Grading RED footage...

[Graeme Nattress]

Yes, very much so like a zone system. A bit different, but it's that kind of idea.

Graeme

 

[James Press]

Thank you all. This is a MASSIVELY helpful (informative and comprehensible) thread.

 

[JustinGD]

I think I understand how to expose, but what about processing the footage. I remember from MacGregor's thread on the subject, it was suggested to process the footage at ISO200, but that doesn't make sense to me now, so I'm a little confused.

If I shoot a high contrast scene at ISO500, and it is exposed properly and I didn't clip any highlights, do I now process it at ISO500 in RedCine or Red Alert? or would I want to use 400 or 320 for any reason?

 

[Michael Ragen]

The idea behind processing at 200 or 250 is you adjust your iso setting in post until the right side of the histogram gets pulled into view and all your highlights are preserved.

If you shoot a high contrast scene at 500 some areas may appear to be clipped in camera viewing, but when you adjust the raw file you will find that there is still more information in the highlights.

When you reset the iso to 200 you are lowering the noise floor as well as the exposure, but you will want to bring back the exposure through use of curves or brightness as that does not clip data and you can throw a shoulder in to roll off the highlights.

Hope that makes sense.

 

[JustinGD]

So I want to stay at a high ISO, as long as nothing is clipped. Got it.

If I am processing at 500, and none of my highlights are clipped, that will give me the most detail in the highlights. But if my subject is wearing a black suit, and it looks grainy, I could process at 320, lose a few stops in the highlights, but gain detail in the suit without grain? Or would I be better off maintaining my range, and crushing the blacks?

 

[Dj Joofa]

If I shoot a high contrast scene at ISO500, and it is exposed properly and I didn't clip any highlights, do I now process it at ISO500 in RedCine or Red Alert? or would I want to use 400 or 320 for any reason?

I think this is an important question. I am assuming that Red's manipulation of ISO works like a digital gain.

I think it is okay to expose properly for 500 ISO but use 320 ISO in post, so that some extra highlights are pulled in and some blacks are crushed.

Exposing it at 500 ISO and processing it at 500 ISO might knock out some highlights that may have been recoverable at exposing at 500 ISO and processing at 320 ISO.

This is my understanding. But I could be wrong. I don't think Red is doing a good job in explaining how exactly is ISO working. There are two variables here. Raising ISO and corresponding metering to go with it. In most of the discussion I have seen so far people are mixing up the two into one. It is best if Red explains that what effect is on the Raw footage if only ISO is raised regardless of any corresponding metering -- best, if exposure via metering is kept constant and the effect of raising/lowering ISO is explained with pictures.

 

[Rudi Herbert]

Exposing it at 500 ISO and processing it at 500 ISO might knock out some highlights that may have been recoverable at exposing at 500 ISO and processing at 320 ISO.

I'm sorry, but my understanding is that it works in exactly the opposite way. Wouldn't those preserved highlights at 500 ISO actually clip at 320?

 

[Dj Joofa]

I'm sorry, but my understanding is that it works in exactly the opposite way. Wouldn't those preserved highlights at 500 ISO actually clip at 320?

My understanding is like this. Suppose you have metered exposure for 500 ISO and it gave you some raw signal values when you shot footage. Now after shooting, when the raw data has been acquired, you have two options in software:

(1) Process at 320 ISO
(2) Process at 500 ISO.

Assuming that raising ISO in software provides higher digital gain, then if I process at 500 ISO it shall try to boost up the acquired raw data / signal. The boosted signal has a higher chance of clipping. However, the Raw data is intact. Only the boosted signal that shall go into a final standard format might get clipped. This is how I see it.

 

[Michael Ragen]

"I'm sorry, but my understanding is that it works in exactly the opposite way. Wouldn't those preserved highlights at 500 ISO actually clip at 320?"

Processing at a lower iso moves the histogram to the left, pulling in highlights that may be past clipping when processing at the higher iso that you shot at.

 

[Dj Joofa]

Rudi, here is some explanation of my understanding, which I said, could be wrong, especially in the case of Red camera.

The confusion is arising because a typical camera system has two broad components: optics control (we are dealing with exposure only here, and discounting stuff such as focus, etc.) and film / sensor.

Therefore, the model for our incoming light to process is given as:

Incoming Light ----> Camera (optics + film/sensor) ----> Processed signal output.

People from film background are oriented towards the overall or lumped response of these two components. Where as we want to see only the effect of second component (ISO) when, especially the first one (optics) is kept constant.

Now the first component, optics, is increased / decreased in light strength via exposure control knobs such as aperture size, exposure time, etc.; the second component is gain via ISO setting, whether via a different film stock or via analog/digital gain in camera/software.

Therefore our model becomes:

Incoming Light ----> Increase/decrease strength (via exposure) ---> Gain via (ISO setting) ---> Processed signal

Now, one thing is immediately obvious that the magnitude of the ISO gain module should not be more than the optics module responsible for increase/decrease of light strength; otherwise it will undo its effect.

Hence, regardless of your optics setting via exposure, which hopefully is metered according to taste, any increase in ISO shall try to boost up the signal. And, assuming the boosting up is less in magnitude than the light reduction through optics, the overall lumped response is lower -- i.e., more highlights are now pulled in due to this overall reduction via lumped response.

 

[Rudi Herbert]

Joofa,

I see the logic in your explanation, but from my background as a film and digital stills photographer, I feel the opposite is what would happen. Having said that, I have yet to play with any RED footage in Redcine, so I'm not aware of how those files behave, but I would assume it'd be similarly to how all other RAW files perform, which I have worked with extensively in Photoshop.

I start from the assumption that we're shooting at a certain ISO,or better yet, metering to that ISO so that no clipping occurs in the highlights. Now, if no clipping has occured, processing that footage at a lower ISO is essentially the same as increasing the exposure, correct?, since we are assigning the light values of a "faster" stock to a "slower" stock. The faster stock needs less light to yield the same exposure as the slower one, so moving those values one stop down from 500 to 320 actually overexposes the slower stock by one stop, thus moving the grey point to the left, and pulling in more highlight information into the histogram, but this "information" is not detail, is luminance that should make the sensor clip. Excuse the analogies to film photography, but this is exactly what happens when I develop RAW images as well. Where is it that I'm missing the point?

Thanks,

 

[Dj Joofa]

Firstly, on a conceptual level the behavior between how a film and digital system should behave are the same, though I have seen comments on this forum notwithstanding. E.g., people saying that this is video, this is digital, this is film, they don't mix, etc., without fully realizing the underlying semantics of what exactly is happening.

Now coming back to your question.

Joofa,


I start from the assumption that we're shooting at a certain ISO,or better yet, metering to that ISO so that no clipping occurs in the highlights. Now, if no clipping has occured, processing that footage at a lower ISO is essentially the same as increasing the exposure, correct?, since we are assigning the light values of a "faster" stock to a "slower" stock. The faster stock needs less light to yield the same exposure as the slower one, so moving those values one stop down from 500 to 320 actually overexposes the slower stock by one stop, thus moving the grey point to the left, and pulling in more highlight information into the histogram, but this "information" is not detail, is luminance that should make the sensor clip. Excuse the analogies to film photography, but this is exactly what happens when I develop RAW images as well. Where is it that I'm missing the point?

Thanks,

I have highlighted the words "less light" in your comments above. This is the missing link. What you are saying is correct if we are allowed to change the light levels. However, I do think that lowering the ISO in software (not the exposure via optics) is tantamount to reducing exposure after the exposure via optics has been held constant.


However, once a Raw file has been acquired at whatever exposure, the data in that file has now solidified. Hence, we can't go back an make the light less now. So say after exposing for 500 ISO you acquired some raw data, which you exposed correctly using a metering scheme. Nothing is clipping at this stage as far as Red is concerned, because, unlike other cameras, which provide instant analog gain via ISO setting, Red will relegate it for later consumption via software digital gain.

So now you have liberty to process that data at the 500 ISO at which you captured it. However before we do that we need to assume that ISO increase results in data value increase (without changing the raw values) and ISO decrease lowers the value. Also lets assume when you metered the exposure at 500 ISO there were enough bright objects to utilize the full range of raw signal values, that is you used all the range of 12 bits.

Now with the assumption though we exposed the scene for 500 ISO, but later 500 ISO in software will boost signal values more than 320 ISO and that we utilized full range of 12 bit values as the scene had enough bright objects, then 500 ISO setting in software (not when exposed) might knock those top numbers out, which 320 ISO might retain.

 

[Troy Smith]

Hi,

I have a newb question, I noticed that red alert seems to have more tools
for grading then redcine, will redcine in the future have the same tools,
or is red alert always going to be a more complete grading tool, and if so
if one is on PC, what would be the best programme to use in it's place on
the PC side.

One more thing, I know scratch is full on grading tool compared to red alert, but does it have the same ability to adjust the exposure etc like red alert does, or does one need to do the basic exposure etc in red alert before going to scratch?

If anyone has the time to answer these couple of questions I'd very much
appreciate it.

Thanks
Stricko

 

[SF Geek]

Here it is. The confusion is in that there is setting the ASA in camera and in post. The reason Jim says shoot with "camera ASA at 500" is that the in camera histogram will show clipping at the point of actual Raw clipping. I just use the camera 320 setting and know that I have an extra half stop I can open beyond the first sign of clipping in the histogram. It works the same way. I just feel that camera 320 represents the mid tones more accurately in the monitor than camera 500 and doesn't give me the false sense that I have more shadow detail then I really have.

Now there is post processing ASA settings. We need to be clear about which one we're talking about in our posts. This creates the most confusion and the Red team needs to distinguish between the two or they will just create more confusion.

When bringing a high contrast scene into Redcine there will be some clipped highlights, especially if your camera ASA was set to 500. If you set your Post ASA to 200 you will find that you recover the information in those clipped highlights and then you can use the curves to adjust the contrast or histogram to your liking. I'm sure there are other ways to recover clipped highlights, but curves is not one of them. Changing the Post ASA was the way I found that works. One easy way to see if you can recover your highlights is to click on the color button on the top toolbar in Redcine to see the Raw image with no gamma applied. If your highlights are clipped in this view that means that they were clipped at the sensor and can't be recovered.

Remember in your posts about ASA to distinguish between Camera ASA settings and Post ASA settings. This should clear up a lot of confusion.

 

[Paul Leeming]

Rudi,

the big difference is that Red's RAW is EXACTLY the same data no matter what you set the camera's "ISO" to, because it doesn't apply any gain in camera to the image itself, only to the monitoring outputs.

With DSLRs the RAW image gets analogue gain applied to it when you adjust the camera's ISO value, so in that sense what you are used to would apply to Red if it behaved the same way, but it doesn't.

So then the issue becomes one where setting ASA500 on Red merely pushes the monitoring exposures up and tells the metadata to record it as ISO 500. The exposure histogram as it works in Build 15 is based on Rec.709 output, so since you are "gaining" up the monitor outputs it responds accordingly to show your new clipping values as if the camera were rated to that value. But of course the camera is still recording the exact same RAW image since there's no sensor gain, so you've effectively underexposed by two thirds of a stop (based on the rated sensor ISO of 320), generating more shadow noise but giving higher amounts of highlight tonality.

Now in Redcine, the metadata initially reads in as ISO 500 and the image is processed that way, but you can re-rate it to ISO 320 and effectively "underexpose" it again to bring back those values that were always there because the camera only records at the one fixed "real" ISO anyway.

Hope that helps, it seems a bit counter-intuitive at first until you remember that the camera has a fixed ISO and doesn't gain it as DSLRs do.

Cheers,

Paul

Joofa,

I see the logic in your explanation, but from my background as a film and digital stills photographer, I feel the opposite is what would happen. Having said that, I have yet to play with any RED footage in Redcine, so I'm not aware of how those files behave, but I would assume it'd be similarly to how all other RAW files perform, which I have worked with extensively in Photoshop.

I start from the assumption that we're shooting at a certain ISO,or better yet, metering to that ISO so that no clipping occurs in the highlights. Now, if no clipping has occured, processing that footage at a lower ISO is essentially the same as increasing the exposure, correct?, since we are assigning the light values of a "faster" stock to a "slower" stock. The faster stock needs less light to yield the same exposure as the slower one, so moving those values one stop down from 500 to 320 actually overexposes the slower stock by one stop, thus moving the grey point to the left, and pulling in more highlight information into the histogram, but this "information" is not detail, is luminance that should make the sensor clip. Excuse the analogies to film photography, but this is exactly what happens when I develop RAW images as well. Where is it that I'm missing the point?

Thanks,

 

[Rudi Herbert]

Ok,

Thank you all. I see where I am "over thinking" and making the mistake, but I guess certain apps handle this issue in a different way, such as Photoshop RAW, where decreasing your ISO will effectively increase your exposure. Anyway, I guess the minute I slide a handle in Redcine and it yields the opposite of what I'm used to, I'll simply learn to slide it the other way

Thanks again,

 

[Jason Sinclair]

I don't have a camera yet, but it seems to me that you can pull information both ends of the scale. Here's a shot i downloaded from redrelay:

exporting from red alert i changed the iso from 320 metadata to 200 and exported as a tiff and then played in photoshop:

I don't really know what to look for in noise but here are close ups (200%) of the clouds and darker areas:

 

[Jason Sinclair]

Here is the close ups (all x 200%) of highlight areas

There is detail in the clipping and there is detail in the shadows.

Apart from my terrible grading, I don't know how this pans out in a real world circumstance but theoretically there's more information there that meets the eye.

 

[Jannard]

Build 16 addresses to tendency for people to over-expose. The fix for over-exposure is to lower the ISO (or exposure) in RED Alert! It isn't that ISO 320-500 is wrong, it is that the exposure was missed at that setting. Thankfully, RAW helps a lot of mistakes.

Jim

 

[Lauri Kettunen]

Shooting in extreme conditions and grading

Shooting in extreme conditions and grading

While Jim started this thread, I spent time next to the Arctic circle shooting outdoors in rather extreme conditions and having a very good chance to challenge the dynamic range of the camera. All images shot with the Nikon adapter.

I had the camera all the time set to ASA 320, but have set zebras, histograms etc. such that have all the time a good idea of the highlights. As most wildlife shooters I try to perserve the highlights unless there's a good reason to blow them off.

Here's the most difficult situation: a pair of Golden eye ducks in water reflecting bright light. It's worth to notice the green tone on the cheek of the male. This is simply amazing!

The black grouse: Typically if the white of tail is not blown out, the rest of the bird is black. Notice how Red One picks the blue and brown tones, which have been very difficult to capture before.

Bright light reflecting on the ice: At this time of year, there's a lot of light. Again, a situation which has always been difficult, or, in fact, impossible to manage. There's deliberately a slight overexposure in the middle of the image.

Raven's nest: This is my favourite. The clip was taken half past ten in the evening with a 500mm/f4.0 lens. There wasn't enough light, so I set aperture to f4.0. This corresponds to about ASA 600, but the blacks are there without any problem.

Evening light: Warm tone agains rain clouds in the back. Again high constrast but the outcome is marvellous.

Summing up, Red One is a wonderful tool.