clipping at less than 100 IRE?

[Tyson Birmann]

I'm hoping I can get some help understanding something I'm seeing in the footage.
In a RAW file, the RMD settings on affect the display of the image not the recorded image data.... Correct?
If I want to see the best representation of what is actually being recorded by the sensor and RAW file... I should look at REDLogFilm an "zero out" all other settings... RiFor now let's assume both of those ideas are true..
My question is regarding IRE. It would seem to me, overexposed = +100 IRE. Meaning, if I have NO detail because of clipping... That should read 100 IRE in REDLogFilm (or any gamma) assuming I haven't messed with the curve and all settings are neutral. As long as I am not changing the signal power input or output levels clipping should always show as 100 IRE, yes?
My problem is, I'm seeing my clipped portions of the image as flat lines that plataeu at around 60 IRE. How can the sensor lose detail at 60 IRE?
This makes me think that the R3D file isn't a true RAW file and that the RMD settings that were used in camera were actually changing the recorded data out of the sensor.
I guess the real question is: are there R3D settings (Look setting in camera) that superseed the RAW data? Are there settings that are actually changing the latitude, exposuere, etc... Of the recorded file?

 

[David Battistella]

The DSMC manual is a very good place to start.

Hi Tyson,


There are many ways to gauge exposure, I think a light meter and ISO rating are a great start.

If IRE has been your preferred method on other systems I might suggest that this is not the best way to judge exposure with RED RAW cameras. The histogram, exposure tool, and RAW VIEW, Goalposts, and STOPLIGHTS all tell you MUCH MORE about what is going on with your image.

IRE is well, not reading from the RAW file for starters, so not the best choice.

LINK TO DSMC MANUAL

Battistella

 

[Tyson Birmann]

I agree about IRE as an exposure tool. The problem is, I'm not the one shooting it. It is being shot and the delivered to me. I just feel like I'm getting truncated results in making dailies. It looks fine as long as I stick to the RMD values... But once I try some changes to match shots, I feel like I'm losing data quicker than I should.
How am I to judge the values of the image if not with IRE and a waveform? To that end, what Debayer / Gamma settings (in post) will show me true values of the sensor data?

 

[Graeme Nattress]

So, R3D is pure RAW. No camera image processing setting effects what gets recorded into the R3D.

I think your confusion comes from where we define clipping. Clipping is when the sensor fails to see any more information or produce a brighter code value upon input of more light to the sensor. Clipping is only defined in the RAW domain at the sensor level.

The next piece of information that is critical for you to know about is that ISO is done in image processing, not in hardware as baked in analogue gain. That means (assuming the image processing itself doesn't clip the data) that ALL ISO settings have the very same dynamic range. The REDLogFilm curve is designed (because it matches and is compatible with the Cineon curve of a film scan) to handle a wide range of brightnesses of image data. It is not expected for it to reach full range (100%) unless you input really bright image data, and that would mean really really high ISO. Similarly if you knock the ISO right down low, the upper extent of the brightest image values will be no-where-near 100%. It's a wide range container curve and there's no magic value within it that means "clip". Our gamma curves work similarly because they must work well over a very wide range if ISO developments, and are all based upon an underlying REDLogFilm curve.

Graeme

 

[David Battistella]

Tyson,

I thought you were talking about using IRE in camera.

Can an you share an r3d? Or some frame grabs. Hard to judge what you are talking about.

First thing ink you should do is open the r3d, set it to 800iso, dragon color, Redlogfilm
if the histogram favours the left its probably underexposed.

If if you can't share an image AHAE an image of the histogram with these settings

Battistella

 

[Keith Putnam]

If you've clipped your sensor, it's perfectly easy for that clip point to be placed at any arbitrary IRE level on a waveform monitor... If you keep cranking your ISO down, or the brightness, or exposure, or any control that brings down the highlights in your image, that "clip shelf" is just going to drop lower and lower on the scale. If you've recorded clipped data it's never going to "unclip"; it'll just get darker as you grade it down.

 

[David Mullen ASC]

Often when viewing a log file of an 11-step chip chart, the black square might be around 10 or 15 IRE and the white square of the chart might fall at 65 or 70 IRE, but that doesn't mean the clip point is at 65 or 70 IRE, it just means that the camera can capture information beyond the range of an 11-step chip chart, what we might call super-white information that exists in the real world, things more than five stops brighter than middle grey.

 

[Peter Moretti]

I'm hoping I can get some help understanding something I'm seeing in the footage.
In a RAW file, the RMD settings on affect the display of the image not the recorded image data.... Correct?
If I want to see the best representation of what is actually being recorded by the sensor and RAW file... I should look at REDLogFilm an "zero out" all other settings... RiFor now let's assume both of those ideas are true..
My question is regarding IRE. It would seem to me, overexposed = +100 IRE. Meaning, if I have NO detail because of clipping... That should read 100 IRE in REDLogFilm (or any gamma) assuming I haven't messed with the curve and all settings are neutral. As long as I am not changing the signal power input or output levels clipping should always show as 100 IRE, yes?
My problem is, I'm seeing my clipped portions of the image as flat lines that plataeu at around 60 IRE. How can the sensor lose detail at 60 IRE?
This makes me think that the R3D file isn't a true RAW file and that the RMD settings that were used in camera were actually changing the recorded data out of the sensor.
I guess the real question is: are there R3D settings (Look setting in camera) that superseed the RAW data? Are there settings that are actually changing the latitude, exposuere, etc... Of the recorded file?

Tyson,

Frame rate, shutter angle, aperture and lens filtration are the only things related to the camera that can change the real exposure of the recorded file.

I'd suggest posting some screen grabs and source files, so we can get a better look at what you're seeing.

 

[Marc Wielage]

My question is regarding IRE. It would seem to me, overexposed = +100 IRE. Meaning, if I have NO detail because of clipping... That should read 100 IRE in REDLogFilm (or any gamma) assuming I haven't messed with the curve and all settings are neutral. As long as I am not changing the signal power input or output levels clipping should always show as 100 IRE, yes? My problem is, I'm seeing my clipped portions of the image as flat lines that plataeu at around 60 IRE. How can the sensor lose detail at 60 IRE?

My advice would be to pull up the R3D file and apply DragonColor and RedGamma4 to it and see where it comes up in terms of a normal waveform monitor. Check that vs. (say) RedColor3 / RedLogFilm and see how it looks. The only way I can imagine it clipping at 60ire would be if some weird LUT is involved. Are there any input LUTs or Display LUTs in the signal path?

It always helps to have the DP shoot test charts prior to production, or (barring that) shoot test charts in the morning of each production day. In a perfect world, you'd get a color chip chart and/or gray scale chart plus a framing chart, with color charts for each different lighting setup. Even five seconds of chart is better than nothing. The chart will give you clues as to where or whether your pipeline is going wrong.

BTW, what program are you using to check the image: RCX, or Resolve, or Scratch, or what? The config and setup for these programs is fairly critical, and it's easy to go off the cliff if there's a single parameter set incorrectly.