9 1/2 stops...

[Jannard]

Just foolin' around. Background T32, face in covered, heavily shaded doorway T1.7 (ISO 400).

1st one is not worked much... nothing clipped and nothing crushed pure black. This test is the "hard way"... putting the subject as the darkest object in a high range scene. We are not debating whether or not the details or noise in the shadows of the trees are good enough or not.

2nd one is worked with highlight recovery only to show that the info is really there. Not as a grading option.

Metered with Sekonic Cine.

Jim

http://www.reduser.net/forum/uploaded/9_9.5stops1.jpg

http://www.reduser.net/forum/uploaded/9_9.5stops1xx.jpg

[Matt Gottshalk]

Sweet.

[Nathan Garofalos]

What build are you on?

[Kyle Presley]

Is this build 16, Jim?

[Jannard]

This is Build 16. Build 15 will do this as well, just won't look quite as good... :-)

This is more about grading technique and not throwing away info.

Jim

[SF Geek]

I can see that there is information there in the second image, but I have to say that it looks terrible. It's not an acceptable image. I'm not trying to jump on you, but it doesn't seem like a very good grade. Is this how the new noise looks?

[Jannard]

I can see that there is information there in the second image, but I have to say that it looks terrible. It's not an acceptable image. I'm not trying to jump on you, but it doesn't seem like a very good grade. Is this how the new noise looks?

Maybe I should delete that 2nd one? It is not a useable grade. It is just hyper-extended to show the info is there. Or were you kidding?

Jim

[Mike Prevette]

The noise looks much better. Far less patterning, I'll need to hold out to see it in motion though.

[SF Geek]

I'm not kidding. I would take the second one off. Do you have a different grade, something in the middle? I think that the second one just makes the camera look noisy.

[SF Geek]

I can delete my post as well if you'd like.

[Brandon Fraley]

i think you should leave it. I agree that it's "unacceptable" but it shows the literal range. I think there's too much discrepancy between literal dynamic range and the more loose "usable" range.

I don't know much about film, but it seems everyone continuously throws out 13 stops all the time for film, just as a given, its a rule, film has 13 stops, this is apparently common knowledge everyone knows (whether they've ever shot a frame of film or not). But then ppl test RED and go, "eww, well you get detail but its not usable, i mean really, you can use about 7.5 stops..."

I'd like to see both literal and usable DR tests from all the major players to really get an idea of whats what. RED ONE, Vision2, F23, F900, Genesis, 16mm, even DVX100 so I can compare with what I've shot a lot with.

Obviously I'm not capable of providing these test, and who knows if we'll ever see them, but i think put an end to a lot of the confusion.

[Jannard]

I'm not kidding. I would take the second one off. Do you have a different grade, something in the middle? I think that the second one just makes the camera look noisy.

Just consider that the 2nd one is a good 3 stops open from meter. So that's what a 12 1/2 stop grade looks like.

Jim

[James Brundige]

I can see that there is information there in the second image, but I have to say that it looks terrible. It's not an acceptable image. I'm not trying to jump on you, but it doesn't seem like a very good grade. Is this how the new noise looks?

It's not a cosmetics commercial, it's a look at an extreme contrast situation. Thanks, Jim.

[Brandon Fraley]

Just consider that the 2nd one is a good 3 stops open from meter. So that's what a 12 1/2 stop grade looks like.

Jim

forgive my ignorance: why do you call it 9.5 stops in the title, but then say its a 12.5 stop grade?

[James Brundige]

forgive my ignorance: why do you call it 9.5 stops in the title, but then say its a 12.5 stop grade?

Because he boosted the shadows up 3 stops in grading.

[SF Geek]

I know it's not a cosmetics commercial First light. I'm just saying that it might not go over so well with the Red skeptics. I'm leaving it at that.

[Brandon Fraley]

so what does 9.5 stops refer too?

[Brian Ferguson]

T32 -1.7
To pull out anything like that is to my calculations 9 stops under exposed and hold that much detail while the background stays almost the same. Damn. I have been to hundreds of telecine sessions and that would be a good grade in anyones book. Obviously this is showing a more docu situation but if you have some fill you are golden. This blows my socks off. I have seen Jim grade images before and I don't think most people understand what he knows... Curves. To anyone who thinks that the second image should be deleted I wonder how many times you forgot your light meter and needed to have a colorists save your .... project. If this was a docu and not a larger project with more money for fill etc. I would be really pleased.

[Jason Sinclair]

same here. Very impressive given those extreme circumstances. Film is going to have a hard time there as well (not that I'm comparing) How many people would choose to shoot in those circumstances even if they were doing a doco? Not many when you could just move subject and camera around 30 degrees.

[Mark Andersen]

[QUOTE=Jannard;224485]J (ISO 400).

Is that a hint at the new B16 ISO?

[Rudi Herbert]

Well, I also feel that the lady's face is unnaceptably noisy, BUT, and this is BIG but and one that makes ALL the difference, all you need is a little fill light on her and you're set, great image. When shooting something like that, be it on film, F-35, Dalsa, whatever, you better throw some fill at that face or you'll just get different versions of noise on her, more or less pleasant depending on your taste, but noise/grain just the same. That's good stuff.

[Jeff Brue]

hmm I like that noise pattern jim, looks like it would be easier to reduce than currently what we're seeing with 15. Guys as far as the noise threshold remember there are a ton of tools out there to reduce that while maintaining great detail. Random noise is a lot easier to kill than mish mashed wavlet reduced noise.

[Sudhir Chaudhary]

1.7 - 32........... What was the exposure ? I wish to know how many stops under the girl is?

[Fredrik Callinggard]

I know it's not a cosmetics commercial First light. I'm just saying that it might not go over so well with the Red skeptics. I'm leaving it at that.

Hmmm I think that it's a general deception, thinking that the ones talking about the DR are skeptics. I think the ones that are, are actually the ones that cares, they want the camera to evolve.

Great example Jim. Finally we're talking about something. I love the noise in build 16. It's less compression noise and more... just noise. It's a great step forward. As for me it also shows what I'm very often talking about. Were a shot is acceptable or not (to me). That's why I'm saying that you should stick with around 8 stops, that's in build 15, of course. Within those 8 stops I know I have salvageable footage, within my acceptance level. I think it's important for people to understand that it is an opinion and actually listen to it. It's these thoughts, from different users that can help you build your own acceptance level, but to ignore it is foolish.

Now as for my general thoughts, I'm happy that the noise level (or pattern) in shadows is improving. That has been my general concern. There's very often in my real life shooting and especially in my test shooting that I've found it to be (at times) disappointing. To tackle this problem, not by only giving some more DR, but mainly by making the noise "look" more acceptable, is very important.

This is a step in the right direction!

PLEASE! KEEP AT IT! Don't stop here :innocent:


Fredrik Callinggard

[Mark Crabtree]

Noise is the biggest issue with build 15. If build 16 can reduce noise as promised I will be so happy rating faster than 160. Since I mostly shoot for TV broadcast I choose low noise over detailed highlights. I'm looking forward to having both.

[James Brundige]

I know it's not a cosmetics commercial First light. I'm just saying that it might not go over so well with the Red skeptics. I'm leaving it at that.

No offense intended. I think the new technology will thrive or not on it's real merits, not the stuff we bat around on this board. For example, on the Che discussion, no one observed that the camera was facile enough to capture footage that would have been very hard on another platform. So, using an early camera and build, SS blows out a white shirt in the jungle. Big deal. The bottom line is that, if shooters get results they like (with whatever tech specs we carry on about), they'll use it. No need to edit ourselves here.

[David Mullen ASC]

Noise is the biggest issue with build 15. If build 16 can reduce noise as promised I will be so happy rating faster than 160. Since I mostly shoot for TV broadcast I choose low noise over detailed highlights. I'm looking forward to having both.

If you're shooting for TV broadcast, then how are you seeing much noise when shooting at 320 ASA to 500 ASA on the RED? With 4K images downconverted to HD, the noise should be quite small and hard to see.

I have seen some noise when looking at the live 720P output, but that's partly because it is a quick and dirty live debayering.

[Mitch Gross]

Interesting. Looks like the new DNR sceme in Build 16 is effective. Wonder how it reacts to tungsten balance. So what was the over/under on that shot? I'm guessing around a Polanski (T11/16 split), maybe a touch more open.

[Brook Willard]

Jim's post and people's reaction also brings something to mind that I've posted about before: misinterpreting the meaning of dynamic range.

She is 9.5 stops under the background. The camera has more dynamic range than is in this image, but not by much. Having 9.5 stops of dynamic range in an image does not mean that you will be able to pull the shadows up into a normal exposure without paying the price.

There is no format in the world where you'd let somebody's face fall 9.5 stops below the majority of your image with the mentality of "yeah, I'll just bring her up in post." That's not how dynamic range works. There is flexibility in these files, but no amount of flexibility will let you truly light in post.

Her face is several stops underexposed. To bring her face up to a "normal" level for the 2nd picture, she had to be processed at a different relative ASA. The noise you're seeing on her face is what you should expect to see when you process something at a 2-4,000 ASA equivalent. Of course it's noisy! It's a deep, deep shadow that he brought up for us to see.

So dynamic range does not mean that anything in that range can be pushed around to your heart's content. You can push stuff around a little, yeah... but not that far. Shadows are shadows, highlights are highlights and midtones are midtones. You can get away with more on the highlight side [if you have not clipped], but if your shadows [or her face...] are too dark for you... point a freakin' light at it.

Jim knows this - obviously he didn't shoot her like that with the intention of bringing her face back in post. It's just another good opportunity to point out that if you're seeing unacceptable noise levels in the 320-400 ASA range... you've messed up.

[Graeme Nattress]

Correct Brook. I don't know of any camera where you can have the subject nearly 10 stops below the highlight and bring it back up to proper exposure in post and have it look noise free.

Graeme

[Mark Crabtree]

At ISO 320 the picture is relatively noise free in all but the shadow areas. It is very noticable on my Panasonic SDI 50 inch plasma. Rating at 160 is like magic to make the noise in the shadows go away. At NAB I attended the screening of Rosencranz and Gildenstern are undead. The DP was answering questions for the audience and he said he rated the camera at 320 but set his key one stop hot. How is that any differant than rating at 160? The footage looked great with very low noise.

[Jannard]

The only time I would ever consider shooting ISO 160 is if it was a low contrast scene with enough light. Then it is a good choice. But more often, I shoot ISO 400-500 and get very noise(almost)free images. But you have to nail the exposure.

Jim

[Jannard]

BTW,

I purposely made the subject the darkest thing in the wide range image. Too often, people make the subject well lit and then argue whether or not the information in the shadows is OK or not. It is much easier to pass detail in the weeds than a face. I tried to make this as hard a test as possible... given shooting, grading and posting in less than 5 minutes. :-)

Jim

[ATF]

Sweet!

Can't wait for B16!

ATF

[Jannard]

Sweet!

Can't wait for B16!

ATF

I guess the lost point here is that you can do this with Build 15. It is a grading issue when people talk about only getting 6-8 stops from a RED ONE.

Jim

[Fredrik Callinggard]

I guess the lost point here is that you can do this with Build 15. It is a grading issue when people talk about only getting 6-8 stops from a RED ONE.

Jim

I completely disagree Jim. I did my tests together with some of the most skilled people in the business (some of them actually took part in building the whole DI process) and they should know how to grade, even if it's RAW.

As I said the 8 stops is about a tolerance level and a lot of us found that 8 stops was ours. We're not saying that in BUILD 15 you couldn't push more out of it. We're simply saying that it didn't look good and that the tolerance level for the RAW file before it "truly breaks" was within that range (in that meaning that at the edges within our opinionated DR it was breaking already, just not as much).

As a lot of users seem to say in this thread - the biggest problem was the compression noise. With 16 that seem to be A LOT better.

So it's not a grading issue. It might be now in the future but with build 15 I disagree with you.


Fredrik Callinggard

[Jannard]

I completely disagree Jim. I did my tests together with some of the most skilled people in the business (some of them actually took part in building the whole DI process) and they should know how to grade, even if it's RAW.

As I said the 8 stops is about a tolerance level and a lot of us found that 8 stops was ours. We're not saying that in BUILD 15 you couldn't push more out of it. We're simply saying that it didn't look good and that the tolerance level for the RAW file before it "truly breaks" was within that range (in that meaning that at the edges within our opinionated DR it was breaking already, just not as much).

As a lot of users seem to say in this thread - the biggest problem was the compression noise. With 16 that seem to be A LOT better.

So it's not a grading issue. It might be now in the future but with build 15 I disagree with you.


Fredrik Callinggard

You certainly have the right to disagree with me, but we have found MANY trained professionals that were incorrectly grading RED footage. I have done this more times than I want to count (with Builds prior to and including 15). And just because you found people that couldn't do it doesn't mean it can't be done. Graeme has seen me do it, as has Jarred, Deanan, Rob and Ted. Actually, all of them can do it. But you telling me it can't be done just doesn't hold water. Now if you (and your pros) would like to learn how to do it... that is another topic.

The good news for all is that Build 16 makes it very simple to grade RED footage. We have done all the work for you (actually, Graeme and the team have)...

Jim

[I Bloom]

I guess the lost point here is that you can do this with Build 15. It is a grading issue when people talk about only getting 6-8 stops from a RED ONE.

Jim

I was in a situation like your posted image today, where reaching this deep into the shadows would be preferable to clipping. Looking forward to seeing more.

So far BRAVO!

IBloom

[I Bloom]

Now if you (and your pros) would like to learn how to do it... that is another topic.

Jim

I'd like to know as much as I can about grading this footage. Sign me up.

IBloom

[Jannard]

I'd like to know as much as I can about grading this footage. Sign me up.

IBloom

I'll post it in a new thread in a bit...

Jim

[Steve Sherrick]

You certainly have the right to disagree with me, but we have found MANY trained professionals that were incorrectly grading RED footage. I have done this more times than I want to count (with Builds prior to and including 15). And just because you found people that couldn't do it doesn't mean it can't be done. Graeme has seen me do it, as has Jarred, Deanan, Rob and Ted. Actually, all of them can do it. But you telling me it can't be done just doesn't hold water. Now if you (and your pros) would like to learn how to do it... that is another topic.

The good news for all is that Build 16 makes it very simple to grade RED footage. We have done all the work for you (actually, Graeme and the team have)...

Jim
Jim, this seems like a good time to make a suggestion for the RED team to do a video on exposing RED properly in Build 16. In fact, if you are only working out bugs right now with 16 and the build won't change dramatically from now until release, perhaps you could put this together before the beta is even released.

To me, it's of the upmost importance to have something that will show people the right way to approach this. Not a video that teaches people how to light and expose a digital image, but rather one that shows to approach your exposure when shooting with RED and a RAW workflow. I think this could really diffuse some of the speculation and "wrong" approaches people are taking.

You guys know the camera inside and out and should be the authority on showing people how to use it. Let's end this back and forth stuff going on, and have a video that leaves very little to the imagination and shows best practices with working with RED footage.

What do you think?

[Jannard]

One of the problems with previous builds is that it was not easy to nail proper exposure. Then grade. Build 16 is "magic" in that it makes it MUCH easier to expose properly and grade more easily.

I'm preparing to start a new thread that shows how to grade RED Build 15 footage and then show that we do it for you in Build 16.

Jim

[Gunleik Groven]

These have been very valuable secrets... :)

Cheers!

Looking foreward to build 16

[Sanjin Jukic]

Thanks Jim.

B16 is going to be a great lecture to grade RED footage for all of us even for 'professionals'...

[Evin Grant]

Jim, the texture in that B16 frame is stunning! I can't tell you what a relief that is. There was something about the first few builds that always struck me as more organic that those of 12 and higher. It looks like you've done those even better. The first thing I shot on Red, literally the second or third day after the first 25 shipped was a music video. (Silly latin piece remember) well I rated the camera at ISO 640 for that shoot and I was totally happy with what came out in the grade. It certainly had texture but it seemed granular and beautiful. Then, somewhere around build 12 it changed very subtlety and I stopped pushing the camera past ISO 400. Not because the noise was too much but because I felt something different, I'm not sure I even realized it till I looked at that frame at the top of the thread.
RC36 did help quite a bit too and now my standard shooting ISO in B15 is 400 ISO but beyond that it still seems to give me a more "Mosquito" pattern that you have mentioned before. Personally I easily and repeatably get 9 solid stops of DR out of B15 at ISO 320-400 and I never really notice the noise, at least not in any way that bothers me. But the example you have posted above means that ISO 640-800 may be a reasonable option again for those of us who are not afraid to trade organic texture for sensitivity and DR.

[Álex Montoya]

Great example, Jim.

Tho' many here are nor understanding what you try to explain.

[Jannard]

Great example, Jim.

Tho' many here are nor understanding what you try to explain.

Just posted a new thread with the "secrets"... although I posted the same thing about a year ago. :-)

Jim

[Evin Grant]

Here is a frame from a short I shot and am beginning post on...
http://www.4kninjas.net/pics/BrumbyHDR.jpg
The top frame here is properly graded, the bottom pic is what I saw when opening the R3D. I exposed this scene at ISO 500 knowing that those building in the background were 4+ stops hot and I wanted the chance to recover them. Experience told me I could do that and still light for the face at ISO 500. All I had was two 1200W HMI's bounced into foam core and a 200W Joker for an eye light. There is some noise in the lowest tones but it's not distracting at all once they are dropped into their proper range. I think this shot is gorgeous! and I can't wait for B16 so I can push it even farther.

[Jan Reiff]

looks fine. looks good. looks professional recorded.

[Jonas Nyström]

Nice! Is shot in 4K? What shutter/iris?

[Jannard]

Great example Evin... we are finding that way too many graders would NOT have recovered the detail in the building.

Jim

[Evin Grant]

Shutter was 1/48th (180º)
Iris was T5.6 on a Cooke 25-250 T3.9 MII
ND .6 if I recall correctly.
Here is the Red Alert Grading...
http://www.4kninjas.net/pics/BrumbyRA.jpg

[Evin Grant]

Thanks, Jim.
This has been one of the most rewarding shoots I've done on my Red so far. I feel I've really able to flex a creative muscle that has been hard to move on all the bread and butter commercials and music videos I shoot. Hopefully we'll have a cut and finish by mid summer, there is a lot of effects work too.

[Jannard]

Looks like you are a "curve professional"!

Jim

[James T Mather]

I think that the second one just makes the camera look noisy.

You've clearly never shot anything on 3200 ASA on film then.

[Evin Grant]

Looks like you are a "curve professional"!

Jim

Thanks, again.
Most of that can be chalked up to 15 years of Photoshop. I started out in still photography. Curves and I have been close friends for quite a while now. That may be why some people seem to get it and other have a hard time understanding the grading process. Even though they are seasoned professionals they may have come from Film or HD video which both have their own completely different tool sets. Red's closest cousin is the DSLR and Adobe Camera Raw is the spitting image of Red Alert (or vice versa). I've been pushing the limits of curves and DR on DSLRs sine the Kodak DSC760c which I shot product photography with in 1998 right out of Brooks.

Here are some of the things curves can do in CS3...
http://farm1.static.flickr.com/134/318377489_78fe8268ac_o.jpg
http://farm1.static.flickr.com/125/317498708_8d67571938_o.jpg
http://farm1.static.flickr.com/129/317498784_bfa3d84a93_o.jpg
http://farm1.static.flickr.com/124/318196930_6746030cca_o.jpg

[Jannard]

I see you have pet zebras in your back yard...

Great stuff.

Jim

[Álex Montoya]

Great, great shot, Evin, but why are you opening the footage in REDalert! as ISO200 and no as ISO320?

AFAIK if you would open it as ISO500 you would loose highlight info but I thought at ISO 320 all the info was preserved...

[Uli Plank]

This is one of the most educational threads I've seen up to now on Reduser!

Plus, Evin, you have confirmed my attitude to those who ask me what you can get out of a Red before deciding to order one. I keep telling them "Get a Canon D-SLR, shoot RAW and play with Photoshop" (well, I suppose others will do, I just love my 10D).

Talking about noise: I've seen catching documentaries shot under available light with 16mm 500 ASA pushed by two stops. Grainy as hell, but if the story is really touching you, you don't care. It just needs to be organic noise, and this build 16 noise looks very organic to me – I hope it won't do strange things in motion. And Graeme and his team are really pushing the envelope here, since 9.5 stops under plus pushing 3 stops is bringing us down into the limits of 12 bit digitization, we're getting into quantization noise here, aren't we?

When talking about DR, we have to keep in mind that beauty is in the eye of the beholder, if it's film or electronics. Yes, film negative can take heaps of light, it just gets more and more flat in response getting ever closer to a flat line. Similar, but more critical due to grain in shadows, that's why you expose film to protect your shadows. But you decide at which point this curve becomes unusable when grading. For electronics, only saturation is clearly defined (so we expose for highlights), but when an image is lost in the mud of noise is again your decision, and the content dictates where you draw the line.

So, yes, that face woud be unacceptable for a cosmetics commercial. But if you shot her in a documentary, and you cought her when her face was telling the audience in this very moment what she went through, let's say in any kind of atrocity against her (or by her) humans are capable of? Wouldn't you use it as long as the noise isn't looking too synthetical?

Best regards,

Uli

[Evin Grant]

Great, great shot, Evin, but why are you opening the footage in REDalert! as ISO200 and no as ISO320?

AFAIK if you would open it as ISO500 you would loose highlight info but I thought at ISO 320 all the info was preserved...

Graeme may correct me on this but I didn't think there was any difference between setting ISO 320 and using the exposure slider to bring back a highlight or using the ISO directly. Either way the scene was still clipping at 320.

I don't ever think of the ISOs in the RA or RC as needing to be correlated to anything specific. I exposed at ISO 500 (Camera setting), chose consecutively lower ISO's in RA until the right side of the histogram pulled behind the wall, then used the exposure slider to fine tune the highest value to what looked right.
I then applied the curve using the shoulder to roll the highlights up smoothly, the knee to keep the skin tone correct and the toe to deepen the shadows without crushing them too much. A little DRX to make all the channels in the top highlight line up and voilla!

A finely baked R3D.

[Joe D'Arcy]

Great thread Jim and informative contributions by Red users. As Steve Sherrick, mentioned, 'Would you (Red team) be willing to put out a 'How to' video for post production red footage?

[Emanuel A.]

Shutter was 1/48th (180º)
Iris was T5.6 on a Cooke 25-250 T3.9 MII
ND .6 if I recall correctly.
Here is the Red Alert Grading...
http://www.4kninjas.net/pics/BrumbyRA.jpg


looks fine. looks good. looks professional recorded.


Looks like you are a "curve professional"!

Jim

L O L

[Emanuel A.]

Well done, Evin.


Thanks, again.
Most of that can be chalked up to 15 years of Photoshop. I started out in still photography. Curves and I have been close friends for quite a while now. That may be why some people seem to get it and other have a hard time understanding the grading process. Even though they are seasoned professionals they may have come from Film or HD video which both have their own completely different tool sets. Red's closest cousin is the DSLR and Adobe Camera Raw is the spitting image of Red Alert (or vice versa). I've been pushing the limits of curves and DR on DSLRs sine the Kodak DSC760c which I shot product photography with in 1998 right out of Brooks.

Here are some of the things curves can do in CS3...
http://farm1.static.flickr.com/134/318377489_78fe8268ac_o.jpg
http://farm1.static.flickr.com/125/317498708_8d67571938_o.jpg
http://farm1.static.flickr.com/129/317498784_bfa3d84a93_o.jpg
http://farm1.static.flickr.com/124/318196930_6746030cca_o.jpgAs usual.

[Alberto Caprioglio]

let's say in any kind of atrocity against her (or by her) humans are capable of

You must know her very well to know such secrets. I didn't realized there was that much information in the shadows.
I don't know how could one tell that, from that simple smile.
Was it the first, thought that came to your mind when you saw the young lady?
I can't believe the public is so automatic to link dark pixels on a girl to monstruous atrocity. Okay, okay, I'be quiet.
The superior world of Professionals...

[Graeme Nattress]

In RedAlert, altering ISO is the same as altering exposure.

Graeme

[Hrvoje Simic]

Not being a smartass here...but there's even more stuff there. Taken from the first dark example JPEG. With the 4K RAW/tiff route results could be even better. Also NOT meant to be usable.
Sorry Jim for fiddling with your pic.

[Andrew M.]

I don't understand one thing the way we measure DR in terms of stops on film.
9 stops will be the highest exposure value minus the lowest or it is how far we can push the whole image frame in terms of stops and still be happy with the highlights and shadows in the whole image?

[Uli Plank]

I don't know how could one tell that, from that simple smile.


Isn't it kind of sardonic ? Just kiddin'

If she'd be hit by beautiful lighting, would a client selling cosmetics buy it?

Regards,

Uli

[Graeme Nattress]

As far as I see it, lattitude is how many stops an over-exposed image can be pulled back and still have detail (not blown out) plus how far an under-exposed image can be pulled up and still not be too noisy. Dynamic range is therefore equal to the range of tones in the image plus it's lattitude. So if you're shooting a macbeth, who's greyscale is about a 5 stop range, and you can successfully put it 2.5 over and 2.5 under, you'd have a lattitude of 5 stops and a dynamic range of 10 stops. Or at least, that's the way I look at it.

With a digital system, you can choose where you put mid grey in code values. That chooses your over-exposure lattitude and under-exposure lattitude.

If you test with not the whole chart, but the single 18% grey chip, then practically your lattitude is the same as your dynamic range as you're now testing how far above and below a single tone can still be reproduced at.

That's partly why the subject is so confusing.... If David Mullen is reading this thread, I'd be keen to hear his take on this subject.

Graeme

[Andrew M.]

I understand the concept of the 18% grey chip DR test.

I like your explanation of DR using 5 stop macbeth.

I think macbeth based DR estimation will be more close to the real life type of scenarios.

Thanks! Graeme.

[Graeme Nattress]

Thanks Andrew. In both cases, the DR is the same, but the lattitude is different as the "object under test" has a different inherent range in it's tonal values.

Graeme

[James Brundige]

I'll post it in a new thread in a bit...

Jim

This is essential information. Some version could be put on the support pages, so it can be found without combing the user boards. Which leads to a couple more suggestions.

When the manpower is available (I know you're crunching on Build 16), please put more people on Red Cine. Because it can't export RD3 to HD broadcast, I'm stuck with Red Alert, which is much weaker in the grading tools.

And that would be a good opportunity to upgrade the video manuals on Red Cine. They cover only the basics. It would be great to touch some of the issues you are talking about here.

Your long hours and great work are much appreciated. I'm just asking for the tools to do what you suggest.

[Jason Sinclair]

It really is true what they say: Images speak louder than words...

[George A.]

Interesting. Looks like the new DNR sceme in Build 16 is effective. Wonder how it reacts to tungsten balance. So what was the over/under on that shot? I'm guessing around a Polanski (T11/16 split), maybe a touch more open.

Mitch,

What exactly is a "Polanski"? Any relation to the director?

[purefilm]

I would like to see a dedicated education page for users to discuss there methods and tips using the camera.

[Andrew M.]

Not being a smartass here...but there's even more stuff there. Taken from the first dark example JPEG. With the 4K RAW/tiff route results could be even better. Also NOT meant to be usable.
Sorry Jim for fiddling with your pic.

For those that do not understand how you can get the extra stops from scaling down (4K->2K)
and why we are getting extra full stop scaling by 1/2 from math point of view, here is the best example.

Look on this picture as an icon (small version at the bottom of the post) and full blown verssion by clicking on this icon and opening in the next page..
In the small version the noise level is not visible.
Click on it and you see the noise.

[Antoine Fabi]

Interesting. I'm pleased to see 9 1/2 stops difference between background and the subject, and still details in the shadows!

Jim,

I'd be curious to see a CC of this image in RedCine, but just with a tad less luma "lift" on her face. Maybe just in between image 1 and 2. I think it would be very usable...and if we could keep only the very low shadow values and "slightly" raise the shadows WHILE keeping the absolute black value. I'm curious if we can avoid banding while streching the shadows a little this way..

anyway..., very promising. Good job guys!

[Michael Lindsay]

Mitch,

What exactly is a "Polanski"? Any relation to the director?

yes... look at the split he refers to and why Polanski has not set foot in the USA for a while.

[David Battistella]

T32 -1.7
To pull out anything like that is to my calculations 9 stops under exposed and hold that much detail while the background stays almost the same. Damn. I have been to hundreds of telecine sessions and that would be a good grade in anyones book. Obviously this is showing a more docu situation but if you have some fill you are golden. This blows my socks off. I have seen Jim grade images before and I don't think most people understand what he knows... Curves. To anyone who thinks that the second image should be deleted I wonder how many times you forgot your light meter and needed to have a colorists save your .... project. If this was a docu and not a larger project with more money for fill etc. I would be really pleased.

I am with you on this.

Also. This is a great example of the open development. I like that we are seeing some build 16 images and that it is being testing under awful conditions vs. perfect lighting that we might see from film companies or others.

It's great to see the words 2.0 Beta in the RED ALERT grabs and now we know there are a few more tools for us to use.

Jim, Graeme and Co. have improved on the image and preserved a very important feature in RED FOOTAGE and that is the AESTHETIC of the image it produces which is far more CINEMA and much less DIGITAL than other offerings. Look at how those highlights roll off!

Thanks guys.

David

PS: The naysayers will always chirp, while people will just continue to SHOOT!

[David Battistella]

Not being a smartass here...but there's even more stuff there. Taken from the first dark example JPEG. With the 4K RAW/tiff route results could be even better. Also NOT meant to be usable.
Sorry Jim for fiddling with your pic.

This almost looks like Vision 320 stock!

d

[Radoslav Karapetkov]

Not being a smartass here...but there's even more stuff there. Taken from the first dark example JPEG. With the 4K RAW/tiff route results could be even better. Also NOT meant to be usable.
Sorry Jim for fiddling with your pic.


Wow.

Amen omen! :)

How much do you charge per hour?

[C.H.Haskell]

Thanks for sharing Jim, great example. The noise seems improved in 16, the grain seems more like film stock. cheers.

[Eryc Tramonn]

Maybe I should delete that 2nd one? It is not a useable grade. It is just hyper-extended to show the info is there. Or were you kidding?

Jim

Don't delete the second image, I (others) understand the point of what you're showing.

[Eryc Tramonn]

T32 -1.7
To pull out anything like that is to my calculations 9 stops under exposed and hold that much detail while the background stays almost the same. Damn. I have been to hundreds of telecine sessions and that would be a good grade in anyones book. Obviously this is showing a more docu situation but if you have some fill you are golden. This blows my socks off. I have seen Jim grade images before and I don't think most people understand what he knows... Curves. To anyone who thinks that the second image should be deleted I wonder how many times you forgot your light meter and needed to have a colorists save your .... project. If this was a docu and not a larger project with more money for fill etc. I would be really pleased.

Extremely well put. As a matter of fact, all the griping and such makes me wonder how many people on this board actually shoot. This is a no-brainer guys.

[Gavin Greenwalt]

You can take it really far in a few minutes if you're wiling to use a garbage matte and venture into uncanny valley. It just starts get pretty splotchy pretty quick. :D

http://i111.photobucket.com/albums/n134/im_thatoneguy/QuickGrade.jpg

[Jannard]

I guess it frustrates me a bit when people say we have 6-8 stops of latitude. But RED is a different animal. If you know a few secrets, the image is incredibly elastic. It is not film. It is RED. RAW is your friend.

Jim

[Eryc Tramonn]

Here is a frame from a short I shot and am beginning post on...
http://www.4kninjas.net/pics/BrumbyHDR.jpg
The top frame here is properly graded, the bottom pic is what I saw when opening the R3D. I exposed this scene at ISO 500 knowing that those building in the background were 4+ stops hot and I wanted the chance to recover them. Experience told me I could do that and still light for the face at ISO 500. All I had was two 1200W HMI's bounced into foam core and a 200W Joker for an eye light. There is some noise in the lowest tones but it's not distracting at all once they are dropped into their proper range. I think this shot is gorgeous! and I can't wait for B16 so I can push it even farther.

Yes sir, now you're talking. Excellent work. So what's your day rate? :biggrin:

[Eryc Tramonn]

Great example Evin... we are finding that way too many graders would NOT have recovered the detail in the building.

Jim

Well, Evin sounds like he truly understands how to expose an image for a given lighting condition. That first step makes it a tad easier to manage a scene such as this in post. If your background is super hot, and you know you don't have a way to physically knock it down on set...your other option is to reduce the contrast ratio between the subject and BG - which is what he did. He brought the subject up to a reasonable point, so a proper grade could be accomplished later.

As simple, and as difficult as that.

[Eryc Tramonn]

I've been pushing the limits of curves and DR on DSLRs sine the Kodak DSC760c which I shot product photography with in 1998 right out of Brooks.



Ahhh, so you went to Brooks. No wonder.

[Eryc Tramonn]

I guess it frustrates me a bit when people say we have 6-8 stops of latitude. But RED is a different animal. If you know a few secrets, the image is incredibly elastic. It is not film. It is RED. RAW is your friend.

Jim

Don't let it frustrate you...it seems to me people just need to USE the camera, read a few books on proper exposure technique, or as many have suggested - play with a DSLR, and then grade the images in Photoshop. That ability in and of itself is revolutionary. Jim has tried to make a professional grade camera accessible to the mainstream, by basing the fundamentals on techniques available to most of the population.

In school, I remember exposing Super 8 reversal and being frustrated with the results at first. Now, I understand why those lessons were so important. Talk about no latitude. If you can get that right, you can shoot damn near anything and get usable exposures.

[Gavin Greenwalt]

The great thing about RAW is that if something went wrong, it's pretty easy to figure out if it was user error or faulty data. No need to call a post house and ask them to see if they can't stretch it further you can just check yourself.

[Karl Roeder]

With all the talk about getting the exposure right, what was the exposure?
Reference T1.7 - 32. Are the highlights +2? +3?

[Antoine Baumann]

I guess it frustrates me a bit when people say we have 6-8 stops of latitude. But RED is a different animal. If you know a few secrets, the image is incredibly elastic. It is not film. It is RED. RAW is your friend.

Jim

Just shot a test yesterday, found 10.5 stops of latitude.

thanks,
antoine.

[Antoine Fabi]

I guess it frustrates me a bit when people say we have 6-8 stops of latitude. But RED is a different animal. If you know a few secrets, the image is incredibly elastic. It is not film. It is RED. RAW is your friend.

Jim

Yep,

There will always be some people to say that sort of things.

This camera is a monster, it's strong, it can save you and do wonders, but without proper care or basic knowledge, it can bite you.

This is a purely MANUAL camera, in the best possible way. Love it!!!!

Now please bring that @#@# beta build 16. :)

Cheers.

Antoine

[Álex Montoya]

It sure is an amazing tool.

[Tai Wah Lim]

In RedAlert, altering ISO is the same as altering exposure.

Graeme

Graeme, you lost me here. Evin - "I exposed at ISO 500 (Camera setting), chose consecutively lower ISO's in RA until the right side of the histogram pulled behind the wall, then used the exposure slider to fine tune the highest value to what looked right"

Exposure slider for fine tuning?

Lim

[Graeme Nattress]

Yes, it's for fine tuning.

Graeme

[Andrew M.]

Just shot a test yesterday, found 10.5 stops of latitude.

thanks,
antoine.

Now do some creative secondary color correction and you get extra stop.
It is what I see but it is by measuring absolute pixel values that may not be always related to the visual perception.

I have to do the film versus RED test, but when I will do it I will screw up the film test big time, since not much up to date experience here:-)

If David will make one together with Jim, I will know that they got the best of both.

[Radoslav Karapetkov]

Just a thought.

If one wants grain in his final picture, he could just play with ISO and processing and achieve a pleasant level of noise [+ (possibly) extra DR].

Without the need for additional grain, later in post.

Not much control, though. Of the noise structure, etc.

[Jannard]

Just shot a test yesterday, found 10.5 stops of latitude.

thanks,
antoine.

It doesn't surprise me... :-)

Jim

[number6]

There have been times over the past year when I thought "Boy, I wish I knew more about this dopey DOP stuff and this grippy and gaffy stuff... now, I think I may have been better off starting with a clean slate (read: blank mind). I am confident with what I've heard about Build 16 that I WILL be able to shoot passable footage, and WILL be able to grade that footage.

I go forth with no fear. WHERE'S MY STUFF!!!

[Jason Sinclair]

Is it fair enough to say that the DR if properly tested using this camera must be taken into account using the full process of the raw workflow and some dp's have only been testing using the image directly out of the camera, and therfore all testing must take into account the processsing (just as it does in film) if it is to be representative in a reality workflow? Or am i just on a different planet?

[Chris Nuzzaco]

I'm anticipating/hoping for 11.something stops when I get under the hood tomorrow. As for how much is usable, thats debatable, especially given that this camera's compression is improving more and more.

By the way, I finally jumped into Red Alert! for a more critical analysis of some Red footage. I'm feeling much more oriented now that I've been able to get a good look at some files set to Linear Light with the DRX code turned off, as well as the matrix and curves disabled. OH... and yes, I dialed in ISO 100 too..

Everything makes so much more sense now, but I have to say, its hard to explain it. Seems very similar to Andromeda with regard to how the histogram looks and where channels clip, etc...

[Lachlan Ward]

This stuff Evin. Good Job. Good luck with the film.

[Jannard]

I'm anticipating/hoping for 11.something stops when I get under the hood tomorrow. As for how much is usable, thats debatable, especially given that this camera's compression is improving more and more.

By the way, I finally jumped into Red Alert! for a more critical analysis of some Red footage. I'm feeling much more oriented now that I've been able to get a good look at some files set to Linear Light with the DRX code turned off, as well as the matrix and curves disabled. OH... and yes, I dialed in ISO 100 too..

Everything makes so much more sense now, but I have to say, its hard to explain it. Seems very similar to Andromeda with regard to how the histogram looks and where channels clip, etc...

You may be a good card player, but don't expect the "river" to be kind to you if you expose at ISO 100 and expect 11 stops.

Jim

[Shawn Nelson]

Im still so confused. Back on Red day in August of 07 there was "10-1/2" stops of latitude, then there were further upgrades (such as the 'daughter board upgrade') that were supposed to noticeably improve DR further, so as far back as build 11 we should comfortably be over 11 stops DR. So why the heck do I keep hearing 9'ish now? Heck, it shouldnt even be at 10. If Build 16 is a massive improvement over a previous 11 stops then I'd guess a solid 12 a minimum. Or am I missing something? Not trolling here, I'm genuinely confused.

EDIT: Is this simply a difference of Red saying the noise in the lower two stops are acceptable to them (thus the Red Build 15 11+ stops) vs others saying only 8/9 because those lower two stops are too noisy? And then Build 16 will make the lower two stops "prettier" and thus hopefully get everyone else to agree with Red's existing definition of 11+ stops?

[Chris Nuzzaco]

You may be a good card player, but don't expect the "river" to be kind to you if you expose at ISO 100 and expect 11 stops.


I'm not really exposing at ISO 100, I'm simply using that setting to process at in Red Alert! When I'm done, I'll post some results, and hopefully some video showing the process I use to custom ISO rate cameras. I'm fully aware that my methods are somewhat "unorthodox" and even hard to explain at times, but I think a little video will make everything more easy to understand. There's lots of ways to achieve one result.

[reality]

EDIT: Is this simply a difference of Red saying the noise in the lower two stops are acceptable to them (thus the Red Build 15 11+ stops) vs others saying only 8/9 because those lower two stops are too noisy? And then Build 16 will make the lower two stops "prettier" and thus hopefully get everyone else to agree with Red's existing definition of 11+ stops?

It's people arguing about the difference between dynamic range which is measurable and latitude which is purely subjective. One persons 8 stops depends on his/her tolerance for noise and might be completely different from another persons' 5 stops or 11 stops. It's the same with film. Someone who wants butter smooth with very little grain is going to underrate the film and be much less tolerant than someone who loves grain.

[Esteban Sosnitsky]

Jim, when I saw those jpgs I felt like when you come into the transfer room, sit in front of the telecine and start color correcting your images. Your first look to the images is what correct exposure was at a set ISO. But then the images come to life with contrast and color. I love how this camera is so film based!!! But I think the problem is that besides exposure, camera owners have now to deal with grading and color correcting with a film based environment. And not everybody knows how to do that or what to expect from a RAW image.

Still, I wonder what would happen if you put a step wedge with 12 or more steps/stops evenly lit in front of the camera with a waveform monitor rating the camera at different ISOs? I just wish I had my camera so I could do this tests :P

I saw at nab this new chart that the Japanese are using to test their sensors, looked very interesting.

Still patiently waiting...

[Esteban Sosnitsky]

It's people arguing about the difference between dynamic range which is measurable and latitude which is purely subjective. One persons 8 stops depends on his/her tolerance for noise and might be completely different from another persons' 5 stops or 11 stops. It's the same with film. Someone who wants butter smooth with very little grain is going to underrate the film and be much less tolerant than someone who loves grain.

I dont think latitude is subjective!!!
Its a term referred to the log curve in which film chemically responds. Its measurable!!!!

[Jannard]

Im still so confused. Back on Red day in August of 07 there was "10-1/2" stops of latitude, then there were further upgrades (such as the 'daughter board upgrade') that were supposed to noticeably improve DR further, so as far back as build 11 we should comfortably be over 11 stops DR. So why the heck do I keep hearing 9'ish now? Heck, it shouldnt even be at 10. If Build 16 is a massive improvement over a previous 11 stops then I'd guess a solid 12 a minimum. Or am I missing something? Not trolling here, I'm genuinely confused.

EDIT: Is this simply a difference of Red saying the noise in the lower two stops are acceptable to them (thus the Red Build 15 11+ stops) vs others saying only 8/9 because those lower two stops are too noisy? And then Build 16 will make the lower two stops "prettier" and thus hopefully get everyone else to agree with Red's existing definition of 11+ stops?

There is a technical measurement for DNR (Graeme can explain better than I can). Basically, if you can tell differentiation it counts. But the noise levels of the science test are not likely to be acceptable to a critic. Hence the difference in numbers.

My 9.5 stop test was not designed to show maximum capability. It was a 5 minute test (including grading) to dispel the "I think RED is really 6-8 stops" comments I was hearing. It was also to point out that grading plays a huge part in the results people are getting.

Someone just posted that they measured 10.5 stops. I assume that is useable.

Jim

[Jan Reiff]

our tests with red earlier this year result an average of 9.5 stops, so we are on the same point.

[Radoslav Karapetkov]

We need a few really IN-DEPTH sticky threads on these matters... or a much more detailed pdf manual.

This cannot be a knowledge scattered among 1000 threads, all around Reduser.

[Kolja1981]

We get back the details captured in the shadow area of the picture. But what about the details in the highlights of your picture. If you overexpose a shot , you won't get back most details in the highlights. You agree with me ?

[Antoine Baumann]

Someone just posted that they measured 10.5 stops. I assume that is useable.

Jim


Yes it is useable, however it was a DR test, cloudy day exterior without additional ligtht. The shadows were at T0.5 1/2, the bright side of the face at T1.4 and the "super" highlights at T22.


Don't have the image right here but I can post it here if people would like. Anyway it will be in my personal white paper on red
(if build 16 doesn't come out before I finish the white paper ;-)

ciao,
antoine.

[Fredrik Callinggard]

There is a technical measurement for DNR (Graeme can explain better than I can). Basically, if you can tell differentiation it counts. But the noise levels of the science test are not likely to be acceptable to a critic. Hence the difference in numbers.

My 9.5 stop test was not designed to show maximum capability. It was a 5 minute test (including grading) to dispel the "I think RED is really 6-8 stops" comments I was hearing. It was also to point out that grading plays a huge part in the results people are getting.

Someone just posted that they measured 10.5 stops. I assume that is useable.

Jim

Actually Jim most of us are saying 8 stops as workable stops. I also said it as a preference to acceptance on build 15. Secondly I said that it looked very nice (in build 16) with the "noise" as you lifted the shadows on her face and that it seems to be an improvement. To be honest your test shows exactly what I mean. Don't get me wrong I love your shot and I would shot it exactly the same but what I'm talking about is to be able to manage that shadow. In build 15 it would have to stay like that (or maybe be slightly darker), In my opinion there would be nothing there to "lift" it would break immediately. Now in 16 I see there's that possibility, though it still is noisy as soon as you do it (which is not wrong, it would most probably be on film as well).

I don't see how we're miles apart when you're talking 9.5 and I'm talking 8. All I'm saying is that 8 stops is what should be considered for you as workable stops (in build 15), if you ask for our opinion. We can of course agree that in ultimate condition you can find more range, but most of the times it wouldn't be acceptable to us - I'm now also talking grading. I would not lift "her" in grading because it would not look acceptable to me or my clients (and I believe your pic shows that). I would though shoot it like that, if I knew that it was exactly how I wanted it to look and believe me that I would and I have. During my DP years I was banned at least ones from every single mayor record company and a few agencies, because I shot it the way it should be and they felt it was to dark and couldn't do much about it. That it generally led to be some of their most successful music videos or ads didn't matter).

As for you others who are saying great shot all it needs is a little fill are kind of saying what I'm talking about.

I was just not agreeing with you Jim that in build 15 you could grade her face to "our" acceptance (I believe she would look one stop darker (in build 15) and therefor it wouldn't work - this is of course purely subjective and impossible to prove). Maybe in 16 and that is perfect. So this is just what we want to see and hear.

So Jim stop being grumpy with us. We're not trying to put you down. I love RED. I've been shooting film for 15 years. In the start of my career I made myself a promise, that I would rather starve then touch video. I kept that promise until know (I know RED is digital, not video, so it effectively doesn't count but still it sounds good :innocent:). The RED camera does deliver brilliant pictures and shot right they're amazing. I know since I have done 10 shoots with it all over the world and they all have turned out great. It's impressive but it's not without flaws.

So you can either see our opinion as a good tip to other RED users. Or keep feeling we're ruining your mojo. Personally I'm not trying to talk down on the camera, on the contrary, I'm one of the huge fans, selling it out there all the time. Why? Well it's not to make money. It's because it delivers freedom. I can use it more freely than with film and that is a great advantage.

I think that you guys need to read between the lines from some posts. There's a lot of fanboys out there with both feet on the ground and we want to discuss the flaws of the camera out of :love:

PS I'll ask the "pro's" if they can have a grading class with you after they've finished the DI on a RED feature :)


Fredrik Callinggard

[Álex Montoya]

Well, Fredrik, what we mean is that when people claim 9 stops in the EX1, they are not applying the clean, completely usable stops concept you are using for the RED.

What Jim tries to show is that the RED has a one or two more stops than most of them.

[Fredrik Callinggard]

Well, Fredrik, what we mean is that when people claim 9 stops in the EX1, they are not applying the clean, completely usable stops concept you are using for the RED.

What Jim tries to show is that the RED has a one or two more stops than most of them.

I understand that and I would say that it looks like 9.5 stops to me from Jim's little test, but that was build 16. I was not agreeing with him that I believe he would have got the same result in 15. He even said it himself in this thread, "you could grade it in 15 but not with the same result".

It's always been clear that most of us that are talking about a lower DR are talking about actually useable (when it breaks it breaks and we wouldn't count that as a DR stop). We've been saying this because most discussions has always started with someone asking how they should interpreted RED's DR - that is a subjective discussion, not a scientific.

You can go out there shoot 8000 ASA for all I care, but if someone asks me how I think you should rate build 15. I'll say 320 ASA and consider 8 stops (most of the time I say between 8-9, maybe 9.5 stops, it just happened to become 8 only,for some reason, in this thread. I would never say 11 though). Why? Well because there's times where the camera will act "poorer" then other times, such as tungsten, or real low light environments with "tinted" shadows etc. Understand that most of the time it's not necessarily noise that is the problem. It's compression artifacts in the noise (especially in the blue channel), making it hard to fix in post, leaving you with crushing the blacks as the only solution.

I've also never had a problem with the clipping looking like video, which either leaves me to be very lucky or I do know how to shoot and grade? :innocent:

Yes it might be a safety measurement but I also think that most the people talking about 8-9.5 stops (NOT 6-8. Jim is referring to the extremes there) are coming from film background. We might be spoiled yes, but I do believe that when all of us saying this, we put it on a general level not when it's ultimate.

Furthermore us from film background would definitely not rate EX1 with 9 stops :usd:. So maybe there's a cultural clash here?

All in all I guess what I'm saying is that the people saying have 8-9.5 stops in mind (I say around 8) are only trying to be helpful. Saying that the other stops are only wishful thinking, yes you can find them if you want but it's very very likely that they won't look good.

So yes if you really want to say it, I guess that RED can be regarded to have more DR, if that is important to you. I don't on the other hand recommend it.

It's also important to say that I'm talking about build 15 and not 16. So that it has improved with build 16 pleases me as much as it does with anybody else.


Fredrik Callinggard

[Graeme Nattress]

First, are we talking dynamic range, the difference between the brightest and darkest thing you can record and still retain detail? Or are we talking lattitude which, to me, is by how much you can under or over expose an object and not clip or crush it.

As Jim points out, on a test chart, I can measure (ie see the "detail" over a range of 11.3 stops). Now, are all those stops equal "quality". No they are not. And they're not equal quality on any camera, or film for that matter. As they get darker, they get noisier. Film complicates things with a curve that actively squashes highlights and shadows. What we record is the linear (in the true, straight line sense of the word) light data from the sensor.

Next, how do you measure the light hitting the sensor? Sometimes tests are performed by using the stops on the lens, and ND, to produce an exposure ramp. This is an exercise fraught with the possibilities of human error and innacuracies in the lens markings, how closely and repeatably you can set them, and in the ND. These all add up.

A light meter reading is good, because then you can measure the range in a single shot. It can also be a real world shot - not artificial. This is what Jim did and illustrated well. Jim showed a subject 9.5 stops below non-clipping highlights. That means the subject was under-exposed by at least 9.5 stops, probably more. The subject itself has a dynamic range. We can therefore see that the range the camera recorded must be greater than 9.5 stops.

Finally, for repeatably testing, we use a 41 step Stouffer wedge. This is great as it can produce a repeatable, if boring scene of nearly 14 stops of dynamic range with 1/3 stop steps, numbered for ease of use. This will tell you a lot, and it's accurate, but it's not a real world scene.

Graeme

[Fredrik Callinggard]

That means the subject was under-exposed by at least 9.5 stops, probably more. The subject itself has a dynamic range. We can therefore see that the range the camera recorded must be greater than 9.5 stops.

You're right Graeme I should most probably talk about it as latitude more than DR. If I do I definitely stand with 8 stops (build 15), my apologies.

As for the underexposure. I don't think that Jims stop was 32+.



Fredrik

[Graeme Nattress]

8 stops lattitude, ie, 4 over and 4 under, for a real world object, is quite reasonable. 8 stops dynamic range is not.

I've seen B13 shots that captured a full 12 stop range, with detail all the way into the shadows and no clipping. If you keep the shadows dark, and in the shadows and don't try to brighten them to daylight, they hold up remarkably well. I don't think it's reasonable to expect to take some object that is 10+ stops under-exposed (compared to clipping) and try to put them at the level of mid grey. I don't know any imaging device that would successfully do that.

Graeme

[Chris Nuzzaco]

I've seen B13 shots that captured a full 12 stop range, with detail all the way into the shadows and no clipping. If you keep the shadows dark, and in the shadows and don't try to brighten them to daylight, they hold up remarkably well.

Great point Graeme. I think way too many people are looking at it in the opposite manner and trying to take shadows and make them mid tones and highlights. Thats just a bad idea, and with that, I'm off to my first Red shoot.

:)

[Brian Harbauer]

Good luck Chris! Loved reading what you had to say over on Reel-stream. Good to see you over here. Looking forward to seeing some of your footage on Red!

-Brian

[Antoine Fabi]

Great point Graeme. I think way too many people are looking at it in the opposite manner and trying to take shadows and make them mid tones and highlights. Thats just a bad idea, and with that, I'm off to my first Red shoot.

:)

Exactly !!!
With digital, we try to believe ourselves in that everything is possible in post...
Wrong, post work can't change the nature of things. Sure it can improve, recover what is there, enhance, adding a mood, a texture, but certainly not changing everything.


Even a 12 bit file will crash with too much manipulations. ...and the same happens will film.
I remember when i did photo development, it took me a lot of time to get reasonable results (flashing etc...)....just... well... reasonable...

...and now we want to dramatically modify a single portion (shadows) of the image...it just won't "stay together".

I made some tests with the RED in a sunny day in a park with a bright white shirt, and i can assure you that i perfectly see the frabrics details AND white buttons AND the shadows under the trees on the same frame.

That says a lot about this camera.

But it's not a free ride.

[Radoslav Karapetkov]

I have another crazy idea which I won't hesitate to share :).

Can you grade different parts of a given shot in a different way.

Say, you motion track the face of the subject and apply a certaing CC just to the face, by actually isolating it from the rest of the image.

Then you "isolate" the background and work on it applying different kind of CC.

And the same with whatever element you want to change.

I guess this will be tricky and time-consuming, and if there's a lot of movement in the shot, it probably won't work.

But, what do you think?

Does this make any sense, or will it result in a Frankenstein of an image. :)

[smelni]

this is common and called secondary corrections

[Shawn Nelson]

It was a 5 minute test (including grading) to dispel the "I think RED is really 6-8 stops" comments I was hearing.

I hadn't heard "6-8" but I've been hearing an aweful lot of "8/9 stops" which confused me repeatedly. I guess those people just have different standards. Is there no solid benchmark for noise? Such as, if Red footage was compared to well exposed Vision 2 5218 and the noise levels made comparable, how many stops does each have relative to each other?

[Graeme Nattress]

Yes, secondary colour correction is something that you have to be good and careful with to get the best out of RED footage. Same goes for any high dynamic range source.

Graeme

[Gunleik Groven]

I have another crazy idea which I won't hesitate to share :).

Can you grade different parts of a given shot in a different way.

Say, you motion track the face of the subject and apply a certaing CC just to the face, by actually isolating it from the rest of the image.

Then you "isolate" the background and work on it applying different kind of CC.

And the same with whatever element you want to change.

I guess this will be tricky and time-consuming, and if there's a lot of movement in the shot, it probably won't work.

But, what do you think?

Does this make any sense, or will it result in a Frankenstein of an image. :)


... try Color...

[jbeale]

Is there no solid benchmark for noise? Such as, if Red footage was compared to well exposed Vision 2 5218 and the noise levels made comparable, how many stops does each have relative to each other?

There are visual noise measurement standards, but in my opinion there is no solid benchmark for noise. Various film stocks and video sensors each have unique qualities, so the noise texture looks and feels and moves differently, even if you match up measurements like RMS and peak noise amplitude.

[Damon Meledones]

With all this talk about "latitude" and "dynamic range" I have to say that I subscribe to the school of thought that there is no such thing as "exposure latitude."

If you fail to place a scene brightness value on the appropriate point of your capture medium's curve, you may still be able to recover that brightness value, but you have sacrificed values at the edge of your scene. If the contrast ratio of your scene is lower than that of your capture medium, and the medium is linear (as is the case with digital), there may be no ill effects, but the real world is rarely so flat.

With film there is often a consistent relationship to brightness values and curve points, dictated by the nature of its response curve, and processing and printing techniques. In our digital RAW world, we have the option of push or pull processing every take, and the approach is much similar to that of the Zone System, where the "correct" exposure is one that utilizes, as much is as possible, the full range of the sensor. That's not to say that the tradition approach of matching scene values to fixed curve points is "wrong" in a digital RAW work flow, simply that it limits, to some extent, your processing options.

[Antoine Fabi]

...and what they do with film in a scene where the subjects are in the shadows ?

They expose for highlights and:
1) they use reflectors or daylight spots
or

2) at least they use a side lighting to "pop" the edges.

..to bring some of the shadows within the recordable lower range of the media.

...so in the end...just to "believe" the scene is real.
You dont see the reflecting light, but it is essential in many occasions.

This seems to be true for any media.

this is a fantastic subject isn't it ?

[arun]

Red red red i know this is not the place to ask this but jimm can we expect one more miracle from u?..............................................RE D PHONE......
with 1080p resolution... lol i think am dreaming little more pardon me..lol have a nice time u Guys ......

[Hrvoje Simic]

I guess this will be tricky and time-consuming, and if there's a lot of movement in the shot, it probably won't work.


It works but it's tricky and time consuming. :usd:



Does this make any sense, or will it result in a Frankenstein of an image. :)

Depends on how far you push it.

[Radoslav Karapetkov]

Thanks. :)

I'm a noob [in CC].

:).

[Andrew M.]

Karapetkov, I have seen clips of Gunleik and asked him once, how he did get all these nice colors on his shots.
He didn't answer, or he didn't see my question. Well I use secondary color correction on some HDR material and I just got the right one for moving pictures and voila, got 75% where Gunleik is.
It took me one month to get to 75% there and probably another 3 years to get where he is now, or never since he will be much further then:-)

Good secondary CC will let you lock on the color in small bracket and on the zone with ability to track the zone.
Results, ....... priceless.

[Jason Sinclair]

Red red red i know this is not the place to ask this but jimm can we expect one more miracle from u?..............................................RE D PHONE......
with 1080p resolution... lol i think am dreaming little more pardon me..lol have a nice time u Guys ......

Please this is a serious forum.... and besides it has already been done:

http://www.youtube.com/watch?v=TYIOIM6hHBk

[el_stupido]

Maybe I should delete that 2nd one? It is not a useable grade. It is just hyper-extended to show the info is there. Or were you kidding?

Jim

Maybe you could show another inbetween image that shows an acceptable lift of the foreground subject?

I just showed people at my work this and they got pretty excited.

El Stupido

[el_stupido]

I'd like to know as much as I can about grading this footage. Sign me up.

IBloom

yee haw do a mythbusters episode. how good are those little vids?

El Stupes

[Dan Hudgins]

This seems to be true for any media.



Although one can "help" make the contrast and color higher in flat parts of the image that is within the range of the sensor, having a long scale sensor will not "fix" un-dramatic lighting.

IMHO it is better to make it look good in the viewfinder by using lighting to bring the subject contrast up and to hold it in the center of the tonal range since, even if the camera is 12 bits, the TV sets being used to watch the result on will be between 8 to 10 bits or less in effective contrast range. You need to light and expose with the whole image chain in mind from subject to end projection or viewing device. Best to shoot for the lowest comman denominator since if you don't you will need to tie your shoes...

[Jason Wingrove]

Also it's not right to judge the noise on a still image. On the move the noise will seem a little less anyway!

Obviously keep the shots up purely to show the amazing amount of information in there.
Jeez tough crowd in here Jim! :sarcasm:

[Uli Plank]

Noise can be less to the eye on moving images IF it's completely random. If not, it can be more annoying. That's where I hope that build 16 will show improvements.

Regards,

Uli

[number6]

Best to shoot for the lowest comman denominator since if you don't you will need to tie your shoes...

DANCAD3D, even if anyone asks, please don't explain that statement... I need something to reflect on:usd:

[Florian Stadler]

First, are we talking dynamic range, the difference between the brightest and darkest thing you can record and still retain detail? Or are we talking lattitude which, to me, is by how much you can under or over expose an object and not clip or crush it.

8 stops lattitude, ie, 4 over and 4 under, for a real world object, is quite reasonable. 8 stops dynamic range is not.

Graeme

Graeme, I am a little confused.

First I see: finally someone tries to define the difference between dynamic range and lattitude. The source of many misunderstandings.

Then you go on to make a totally unreasonable claim in your next post.

Are you claiming that you can pull up a normal subject that has been underexposed by 4 stops and pull down a subject that has been overexposed by 4 stops also? I don't think so. 1 1/3 underexposed and 1/2 stop overexposed is more like it imho.

Let's make the definitions again to clarify:

-Dynamic Range is the mediums ability to capture differences between lightest and darkest object within the frame and retain detail.
-Lattitude is the mediums ability to be under/overexposed and brought back to normal in printing or digital post manipulation and retain detail.
-Both of these measurements are subjective: How much grain/noise is acceptable to you and what do you consider retaining detail for the specific look you are going for.

A lot of times - as with Stouffer charts - the combination of Dynamic Range and Latitude are measured. This is mainly for marketing purposes and has little real life value to someone creating an image.

Where is the discrepancy between our definitions?

[Graeme Nattress]

Let's take a real world example of a macbeth chart. The greyscale on that is say a 5 stop range, so we have an object under test that has it's own DR of 5 stops. Say our camera, for instance (not being specific to any camera here) is good for a 10 stop dynamic range. Therefore, your exposure lattitude for that chart on that camera would be +-2.5 stops. Or 5 stops of lattitude in total. You may even be able to get more on the bottom end with the expense of some noise. The most exposed image without clipping will have the best image though.

Graeme

[Florian Stadler]

Let's take a real world example of a macbeth chart. The greyscale on that is say a 5 stop range,
Graeme

Ah that's where we are off. I consider a real life subject with a "normal" tonal range to be about in the seven stop range...

[Graeme Nattress]

A scene will surely have >7 stop range, but objects within that scene will usually be less. That's why I was specifically talking about exposing an object, rather than exposing a scene. Moving to a specific object like a Macbeth with known dynamic range is helpful here.

But a scene could have still a greater dynamic range. At that point, your lattitude is close to, if not zero as there would be only one perfect exposure, if you're lucky, that can take in the entire scene.

[Dj Joofa]

Ah that's where we are off. I consider a real life subject with a "normal" tonal range to be about in the seven stop range...

7 or 5 is immaterial, basically what Graeme is saying is that:

Dynamic range (on camera, in stops) = number of stops from pitch black to the darkest tone in the scene + number of stops for the scene + number of stops from the highest highlight in the scene to the max the sensor can have.

Therefore,

tonal range (scene) = number of stops for the scene = number of stops between highest highlight in the scene and darkest tone in the scene.

Latitude = number of stops from pitch black to the darkest tone in the scene + number of stops from the highest highlight in the scene to the max the sensor can have.

So, if latitude > 0 (i.e., the tonal range in the scene is less than the full dynamic range on the camera), then you have that wiggle room to move your tonal range up and down the dynamic range as a slider, with some consequences, of course.

[Graeme Nattress]

Precisely! Latitude is totally scene / object dependent, as well as dependent on your noise threshold / preference.

Graeme

[Florian Stadler]

Joofa,

the aspect you are not taking into account in your explanation is the display medium. All this extra information you have at your disposal has to be squeezed into a bottle like a genie before it gets to the viewer (color correction!)...

With film your print could replicate a 7-8 stop range linear. It was a known quantity. For TV broadcast a 7 stop range would be more normal. That's why I "randomly" picked that 7 stop range and why it matters...

That's why some people say there's 7-8 stops of DR plus and 2-3 stops of latitude while others define it as 9.5-11 stops of "dynamorangeitude":nerd: (dynamic range & latitude).

It's a matter of definition. I think we're all talking about the same thing and all I'm trying to do is define what we're talking about so that ultimately we would understand each other.

[Graeme Nattress]

Usually, for a grade to look "good", on an average display you need to squash the range to nearer 5 or 6 stops though. However, this is in no way related to what or how you shoot, because if the range you shoot is much greater than those 5, 6 or 7 (whatever) stops, you have to get into secondaries to correct it so all the information you want the viewer to see lies within that limited range. I don't see how that relates to either dynamic range or latitude of the capture device, just that you're going to have to do some grading afterwards for best results.

Graeme

[Dj Joofa]

Florianstadler,

I understand what you are saying. I think partly the confusion is arising that the word "dynamic range" is being used for both:

(1) Full range of the sensor i.e., 12 bits
(2) the tonal range of the objects in the scene, that may or not may not occupy the full 12 bit range. If it does not then the difference is latitude.

We need to stick with one meaning of DR. However, there is no clear consensus that should DR mean (1) or (2) as far as definition is concerned. Though, as mostly happens, one typically knows by context that by DR one is referring to (1) or (2), and then some of confusion gets eradicated.

[Florian Stadler]

I don't see how that relates to either dynamic range or latitude of the capture device

Graeme

It relates to the testing you do before you start a film where you test your entire IMAGE CHAIN in order to come up with the best way to expose and light your Negative (be it digital or not) to achieve the look you are after.

[Uli Plank]

With film your print could replicate a 7-8 stop range linear. It was a known quantity. For TV broadcast a 7 stop range would be more normal. That's why I "randomly" picked that 7 stop range and why it matters...

A perfect film in an average cinema (with 'Exit' lamps and enough diffuse lighting to identify your popcorn) can reproduce about 10 stops at max.

Some home theaters could reproduce more than 14 stops these days, if you paint the walls black, have black velvet seating and wear something black while enjoying images graded for a JVC DLA-HD100.

Hard to find a camera with a 14 stops sensor…

A TV in an ordinary living room (not a Pioneer plasma in the home theater described above) will be proud to reproduce 5-6 stops!

A scene can have any contrast from 2 stops in dense fog to nearly infinite when filming straight into sunset/sunrise.

The art of shooting plus the art of grading – hopefully two versions, one for cinema and one for TV – is finding the best compromise to make any of those scenes look convincing for the audience.

Regards,

Uli

[BradWright]

Precisely! Latitude is totally scene / object dependent, as well as dependent on your noise threshold / preference.

Graeme

Hi G:

Instead of just saying dynamic range, what about Recordable Dynamic Variation. This would take into account that Silcon is much more linear than film. A recordable change in the level grey for the medium would be a Recordable Dynamic Variation. For example, film doesn't have shaddow detail while silicon does. Maximum Dynamic Range would be better fitted to film. Here's an example of what it could look like:

Red:
RDV 12 bit (12 stops)
MDR 12 bit (12 stops)
Film:
RDV 10 bit (10 stops)
MDR 16 bit (16 stops)

[Graeme Nattress]

I think that might add a whole new layer of confusion. I know what you mean though....

Graeme

[BradWright]

To Be less confusing, it could be simplified into a ratio. For each shade of grey, it represents how much on average is recorded.

MDR/RDV= MLR

Media Linear Ratio:

Red: 1.0

Film: .625