or Connect
AVS › AVS Forum › Display Devices › Display Calibration › How power law gamma calibration can lead to crushed blacks
New Posts  All Forums:Forum Nav:

# How power law gamma calibration can lead to crushed blacks - Page 7

It affects the gamma curve - the rate of delta in luma from one step in stimulus to the next. We check this in greyscale but it affects all gradations of colour.

### AVS Top Picks

Quote:
Originally Posted by turboman123

Sotti, thanks for the advice on BT.1886. One more question:
I thought BT.1886 only affects greyscale, mainly making the picture brighter at low IRE.
But I am surprised you say BT.1886 also affects gamut and gamut saturation. Can you provide a short easy explanation?

Using standard power gamma I'll show how.

a 25% saturation at 75% stim for blue is 141,141,180 or 57%, 57%, 75%

With a power of 2.2 you have

red .57^2.2 = 29% linear red
green .57^2.2 = 29% linear green
blue .75^2.2 = 53% linear blue

So the actual color is a mix of 29% red, 29% green and 53% blue. Flip the exponent to 2.4 and you get

red. .57^2.4 = 26%
green .57^2.4 = 26%
blue .75^2.4 = 50.1%

So not only does luminance change with different gammas, but the ratio between the red, green and blue change as well, so anything that isn't either 0% or 100% saturation will have a hue shift.
^^^

OTOH, for the practical implementation (actually adjusting a CMS) one will note that red and green are still present in equal amounts. ... For me, I see this as a situation that "takes care of itself," provided you do things in the right order.

IOW, by definition, setting the gamma curve for greyscale, is going to set the gamma curve for for each of red, green and blue ... unless you've really screwed up somewhere in the greyscale.

Perhaps I've missed something ....
Quote:
Originally Posted by HDTVChallenged

^^^
OTOH, for the practical implementation (actually adjusting a CMS) one will note that red and green are still present in equal amounts. ... For me, I see this as a situation that "takes care of itself," provided you do things in the right order.
IOW, by definition, setting the gamma curve for greyscale, is going to set the gamma curve for for each of red, green and blue ... unless you've really screwed up somewhere in the greyscale.
Perhaps I've missed something ....
Yes R&G are in equal amounts but those values differ depending on gamma. Blue also varies depending on gamma but in one case blue is more desaturated than in the other case. So there is a hue shift and a saturation shift and a luminance change accompanying a gamma change.
Quote:
Originally Posted by Geof

Yes R&G are in equal amounts but those values differ depending on gamma. Blue also varies depending on gamma but in one case blue is more desaturated than in the other case. So there is a hue shift and a saturation shift and a luminance change accompanying a gamma change.

Again ... not disputing that. I just don't see how it matters, provided you adjust things in the right order.

Now, if for some reason you "need" to adjust the saturation of of blue at 25% saturation, 75% stimulus independently of everything else going on with the display, then yeah ... I guess it's helpful to have a "target," otherwise, the properties of linear algebra are pretty much going to take care of things auto-magically.

See also: sRGB definition. In that system, gamma is technically applied to each primary color individually but it makes no difference mathematically when all you need to adjust is luminance.

It's true that 1+1+1 = 3, it's also true that 3=3.
Quote:
Originally Posted by HDTVChallenged

Again ... not disputing that. I just don't see how it matters, provided you adjust things in the right order.
The right order......key words....for example, we shouldn't change gamma after a 125pt cal cause that is the wrong order.

But we digress, Joel was answering this: "But I am surprised you say BT.1886 also affects gamut and gamut saturation. Can you provide a short easy explanation?" which I thought he did in a clear and concise manner.
Quote:
Originally Posted by Geof

The right order......key words....for example, we shouldn't change gamma after a 125pt cal cause that is the wrong order.

Why not? The "gamma" change should auto-magically carry any saturation changes along with it. Again, I don't see the a problem here, unless you're saying we should try to hit gamma=2.2 targets when using gamma=2.4, but then would have destroyed the gamut's "linearity."

Hue angles should not be affected at all, which was the point I was trying to make: There's no "hue shift." There is a Luminance shift which also causes a Saturation shift, which is exactly what sotti's example demonstrates.
Quote:
Originally Posted by HDTVChallenged

Why not? The "gamma" change should auto-magically carry any saturation changes along with it. Again, I don't see the a problem here, unless you're saying we should try to hit gamma=2.2 targets when using gamma=2.4, but then would have destroyed the gamut's "linearity."
Hue angles should not be affected at all, which was the point I was trying to make: There's no "hue shift." There is a Luminance shift which also causes a Saturation shift, which is exactly what sotti's example demonstrates.
Because the luminance and saturation targets for the 125 point autocal are calculated using the gamma targets. If the gamma target varies from the actual gamma value then luminance and saturation targets will incorrect (ie, if one changes gamma after a 125 point autocal the luminance and saturation values will still be based on the original gamma target values not the new gamma).
Quote:
Originally Posted by Geof

Because the luminance and saturation targets for the 125 point autocal are calculated using the gamma targets. If the gamma target varies from the actual gamma value then luminance and saturation targets will incorrect (ie, if one changes gamma after a 125 point autocal the luminance and saturation values will still be based on the original gamma target values not the new gamma).

I guess the question here is what device is providing the "125pt" calibration ... I'm assuming not the display itself.

Honestly, I keep turning visions of Gamut spaces (and planes) over in in my head, and I still can't visualize a display "problem" here, I agree that it might cause issues for calibration software that's looking for a specific target in RGB space.

My only reason for wading in here, is I'd hate for "casual" readers to walk away from this thread with the idea that BT1886 is going to 'mess up all my colors, so I should avoid it.'
One of the problems with many calibration software systems is they need to know in advance the colour space and gamma of the target display.

This is a fundamental limitation of such calibration systems.

Other calibration systems do not need to know this in advance, ans as you say, the color information will be accurate in the final calibration regardless of the gamma value selected as the final desired target.

This is a key capability of LightSpace CMS for example.
But, this capability is rather unique.

If you don't have a calibration system that is capable of profiling without needing to now the target colour space and gamma in advance you will have to re-profile if you decide to change the end target.

Steve
Their isn't an issue with having different gamma's and 3D cubes.

It's about having a a 3D-LUT + 3x1D LUTS.

You can't tweak 1D LUTS after calibrating a 3D-LUT. If 75% Red is a little high in the gray scale after doing a 125 point calibration, you can't turn down red 75% because that will effect.

75,0,0
75,0,25
75,0,50
75,0,75
75,0,100
75,25,0
75,25,25
75,25,50
ect....

So if the cube had correctly calibrated any of the points that use 75% their values will be altered by tweaking the 1D-LUT.

Gamma changes chromaticity, but the correct chromaticity is a by product of the gamma formula. But calculating saturation with a black offset is a bit trickier than it would seem, that was why I mentioned that the software needs to be up to snuff.
Quote:
Originally Posted by HDTVChallenged

I guess the question here is what device is providing the "125pt" calibration ... I'm assuming not the display itself.
Honestly, I keep turning visions of Gamut spaces (and planes) over in in my head, and I still can't visualize a display "problem" here, I agree that it might cause issues for calibration software that's looking for a specific target in RGB space.
My only reason for wading in here, is I'd hate for "casual" readers to walk away from this thread with the idea that BT1886 is going to 'mess up all my colors, so I should avoid it.'
The device performing the 125 pt calibration is the Lumagen Radiance.

To be clear, this is not about BT1886 screwing up colors. Joel has aptly shown that gamma affects luminance and hue. That's a given whether one is calibrating to 2.2 or 2.4 or BT1886.

Up until the cube cal became possible all anyone could do is calibrate at either 75% or 100% saturation points. The other points simply could not be calibrated. If the display is linear then the other points end up where they should be and saturation tracks well (the 25%, 50%, 75%, and 100% points all align as predicted). If the display does not track well (and many do not) then one could calibrate at 75% but find significant errors at 25%, 50% or 100% (any or all could be good or bad or anywhere in between).

In the case of a 125 point calibration each of the 25%, 50%, 75% and 100% saturation points can be calibrated. This is beautiful because it can correct for a displays non-linear saturation tracking. For any given 25%, 50%, 75% or 100% saturation level both hue and luminance need to be correct to achieve the desired results. But the correct hue and luminance values are dependent on gamma (ie, gamma directly affects luminance and hue as Joel shows above). Ergo, to successfully perform a 125 point calibration calman and chromapure calculates the correct hue and luminance values for each saturation data point and then it diddle values in the Radiance to achieve those values. If gamma is then changed after this calibration all of those calculations are affected and all of the existing locations are incorrect . IOW, for any given gamma value there is a unique hue and luminance value for any given saturation level.

I probably confused instead of helped....
You might want to have a read of this, and the associated pages:

http://www.lightillusion.com/home_cinema.html
(It is a work-in-progress, so any errors are of my making as I have not yest checked everything for 100% accuracy!)

You should be able to understand that with a calibration system that separates profiling from the target Colour Space there is no issue.

LightSpace CMS can generate full 3D LUTs for the likes of the eeColor LUT Box, or 1D and 3D combinations for the Radiance.
Deciding to alter the gamma or even the gamut you intend to calibrate for after having performed a profile is no problem.

I think this is what you are discussing, is it not?

Steve
Quote:
Originally Posted by Geof

The device performing the 125 pt calibration is the Lumagen Radiance.
Alrighty then ... clearly something I don't need to worry about with my lowly 2.5D CMS, or even a full 3D CMS if I had one.
Quote:

To be clear, this is not about BT1886 screwing up colors. Joel has aptly shown that gamma affects luminance and hue. That's a given whether one is calibrating to 2.2 or 2.4 or BT1886.

I think part of the problem here, is that we may be using different definitions of Hue, Saturation and Luminance. To me Hue is an angle, while Saturation and Luminance are scalar components of a Vector relative to a given white-point in a 3D gamut space (specifically CIE1931 coordinates / REC709.) I don't see how changing gamma could possibly affect either Hue or Saturation in this coordinate system. What it maps to in RGB or R'G'B' space is a different issue. Again ... maybe I just haven't had enough coffee or perhaps I'm having a lengthy senior moment.
Quote:
Up until the cube cal became possible all anyone could do is calibrate at either 75% or 100% saturation points. The other points simply could not be calibrated. If the display is linear then the other points end up where they should be and saturation tracks well (the 25%, 50%, 75%, and 100% points all align as predicted).

Yes exactly my point.
Quote:
Originally Posted by HDTVChallenged

I don't see how changing gamma could possibly affect either Hue or Saturation in this coordinate system. What it maps to in RGB or R'G'B' space is a different issue.

Let's take "Orange" from the color checker chart (the color checker is actually defined spectrally, but we'll short cut that), for a gamma of 2.2 it had the triplets 202,119, 51

This gives it targets of
202 stim = 0.6982
119 stim = 0.1902
51 stim = 0.0177
so...
x: .5146 y: .4095

L*a*b*: 60.4, 32.1, 58.7
L*a*b* sat: 66.9
L*a*b* hue: 61.3

Gamma 2.4:
202 = 0.6757
119 = 0.1636
51 = 0.0123

x: 0.526 y:0.4054

L*a*b*: 58.18, 34.9, 59.1
Sat: 68.7
Hue: 59.4

So we've changed hue and saturation and luminance targets for the color by changing gamma formula.

If a color has equal parts of two components it won't shift hue, but most colors aren't on the diagonals from the primaries through white to the other side. Even then gamma is still effecting saturation.

If you don't understand it, try opening up CalMAN (Even the demo version) and play with the color checker chart changing gamma formulas.
Quote:
Originally Posted by sotti

but most colors aren't on the diagonals from the primaries through white to the other side.

But the only ones that I can control directly *are* on those diagonals/axes. The rest of them are for the display to sort out

All I'm saying is that changing gamma is *not* going to have any effect on how I set up my six axis 2.5D (or 3D) CMS. If I have the six primary/secondary Hue angles as correct as they're going to get, I'm calling it a day. I've got too much DVR backlog to worry about it further.

My problem here, is you're starting with an RGB triplet ... and I just don't care about that. As far as I'm concerned, a color starts with YCrCb that was encoded using some unknown gamma/EOTF, I just need to get that data converted linearly and consistently to something that "drives" my display panel.

I'm not questioning your math, I just don't see the significance from a practical standpoint.
Quote:
Originally Posted by HDTVChallenged

But the only ones that I can control directly *are* on those diagonals/axes. The rest of them are for the display to sort out

I'm not questioning your math, I just don't see the significance from a practical standpoint.

Sure, but it's my job to be able to measure and quantify the error of your display.

With Color Cube 3D LUts, some people can fix those issues. Even just calibrating at 75% saturation helps with those kinds of errors. When you can't fix it that kind of error is valuable to report.

You'll have to forgive me for being sticky about the details. But since with any CMS you can now calibrate at 75% saturation you need to pay attention to how gamma effects saturation and hue. Granted CalMAN just takes care of all the retargeting so it's just easy for you.
Quote:
Originally Posted by sotti

Even just calibrating at 75% saturation helps with those kinds of errors..

Absolutely ... 100% stim, 75% sat is mandatory on my display since the 100% sat points (at least the primaries) are immutable. As long as we are in agreement that gamma change isn't going to affect hue or saturation along any of the six primary/secondary axes, I'll let the matter rest.
Quote:
You'll have to forgive me for being sticky about the details. But since with any CMS you can now calibrate at 75% saturation you need to pay attention to how gamma effects saturation and hue.

There you go again . 0.75 is a fixed scaler distance from D65 in the x,y plane ... how will changing gamma change "saturation?"

Clearly we are mis-communicating a/o looking at things from a different reference points. I'm just getting a headache at this point.

PS: At this point the closest I can get to what you may be trying to say is we need to compensate between the "gamma" value used to encode a 75% saturated target on a disk vs. the measured/calibrated gamma/EOTF on the display. No? ...
Edited by HDTVChallenged - 11/7/12 at 12:53am
Quote:
Originally Posted by HDTVChallenged

There you go again . 0.75 is a fixed scaler distance from D65 in the x,y plane ... how will changing gamma change "saturation?"

Because the 75% pattern is a baked on the disc as a YCbCr value.

The reality is that once you change from a target gamma of 2.2 to anything else the RGB triplet for 75% changes. So one of two things happens either you need a new RGB triplet for 75% or you need to recalculate the target based on the saturation change of the encoded triplet.

You can find documentation about the AVS.709 disc for instance. It's RGB values for 75% RGB are red 90.4%, green 26.5%, blue 26.5%.

So the actual RGB triplet of the pattern is 214, 74, 74, or in YCbCr space 104, 112, 198.

First you decode the RGB triplet to percent.
Then you degamma the RGB percent to linear percent <-- gamma formula.
Then you convert the linear RGB to XYZ via a 3x3 matrix.

So any time you change gamma, the target XYZ for a value can change. In fact only White, Red, Green, Blue, Cyan, Magenta, Yellow and Black don't change (100% luminance, 100% saturation). So 75% in x,y never changes, but a pattern labeled 75% will not actually be 75% between White and the Color unless you use the specific gamma the disc was encoded with.
Quote:
Originally Posted by sotti

Because the 75% pattern is a baked on the disc as a YCbCr value.
The reality is that once you change from a target gamma of 2.2 to anything else the RGB triplet for 75% changes. So one of two things happens either you need a new RGB triplet for 75% or you need to recalculate the target based on the saturation change of the encoded triplet.

Alright, I'll buy that. I was looking at things from the opposite direction. For example:

1) Adjust greyscale to D65 with gamma 2.2
2) Adjust CMS with 75% Sats
3) Re-work greyscale to BT1886

In that order, I don't see the need to re-do the the CMS, everything *should* be carried along with the greyscale/gamma change. Perhaps it's fortunate that I don't really have any direct control over saturation with the CMS on my set (just Hue angle and Luminance.) Like I said before, I just line up the six Hue angles, then make sure the Luma's haven't gone rogue and call it a day (or early morning.)

If you go looking for trouble, you'll probably find it somewhere.
Here is my result
The first one is 2.35 gamma without isf calibration and the second one is ISF calibrated from 2.4 gamma point. I prefer the old one (2.35 non calibrated) because of deeper shadow detail and non crushed blacks.

What are your Y mins and maxes?
I couldnt receive any data after calibration because the calibrator has lost the data on broken laptop. But he said " after the isf night calibration, the grayscale, colour gamut and white balance are very successfull."
at the end, the shadow detail has almost gone as you can see...
Edited by paulbroke - 11/9/12 at 3:18am
Quote:
Originally Posted by paulbroke

Here is my result
The first one is 2.35 gamma without isf calibration and the second one is ISF calibrated from 2.4 gamma point. I prefer the old one (2.35 non calibrated) because of deeper shadow detail and non crushed blacks.
http://www.avsforum.com/content/type/61/id/87645/width/500/height/1000
http://www.avsforum.com/content/type/61/id/87646/width/500/height/1000
If it is not calibrated, what makes you think the first image is 2.35 gamma?

Photographs can be deceiving, but it would appear that the calibration is either very poor, or you are running 2.40 gamma on a display which doesn't have nearly enough contrast to support it. (which would also be a bad calibration)

Seems to be very convenient that the data is missing...
Quote:
Originally Posted by turboman123

I thought BT.1886 only affects greyscale, mainly making the picture brighter at low IRE.
But I am surprised you say BT.1886 also affects gamut and gamut saturation. Can you provide a short easy explanation? .
Quote:
Originally Posted by sotti

Using standard power gamma I'll show how.
a 25% saturation at 75% stim for blue is 141,141,180 or 57%, 57%, 75%
With a power of 2.2 you have
red .57^2.2 = 29% linear red
green .57^2.2 = 29% linear green
blue .75^2.2 = 53% linear blue
So the actual color is a mix of 29% red, 29% green and 53% blue. Flip the exponent to 2.4 and you get
red. .57^2.4 = 26%
green .57^2.4 = 26%
blue .75^2.4 = 50.1%
So not only does luminance change with different gammas, but the ratio between the red, green and blue change as well, so anything that isn't either 0% or 100% saturation will have a hue shift.

Here's a way I think will help many users visual luminance in the gamut

normally, you see something lke this to represent the color space which doesn't help..

Here is what I do (709D65 10^3 sampling) - VIDEO BELOW SHOWS IT BEST

I rotated for another view

Represented as the meter sees it

If you need another format, use an online video converter: http://www.zamzar.com/

Hope that helps....

.
Edited by turbe - 11/29/12 at 7:42pm
I read through this thread . Most of it is beyond my comprehension. I read the part the mentioned that 1886 is better for displays the don't have good black levels. How would this related to the VT50? Which is better for the VT; power law or 1886 and why? Thanks
Edited by TWD - 11/29/12 at 9:36pm
Quote:
Originally Posted by turbe

Here's a way I think will help many users visual luminance in the gamut
normally, you see something lke this to represent the color space which doesn't help..

All of us used to looking at gamut plots for Photoshop and printing applications find the plots used at the top of your post rather useless. We're also used to a much more dynamic colour management system where the gamut of the output device is calibrated then measured and a dynamic CMM is used to map from the source gamut (also measured with the measurements transported along with the file) to the device gamut for best colour reproduction with defined protocols for handling out-of-gamut colours. All this hard-wiring used in video is painful.
We have done a lot of work with ICC's (CMM's), and have found time and again just how inaccurate they can be...

Let's face it, any calibration process that has a 'Use Black Point Compensation' as part of its options isn't really looking like an accurate calibration method.

Seriously, we have been called into facilities time and again to rescue them from issues caused by trying to use ICC profiles for accurate calibration.
Quote:
Originally Posted by TWD

I read through this thread . Most of it is beyond my comprehension. I read the part the mentioned that 1886 is better for displays the don't have good black levels. How would this related to the VT50? Which is better for the VT; power law or 1886 and why? Thanks

For your VT50, assuming you have brightness set correctly and are achieving the black levels this display is capable of, then stick with a pure power law calibration. BT.1886 is a pure power law in the limit of 0 mll.
sorry for kicking this up when the dust has settled. Why is it indeed, that "gamma" affects the gamut?
New Posts  All Forums:Forum Nav:
Return Home
Back to Forum: Display Calibration
• How power law gamma calibration can lead to crushed blacks

### AVS Top Picks

AVS › AVS Forum › Display Devices › Display Calibration › How power law gamma calibration can lead to crushed blacks