Dynamic Range

[Jan Tore Soerensen]

It recently came to my attention that most panels today can't display more than 10-11 stops of dynamic range. I read this in a blog post, so I don't know if it's factual.

 

How can we then see a huge difference in retained high- and lowlight if the monitor display isn't supposed to be able to show it? We can obviously see a huge difference between a 5d2 and an URSA Mini.

 

Can anyone elaborate on this?

 

Thanks!

[Tyler Purcell]

I struggle with this quite a bit when coloring because you have to use a program like DaVinci and have a 10 or 12 bit I/O card + 10 bit display with 10 bit electronics, to even see much past 10 stops.

 

A lot of monitors say 10 bit color, but then you research and they have an 8 bit display.

 

So it's less about dynamic range, it's more about color depth then anything else.

 

Then, where do you even get a 10 bit source? Most cameras don't even shoot 10 bit natively.

[Albion Hockney]

The range is compressed on screens. A camera can film 14-15 stops of range and then when its put it a color space like Rec709 that latitude is compressed but still present. If the camera couldn't capture the whole range though it will show as blown out or as solid black.

 

I'm sure it's more complicated, but that is the basic idea.

[Bruce Greene]

The display system doesn't need to match the dynamic range of the camera. If you display all 14 stops from the camera, in a linear fashion on the display you will see all the data, but in a lower contrast version.

 

The job of the colorist is to interpret these 14 stops into the range of the display to make something that looks photographic and serves the project. This is usually done by compressing the highlights and shadows (low contrast) and increasing the midtones contrast to something that more resembles the scene.

 

I hope this makes some sense :)

[Robin R Probyn]

There are alot of cameras that can record 10 bit natively aren't there.. even an fs5.. ? AFAIK.. all TV,s are rec709. although HDR sets are around the corner and REC2020... I think the point to shoot log is to get that DR in capture .. then at least you have that data out the window in a shot.. and it can be all massaged back down to REC709.. rather than just shot REC709 gamma and only have 7 stops DR in the first place .. the problem is a very old tv broadcasting standard .. REC709.. forget 3D and 8K the best thing that can happen with TV,s is HDR REC2020..

[Max Field]

When I first got in this Sony F900 I had the same problem you're having (very concerned with clipped highlights).

Setting the monitor to false color is a savior. My Atomos Samurai Blade does it, not sure if others do.

[Jan Tore Soerensen]

Aha! So we are capturing a higher dynamic range, but that DR isnt preserved all the way through to the final deliver. That makes sense.

 

So a camera that is able to capture, lets say 9 stops, will end up at maybe 6-7 after a light grade. And a 15 stop camera, would end up at around 10, and retain more detail in the lift and gain?

 

Thanks for all the input!

[Robin R Probyn]

Here is an article by our own Sir Philp Rhodes OBE.. & garter.. I hope he doesn't mind, and that I don't find myself in front of a QC at the Old Bailey ..

 

http://www.redsharknews.com/technology/item/3289-what-exactly-is-rec-709

[David Mullen ASC]

Unless you shot slide film, traditionally the contrast of a print off of a negative was higher than the original and held fewer stops of information, which is why many of us when shooting for print roughly worked within a 10-to-11 stop range even though the negative held 14-stops of information. This was because a print had to have a lot of contrast in order to retain rich blacks when projected onto a white screen. To some degree, this has some advantage in allowing us some flexibility in making a print brighter or darker, and later with digital color-correction, to bring out information in select areas of the shadows or highlights.

 

Not sure now with HDR, where all 15-stops of information from a high-end digital camera like an Alexa, gets displayed an HDR monitor, whether that means it is more like shooting with slide film. I guess you can always throw out information, crush the shadows or clip the highlights, for creative reasons but it will look more "obvious" or less natural now on an HDR display, more stylized.

[Robin R Probyn]

From what I've read.. its OLED technology that have allowed HDR TV REC2020 sets to evolve.. but for viewing comfort (not too bright) they are more around 10-11 stops DR not 15.. so anything shot Slog/LogC/ Canon Clog2 (think thats whats its called.. anyway the higher DR Canon Log than before.. or even some of the hypergama curves from Sony or equivalent from other camera,s, will be all set for HDR display.. in theory this will be a much bigger actual visual difference to the ordinary punter.. than 4K ever was.. I think they will push it hard soon to revive flagging TV sales..

[Phil Rhodes]

It's actually quite difficult to make OLEDs bright enough for anything, let alone HDR.

 

There are actually two HDR standards for home TVs, one of which defines a very low black level and a moderate white level, which is intended for OLEDs, and one which defines a moderate black level and a rather high white level, which is clearly intended for LCDs.

 

More or less the only display in the world which can do both the very low black level and the very high white level is something like a Sony BVM-X300, which is ferociously expensive and I have my concerns over the longevity of its blue channel.

 

P

[Tyler Purcell]

Aha! So we are capturing a higher dynamic range, but that DR isnt preserved all the way through to the final deliver. That makes sense.

 

So a camera that is able to capture, lets say 9 stops, will end up at maybe 6-7 after a light grade. And a 15 stop camera, would end up at around 10, and retain more detail in the lift and gain?

 

Thanks for all the input!

 

Forget about the camera, how is the signal being presented, that's the #1 question to ask.

 

Always look backwards, not forwards. Think about your audience, how are THEY getting your signal. Are they streaming an 8 bit 4:2:0 signal from the internet or on BluRay? Are they watching it in a movie theater with a DCI-P3 color space DCP? Are they watching it on television, which is still Rec709. Where is there a decent display system for your viewer, that allows the full dynamic range to be seen? From my point of view, outside of very specialized equipment, there just isn't any way to do that.

 

Shooting RAW with a wide dynamic range, is really only there so you can compress that dynamic range in post production. Most people actually do finishing in Rec709 because until there is a breakthrough on the distribution side, we're all pretty much stuck presenting that limited dynamic range to the consumer.

[Robin R Probyn]

It's actually quite difficult to make OLEDs bright enough for anything, let alone HDR.

 

There are actually two HDR standards for home TVs, one of which defines a very low black level and a moderate white level, which is intended for OLEDs, and one which defines a moderate black level and a rather high white level, which is clearly intended for LCDs.

 

More or less the only display in the world which can do both the very low black level and the very high white level is something like a Sony BVM-X300, which is ferociously expensive and I have my concerns over the longevity of its blue channel.

 

P

 

 

This is wot I read a while back.. OLED as opposed to LCD.. LCD never being able to be dark enough as they are never truly black..

 

An OLED display on the other hand uses an array of individual LED emitters that are totally dark when they are off and can be very bright when all the way on. As a result they are capable of displaying a far greater dynamic range. TV and monitor manufacturers are now producing displays that are capable of showing dynamic ranges well in excess of the normal 6 stop range. If we then change the standard of the signals that we send to these displays to include a greater dynamic range that’s when the magic starts to happen. But it’s not just about contrast, it’s also about color. OLED displays can also show much more highly saturated colors.

[Mei Lewis]

Aha! So we are capturing a higher dynamic range, but that DR isnt preserved all the way through to the final deliver. That makes sense.

 

So a camera that is able to capture, lets say 9 stops, will end up at maybe 6-7 after a light grade. And a 15 stop camera, would end up at around 10, and retain more detail in the lift and gain?

 

Thanks for all the input!

I don't think that's quite right.

There are 3 places along the process of image capture and display whose dynamic range we primarily have to worry about:

1) The scene we're filming - this could have very many stops of range

2) The camera (digital sensor or film) that is recording the scene - 9 or 15 stops say

3) The output image - 6 stops?

Going from (1) to (2) anything outside the camera's dynamic range is lost. When the transition is made between (2) and (3) no tonal range of the image is necessarily lost (unless highlight or shadows are clipped for creative reasons or by mistake), but the range is scaled down to fit into the available dynamic range of the display.

So a 9 stop camera would record 9 stops of the real scene, and those 9 stops would be shown as 6 stops on a typical display.

A 15 stop camera would record 15 stops of the real scene, and those 15 stops would be shown as 6 stops on a typical display.

The range between the brightest and darkest parts of the output picture is the same for both cameras, because they are both being shown on the same display.

The difference is that the lower dynamic range (9 stops) camera won't have recorded as much into the highlights and shadows, so detail won't be visible there, and so the mid tones of the image are spread over much of the 6 stops of display range. For the 15 stop camera, all 15 stops have to fit into 6, so the mid range is compressed into fewer stops to allow the highlight and shadow stops to fit, and the mid range loses contrast.

You *can* display HDR images on a normal dynamic range display. This is what is happening in the images you see if you google image search for "hdr photography". I think it's a pretty ugly look, as do many people, which is probably why it's not used very often, particularly for video.

This is also basically what's happening when you look at log footage on a normal display without the appropriate LUT, and it looks very 'flat' (lacking contrast) .

[Jan Tore Soerensen]

 

I don't think that's quite right.

There are 3 places along the process of image capture and display whose dynamic range we primarily have to worry about:

1) The scene we're filming - this could have very many stops of range

2) The camera (digital sensor or film) that is recording the scene - 9 or 15 stops say

3) The output image - 6 stops?

Going from (1) to (2) anything outside the camera's dynamic range is lost. When the transition is made between (2) and (3) no tonal range of the image is necessarily lost (unless highlight or shadows are clipped for creative reasons or by mistake), but the range is scaled down to fit into the available dynamic range of the display.

So a 9 stop camera would record 9 stops of the real scene, and those 9 stops would be shown as 6 stops on a typical display.

A 15 stop camera would record 15 stops of the real scene, and those 15 stops would be shown as 6 stops on a typical display.

The range between the brightest and darkest parts of the output picture is the same for both cameras, because they are both being shown on the same display.

The difference is that the lower dynamic range (9 stops) camera won't have recorded as much into the highlights and shadows, so detail won't be visible there, and so the mid tones of the image are spread over much of the 6 stops of display range. For the 15 stop camera, all 15 stops have to fit into 6, so the mid range is compressed into fewer stops to allow the highlight and shadow stops to fit, and the mid range loses contrast.

You *can* display HDR images on a normal dynamic range display. This is what is happening in the images you see if you google image search for "hdr photography". I think it's a pretty ugly look, as do many people, which is probably why it's not used very often, particularly for video.

This is also basically what's happening when you look at log footage on a normal display without the appropriate LUT, and it looks very 'flat' (lacking contrast) .

 

But we don't necessarily know the DR of the display we are showing at. It's not like we have a slider in Premiere that says, 5-20 stops. When we export from premiere, let's say we have an image with 10 stops. If that's shown on a projector that can handle 8 stops - then that 10 DR image is projected at 8 stops. Is this correct?

[David Mullen ASC]

Keep in mind that it is not really accurate to talk about number of "stops" that can be displayed -- after all, if you look at a log signal on a Rec.709 monitor from an Alexa, you can see all 14+ stops displayed. The reason why you generally only see 10 to 11 stops in Rec.709 is because if we want something that looks like it has "normal" contrast with good blacks and whites, that's about how many stops of information we can display. If images were all strictly linear, then maybe we could just talk about stops of a display but images often aren't strictly linear.

[Michael Rodin]

But we don't necessarily know the DR of the display we are showing at. It's not like we have a slider in Premiere that says, 5-20 stops. When we export from premiere, let's say we have an image with 10 stops. If that's shown on a projector that can handle 8 stops - then that 10 DR image is projected at 8 stops. Is this correct?

No. Stops refer to exposure range. What's coming out of a digital camera doesn't have stops, it has (either right away like from an HD-SDI out or after debayering) two fixed - black and white clip - code values, where the captured exposure range of whatever stops is crammed inbetween using some gamma curve.

Your display device doesn't care how much exposure range there is in the image - if it handles the bit depth and encoding, it'll pretty much display it all. As Mr Mullen said, you can feed Log C to any '709 monitor or TV and all the range captured will be there. You won't likely see all the shadow detail though because it'll be buried in noise.

Display devices do differ in DR, but that DR has nothing to do with the exposure range you captured in camera. Display DR is what's otherwise called contrast ratio and it's responsible for deep blacks, not how much stops you can show on it. When some display technologies reached crazy contrast ratios, it made sense (not to everybody though...) to feed them with lower contrast imagery and call it HDR. It has a potential of showing clearer highlight/shadow detail, but it's not that it allows you to display more "stops" from your footage.

And you don't need to know the exact display contrast ratio when making a master/deliverable, and you shouldn't, as every make of projector or TV has different contrast performance. You just need your grading suite to have a monitor setup to the same gamma/color encoding as viewers' and lighting conditions similar to theirs. They'll see (almost) all the exposure range you and colorist see - almost, because TV viewers have their noisy glare-picking screens setup by ignorant salesmen, projectors will muddy up blacks to varying extents, etc. That's why we have standarts for gamma and color like '709.

[Mei Lewis]

This is my current understanding of what's going on here.

 

When an image is displayed white in the image data is usually mapped to white on the display, and black to black, so CU=DU and CL=DL, but that doesn't have to be the case.

If CU is mapped some value higher than DU then clipping of whites happens. If CL is mapped to somewhere lower than DL then black clipping happens.

[Mei Lewis]

Keep in mind that it is not really accurate to talk about number of "stops" that can be displayed -- after all, if you look at a log signal on a Rec.709 monitor from an Alexa, you can see all 14+ stops displayed. The reason why you generally only see 10 to 11 stops in Rec.709 is because if we want something that looks like it has "normal" contrast with good blacks and whites, that's about how many stops of information we can display. If images were all strictly linear, then maybe we could just talk about stops of a display but images often aren't strictly linear.

 

 

I think you _can_ talk about the number of stops a display has, it's possible to measure it as the difference between its black and white levels using a light meter. But this range doesn't really have much to do with the range of the camera or the original scene, and I think that's where Jan is having difficulty.

[Michael Rodin]

It's more like this...

Note that "encoded low-con video" displays virtually all the exposure range of the "digitized from sensor" data.

Edited August 11, 2017 by Michael Rodin

[Mei Lewis]

Why are you showing "Low-con encoded video", "<<ITU709 video>>" and "ITU709 display" as having the same range?

 

Aren't "Low-con encoded video" and "<<ITU709 video>>" just data, which doesn't have a range? (it's inside a computer).

[Michael Rodin]

It's not just data :) - they're video signals. They have a range of code values. Which are then mapped to display brightness levels in a pretty standartized way.

 

The chart assumes they're all the same bit depth, so they all share codes between 0 and 255. I won't get into YUV vs RGB and video vs data ranges as it's unnecesarilly complicating the talk...

[David Mawson]

It recently came to my attention that most panels today can't display more than 10-11 stops of dynamic range. I read this in a blog post, so I don't know if it's factual.

 

How can we then see a huge difference in retained high- and lowlight if the monitor display isn't supposed to be able to show it? We can obviously see a huge difference between a 5d2 and an URSA Mini.

 

Can anyone elaborate on this?

 

Thanks!

 

From my experience with stills, one issue is that cameras have flat DR response and a curve seems natural to the human eye. Getting an acceptable curve from a flat line means throwing away DR.

 

Another issue is that engineering DR and practical DR are different things - you'll probably have a lot of noise and reduced resolution in dark areas of your frame shooting at full DR.