or Connect
AVS › AVS Forum › Display Devices › Display Calibration › LightSpace CMS Now Supports Lumagen + eeColor 3D-LUT 4 All
New Posts  All Forums:Forum Nav:

LightSpace CMS Now Supports Lumagen + eeColor 3D-LUT 4 All - Page 12

post #331 of 710
Quote:
Originally Posted by Light Illusion View Post

I had our colour scientist (who is also a mathematician) look over the various comment and assertions regarding BT1996 on this thread, and after he spat his coffee over his keyboard (lucky they are cheap to replace) he gave the following response:
BT1886 wink.gif
post #332 of 710
Quote:
Originally Posted by Light Illusion View Post

I had our colour scientist (who is also a mathematician) look over the various comment and assertions regarding BT1996 on this thread, and after he spat his coffee over his keyboard (lucky they are cheap to replace) he gave the following response:

Steve,

You have already admitted that LightSpace does not use the annex formula and went as far as to say you would now include it as an option, therefore LightSpace is currently using the wrong formula.


1. Correct: L = a(max[(V + b),0])^γ

The correct terminology for this function is power law with black level offset (you really need to read Poynton at some point)

2. Incorrect: L = V^γ

The correct terminology for this function is power law (colloquially known around here as pure power law)

Now as simply as I can put it - Which equation does LightSpace use, 1 or 2?
post #333 of 710
Quote:
Originally Posted by zoyd View Post

Steve,

You have already admitted that LightSpace does not use the annex formula and went as far as to say you would now include it as an option, therefore LightSpace is currently using the wrong formula.


1. Correct: L = a(max[(V + b),0])^γ

The correct terminology for this function is power law with black level offset (you really need to read Poynton at some point)

2. Incorrect: L = V^γ

The correct terminology for this function is power law (colloquially known around here as pure power law)

Now as simply as I can put it - Which equation does LightSpace use, 1 or 2?

dude, what planet are you on ?

Nobody has "admitted" anything. He has stated that LS - ON PURPOSE - implements BT.1886 in a way they see fit. As per the specification.

Now, I agree that the label might be confusing for users who expect what you expect, but not for user who expect the pure 2.4 implementation.

The annex VERSION, which is a VERSION of the specification, will be added... maybe they'll change the label so u're not so confused anymore... cool.gif

It must be frustrating for you trying to find dig up some dirt about LS, heh ?

maybe use ur time and help out in building a GUI for Argyll........ rolleyes.gif
post #334 of 710
Quote:
Originally Posted by Manni01 View Post

My Discus reads the black levels of my rs45 fine and CM does read black levels to calculate the correct targets when working on gamma with BT1886 (which is the default since 5.2.x).
I would never use my i1pro2 to do some work on gamma, so I don't see why anyone would use a Jeti 1211 to do so. You need to train a sensitive enough probe to the 1211 to get reliable black readings. The Klein K10 should be fine.

Yes I understand all of that I have a K10-A, my question was to Chad and how he goes about it.knowing he uses a Jeti.

I myself don't care to use 1886 on my VT60, but maybe that's because I know how to use 709 with a power gamma of 2.2 for my display when running a 17^3 profile.
The K10-A is the most sensitive color meter there is and as far as I know the most accurate color meter. I use to have a C6 but found it not to be reliable when reading very low light..

If a high end meter, be it a color or spectro is not repeatable then imo you have a problem. .My weak link is with my I1Pro 2 and how it reads Red, I just sent back two new Jeti's 1201 because they were guessing about reading Red more than my I1Pro 2,. I use it (spectro) to profile my Klein.

Yes I have the offsets from Klein when they were kind to me and compared my I1Pro 2 using the CS-2000 on 6 different displays. One being a older Panasonic Plasma, but still those offsets don't really help that much.

ss.
post #335 of 710
Quote:
Originally Posted by Iron Mike View Post

implements BT.1886 in a way they see fit. As per the specification.


Those are contradictory statements, you can't implement a specification "as you see fit".

ITU.BT1886
Quote:
recommends
1 that the reference EOTF for displays used in HDTV production and programme interchange
should be the one specified in Annex 1;

Quote:
Annex 1
Reference electro-optical transfer function
The reference EOTF is specified by the equation:
L = a(max[(V + b),0])^γ


You can't have it both ways, either call it what it is, Rec709(2.4), or fix it.
Edited by zoyd - 12/11/13 at 4:50pm
post #336 of 710
You guys need to stop banging heads and work together to find out what the intent of "Annex" means.

Apparently, it's up to interpretation which it shouldn't be.

Contact the authors and see what they say.
post #337 of 710
What is it about the word "specification" that you don't understand? They are not open to interpretation, do you think the primary chromaticities of Rec709 are open to interpretation? A specification has a number or formula that is "specified" to be used. The document clearly states the actual, real-life, honest-to-god specified formula, otherwise it wouldn't be a "specification". You'll note the text of the recommendation shown above even uses the word "specified" so that you as a reader can be certain that yes, this is the specification. jeez.
post #338 of 710
Quote:
Originally Posted by sillysally View Post

Yes I understand all of that I have a K10-A, my question was to Chad and how he goes about it.knowing he uses a Jeti.

I myself don't care to use 1886 on my VT60, but maybe that's because I know how to use 709 with a power gamma of 2.2 for my display when running a 17^3 profile.
The K10-A is the most sensitive color meter there is and as far as I know the most accurate color meter. I use to have a C6 but found it not to be reliable when reading very low light..

If a high end meter, be it a color or spectro is not repeatable then imo you have a problem. .My weak link is with my I1Pro 2 and how it reads Red, I just sent back two new Jeti's 1201 because they were guessing about reading Red more than my I1Pro 2,. I use it (spectro) to profile my Klein.

Yes I have the offsets from Klein when they were kind to me and compared my I1Pro 2 using the CS-2000 on 6 different displays. One being a older Panasonic Plasma, but still those offsets don't really help that much.

ss.

I was primarily responding specifically to your statement that Calman did not take any black readings, which is no true.
I can't speak about the C6 as I never had one, but I agree with you that the i1d3 is not reliable in very low light, that's why I sold mine and got a Discus. However, these meters are good entry-level meters for non professionals, otherwise I am sure that LS would not be bundling i1d3s with their software smile.gif.
While my i1pro2 and my Discus are not reference meters (I need to upgrade my display before upgrading my meters or my software can make any kind of visible difference), they are 100% repeatable as long as you know how to use them (which means amongst other things observing warm-up time for both meters and display, taking dark readings when needed and making sure the meters have the same FOV when profiling them). It also helps if you know how to optimize the settings to get a decent profile, and if you know how to check it.
Feel free to do anything you want with your equipment smile.gif. If you're happy with a power gamma of 2.2, I assume you are watching your display in a room which is not dedicated and/or has some ambient light. Because in a dedicated room with no ambient light, 17x17x17 LUT or not, this is not in my opinion the way to get the best results from a decent display (which the VT60 clearly is).
post #339 of 710

Hi

 

There seems to be some frustration around the implementations of the BT.1886 standard.

 

As Zoyd correctly states, the standard is defined in Annex 1.  The information in Appendix 1 is informative and there for not part of the standard.  There may be a mixup in the use of the words Annex and Appendix, which is causing all this frustration.

 

White the document is the standard, it must be correctly implemented within the structure of any tool (software or hardware) for it to be valid.  The implementation is not defined in the standard.

 

BT.1886 as the standard, defines a pure gamma function (power function call it what you will), offset by the mimimum black signal and scaled to the maximum white signal.  If your system already automatically removed black offset and scales to white, then BT.1886 is correctly implemented as a pure power / gamma function.  If your system needs manual offset and scale, then you need the full equation.  Both are absolutly correct and both are to standard, just different due to implementation architecture.

 

Smedith

post #340 of 710

Hi

 

Its not a case of which should be used, they are identical within the architecture of the implementation.

 

BT.1886 has its place, but it is not the be all and end all.

 

It is important that it is correctly implemented.  That does not mean blindly reading and copying the standards document, but implementing it correctly within your own environment.

 

Smedith

post #341 of 710
But in the context of this discussion (LS + consumer displays), consumer displays are unable to implement the shift and scale on their own, so it must be done in software using the formula specified.
post #342 of 710

Hi

 

Why would a user display implement any shift and scale at all?  If the display has a gamma 2.4 between its minimum luminance and maximum luminance, then it is a BT.1886 standards device (as far as the  luminance transfer function goes).

 

The BT.1886 standard does not tell you to set these values, mearly to allow for them when fixing the pure gamma 2.4 between them.

 

In the case of any calibration, be it LS or anything else you need to take account of these values and get the best responce between them.

 

I would assume that your display had been optimised in hardware to give the best contrast ratio etc.  Then you would use a software tool to profile that display to get the very best responce to what ever target you desire.

 

Any form of software LUT can only reduce the output of a display.  In the best case it removes  colour responces that are not valid for your desired target.  Setting ther hardware first gives you the maximum raw responce.  This then gives you the biggest gamut from which to cut out the desired responce.

 

I am happy to read your argument, but I see no real reason to setting offset and scale in software to help achieve a BT.1886

 

Smedith

post #343 of 710
Quote:
Originally Posted by Manni01 View Post

I was primarily responding specifically to your statement that Calman did not take any black readings, which is no true.
I can't speak about the C6 as I never had one, but I agree with you that the i1d3 is not reliable in very low light, that's why I sold mine and got a Discus. However, these meters are good entry-level meters for non professionals, otherwise I am sure that LS would not be bundling i1d3s with their software smile.gif.
While my i1pro2 and my Discus are not reference meters (I need to upgrade my display before upgrading my meters or my software can make any kind of visible difference), they are 100% repeatable as long as you know how to use them (which means amongst other things observing warm-up time for both meters and display, taking dark readings when needed and making sure the meters have the same FOV when profiling them). It also helps if you know how to optimize the settings to get a decent profile, and if you know how to check it.
Feel free to do anything you want with your equipment smile.gif. If you're happy with a power gamma of 2.2, I assume you are watching your display in a room which is not dedicated and/or has some ambient light. Because in a dedicated room with no ambient light, 17x17x17 LUT or not, this is not in my opinion the way to get the best results from a decent display (which the VT60 clearly is).

No I never said CM would not take dark readings, I said that the Jeti would not.. Matter of fact I was agreeing with you. wink.gif

No, I view in a dedicated room, no ambient light unless I want it. I also use bais lighting (12W-65K) when viewing. 99% of what I view are HQ Blu Rays, and as we know most all of these Blu Rays are master in 709 not 1886 so maybe thats why I like 709 over 1886.

Understand I don't paint by numbers anymore, all though I used to do that. iow I don't necessarily shoot for the best gamma line or the best dE's. What my goal is the best PQ that my display will give, so to me what that means is working with what the display will give me not trying to force the display to do something it really can't do just so I can have bragging rights and post my charts and numbers. I have been there and done that. . cool.gif

ss
post #344 of 710
Quote:
Originally Posted by smedith View Post

Hi

Why would a user display implement any shift and scale at all?  If the display has a gamma 2.4 between its minimum luminance and maximum luminance, then it is a BT.1886 standards device (as far as the  luminance transfer function goes).

The display is unable to reproduce this response over the full range of input values. The previous example used a typical minimum luminance of 0.05 cd/m^2 and maximum of 120 cd/m^2. For this example it is only compliant with BT.1886 for the input level range from (0.05/120)^(1/2.4) = 3.9% to 100%. Everything below 3.9% will be clamped to minimum luminance without an offset adjustment. You are throwing away codes 17-25 and under the right viewing conditions the compression of shadow detail in this situation is not acceptable.

Quote:
I am happy to read your argument, but I see no real reason to setting offset and scale in software to help achieve a BT.1886

The argument is not about whether one should use it, but given that one has weighed the trade-offs and decided to do so for whatever reason that it be done using the recommended formula.

Perhaps you missed most of the context of the discussion (aka brouhaha). The software in question (LS) advertises a BT.1886 EOTF emulation for the LUTs their software produces for use in home theater equipment such as the eeColor box. However, as Steve has already admitted they do not use the annex formula to compute the emulation mapping, they use a gamma 2.4 function with no accommodation for offset and scaling factors in the specification. I questioned the validity of this approach since it will not lead to a BT.1886 emulation on any display with a non-zero luminance response at input level 0. And here we are.
Edited by zoyd - 12/9/13 at 1:44pm
post #345 of 710
Quote:
Originally Posted by sillysally View Post

No I never said CM would not take dark readings, I said that the Jeti would not.. Matter of fact I was agreeing with you. wink.gif

No, I view in a dedicated room, no ambient light unless I want it. I also use bais lighting (12W-65K) when viewing. 99% of what I view are HQ Blu Rays, and as we know most all of these Blu Rays are master in 709 not 1886 so maybe thats why I like 709 over 1886.

Understand I don't paint by numbers anymore, all though I used to do that. iow I don't necessarily shoot for the best gamma line or the best dE's. What my goal is the best PQ that my display will give, so to me what that means is working with what the display will give me not trying to force the display to do something it really can't do just so I can have bragging rights and post my charts and numbers. I have been there and done that. . cool.gif

ss

You obviously don't know me or you wouldn't have written the above smile.gif I recently posted in the Sony 500ES thread saying that I wouldn't personally try to calibrate the unit I was testing with my i1pro2/discus because I had no idea whether I would get the display closer to reference or simply produce nicer looking charts. I even said that I would need a Klein K-10 profiled to a Jeti 1211 at a minimum to stand a chance to not make things worse.

You clearly have no idea what you are talking about re 709 vs BT1886, so I'm not going to discuss the subject with you any further. Until about four years ago, most blurays were mastered in SMPTE-C and not even in rec-709. Gradually, as post-prod houses got rid of their CRTs and started using digital monitors, they moved to rec-709, but Rec 709 doesn't specify any gamma function, so anything could be used: power 2.2, power 2.3, power 2.4, power 2.5... Absolutely anything. BT1886 is the first standard that specifies a colorspace (rec709) AND a gamma function (2.4). And even now, not everyone seems to be using it. Some say it's the new standard, others say it's already obsolete. Go figure.

There is absolutely nothing in rec 709 that says a power gamma of 2.2 should be used, so if you think you are closer to what was mastered simply because you use rec 709 + power gamma 2.2 you are hugely mistaken.

Anyway, someday you should try a proper BT1886 implementation and compare it to a power 2.2 in your dedicated room, who knows, you might like what you see smile.gif.

I'm out now.
post #346 of 710
Just as a general reply, as there is a limit to the point of trying to explain things...

The naming of the function as a power law with offset is just semantic. It can be referred to by many different names.

To answer specifically, equation 1

L = a(max[(V + b),0])^γ

Is identical to equation 2

L = V^γ

If you pre-correct for the black offset and white gain, which is how LightSpace works and has always worked.
So we use both as in use they are identical.

Beyond that, it's worth realising (and I have tried repeatedly to explain this) LightSpace has always measured minimum and maximum luminance.
BT.1886 specifies minimum and maximum luminances and then fits the gamma 2.4 between them.
Within LightSpace we are exactly correct, as we automatically deal with minimum and maximum offset and scale, exactly as BT.1886 specifies.
LightSpace has no need to specify the min and max values manually.

The BT.1886 standard is the standard, it is the annex informative bit that is not part of the standard.

With in the implementation of LightSpace BT.1886 and REC709 +gamma 2.4 are and should be identical

BT.1886 does not specify a contrast ratio for the display nor does it specify the minimum and/or maximum display luminances.
If your contrast ratio is too low then you may get crushing in the blacks.
This is because the display is not of a high enough quality and can not resolve these variations in luminance - the calibration is not the problem though.
If this is the case with a consumer display, you can manually compensate by adjusting the transfer function (contrast, gamma what ever you call it) to be non-standard.
This makes better looking (non crushed) images, which is the point, but it is not the calibration standard (but that is fine).
There is a visual compromise between the contrast look of the image and dark detail.

The bottom line is the specification is the specification.
The Annex is just that - an annex with additional information that is NOT part of the specification.
As there seems to be interest in this we will add the capability...
But it is still not part of the specification.
But with bad displays it may have some user validity in making pretty images - NOT calibrated images.

And as stated - the annex info is not used by any mastering houses that we know of, as it is not part of the specification!
If you want to match the graded intent of the film you watch do not use the annex info.

Hopefully this helps, but this will our last comment on the subject.
We have better things to be doing wink.gif
post #347 of 710
Quote:
Originally Posted by Light Illusion View Post

To answer specifically, equation 1

L = a(max[(V + b),0])^γ

Is identical to equation 2

L = V^γ

Lets just stop here.

This is bold face incorrect. To try and tell an intelligent community that these two things are equivalent, it demonstrates a serious lack of understanding.
Edited by sotti - 12/9/13 at 10:04am
post #348 of 710
Quote:
Originally Posted by Manni01 View Post

You obviously don't know me or you wouldn't have written the above smile.gif I recently posted in the Sony 500ES thread saying that I wouldn't personally try to calibrate the unit I was testing with my i1pro2/discus because I had no idea whether I would get the display closer to reference or simply produce nicer looking charts. I even said that I would need a Klein K-10 profiled to a Jeti 1211 at a minimum to stand a chance to not make things worse.

You clearly have no idea what you are talking about re 709 vs BT1886, so I'm not going to discuss the subject with you any further. Until about four years ago, most blurays were mastered in SMPTE-C and not even in rec-709. Gradually, as post-prod houses got rid of their CRTs and started using digital monitors, they moved to rec-709, but Rec 709 doesn't specify any gamma function, so anything could be used: power 2.2, power 2.3, power 2.4, power 2.5... Absolutely anything. BT1886 is the first standard that specifies a colorspace (rec709) AND a gamma function (2.4). And even now, not everyone seems to be using it. Some say it's the new standard, others say it's already obsolete. Go figure.

There is absolutely nothing in rec 709 that says a power gamma of 2.2 should be used, so if you think you are closer to what was mastered simply because you use rec 709 + power gamma 2.2 you are hugely mistaken.

Anyway, someday you should try a proper BT1886 implementation and compare it to a power 2.2 in your dedicated room, who knows, you might like what you see smile.gif.

I'm out now.

What are you talking about, of course I don't know you and why would I want to know you. All you keep doing is misquoting me.

I generally talk in term others can understand, not just the geek hobbyists.

I said 709 nothing about power 2.2. Please reread my post that you quoted from..

As far as a "proper BT1886 implementation" goes you know nothing about LS and how a Profile works. I suggest you start reading up on First what a proper implementation and how a 17^3 (17x17x17)
profile works. Please look at the picture below, and maybe you will start to understand.wink.gif



As you can see from the picture above, all you do is take one of your profiles you have already done. Put that profile in the box that says destination color space (Prof2-1126) then you choose from the drop down menu what you want to use like REC-709, ITU-R BT1886 or whatever Then you have three other choices, peak luma, peak chroma or fit chroma. Once you have made your choices you hit create new and in about 3 seconds you have your LUT. So my point is I can create a REC-709, ITUR BT1886 or whatever using the same profile. Load both LUT's into my eecolor and compare by just switching between the two. Can you do this using the same profile/calibration.???
So please don't tell me to compare because I have, using the same 17^3 profile.
If you don't understand what a 17^3 profile is, ask me and I will show you a picture and what it means.cool.gif

All REC-709, ITU-R BT1886 or whatever is a personal choice, I explained why I like 709 for my display. If you like 1886 then I say more power to you. Its really that simple. smile.gif

ss
post #349 of 710
Quote:
Originally Posted by Light Illusion View Post

Chad, if the probe/display combination can read the black level 'once' LightSpace will use that value.
There is no need to input that value manually as it will have been read.
The Average Low Light option in LightSpace makes sure if the probe can read a black level it will get used.
But, if no valid level is available you can't enter it manually. other than via guess work as no valid reading will have been taken ever...
So the only option is to 'guess' it, or let the maths calculate what the value would have been.
And that calculation is far more accurate than a guess...

cool.gif
Quote:
Originally Posted by Iron Mike View Post

In addition, if u know LS and it's formats, u can enter a black level if u really want to... but what's the point, if ur probe can't read black u're only gonna enter a best guess...

it also depends on the settings and mode You're using. As a LS HCL customer + TEDs disc I use DIP mode with meter integration time of 1.04s (DIP time 4s). With this settings the probe can not read the mll of my display, but it accurate enough for stimulus levels >10% (maybe 5% too). To get the meter reading the mll correctly I have to use integration time of 6s but than I will loose the speed advantage completely.
So for this particular case it would make sense to enter the black level manually, without guess work.

And regarding gamma, I can not understand why You all stick on 2.2. The main reason I purchased LS was to calculate several targets without running a calibration over and over again, so I have 4 different gamma targets in my Radiance and just use a gamma setting the movie looks best to me (most times 2.25 or 2.3). rolleyes.gif
post #350 of 710
Quote:
Originally Posted by sillysally View Post

What are you talking about, of course I don't know you and why would I want to know you. All you keep doing is misquoting me.

I generally talk in term others can understand, not just the geek hobbyists.

I said 709 nothing about power 2.2. Please reread my post that you quoted from..

As far as a "proper BT1886 implementation" goes you know nothing about LS and how a Profile works. I suggest you start reading up on First what a proper implementation and how a 17^3 (17x17x17)
profile works. Please look at the picture below, and maybe you will start to understand.wink.gif



As you can see from the picture above, all you do is take one of your profiles you have already done. Put that profile in the box that says destination color space (Prof2-1126) then you choose from the drop down menu what you want to use like REC-709, ITU-R BT1886 or whatever Then you have three other choices, peak luma, peak chroma or fit chroma. Once you have made your choices you hit create new and in about 3 seconds you have your LUT. So my point is I can create a REC-709, ITUR BT1886 or whatever using the same profile. Load both LUT's into my eecolor and compare by just switching between the two. Can you do this using the same profile/calibration.???
So please don't tell me to compare because I have, using the same 17^3 profile.
If you don't understand what a 17^3 profile is, ask me and I will show you a picture and what it means.cool.gif

All REC-709, ITU-R BT1886 or whatever is a personal choice, I explained why I like 709 for my display. If you like 1886 then I say more power to you. Its really that simple. smile.gif

ss

I have first hand experience of LS, and I do understand what 17x17x17 LUTs are, thank you very much.

You are still very much confused and this is getting silly, Sally. As I said I'm not discussing this further, at least not with you.

I sincerely hope that this is just a hobby for you and that you are not selling your services professionally smile.gif
post #351 of 710
Quote:
Originally Posted by sotti View Post

Lets just stop here.

<<>>.

Joel, as a long term user and a very good customer of CM. The above is just plain nasty, and has no place coming from someone like you a representative of CM in a LS board.

I took out what you said in my quote incase you decide to delete your post. I will do the same.

ss
post #352 of 710
Ignoring the stupidity in Sotti's post, the reality is rather obvious.

In a very simple case you can set a = 1 and b = 0. In this simple case,

L = 1.0 x (V + 0.0) ^ y

Which is the same as :

L = V ^ y

The value of a is simple a scalar value, so L = aV ^ y, is identical, just scaled.

The b is simply an offset, it just moves the gamma / power curve up and down, by subtracting black

As stated initially, all values in LightSpace are normalized by removing a black offset, and scaling the result so max luminance is 1.0

This process is EXACTLY the same as that used in BT.1886, so within the architecture of LightSpace, the 2 equations are identical.

If you don’t quote the whole response then it does not make sense and is easy to deliberately miss interpret.
But then that was obviously the intent from someone who really doesn't understand the underlying mathematics.

Sotti, you really should have known better!

frown.gif
post #353 of 710
Quote:
Originally Posted by DrFaxe View Post

So for this particular case it would make sense to enter the black level manually, without guess work.:

If you export the XML profile data you can easily manually edit the black reading biggrin.gif
post #354 of 710
Quote:
Originally Posted by Light Illusion View Post

In a very simple case you can set a = 1 and b = 0. In this simple case,

L = 1.0 x (V + 0.0) ^ y

Which is the same as :

L = V ^ y

So any two formulas that have an intersection where they produce the same result are the same?

Please enlighten us.
post #355 of 710
Quote:
Originally Posted by Light Illusion View Post

Ignoring the stupidity in Sotti's post, the reality is rather obvious.



The b is simply an offset, it just moves the gamma / power curve up and down, by subtracting black

b is not the black level,

b=LB^(1/γ)/(LW^(1/γ) - LB^(1/γ))

Quote:
Originally Posted by Light Illusion View Post

If your contrast ratio is too low then you may get crushing in the blacks.
This is because the display is not of a high enough quality and can not resolve these variations in luminance - the calibration is not the problem though.

I really find your attitude shocking. You are saying that instead of providing your users the clearly specified target EOTF and letting them decide whether the additional dark detail is worth the trade-off in color calibration accuracy, they should go out and buy a better TV?
Edited by zoyd - 12/11/13 at 4:52pm
post #356 of 710

Hi Zoyd

 

 

Thanks for your response.  I understand your maths, but I think you have got this quite wrong.

 

A display has a minimum and maximum luminance.  Un-calibrated by any external LUT, minimum code (0, 8bit RGB) will produce minimum luminance and maximum code (255, 8 bit RGB) with produce maximum luminance.  If the response curve between these 2 points is a power curve of 2.4 it is BT.1886 standard.  That is a fact, but it does not mean the display will produce a nice image or that the blacks are not visually crushed or clipped.

 

Your maths

 

(0.05/120)^(1/2.4) = 3.9%

 

Is not correct this assumes 0 code value should correspond to 0 CdM-2 luminance.  No display produces 0 luminance and the standard for BT.1886 does not expect this.  To calculate this as 3.9% in this way is simply incorrect maths

 

BT.1886 states that 0 code values should map to minimum luminance not 0 CdM-2, which in your example would correctly be 0.05.

 

The problem with clipping and crushing is down to the contrast ratio.  If this is too low, you can not visually resolve the luminance difference between codes starting at 0.  The original design of 8 bit digital video set the visual perceived difference in luminance of 1 bit above black as just not visible.  At 2 bits above black, the difference should be just seen.  If this is the case, you have a good system.  You can find this in Poynton.

 

When you activate a colour management LUT it operates on the data code values not the light levels.  As such, even through a LUT, the code value 0, should still map to the minimum luminance the display can produce (unless you deliberately lift it, which would be madness).  The code value 255 will map to a luminance of just lower than the maximum luminance of the display (as you will lose some light power in 2 of the 3 channels to achieve calibration).  As such, the LUT must be operating on the display above the minimum offset, so the “b” parameter can be correctly set to 0.  The LUT will change the over all scale, but the gamma curve from 0 to 255 will still be 2.4 (or what ever else you desire).

 

In short BT.1886 is a mapping between a normalized signal into a display and the light out of the display.  It is a simple power function of 2.4, mapped between minimum luminance and maximum luminance.  If you measure a display and subtract a black offset and scale by the peak luminance, the functions of BT.1886 simplifies correctly to a power function of 2.4.  If LS is operating by subtracting black and scaling, then it is correctly implementing BT.1886.  If it is working on absolute luminance values, then is wrong, simple as that.

 

Just noticed your reply latest reply, so Ill add some.

 

The value b is a simple offset, the maths you quote simply moves it from the luminance domain to the code domain to simply code level implementation.  You can do the proof to show this.  I don’t mean to defend LightSpace specifically, but your statements are only true from one perspective and that is of absolute luminance values.  If that’s how CM works, then I understand the confusion, but it is not the only way.  If as I suspect, LS uses relative offset and scaled data, then it too is absolutely correct.

 

 

Smedith

post #357 of 710
Ah - it has just been pointed out to me that I did make a mistake!

I used 'Annex' when I actually meant 'Appendix'.

lol

That may have caused some potential confusion, but anyone who has actually read the specification will have spotted that and understood it was a simple mix-up of words rolleyes.gif
post #358 of 710
Quote:
Originally Posted by smedith View Post

In short BT.1886 is a mapping between a normalized signal into a display and the light out of the display.  It is a simple power function of 2.4, mapped between minimum luminance and maximum luminance.  If you measure a display and subtract a black offset and scale by the peak luminance, the functions of BT.1886 simplifies correctly to a power function of 2.4.  If LS is operating by subtracting black and scaling, then it is correctly implementing BT.1886.  If it is working on absolute luminance values, then is wrong, simple as that.

So lets just talk about step sizes then.

With a simple offset function the difference between bit 18(.913%) and bit 20(1.826%) with the straight power function is always have a difference of 0.0000545, normalized 0-1.

If we calculate that for a black level of 0.05 and white level of 100 we get the same normalized value and an actual step size of 0.00545 cd/m

Now with BT.1886 stepsize with those black and white levels would be 0.036443 cd/m, that is 7x as big of a step.

There is a substantial difference in shadow detail between the two algorithms, any argument to the contrary is ignorant or malicious.
Edited by sotti - 12/9/13 at 11:39am
post #359 of 710
Quote:
Originally Posted by smedith View Post

Hi Zoyd


Thanks for your response. I understand your maths, but I think you have got this quite wrong.

A display has a minimum and maximum luminance. Un-calibrated by any external LUT, minimum code (0, 8bit RGB) will produce minimum luminance and maximum code (255, 8 bit RGB) with produce maximum luminance. If the response curve between these 2 points is a power curve of 2.4 it is BT.1886 standard. That is a fact, but it does not mean the display will produce a nice image or that the blacks are not visually crushed or clipped.

Your maths

(0.05/120)^(1/2.4) = 3.9%

Is not correct this assumes 0 code value should correspond to 0 CdM-2 luminance. No display produces 0 luminance and the standard for BT.1886 does not expect this. To calculate this as 3.9% in this way is simply incorrect maths

BT.1886 states that 0 code values should map to minimum luminance not 0 CdM-2, which in your example would correctly be 0.05.

Thank you as well. I used those values because that is what LS assumes (not me). By not using the specified offset and scaling factors (a,b) they implicitly assume the perfect display device (BT.1886 in the limit of zero black), I'm sorry I didn't mention that. So yes, the math is correct under that assumption. The offset and scaling factors they use are not correct.
Edited by zoyd - 12/9/13 at 1:48pm
post #360 of 710

Hi Sotti

 

Thanks for your reply.  As you can tell, I am new to this forum, so I am sorry if I have offended your position in some way, but I think your response was a bit harsh.  That said, I am not a novice to this area.

 

Your quoted maths makes no sense.  If you are going to give mathematical proof to back an argument, please can you make it more accurate?

 

In an 8 bit system, going to a ideal monitor with a perfect power function EOTF, then assuming normalized luminance, the code value 1 produces  0.000001617 = (1.0 / 255) ^ 2.4 of the luminance.

 

If it is a real monitor, then code value 0 will not produce 0 Cdm-2, but minimum luminance.  So, code value 1 must produce min +( (max –min) x 0.000001617).

 

This shows that the minimum step in luminance (code 0 to code 1)is dependant only on the physical transfer function of the display, not its minimum and maximum luminance values.

 

If you then use a software LUT to manipulate this further, then this minimum step value from 0 to 1 must get worse due to quantization error.  A BT.1886 calibration is based on the actual characteristics of the monitor, min and max luminance.  The black value “b” is just an offset.  Once the offset has been removed the power function should them be applied.  You could increase the offset in the LUT, but this makes no sense at all!!

 

I would be happy to keep discuss further, but I would appreciate it if you could tone your replies in a more professional way.  If this is the tone your company takes on an open forum with a potential new customer, it makes me question if I should actually purchase your product for my facility.

 

I will sign off this thread now.  As a last note I would say, there are may correct ways to implement BT.1886, which are valid and to the standard.  Just because one tool does it differently to another does not make it wrong just because you don’t understand how they do it.  It serves no professional integrity to just bash on a competitor and shout loud in public just because they are different.  I will now do further research into LS to see if it fits my needs before I commit.

 

regards

 

Smedith

New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Display Calibration
AVS › AVS Forum › Display Devices › Display Calibration › LightSpace CMS Now Supports Lumagen + eeColor 3D-LUT 4 All