AVS Forum banner
1 - 20 of 248 Posts

·
Home Theater Enthusiast
Joined
·
6,041 Posts
Discussion Starter · #1 ·
Display Gamma


Pixar = 2.4

Quote:
Originally Posted by sspears /forum/post/16995827


I can tell you that Pixar calibrates their displays using REC 709 primaries and 2.4 gamma.
Quote:
Originally Posted by darinp2 /forum/post/16961707


And the information I have is that Pixar sets up their in house DLPs to 2.4-2.45 to match their BVMs.



I will add to the list as we find out more from the pros.
 

·
Registered
Joined
·
3,314 Posts
This doesn't tell you anything about how the video was mastered.


For a start the CGI would likely have been lit and textured in linear , the DI would have been 0.6 on a 1.7 display, all this could have been done on a display with a gamma of anything between 1.7 and 2.5. This has been profiled and lutted to take it to any reasonable colorspace the colorist desires.


I sit in front of a display with a 2.2 gamma and a D.65 color temp , what I actually view on it is D.54 gamma 1.7 with a suitable LUT.


The hardware isn't important , the end to end gamma is : and even then people have moved away from defining simplistic gammas and now use transfer functions.


Also professional imagery is created with an eye on robustness. A gamma deviation on the order of 0.3 is within the tolerance of good imagery as you may want to tweak the end gamma depending on display capabilities and environment.


When I'm creating video imagery I view it at 2.2 and 2.5 with and without a hard clip to 16-235 in place to ensure robustness. I actually find retaining detail above 235 to be far more critical than nailing gamma to within a couple of tenths of a point.


Within the envelope of technical video requirements how the actual imagery looks is not an absolute quantity. If you gave two potters exactly the same amount and type of clay and asked them to give you a pot of a desired design each pot will have slight differences. Imagery is a bit like this.


There is no single set of numbers that will convert film imagery into video imagery and give acceptable results for every single shot in the film. This is why you need a human being being paid a lot of money to subjectively decide how best to compromise the film into video.


I assume that something competently mastered for video will look acceptable between 2-2.5. I generally prefer something a bit higher than 2.2. There is little merit in trying to hit an absolute gamma number on your display beyond starting at 2.2 ( although I prefer to hit Rec.709) and apply a reasonable gamma offset accordingly to your preference and situation.
 

·
Banned
Joined
·
3,257 Posts
I say the max is 250:1 thats the said peak density. and said 2.5 w/special calibration? (gamma tweak) things could go exponentially WRONG



To the human eye, it looks like max ideal density of 2.2 or (signal or RGB) uncorrected Y with no prime. Gamma is just new... "new eq" distortion that gets us the last two or 2.22 or even +1.25 camera –eyeball.


Also: RGB i.e. blue is 50% duty cycle!

Thats no accident..?
 

·
Registered
Joined
·
7,998 Posts
Nice post. Strangely reasonable and undogmatic for these parts.

Quote:
Originally Posted by Mr.D /forum/post/16998269


This doesn't tell you anything about how the video was mastered.


For a start the CGI would likely have been lit and textured in linear , the DI would have been 0.6 on a 1.7 display, all this could have been done on a display with a gamma of anything between 1.7 and 2.5. This has been profiled and lutted to take it to any reasonable colorspace the colorist desires.


I sit in front of a display with a 2.2 gamma and a D.65 color temp , what I actually view on it is D.54 gamma 1.7 with a suitable LUT.


The hardware isn't important , the end to end gamma is : and even then people have moved away from defining simplistic gammas and now use transfer functions.


Also professional imagery is created with an eye on robustness. A gamma deviation on the order of 0.3 is within the tolerance of good imagery as you may want to tweak the end gamma depending on display capabilities and environment.


When I'm creating video imagery I view it at 2.2 and 2.5 with and without a hard clip to 16-235 in place to ensure robustness. I actually find retaining detail above 235 to be far more critical than nailing gamma to within a couple of tenths of a point.


Within the envelope of technical video requirements how the actual imagery looks is not an absolute quantity. If you gave two potters exactly the same amount and type of clay and asked them to give you a pot of a desired design each pot will have slight differences. Imagery is a bit like this.


There is no single set of numbers that will convert film imagery into video imagery and give acceptable results for every single shot in the film. This is why you need a human being being paid a lot of money to subjectively decide how best to compromise the film into video.


I assume that something competently mastered for video will look acceptable between 2-2.5. I generally prefer something a bit higher than 2.2. There is little merit in trying to hit an absolute gamma number on your display beyond starting at 2.2 ( although I prefer to hit Rec.709) and apply a reasonable gamma offset accordingly to your preference and situation.
 

·
Registered
Joined
·
6,723 Posts
Greetings


People that can't live life without absolute numbers would not be able to handle this.


regards
 

·
Banned
Joined
·
3,587 Posts
Dr. Ray Soniera owns (or was donated) a Sony CRT mastering monitor that he has measured at 2.2 Gamma.


But as I've been saying over and over again... whatever gamma is either native to the mastering monitor or whatever gamma it has been adjusted to is of little consequence since the software RUNNING the monitior (and the profile driving the monitor) can be tweaked to produce just about any gamma, either by accident (hopefully not, but not all operators are created equal) or on purpose.


It does NOT MATTER what Pixar's mastering monitors gamma is. All that matters is what the gamma is of the final product (again, because the workstation software and monitor profile (more software) can produce any gamma you'd like, regardless of the monitor's "state of tune"). Until somebody can tell us the gamma of the end product, nobody will know what the right gamma is to reproduce what they created.


I'm surprised anybody would say any gamma between 2-2.5 would probably be acceptable because (for me) 2.0 gamma is much too bright while 2.5 is much too dark. You really need something like a Lumagen Radiance to see this at home on the same display since viewing different brands/models with different gamma is very misleading. The Radiance processor will produce a uniform gamma curve from bottom to top no matter what the native or calibrated gamma of the display is (without the Radiance), just like mastering monitors in workstations. I could live with a gamma between 2.1 and 2.35 or thereabouts, but 2.1 makes images look a little dimensionally flat from being too bright through the midtones) and some displays with black levels that aren't nice and dark don't look that great at 2.35 and should be closer to 2.2.
 

·
Banned
Joined
·
3,257 Posts

Quote:
Originally Posted by Doug Blackburn /forum/post/16999881


Until somebody can tell us the gamma of the end product, nobody will know what the right gamma is to reproduce what they created..

Just doesn't seem right in my book



I think I know exactly what the mastering engineers saw in 100% of ALL content production.
 

·
Registered
Joined
·
7,998 Posts

Quote:
Originally Posted by tbrunet /forum/post/16999199


Tom is gamma 2.22 or 2.5? Reference is what?

I have always recommended anything between 2.2-2.35, depending on tastes and viewing environment.


I agree completely with Doug on this. Both 2.5 and 2.0 look obviously wrong to my eyes, and I have the Radiance so I can create any gamma response I want.
 

·
Banned
Joined
·
3,257 Posts
If your not in the same space (Rec 709) you are not seeing a uniform pic. The nominal viewing ambient condition is integrated into the transform function. The space being a complex 3D space.


To recover the original perception, one must use the same default profile and assume the "mastering engineer" knew something about video. i.e.


1) The perceptual space (Rec 709) or video signal

2) Target (CRT @ 2.22) is normalized for unity 100% = 1
 

·
Registered
Joined
·
773 Posts

Quote:
Originally Posted by tbrunet /forum/post/17000414


If your not in the same space (Rec 709) you are not seeing a uniform pic. The nominal viewing ambient condition is integrated into the transform function. The space being a complex 3D space.


To recover the original perception, one must use the same default profile and assume the "mastering engineer" knew something about video. i.e.


1) The perceptual space (Rec 709) or video signal

2) Target (CRT @ 2.22) is normalized for unity 100% = 1

Apart from the fact that the Rec. 709 transfer curve as specified in Recommendation ITU-R BT.709-5 ( how to get it free ) is closer to 1/0.51 (1.96) than 1/0.45. (2.22) So unity' would be 1.96.


How many times do we have to go over this?





I think the EBU's recommendation of an end-to-end system gamma of 1.2 is good, which is ≈2.35 with a tolerance of ±0.15, though I would also agree with what has been said above about 2.0 and 2.5 looking obviously wrong. From my own experiences, I would say the acceptable range is more like 2.2 to 2.4.


I have said this before, but no-one seems to have really tried or given feedback. The EBU's actual recommendation is a system gamma of 1.2, though they refer to this as 2.35 at times. I'm wondering if this is perhaps the same thing that Poynton does when he uses 0.5 as shorthand for the BT.709 transfer function, and that it should be the BT.709 curve with an end-to-end of 1.2.


This would fit in with a lot of people's preferences for a high gamma for mid/bright tones, and a lower gamma as things approach black to retain shadow detail. Here's the BT.709 curve with a system gamma of 1.2 compared to 2.35. (1.96 * 1.2)




What I also thought was interesting, is how Poynton's 1.25 system gamma recommendation matches up almost perfectly with Lstar (perceptually linear) when you use the BT.709 curve rather than the best-fit power curve of 1/0.51.

 

·
Banned
Joined
·
3,257 Posts

Quote:
Originally Posted by andrewfee /forum/post/17000844


How many times do we have to go over this?

Linear tail?


In my book the "offset" is 0V,0V,0V or 16,128,128


For some reason andrew you feel comfortable shifting the concept of pic black level up to 18% (gray), then re-calcualte said 2.4 target?
 

·
Registered
Joined
·
773 Posts

Quote:
Originally Posted by tbrunet /forum/post/17000896


Linear tail?


In my book the "offset" is 0V,0V,0V or 16,128,128


For some reason andrew you feel comfortable shifting the concept of pic black level up to 18% (gray), then re-calcualte said 2.4 target?

How did you get that from what I said?
 

·
Banned
Joined
·
3,257 Posts
 http://www.cs.ubc.ca/~mmt/Papers/MscThesis.pdf

"A linear segment is included for practical reasons and the break occurs where the function equals an L* value of 8, corresponding to a contrast ratio of 100 :1. Obtaining values below 8 is rare in practice and the break is considered the effective limit for video applications, reinforcing the fact that L* is only applicable to LDR images."


Andrew I think it safe to say, this signal level (L*8) is generally understood to represent where the model perceptual defines ~black level of the exposure
 

·
Registered
Joined
·
23,188 Posts

Quote:
Originally Posted by Mr.D /forum/post/16998269


This doesn't tell you anything about how the video was mastered.

Since the display is used for mastering it tells a lot about how to reproduce what the person mastering was seeing.
Quote:
Originally Posted by Mr.D /forum/post/16998269


The hardware isn't important , the end to end gamma is ...

If the goal is to see what the person mastering saw then the end to end gamma is not what is important as far as knowing what to do. It is the display gamma. As Poynton points out, the part before that can be treated like a black box as far as playback. If you want to see what things looked like to them you don't need to concern yourself with source gamma and things, just what the display does. If the displays were identical then what would come off the display would be the same no matter what kinds of tricks were played before the input to the display.


Put another way, in an end to end of x times y (where x is source gamma and y is display gamma) getting the end to end doesn't require knowing the x, it only requires matching the y. If y is matched and the source is the same then the end to end will be the same regardless of what they did to x.
Quote:
Originally Posted by Mr.D /forum/post/16998269


Also professional imagery is created with an eye on robustness. A gamma deviation on the order of 0.3 is within the tolerance of good imagery as you may want to tweak the end gamma depending on display capabilities and environment.

If a person wants to go with a 2.6 gamma for instance, then if the mastering was done on a 2.2 display with 2.2 in mind then they have deviated by .4, but if the mastering was done on a 2.4 display with 2.4 in mind, then they have only deviated by .2, or the same amount as a person doing 2.2 from a source mastered with a display gamma of 2.4 has deviated. And if .3 deviation is within tolerance then 2.0 gamma would be if the master was done with 2.2 in mind, but wouldn't be within the .3 if 2.4 display gamma was what was in mind with the master. This is one reason it is useful to have an idea of how the source was seen, to have some idea of how much the deviation is.


Much like with color. I think it makes a lot of sense to define a display standard of D65. It would make very little sense IMO to just tell people that mastering displays are going to be anywhere from what would be equivalent to D60 to D70 and people should just set things up with that in mind, but no idea of how the mastering monitor was actually setup within that range.
Quote:
Originally Posted by Mr.D /forum/post/16998269


Within the envelope of technical video requirements how the actual imagery looks is not an absolute quantity. If you gave two potters exactly the same amount and type of clay and asked them to give you a pot of a desired design each pot will have slight differences. Imagery is a bit like this.

Completely true. The point here is that if you have your display setup the same as them then whatever choices they make to the signal come through.
Quote:
Originally Posted by Mr.D /forum/post/16998269


There is no single set of numbers that will convert film imagery into video imagery and give acceptable results for every single shot in the film. This is why you need a human being being paid a lot of money to subjectively decide how best to compromise the film into video.

Definitely. But the way to see what they saw is to have the displays the same. Otherwise the choices they made were for their display, not yours (although it gets a little more complicated than this).
Quote:
Originally Posted by Mr.D /forum/post/16998269


There is little merit in trying to hit an absolute gamma number on your display beyond starting at 2.2 ( although I prefer to hit Rec.709) and apply a reasonable gamma offset accordingly to your preference and situation.

Why start at 2.2? I know your stuff is mastered on a monitor that is 2.2 and you try to make sure it looks good between 2.2 and 2.5, but Pixar masters DCI content on displays at 2.6 and consumer HD on displays at about 2.4. For their stuff it seems like starting at 2.4 and then applying a gamma offset like you said would make more sense than starting at 2.2 and then doing that.


--Darin
 

·
Registered
Joined
·
7,056 Posts

Quote:
Originally Posted by darinp2 /forum/post/17002986



Since the display is used for mastering it tells a lot about how to reproduce what the person mastering was seeing.

I think what he ment was, just because the hardware was calibrated to that number, doesn't mean a LUT in software wasn't used to change the end-to-end gamma number on the fly.


I do get your point though. Everything should have a standard that they are encoded to so that the displays have a standard to decode from. So that the important bit, end-to-end gamma can easily be preserved.


Although nearly everything I've read has given gamma a range, even end-to-end gamma because between viewing conditions, and display quality you play a balancing act as a calibrator. So to go into a situation and say gamma must be X, doesn't take the full scenario into account, because unlike the color pallete, the way we percive the gamma curve is effected enviroment. Our pupils dialate and constrict and modify our ability to percive contrast.


I do believe the official answer is a target range, not a percise number.
 

·
Registered
Joined
·
23,188 Posts

Quote:
Originally Posted by sotti /forum/post/17003219


I think what he ment was, just because the hardware was calibrated to that number, doesn't mean a LUT in software wasn't used to change the end-to-end gamma number on the fly.

As long as the LUT was changing the video/software and not the display then it doesn't matter. It's pretty simple algebra that most of us learn in school. I highly doubt that Pixar pays somebody to calibrate all their displays to certain levels (a 2.6 gamma for DCI content and 2.4 for consumer HD content) and then is so incompetent that they have people adjusting the display instead of the source to get the images looking right. The whole point is that you adjust the source at that point. They can change the end-to-end gamma all they want and it doesn't make a lick of difference to the display gamma to play it back if the changes to the end-to-end were done to the source.


And given that Pixar has setups they do for both 2.6 and 2.4 with their digital displays for mastering they should be using LUTs to get to those. It seems like some people are assuming that one person sets the display up so it does 2.4 and then somebody else changes the display settings (like the LUT for it that got to 2.4 in the first place) themselves to get the material looking a certain way. As I said, I doubt that. There might be that much incompetence somewhere, but I doubt at Pixar.
Quote:
Originally Posted by sotti /forum/post/17003219


So to go into a situation and say gamma must be X, doesn't take the full scenario into account, because unlike the color pallete, the way we percive the gamma curve is effected enviroment.

I agree that it is okay to change gamma, but I think there should be a starting point. The fact that people can vary gamma doesn't seem like a good reason not to have a starting point as it just makes it more likely people will deviate more than somebody considers reasonable. I think environment and even light levels like white level can affect our perception of color also. An example is how colors can be mastered differently for 3D content assumed to be watched at closer to 4.5 ft-lamberts than regular 2D content.
Quote:
Originally Posted by sotti /forum/post/17003219


I do believe the official answer is a target range, not a percise number.

I don't believe there is an official answer. Even for the same setup I think different professionals would recommend different gammas just because they have differences of opinion about what the starting point should be. Put another way, if a professional says that gamma should be reduced by .2 because of somebody's viewing conditions then they would end up at different answers whether they started at 2.2 or 2.4. I can understand why Poynton is so frustrated and just wants an official standard for display gamma.


--Darin
 

·
Registered
Joined
·
3,314 Posts

Quote:
Originally Posted by darinp2 /forum/post/17003257


I highly doubt that Pixar pays somebody to calibrate all their displays to certain levels (a 2.6 gamma for DCI content and 2.4 for consumer HD content) and then is so incompetent that they have people adjusting the display instead of the source to get the images looking right. The whole point is that you adjust the source at that point. They can change the end-to-end gamma all they want and it doesn't make a lick of difference to the display gamma to play it back if the changes to the end-to-end were done to the source.



--Darin

When I calibrate monitors ( and I'm usually the person that gets a phone call when they have problems getting a monitor to calibrate... I'm an artist not a tech guy but I know the black magic to maximise hardware) the hardware calibration is only important in so far as optimising the dynamic range whilst keeping the device from being "stressed" ie clipped or crushed or unecessarily over-driven. This is the same for ANY display technology but obviously they all slightly diffirent implications. ( LCD for example have precision limitations but massive peak output so you exploit this)


I have no dea what the gamma or chromaticity is on the end display hardware nor do I care. So finding out that pixar use 2.4 on a particular monitor is meaningless in the context of mastering.


I then profile the display and a color management system generates correctional LUTS to take that display to whatever target standard is required. I use on a regular basis:


Various print stock models (nominally D.54 , gamma 1.7 at the display side)


Rec.709 at gamma 2.2 and 2.5 ; clients generally prefer 2.5 for presentation. I work at 2.2 or even lower depending on content so I can easily see all the "data" of the image. I'll push it about as I require


PAL EBU at 2.2 and 2.5 ( I also have 2.8 because of spec but its never used).


90% of the work I do will get done using 1d LUTs : This only adjusts RGB balance so essentially you are talking about gamma and color temperature, gamut is not critical because usually I'm not responsible for the end look of the material all I have to do is ensure I match the original neg or footage so the colorist gets what he is expecting.


However I will bang the end image around to ensure he won't easily "break" it . This includes going to extremes of grading : usually I'll neutralise any color cast in the source material and grade it up and down and twist the gamma...however ths is just a verification of robustness its nothing I'll bake into the deliverables).


Working with the 1dluts although innacurate in terms of gamut is enough to let me do most of the work transparently to the original material and it entails no processing overhead .


Additionally in the case of film work whilst the 3dLuts of print stocks are more accurate in terms of end presentation they are actually a bit limiting in terms of being able to visually interrogate the data contained in the image. Reason being that they have a fairly rigid truncation of the negative into print colorspace and all this entails for gamut. Whilst this is accurate in terms of looking at the end print what I'm delivering is full "negative" density and the 3dluts are not always conducive to this.


I'm after a workable level of accuracy but also transparency and in terms of delivery to DI as long as I have transparency its someone further down the chain who is concerned with absolute accuracy safe in the knowledge that I haven't thrown anything out.


If I am responsible for the end look of the material; as sometimes happens when the DI can't quite chop up the image precisely enough then I'll usually work with a true 3dlut model of the print stock for the "look". I'll still ensure robustness.


If its 1080p video then Rec.709 2.2 is my target and I'll use a true3d lut for final appraisal but as I've said I will also appraise robustness outside the spec.


My point for us home cinema fans:


Aim for Rec.709 . See what if you prefer it a bit higher than 2.2 on your kit.

Then get on with watching something on it.


I see no practical merit in trying to poll what goes on in different mastering houses in terms of what the "gamma" is on a bit of display hardware. It will only lead to confusion as you will be missing the context and possibly the person telling you may not understand you are actually after the end result.


If you think any of what I've said sounds lapse in terms of mastering let me assure you I am generally regarded in the industry as being quite overbearingly anal in terms of color pipeline and calibration.
 
1 - 20 of 248 Posts
Top