AVS Forum banner

Modifying White Point / Color Space

5K views 88 replies 12 participants last post by  zoyd 
#1 ·
My television's white-balance/color temp controls don't perform well enough to allow me to get my grayscale to reach D65 without significant (gamma and/or chromaticity) error. So, short of new hardware, I am forced to compromise. I believe there are four ways to compromise (i.e., a workflow approach) as follows:

1. Minimize gamma error at the expense of chromaticity.

2. Minimize chromaticity error at the expense of gamma.

3. Seek and accept some amount of error in both gamma and chromaticity that is neither minimal for either but represents some semblance of "error balance".

4. Perhaps controversially/illegitimately "calibrate" to a slightly higher white point. HCFR software currently allows this to be done rather easily.

So a couple of questions follow:

1. Which compromise above is best in your opinion and why?

2. I know the formula for finding out the x and y coordinates of a different white point on the daytime locus. So if, hypothetically and potentially against the advice of the community I nevertheless choose option 4, would the coordinates for the primary colors (RGB) need to also change from the sRGB/Rec709 coordinates? And, if so, where is the formula for backing in to the alternate RGB coordinates based on a manually chosen white point?
 
See less See more
#11 · (Edited)
To answer the need to know the make and model; it is important. Each manufacturers and models have different idiosyncrasies. With Sharp, their displays push about 9000 K ~ 10000 K white point out of the box. Many professionals have had difficulty setting the D65. Most of the time the best that is achieved is around 7500 K but this is not without sacrificing something along the way. It is as if Sharp had not bothered to adjust their displays from the Asian 9300 K white point standard to the REC-BT709 D65 white point standard. Others have stated that on some models the CMS features do not work as they should.

If you read posting #6 of this thread by "johnfull", you will see he states, "I have had 2 Quattrons in the last 3 years, which I have set up with the AVS calibration disc and my own perception of white point". Why did he need to create his own white point? Is it that he could not achieve D65?

Your statement, "Sharp Aquos Quattron LC-70LE847U...although I can't imagine why that would affect the answers. Do the principles of calibration change from display to display? Wouldn't that be like the laws of physics changing from city to city"? Yes, it does effect the answer. The principles of calibration do not change, it is dealing with the quarks of the display to achieve the desired settings of a calibration this makes it more difficult. No the laws of physics still apply from city to city. Many better than you or I would opt out for option #4 from the option list from the first posting. If I were doing such a display, I too would probably settle for #4 .
 
#6 · (Edited)
I'm not an expert, but I have had 2 Quattrons in the last 3 years, which I have set up with
the AVS calibration disc and my own perception of white point. The blue LED that supplies
the primary light for these sets is the culprit for a lot of the problems that they exhibit.
You almost certainly have to use a higher temperature because of the predominance of blue
light source. Compound that with the deliberate use of yellow to complement the excess
blue and you have a chrominance nightmare -- until you get it tamed.
Sharp uses the industry standard blue LED to generate white light, with a phosphor that
radiates in yellow frequencies as well as green and red. They give no tools for directly affecting
the yellow pixels, other than the CMS, which only affects chromacity.
My best efforts are published for my new set and my old set. See what you think:

http://www.avsforum.com/forum/166-lcd-flat-panel-displays/1568122-sharp-uq17u-calibrations.html

http://www.avsforum.com/forum/166-l...arp-lc-70le845u-calibration-including-3d.html

The second listing is for my old set, which is very similar to yours.
Lots of people tried the settings and found them pleasing.
I can say that my first set was much easier to calibrate than the new
one, which more aggressively uses the yellow pixel and is harder
to tame.
 
#7 ·
Thanks but I'm a firm believer that using other people's settings is worse than useless.

I thought my original post would spark a general discussion above the idiosyncrasies of any particular TV. So I'm hoping the responses to the original post are still to come.
 
#8 ·
I understand, but there is general hostility to Quattron among calibrators because of the
hoops you have to jump through. Adding the yellow wavelength of light, in addition to
the yellow generated with R+G, causes hardware calibration equipment to misread too
much blue. This has been disputed, but everyone who uses hardware comes up with
a decidedly green/yellow cast to their white balance and without other explanation.
Anyway, the readings are NOT worse than out of the box and are there for you to use
or not. You can always re-set to factory specs and never mention it again...;)
 
#9 ·
I understand, but there is general hostility to Quattron among calibrators because of the
hoops you have to jump through. Adding the yellow wavelength of light, in addition to
the yellow generated with R+G, causes hardware calibration equipment to misread too
much blue.
How can adding extra yellow energy cause a meter (at least a spectro) to read "too much blue?"

Let's be clear here. The hostility (at least on my part) comes as a reaction to Sharp completely breaking the HDTV/REC709 color system and marketing the results as a "fix" to something that was not broken. Then subsequently offering the calibrator *no way* to undo the so-called "fix."

Sorry for the rant ... first day back on caffeine in a while ... bit twichy. :)
 
#20 ·
So which compromise approach do you recommend and why?
I can't answer that without looking at the set and the calibration results. But of your original choices, I would tend toward Option 2: Sacrificing the gamma. Of course, that would depend on the actual results: i.e. whether that would actually allow you to hit D65 and minimize r.709 gamut errors.
 
#15 · (Edited)
You bring up a point that is very interesting, do you have any documentation for us to read? I am interested in what Sharp is doing. I can not help but feel that Sharp is skirting the standards but this is only a hunch.

Here is something that I came across;

Analysis of Quattron

Color researchers at Queen Mary University of London investigated the Quattron technology and found that although Quattron does have 4 physical color sub-pixels it does not have a fourth primary in the backlight to drive it (yellow is approximately 575 nm). Quattron has a yellow sub-pixel but the manufacturer has not made any provision to produce the yellow light needed to pass through it. On that basis they conclude that it serves no useful function.[11]
The lack of a fourth primary is clearly shown by a spectrogram where the red primary is given separately from the green primary. Yellow may be seen to be simply the sum of red and green. A yellow push may also be observed when yellow is compared with the red and green primaries. The spectral power of yellow is approximately twice that of red and green. Yellow colors will therefore appear more prominently on a Quattron display than red or green colors, but this is because the manufacturer has 'pushed' yellow to a higher luminance rather than having improved the color to more closely approximate a natural yellow (such as the yellow of a sunflower), which has a flat spectral power distribution from lime-green to deep red wavelengths. Green dominates the Quattron yellow and this is confirmed perceptually by color matching.[11] The human visual system is particularly sensitive to yellow, and as a result of the lack of a yellow primary (or a yellow boost to either red or green) there will be difficulty in discriminating between green, red and yellow. The reproduction of a yellow image will as a result always be perceived as slightly green or slightly red (or orange) but never a pure natural yellow.


Criticism

According to an analysis published in MaximumPC Magazine by Raymond Soneira, president of DisplayMate Technologies, a video calibration equipment producer, Sharp's Quattron technology does not have the ability to show more colors than a standard RGB set. He argues that, due to industry-standard color spaces used by content providers, there is no existing source material that contains the fourth color channel. He further states that any "extra" colors displayed must simply be created in the television itself through video processing, resulting in exaggerated, less accurate color
 

Attachments

#17 ·
The white LED backlight generates red, yellow, and green wavelengths from its yellow phosphor. A colorimeter set up to see only red and green will not accurately record the yellow spectrum.
Sharp was onto something with a tetrahedron-shaped color space,
similar to some of the better desktop printers that use multiple
primaries. A benefit was to be pure yellow from the far righthand
margin, and bluegreens from the interior of the colorspace that is'
untouched by conventional RGB devices. Sony used a blue-green
without the compensating yellow pixel and stopped after one year.
You can't have pure yellow and robust cyan on a triangular plot.
When backlighting improves, I hope Sharp will return to Quattron in
the 4K sets and produce a truly wide spectrum -- and one that is
controllable by calibrators! As it is, the yellow adds a great deal of
efficiency to the light output, in particular because of the overbearing
blue LED; B+Y=white, afterall, which boosts the R+G+B=white considerably...
 
#25 ·
I won't quibble over language here -- colorimeters don't actually 'care' about anything.
My question is about the range of each of the three receptors in the equipment.
Yellow lies outside of either red or green, but may or may not be perceived as one or the other or both.
It just depends on the equipment. Sharp is using the yellow part of the spectrum in the phosphor
that would ordinarily be thrown out by red and green filters on an RGB set. It uses this yellow to
alter the white point of the final image.
 
#26 ·
My question is about the range of each of the three receptors in the equipment.
Yellow lies outside of either red or green, but may or may not be perceived as one or the other or both.
It just depends on the equipment. Sharp is using the yellow part of the spectrum in the phosphor
that would ordinarily be thrown out by red and green filters on an RGB set. It uses this yellow to
alter the white point of the final image.
Colorimeters don't have a "range". They have filters which, in the ideal, match the standard observer functions, which in turn are a set of functions that describe how the average human integrates radiant energy across the spectrum.

Thus, a colorimeter, in the ideal, will provide an accurate measure of what "color" the average human sees.

There are typically two sources of uncertainty, however:

1: mismatch between any given colorimeter and the standard observer functions
2: the fact that there is variance within the visual systems of the human population, which means that even an ideal colorimeter will not accurately predict what color every single human will see.

Importantly, both of these sources of error tend to be exacerbated with narrow wavelength sources.

None of what you are saying makes any sense to me, in light of this. You seem to be operating with a few conceptual flaws.
 
#27 ·
Colorimeters are set up around Red, Green, and Blue wavelengths for the sake of economy.
Sensitivity is not Panchromatic, but Orthochromatic in a colorimeter, correct?
If a wavelength of 580 nanometers strikes a colorimeter's pickup, the results will be something
that is iterpreted either as red or green or not seen at all. 570-590 nanometers is the range of
yellow light and there has been, heretofore, no reason to design a colorimeter that is sensitive
to this range or software to take it into account in calibrations. The white LEDs have a large
part of their spectral output in this range, but a RGB set will filter it out. Not so a Quattron.
If there is a mismatch between the sensitivity of the colorimeter and the output of the set, then
the calculations will be thrown off. Spectrometers are panchromatic, but not colorimeters, right?...
 
#28 ·
If a wavelength of 580 nm strikes a colorimeter, it will produce three quantities, which correspond to X, Y (luminance), and Z. These in turn can be transformed into chromaticity coordinates (x,y) which represent the particular color that a human would see if a wavelength of 580 nm struck her eye.

I've never heard of the terms panchromatic or orthochromatic, but a quick google search suggests that they are concepts that relate to types of photographic film, and are not appropriate for discussing colorimeters.

A colorimeter will integrate light across the entire visible spectrum.

I've given you this link before, but I'm not sure if you read it. Give it a read and see if things start to make more sense. I've also given you links to craig blackwell's youtube series on color vision, which is an excellent introduction to the CIE system. Again, here is part 1 (watch the first four parts).

 
#29 · (Edited)
Another thing that bolsters my theory is the trouble that some calibrators are having
with extremely narrow sources of spectrum -- like Quantum Dots. Colorimeters may miss
part of the output by having a peak sensitivity other than the specific wavelength emitted.
This lends credence to the idea that the yellow output of a Quattron is missed by colorimeters
that are sensitive in the red and the green, but not between.
You can find the discussion on Sony Quantum Dots and false readings
that result from the narrow spectra elsewhere.
You're a bit glib about the capabilities of colorimeters.
 
#30 ·
This lends credence to the idea that the yellow output of a Quattron is missed by colorimeters
that are sensitive in the red and the green, but not between.
Quattron's may have issues with narrow wavelengths and poor primary selection (I don't know and I don't care), but you clearly haven't a clue how colorimeters work, and presumably have no wish to learn.

I'm done trying to explain things to you. Enjoy your theories.
 
#31 ·
Sony had a problem with the narrow spectrum of its green and red quantum dots.
I used that as an analogy to the problem of yellow in Quattrons.
Colorimeters are not equally sensitive to all spectra. The cost would be prohibitive.
Most are set up to examine output from RGB devices, which means all other color ranges
of the spectrum are either ignored or rolled into one of the primaries.
You don't care about Quattron and the problems of machine-calibration. Fair enough.
But you are putting out far more heat than light on the subject.
This forum is rife with examples of colorimeters lacking profiles for certain output ranges.
People are sharing info and creating their own profiles, particularly on the Sony Quantum Dot post.
Since there is a phobia against Quattron, calibrations are left to hacks like me, who use the
calibration discs to set up white and black and white temps without the mathematics. Pity us...
 
#34 ·
By reading the numerous postings that you have made, I am having some difficulty following and/or agreeing with your logic.

First, when you reference colorimeter's inability to read various colors I would say, "I can't agree". I feel that the colorimeters work but it is the display that is the cause of the issue. With Sharp's introduction of trying to include a secondary color as a primary has introduce an array of issues. In many opinions this is a bad design. Even with LG's OLED display, LG 55EC9300 with the introduction of a white sub-pixel maybe questionable. This LG display is so plagued with issues of bad designing I would highly recommend, "Run ! , don't walk from this display". IF you look at displays and the issues revolving around them you will always find that the fault is not with the meter but with the engineering of the display. For what ever reason you wish to come up with, "Brain Farts" with the adding of sub-pixels or "Delusions of Adequacy" with regards of making the controls function as they should, the fault seems to always lay with the manufacturer. I have been around long enough to see many ideas from the manufacturers come and go and as for Quantum Dots, the jury is still out. This concept is still filled with a lot of issues. For all in all, the meters work and the science and engineering is sound. Due to the bad engineering of displays a secondary market and product has emerged; Video Processors. These little black boxes are designed to correct the inefficiency of the display. If the displays were designed properly these so called black boxes would be of little value.

The major problems with colorimeters is when individuals use them without profiling them against a spectrometer. You use the term "Phobia", I would use the term "Cautious".

As with most threads, this thread has been high-jacked away from robarivas. Did anyone really answer his question? We all owe him an apology. Instead, individuals come out of the woodwork to argue over non-related subjects like the color of red instead of assisting the one who is in need of assistance.
 
#36 ·
The major problems with colorimeters is when individuals use them without profiling them against a spectrometer. You use the term "Phobia", I would use the term "Cautious".

As with most threads, this thread has been high-jacked away from robarivas. Did anyone really answer his question? We all owe him an apology. Instead, individuals come out of the woodwork to argue over non-related subjects like the color of red instead of assisting the one who is in need of assistance.[/QUOTE]

I offered my quasi-calibration of my own Quattron sets (2) and
he respectfully declined, saying that out of the box is better than
trying to adjust with a calibration disc, but without a colorimeter
and software. He and I disagree on that point and I have more
than a dozen folks who have used my settings with satisfaction.
Meantime, the priest class continues to defend the status quo,
even though it stifles innovation and refuses help with what is
already on the market. Backlight quality is my biggest complaint
with modern LCD displays. Back to CCFL, I would say, but that
puts me in your camp of stifling innovation.
Sharp shrewdly dropped the wide gamut claim when backlight
quality fell, just as they were introducing Quattron. Bad timing!
They switched to claiming better yellows (false) and brighter
whites (true), thanks to the efficiency of phosphor emission filtering.
I would prefer less brightness and more gamut, but that's just me.
As for crawling out of the woodwork, I have been a participating
AVS member for many years, thanks.
 
#37 ·
"...Any perceived errors that result from using a colorimeter are a function of how different your vision is from the standard observer and how different the colorimeter filter functions are from the standard observer, and have nothing to do with an imaginary lack of sensitivity to yellow. These perceptual errors are accentuated by displays with narrow spectral peaks, this is a well known issue in colorimetry."

So, to choose a medium-hi white temperature setting and
smooth out the humps of discoloration with either a 2-point
or a 10-point adjustment is not at all recommended?
I haven't seen a calibration from a colorimeter that comes close
to the clean white that can be generated with the eye on these sets.
Three different folks have posted colorimeter/software calibrations
and all but one of more than a dozen users prefer my settings to
those. Theirs result in a yellowed-white. No one is answering why
the results come out wrong! I would prefer to set up my Movie
setting with 10 Point, properly calibrated and have fudged the
results of those guys' settings with a bit of blue to decent effect.
But then there is the fleshtone problem which they don't even
begin to solve and they resort to turning the Global Tint control (!).
The Quattron sets have been out for 5 years and the calibration
community has preferred to malign rather than to innovate. It's what
comes when innovation is from a relatively small player and is not
adopted by more. It doesn't mean it's not a good innovation but
simply that the market weeds out innovations all the time.
Sharp was a pioneer in flat screens, just as a reminder...
 
#38 ·
So, to choose a medium-hi white temperature setting and
smooth out the humps of discoloration with either a 2-point
or a 10-point adjustment is not at all recommended?
I haven't seen a calibration from a colorimeter that comes close
to the clean white that can be generated with the eye on these sets.
However you deal with perceived problems generated by your colorimeter calibration is of course up to you since it is no longer in the realm of calibration. My comment was only to correct your misinformed view of where the problems originate from. Colorimeters will register a signal for any normally visible color in the spectrum, they are not limited to R,G,B areas of the gamut. If the Quattron has generally agreed upon perceivable errors even though the colorimeter says it doesn't, it's because of the shape of the entire spectrum and not which colors were chosen as primaries.
 
#39 ·
Or it could be what the software does with that data -- if the colorimeter puts out numbers
for the three primaries, what does it put out for yellow? And how is that figured into the
calculation with the vectors? Yellow will definitely make white light mixed with blue, so the
mis-application of the yellow data could still give a false reading that blue is too intense.
Right? SOMETHING is making the SpectraCals read yellow whites and green fleshtones!:confused:
 
#40 ·
Thanks to all responses thus far. While there has been a side discussion going on I'm nevertheless appreciative. 2 responses really stood out for me and I have some follow-up questions/comments on those as follows:


1) john stephens: I really appreciate the hyperlink. And I do think that if I shoot for a different white point, the answers about what to do with the primaries and how to do it are contained in that link. However, I must admit that the content is too far above my head for me to put it to practical use (even with the matrix for D65 to D75 already laid out). Furthermore, I have not heard a great deal of support thus far for choosing a different white point. Indeed I'm wondering if the only value of increasing the white point is to make charts look better.


2) zoyd: Your response about always minimize dE is attractively simple. However, if that is the best course of action then which dE should I be looking to minimize? HCFR has three: "without gamma", "Relative Y", and "with gamma". Using D65 as my white point, I can often get "without gamma" and "Relative Y" dE's below 3. But, using D65, my "with gamma" dE is always above 5 and sometimes as high as 9 or 10.
 
#41 ·
2) zoyd: Your response about always minimize dE is attractively simple. However, if that is the best course of action then which dE should I be looking to minimize? HCFR has three: "without gamma", "Relative Y", and "with gamma". Using D65 as my white point, I can often get "without gamma" and "Relative Y" dE's below 3. But, using D65, my "with gamma" dE is always above 5 and sometimes as high as 9 or 10.
ok, my comment was more for general color calibration. In your case regarding the grayscale I agree with HDTVChallenged that chromaticity (the "without gamma" option) is first priority. Then if you can't reproduce the correct gamma tracking, I don't think changing your white point target will help you. Do you have gamma adjustments? What is your target function? Since the transfer function standard itself has not been well defined in the past and since environment plays a role, perhaps you are being too restrictive as to what would be acceptable.
 
#42 ·
couldn't one make the argument that since

1) there may be a degree of chromatic adaptation, but no "gamma" adaptation

and

2) having the wrong gamma may actually render details invisible, or introduce visible artifacts

that one should sacrifice a bit of accuracy for which particular white point you calibrate to (assuming a consistent grayscale), to gain a more accurate gamma?
 
#44 ·
With 10 pt RGB controls you should be able to get close to the gamma curve you want by raising/lowering all three at the same time at each stimulus level. Usually gamma and white point are not coupled but maybe your display is not typical. Also, sometimes a low gamma average is due to a too high contrast setting.
 
#45 ·
You are correct that the gamma would fall in line to target by lowering all 3 colors (R, G, and B). But the core issue with this display is that at D65 I run out of blue (across all 10 IREs by the way), meaning I can't lower blue any further to get gamma in line. I could lower red and green, which will help with gamma, but then my chromaticity would be out of whack (e.g., think red and green at 94% and blue at 115%). This is the tradeoff. Possible solutions are (1) to get into the service menu to potentially change some setting that would effectively expand my blue range on the minus side, (2) decide on some tradeoff between chromaticity and gamma, or (3) increase my white point.


My contrast tracks the greyscale correctly at its current setting, meaning that changes to, say for example, point 8 on my 10 point control show up on the grey scale step pattern in the actual 80 IRE step rather than step 70, 75, 85 or 90. If I were to change my contrast then my tracking would be off.
 
#49 ·
My contrast tracks the greyscale correctly at its current setting, meaning that changes to, say for example, point 8 on my 10 point control show up on the grey scale step pattern in the actual 80 IRE step rather than step 70, 75, 85 or 90. If I were to change my contrast then my tracking would be off.
This might be preferable to a higher white point, how much of a contrast hit do you have to take to get a reasonable gamma response? Working around misalignment of the 10 pt controls is a pain but it is doable, you should also be able to adjust sub-contrast in the service menu to realign.
 
#62 ·
See FSI's discussion of using a Perceptual approach to calibrating some of these new technology based displays. Proper setting of the white point is key to all subsequent analysis. Browse to their web site and consult with some real Color Theory Physicists!
To get started, read this attached .PDF from FSI:
So this is fascinating. Basically, if I'm understanding the gist correctly, it's saying that some new display technologies (whether they are considered good or bad) do not lend themselves to traditional calibration methods and/or tools. And furthermore, the best way to adjust for this is to use the human eye to make a subjective judgement!! (albeit on the basis of second "normal" display). While this conclusion/approach may be true it is quite shocking for me to hear because it seems to fly in the face of most of the experts on the calibration forum(s) who religiously promote exact standards and impartial equipment (not the human eye). I even think the link seems to sort of bolster what johnfull has been saying, at least to some extent.

I'd be extremely interested to hear what some of the established forum members have to say about the FSI discussion especially as compared to approaches like trying to mess with the contrast or manually shifting the white point.
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top