or what development ISO you set RCX to.

The intricacies of exposing for digital cinema are touched on a bit more in the EXPOSING + NATIVE ISO thread found here:

http://www.reduser.net/forum/showthread.php?112540-Exposing-Native-ISO

 

[David Battistella]

Thanks Adrian,

And we are talking about how we control measured light values hitting a sensor based on a starting point. You need start with an ISO rating and move the middle grey consciously to have your image DR fall into a ZONE by sliding the middle grey around based on the DR you want to capture in your scene.

This is why one base (reference) is need, because a known value needs to be there to move from.

Or maybe you don't and you have to discover it on your own? But every film stock ever releases starts with a "known value", from which we can obviously manipulate response with our own methods.

Battistella

 

[Adrian Jebef]

We all do realize that even though RED's cameras shoot raw they're still video cameras, right? Just because raw data can be interpreted numerous ways into a target gamma/color space (ie: RGB Video) doesn't mean that the "RAW VIEW" mode is useless because it gives us an RGB image instead of a linear-light one. If anything, I'd say that RAW VIEW works pretty well even though it's gamma encoding is a tiny bit misleading.

If anything, and as others have stated, it would be a lot simpler for RED to define the Base Sensitivity of their cameras via an RGB Video encoded Log output. For EPIC this would basically be 320 ISO / RedLogFilm.

We all realize that both ARRI and SONY have done this from the start, right? The ALEXA's Log-C output and the F65 or F55's S-Log output map very well back to their respective raw files (though the F55 still needs a bit of work). Both SONY and ARRI also offer up proper transform LUTs that make working with their cameras very straight-forward (though the F55 does need some work). It would be great to see RED finally embrace this approach as DRAGON seems to be the camera that would benefit the most from it.

 

[Adrian Jebef]

This is why one base (reference) is needed, because a known value needs to be there to move from.

Or maybe you don't and you have to discover it on your own? But every film stock ever released starts with a "known value", from which we can obviously manipulate response with our own methods.

Battistella

And this is, in essence, what creative or intentional exposure is all about. Amen, David!

 

[Mark L. Pederson]

Not that I enjoy disagreeing with you, but...

Adrian - I actually appreciate when you do disagree with me as I make me think about how I'm representing and explaining something

This chart does really show what happens in a nice graphical way - but we are not really disagreeing - I just failed to make the statement clear enough. Which is very bad - since I really don't want any confusion.

"where 18% gray is located within the dynamic range" should be "where 18% gray is located RELATIVE TO the total dynamic range"

That is what changes.

I am very glad you are posting here on Reduser Adrian.

 

[David Battistella]

We all do realize that even though RED's cameras shoot raw they're still video cameras, right? Just because raw data can be interpreted numerous ways into a target gamma/color space (ie: RGB Video) doesn't mean that the "RAW VIEW" mode is useless because it gives us an RGB image instead of a linear-light one. If anything, I'd say that RAW VIEW works pretty well even though it's gamma encoding is a tiny bit misleading.

If anything, and as others have stated, it would be a lot simpler for RED to define the Base Sensitivity of their cameras via an RGB Video encoded Log output. For EPIC this would basically be 320 ISO / RedLogFilm.

We all realize that both ARRI and SONY have done this from the start, right? The ALEXA's Log-C output and the F65 or F55's S-Log output map very well back to their respective raw files (though the F55 still needs a bit of work). Both SONY and ARRI also offer up proper transform LUTs that make working with their cameras very straight-forward (though the F55 does need some work). It would be great to see RED finally embrace this approach as DRAGON seems to be the camera that would benefit the most from it.

Yes. They are all video cameras (because they don't actually shoot film) but the approach with these is much more film like and the language is that way as well.

I think a LUT for RLF/RG3 would be a good place to start.

So from the camera standpoint you'd have

1. PRO VIEW RFL (replacing RAW view, kinda like LOG in other cameras) - designed for operators and cinematographers.

2. Monitoring view with LUT applied (kind like other cameras). For those pesky commercial clients.

3. Exposure tools only working on raw sensor data but with BASE/KNOWN VALUE from the manufacturer because we need the reference point to start from.


easy as 1,2 3

Battistella

 

[Michael J Brennan]

Red make the worlds highest fidelity colour and exposure meter that precisely and automatically compensates for aperture and field of view and needs no cross referencing, tables or conversion factors in relating the measurement to the imaging sensor.

It can measure photons to within at least a 1/3 of a stop accuracy to the DP's sensor.

OK an Epic isn't the worlds cheapest light meter

ISOs were designed as part of the methodology for a chemical process whose results were variable. DPs who were anti anything to do with "video" were keen to keep their meters when Dcinema came along and so manufacturers incorporated features to help them with the transition.

That brings us to present day with new DPs who are not so familiar with either film nor video and who are presented with a smorgasbord of exposure options although the dessert is mixed in with the main course.

When I discussed the concept of false colour for shadow detail with the manufacturer it was at a time when there was little room to manoeuvre tones or exposure in post because there was no such thing as raw or log recording. So an easy to see zebra of a shadow in a viewfinder was a significant feature to nail the lighting and exposure of shadow detail, a part of the curve where light meters were highly inaccurate at low light levels with somewhat noisy HD cameras.

But nowdays this feature seems to be less appreciated than I would have thought.

Using the false colours and combined with the added flexibility in controlling shadows that comes with raw recording there is now no excuse for massive diversions from acceptable exposure…

It seems to me the potential in increased fidelity and advantages of RAW are often squandered by novices. Under exposing one or two stops unequivocally requires the DP to have calibrated the post process into his own "zone system", on set monitoring, his head or his hand held meter.

But for a jobbing, non owner-operator, such calibration often turns out to be less than precise.


In my view the ultimate beauty and power of electronic imaging is the ability to see in fantastic detail, how tones, colours and textures of the subject interrelate under the particular controlled or ambient lighting, lens, makeup, set design, filter and sensor combination.. of course a calibrated monitor plus blackout required!

A hand held meter does not speak to such inter-related subtleties, but a calibrated monitor does… as does, to a lesser extent, an experienced eye viewing the scene.

Of course with raw you can get it in the ball park and fine tune it in post, but this is a destructive process in respect to data and one that can't easily add or subtract localised fill not to mention tweaking the position of the subject.

Here is the well known example of where the meter sees one thing but the monitor reveals another….. (A and B are the same luminance).

Optical illusion - Wikipedia, the free encyclopedia

Thats my rationale on the subject of exposure…. lighten or darken to your taste



Mike Brennan

 

[Mark L. Pederson]

UPDATE on the whole EXPOSURE meter not working as advertised when not in RAW view (the way I want it to work in post #1) -

I have re-tested and verified with an engineer at RED that was it was working almost perfectly in the last 4.0 firmware builds released - but it does NOT work correctly in any of the 5.0 firmware builds outside of RAW view - RED is taking this very seriously and working towards a fix.

It DOES work perfectly in RAW view in all the 5.0 firmware builds.

 

[Graeme Nattress]

If anything, and as others have stated, it would be a lot simpler for RED to define the Base Sensitivity of their cameras via an RGB Video encoded Log output. For EPIC this would basically be 320 ISO / RedLogFilm.

That's pretty much what we do do though..... We map sensor data to linear floating point data, we put that through the RLF curve so as to map 0.18f to mid grey cineon code value which linearizes back to floating point linear light with mid grey at 0.18f via the standard cineon formula (or Nuke log2lin node, or Shake log2lin node).

In other words, the RLF maps perfectly back to linear light, and after compositing you can get right back to the RLF via a lin2log node.

Graeme