I have a Mitsubishi WD-57831 DLP, and while I'm not looking for a professional calibration (yet...), I would like to do some user tweaking with the PerfectColor and PerfectTint controls.

However, I'm not sure what Cyan and Magenta are supposed to look like when calibrated! Is there a place where I can find a correct SMPTE color bar pattern? Or where I can download 1920x1080 calibration images for use on PS3 and 360?

Any help would be MUCH appreciated!

http://www.mediacollege.com/video/test-patterns/colour-bars-smpte.html
http://www.real-time-vision.com/background/smpte_colour_bar_pattern.shtml

Viewing colour bars on your computer monitor isn't going to help unless your computer monitor is calibrated.

Yeah... but now we know what cyan and magneta look like. ;)

Get the HD-DVD add-on for the XBox and use this:
http://www.dvdinternational.com/pd_digital_video_essentials_hd.cfm

Viewing colour bars on your computer monitor isn't going to help unless your computer monitor is calibrated.

I have a 6500k setting on my monitor and am using the sRGB color profile...I figure that's pretty close (have also used the Media Center patterns).

Yeah... but now we know what cyan and magneta look like. ;)



No we most certainly do not.

You can use those colorbar patterns to set color properly, but you certainly cannot use them as a reference to know what they're supposed to look like obviously.

I have a 6500k setting on my monitor and am using the sRGB color profile...I figure that's pretty close (have also used the Media Center patterns).

That is a terrible assumption. The 6500K setting on your monitor is extremely unlikely to be accurate at all.

That is a terrible assumption. The 6500K setting on your monitor is extremely unlikely to be accurate at all.

Any suggestions then, other than getting an ISF calibration? I'm just looking for some basic, do-it-yourself tips and am overwhelmed with the options on the TV.

I have DVE, GetGray, and the THX patterns and glasses, but nothing else.

No we most certainly do not.

You can use those colorbar patterns to set color properly, but you certainly cannot use them as a reference to know what they're supposed to look like obviously.

What is your problem?
Did you not pay attention to the "wink" or do you just like to continually to take jabs at me? Ease off a little... thanks.

Any suggestions then, other than getting an ISF calibration? I'm just looking for some basic, do-it-yourself tips and am overwhelmed with the options on the TV.

I have DVE, GetGray, and the THX patterns and glasses, but nothing else.

Well, without colorimetry equipment, you're basically stuck using your eyes and guessing, which can be fine. You can still use colorbars correctly to set colors fine, but greyscale is different, and the secondaries will depend on that as well.

What is your problem?
Did you not pay attention to the "wink" or do you just like to continually to take jabs at me? Ease off a little... thanks.

I'm sorry, but it wasn't clear to me that you were being sarcastic. My mistake. In any case, we're on the same page.

Greetings

Made me laugh there. I once had a client in LA that made me come back because the calibrated image on his RPTV did not look like his 6500K setting on his computer monitor ... which he somehow took to be correct because it said so. :D

Cyan is between blue and green. Now you have to determine if you want blue cyans or green cyans.

Magenta is between red and blue. Now you have to determine if you want blue magentas or red magentas.

Without test gear ... it just comes down to your "taste."

Regards

The logician in me says:

The primary colors are
red
green
blue

The secondary colors are
"not red" = green + blue = cyan
"not green" = red + blue = magenta
"not blue" = red + green = yellow

I'd assumed that the secondaries are equal parts of the two primaries that make them up. Is that not the case?

http://www.colorado.edu/physics/2000/tv/colortv.html
http://en.wikipedia.org/wiki/RGB
http://www.cybercollege.com/tvp015.htm

The logician in me says:

The primary colors are
red
green
blue

The secondary colors are
"not red" = green + blue = cyan
"not green" = red + blue = magenta
"not blue" = red + green = yellow

I'd assumed that the secondaries are equal parts of the two primaries that make them up. Is that not the case?


Well, sure. But again, how do you establish equal parts of the two primaries? You have to measure that, which is what we're all getting at.

The logician in me says:

The primary colors are
red
green
blue

The secondary colors are
"not red" = green + blue = cyan
"not green" = red + blue = magenta
"not blue" = red + green = yellow

I'd assumed that the secondaries are equal parts of the two primaries that make them up. Is that not the case?

Not if you think of it as

cyan = not red = white - red.

Which is what it is in reality on a gamut diagram - the primaries align with the secondaries thru the white point - secondaries are NOT the mid point of the edge. If Red+Blue+Green is balanced for a 10000K white - then Red+Blue would be more Purple than Magenta. This assumes color/tint are correct - what color/tint does is rotate colors around or push away (towards) the white point - or it may just make colors brighter. You really need gear if you want to fine tune these individually beyond what you can do with AVIA filter checks.

While it will not help with individual color controls which would require calibration gear that you know works well out at the primaries - when using color bars visually I would look at yellow first - much easier to tell if it is lime or orange - or even lemon more than banana yellow. I have heard some criticism of calibrated PJs having a mustard yellow - that is what it is supposed to be - people get used to the limy yellows which is very wrong!

I prefer grey poupon.

http://www.youtube.com/watch?v=zmOHeWrr9V4

Well if you like your yellows desaturated with wine and grit - OK :D Anyways I missed the joke in the faux commercial - sounded like a phone underwater.

Not if you think of it as

cyan = not red = white - red.



What do you mean "not if I think of it as"? That's exactly how I think of it. So I'm confused what you mean by "not".

If my set were perfectly calibrated, I'd have white = 100% red + 100% green + 100% blue. Then cyan would be 0%red + 100% green + 100% blue, right?

So what is it that I'm not "not" thinking of it as? :D
Roy

roy, that's correct. Although in addition, you also have to take into account the fact that for the secondaries to be correct, the primaries also have to be correct.

roy, that's correct. Although in addition, you also have to take into account the fact that for the secondaries to be correct, the primaries also have to be correct.

Yes, and this is the point that is currently causing me the most consternation in my attempts to achieve happiness with my home calibration attempt.

My TV has hue and gain controls for all primaries and secondaries. Unfortunately, gain doesn't desaturate the color, it just makes it dimmer. That makes it of limited value in adjusting my gamut.

However, with the hue control I can move each primary and secondary towards either of the adjacent colors. Doing that, I can get red and blue pretty darned close to the rec 709 points. Unfortunately, this is not true for green.

Zoyd has suggested that the thing to do is to minimize deltaE for the primaries, and I have done this. Minimum deltaE for green is about 15, and occurs when my green value is towards the blue side of the rec 709 value. Although this minimizes the primary deltaE, by tilting the gamut triangle so much by moving green off to the side like that I end up putting yellow considerably outside the triangle. Since I can't desaturate yellow, only move it to one side or the other, it remains far outside the gamut.

So I'm still not convinced that zoyd's advice is the best all around. I still think that the correct place for green is outside the rec 709 gamut but with the same ratio of red and blue to green that the rec 709 green point has. My only problem with experimenting that way is figuring out where that point it. Perhaps it's on a straight line out from D65 that passes though the rec 709 point, or perhaps it's the point that makes the sides of my gamut triangle parallel with the rec 709 gamut triangle (or perhaps those are the same points).

But my picture is still good, so mostly I just spend my time scratching my head.
Roy

What do you mean "not if I think of it as"? That's exactly how I think of it. So I'm confused what you mean by "not".

If my set were perfectly calibrated, I'd have white = 100% red + 100% green + 100% blue. Then cyan would be 0%red + 100% green + 100% blue, right?

So what is it that I'm not "not" thinking of it as? :D
Roy

You are not understanding what this % measure in your software is. Yes it is for a 100% signal coming in - but your display has RGB gain/offset adjustments that rebalances that to move from the native color temperature you get fully maxed with 100% RGB inside the PJ (not in the signal) - down to D65 which usually requires cutting blue and green significantly in UHP lamp based PJ without optical filtering. Generally this adjustment needs to be done with more bits than the video signal has to avoid loosing any source colors.

Every software has a different measure of what this calibrated % means - but in general it means how close the RGB are balanced to achieve the D65 target - it does not mean that the RGB registers inside the PJ have the same value like you are thinking!

So if you want to generate a secondary color using a RGB signal it is 100% of two of the primaries coming in - but when you measure the results in the PJ - their actual color is dependent on the white point - which is dependent on the balance of RGB inside the PJ needed to overcome the optical design to achieve D65. It will not be 100% balanced out of the box in most cases. The secondary colors are also dependent on the decoding of the video to RGB - so it is important to know what the adjustments are adjusting - the RGB display itself - or the video decoding.


If you cannot make your primaries perfect RGB to the REC709 spec, and you cannot get your grayscale perfect D65 to REC709 spec - I would focus on realigning your secondaries to get close to the REC709 spec. Most of the natural colors in video will appear within that secondary gamut so this is a better compromise than trying to figure out where the secondaries should be for your actual white point and actual primary. The most critical for primaries will be getting the hues and brightness correct - an orangy glowing red is worse than a deeper red of correct brightness/hue, and if red is too deep you may even what to cut the red brightness below spec a tad as a visual band-aid (after getting hue aligned - use the dE measure to do that)

At least this way you have most of the signals colors right - with only the deeply saturated primary colors being wrong (which is the case for most displays) and white being wrong (which your eye will adapt to). So it sounds like you can at least avoid glowing incorrect colors even if you cannot correct how deep they will go.

Nope, you're putting words in my mouth (head?) I never even considered the RGB registers inside the projector.

I'm always talking about what's measured by my colorimeter, not the numbers that are being fed into the TV to generate that.

When I say 100%, I mean the exact intensity to equal that component of D65. 100% of the correct value, not the 100% of possible output that you seem to be assuming.

But I'm still unclear exactly which of my points you're saying I don't understand. On my TV, according to HCFR 100% white from the gray scale is (approximately) R=65, G=65, B=65 adjusted for the D65 reference. So I'm presuming that the 100% red window should be R=65, G=0, B=0 if things were perfect, for example. And the 100% cyan window should be R=0, B=65, G=65, also if everything was perfect which of course it's not.

Again, these are the RGB numbers reported by HCFR when the RGB output radio button is selected, not any reference to the 16-235 input values or anything else on the TV; they're just numbers being measured by a display2 colorimeter.
Roy

One should not presume that the full range of the greyscale RGB adjustments you have correspond to the absence of that color or the maximum possible of that color. Often the maximum clipping point must be found experimentally with your sensor as the range exceeds that - and your gains often cannot be turned all the way off to get a black screen. As I said it also depends on your video decoding - which is before you even have RGB to work with.

In the specific case you are assuming that RGB % reported in your sensor means % intensity required to hit D65. It is a balance meter that expects the balance target to be white - if you give it a red - the balance meter becomes completely unbalanced saying it is way off target. The only meaning that RGB% charts have is that when everything is 100% and you have provided a white signal - you are D65 calibrated - anything away from that - has no meaning.


So what I am saying is forget what your sensor or your TV says the RGB numbers are - instead use the xyY measures from your sensor and compare to the standard. xyY (or XYZ) is the only absolute measure you should be looking at - as any RGB % measure depends on too many other things.

If you want to make perceptual compromises and you cannot hit xyY perfection - look at my LCH conversion spreadsheet in the charts&data sticky.

The RGB numbers above aren't percentages (at least they don't appear to be). They're whatever the three values are that HCFR shows in its tables when you click the RGB radio button at the lower right of the window.

It sounds to me like you're saying, "forget the software, get the raw numbers from the colorimeter and work it out yourself."

That seems like a lot of unnecessary work to me. I became a programmer because I'm essentially lazy. I write software to do repetitive translation stuff like this because that's what computers are for. Whyever would I want to ignore all that HCFR can do for me and go back to the xyY numbers?

You simply are not understanding the process of calibration with sensors - you are focusing on the wrong numbers in your tool.

RGB is an abstraction that is impossible for a calibration sensor to measure. Sensors read XYZ normalize it to xyY - then software compares them to SMPTE or other standards for what the test pattern should measure as xyY. If your test pattern is Red - there is a SMPTE spec that says red must be a certain xyY measure. There is no SMPTE spec that says Red must be a certain RGB value or percentage - that spec would have to be self referential saying - Red is ->Red<-!

A display is a device that takes RGB data and converts it to spectra that your eye sees using XYZ tristimulus measures. It does this by assigning each primary an XYZ value thru optical design that hopefully conforms to SMPTE standards. If you want to work that transform backwards from XYZ to RGB - then you must assume a XYZ spec for the primaries and white point. There are many standards that you can use for that http://www.brucelindbloom.com/index.html?WorkingSpaceInfo.html - or you can even use what was measured on your display - but that will change with each adjustment. There are many formats you can use to display the value of RGB as well - different bit depths, floating point, integers, hex, or relative %.

You don't seem to understand you are assigning meaning to an RGB instrument that is telling you it is off target - when the only relevance that instrument has is when the measured thing is on target. If the target is expected to be white but you give it a red and the display measures mostly blue spectrum - then obviously a RGB measure is completely meaningless.

Most calibration softwares do provide some form of RGB measure simply because greyscale is usually calibrated with RGB adjustments - so it has meaning to have a chart that white should be more red, less blue - even if the values themselves do not match your displays adjustment scale/range. But you should only ever interpret those value as more or less - not some absolute. The only absolute measures are the xy values which can be interpreted as a cyanish-reddish and magentaish-greenish color axis if you want to be descriptive. When it comes to measuring colors - the RGB format falls apart because it is constrained to a gamut - you cannot possibly use it to specify a Red more Red than Red - and it is meaningless because Red itself can be anything.

This is why you do indeed need to use the raw XYZ data as the basis of your measures - because with that system it makes sense to say the Red primary contained more reddish (x) and greenish (y) spectrum than SMPTE REC709 specifies and is higher intensity (Y) than allowed. I personally prefer using the CIELCH variant because it maps to perceptual senses - makes more sense to convert and say the lightness is a bit high , and the hue is orangey, and color strength is too deep and bright for an overall deltaE error that is perceivable. But you can just as well have color targets using xyY or Lu'v' or or L*u*v* or L*a*b* CIE variants if you prefer those.


If you want a relevant deviation spec that gives you information for being off target in an absolute sense - then CIE standardized on delta E - which is the distance between two sets of xyY measures after a perceptual weighting. There are several CIE colorspaces variants that you can come up with chart formats you can use to convey that delta E in a meaningful fashion - but at thier very core - those charts will be built on top of the CIE XYZ colorspaces. That is the very nature of calibration - RGB has nothing to do with calibration as it is a dataformat entirely dependent on a display to convert to a physical spectrum.

Sorry if you are not willing to learn from calibrators that know more than you - but this is the way calibration is.

Zoyd has suggested that the thing to do is to minimize deltaE for the primaries, and I have done this. Minimum deltaE for green is about 15, and occurs when my green value is towards the blue side of the rec 709 value. Although this minimizes the primary deltaE, by tilting the gamut triangle so much by moving green off to the side like that I end up putting yellow considerably outside the triangle. Since I can't desaturate yellow, only move it to one side or the other, it remains far outside the gamut.



That's not exactly what I said but anyway minimizing deltaE (for both primaries and secondaries) is the goal of calibrating proper color. Most times you can't get all points to line up so you have to compromise, I tend to favor weighting more toward accurate reds and yellows. In any case, as Krasmuzik recommends don't use RGB "values" for calibrating color, use xyY or u'v' and the CIE diagram to do this.

Sorry if you are not willing to learn from calibrators that know more than you - but this is the way calibration is.

Where exactly do I give you the impression I am unwilling to learn?

As in all human intercourse, first we have to define our terms. You appear to be attempting to instruct me, but the words you use are meaningless to me. You tell me something makes a difference, yet looking at what you said I can't see what that difference is.

The way it appears to me, I say "X", and then you say "Absolutely not. X!", and then I say "What I said appears indistinguishable from what you said", and then you say "Clearly, grasshopper, you simply do not understand."

When I express my lack of understanding, rather than explaining what the difference is you continue to state my lack of understanding in terms that are meaningless to me.

xyY, XYZ, RGB, what's the difference? They're all numbers that are transformable between one another, assuming a D65 reference white and a rec 709 color transformation. And D65 and rec 709 are the standards, so why would I assume anything else? At least that's what I think I understand so far.

As I said before. I'm using a display2 and HCFR. I believe the d2 emits xyY values. HCFR has a wonderful little clicky button that allows me to display the values recorded as RGB, XYZ, xyY, etc. Changing that clicky button doesn't move the points that have been recorded. All it does is change the units in which they are displayed. So why should I care if they're displayed as RGB or xyY? You measure a board in centimeters and I measure it in inches, the board doesn't change length, does it?

In attempting to adjust my gray scale to D65, I see that the smaller my deltaE the closer the RGB values being reported by HCFR are to being equal. So I conclude that when using software set to a D65 standard reference and a "perfect" display whose gray scale is D65, the RGB values recorded for each gray scale point would be equal (R = G = B for each individual point, not between points).

Then I'm told that to calibrate my color decoder I should display a 100% blue window and then adjust color so that the B value measured by the 100% window equals the B value recorded in the 100% gray scale window. Is that not accurate? So ideally the R in the red window and the G in the green window and the B in the blue window should equal the corresponding RGB values from the gray scale window. Right?

So now I asked if the secondaries, which appear to be white with the primary removed, will have their two remaining components match the corresponding values from the gray scale window. Nobody says "yes". Nobody says "No, it should be 2/3s of the gray scale window." (or whatever.) All you say is "Clearly you are too uneducated to understand."

And I'm the one who refuses to learn? Sheesh.
Roy

If you want to educate yourself on the topic of calibration - using a tool to learn how to do it is not going to help much. Visit poynton.com and read his many white papers and FAQs on color to be able to speak the language of calibration or even better - buy one of his textbooks so you can understand the color science in detail. Once you go beyond doing basic greyscale you need to understand not only color science - but video engineering. I am not being elitest and saying you need to be a scientest or engineer - but you do need to learn what they did when they setup video display systems to be able to use advanced calibration features..

If you want to use analogies - what you are trying to do is to use a ruler that has units of measure in twigs to measure twigs. Now while you could define twigs using a standard twig as your reference - that does not work very well when you start whittling on the standard twig and changing your reference, nor does it work to compare measures to someone else with a different set of twigs, or worse - this is like if you try to measure roots using twigs - which lie outside twig space.

maybe the problem is that the RGB arguments (used in the technical sense) you're making assume that your RGB primaries are perfect when in fact they may not be so...

:confused:

The exact value of the numbers might make that assumption, but I'm not sure how anything I'm doing is making that assumption. It might be true that the software package I'm using is incorrect, but I'm not trying to comment on that.

If the RBG primaries are perfect or imperfect might change what number the software produces for R, and for G, and for B, but it doesn't change what the colorimeter returns and it doesn't change where the measured value shows up on the chart. Are you saying that if I set HCFR to display RGB and measure my gray scale I'll get a different result than if I set it to xyY and measure? That doesn't seem to be the case.

All I'm saying is that it appears, empirically, that if I adjust my gray scale to D65, and I display my measured values in HCFR set to the D65 reference standard, then the RGB values reported by HCFR are equal when deltaE is zero. Now, from the math that is being used to convert xyY to RGB, that's exacty what I'd expect. But you say that's wrong. Ok, if that's wrong tell me why.

But what does that have to do with the price of tea in china? That's not how this whole discussion digressed for me. If I may be allowed to recap.

I had the temerity to suggest that you could think of cyan as "not red". While I'm sure that cyan is blue plus green with no red, I'm not sure what intensity blue and green should have. With the primaries, the intensity of the primary for a particular percent color window should match the intensity of that same component of the same percentage gray window. I hope I've got that straight. But it's not clear to me that that should be true of the secondaries. I suspect it's true but I've never heard anyone come out and say it.

In response to that, krasmuzik said that I would understand if I thought of cyan as white minus red. But that's exactly what I said in the first place. The very definition of "NOT x" is "the universe minus x", so for our example that would be white (everything) minus red. So krasmuzik said back to me exactly what I said in the first place, and claimed he understood and I didn't.

When I asked what exactly it was that I didn't understand we degenerated into this discussion of units, which seems total nonsense to me since what units I measure have no bearing on what's happening on my screen. If I tell you a number with no units, and you assume a different unit than what I'm reporting, we have an inability to communicate. That doesn't translate, in my opinion, into "I understand and you don't." Perhaps that is true, but it doesn't follow from this logical argument.

This is not rocket science. The math to convert among the different representations of the measured color is pretty straight forward. Converting to different units makes talking about different aspects of the color easier, but I don't think it equates to differences in level of understanding. If I know you converted from xyY to RGB using the rec 709 transform, then I know what point in color space you're talking about regardless of which set of numbers you give me. I'm only lost if you don't tell me which transform you used.

Where I get unsure is in the subjective experience of color. Given the reality of what my TV can actually calibrate, I have to make compromises. Which compromises result in the best picture? Since I can't adjust green to be close to the rec 709 green, I minimized it's deltaE. The way I have to do that skews my gamut triangle somewhat. So my secondaries no longer lie on the edges of my gamut. How important is that? My measured yellow, and the rec 709 yellow, are now completely outside my measured gamut. Do I still want yellow to be near the rec 709 yellow, or do I want it to be on the edge of my gamut? If the "correct" place is on the edge of my gamut, then by minimizing the green deltaE and skewing my triangle I've actually increased the error in my yellow by placing it further outside my triangle. Which is more important?

I don't know, but my picture looks good to me :)

maybe the problem is that the RGB arguments (used in the technical sense) you're making assume that your RGB primaries are perfect when in fact they may not be so...

:confused:

I don't think I'm assuming that. Because my primaries aren't perfect, I won't be able to achieve a deltaE of 0. That's about all I'm assuming.

There exists a point in color space that my colorimeter is measuring, right? The colorimeter has one set of units in which it reports the measured color and intensity. In my case, I believe that is xyY.

There exists other sets of units for describing that same point, and a set of mathematical transformations between the units. There is no loss of information in those transforms. I can freely convert among all of them. Using different transforms results in different numbers, but they all describe the same point in color space. If you know the units, and you know the point, you can convert back to any other set of units.

So how does which unit you use to report that color point indicate understanding or lack thereof? How is one set of units any more true than any other. RGB seems convenient to me since the physical reality of the device includes a light engine for each of those colors, but it's still a rose by any other name.

maybe the problem is that the RGB arguments (used in the technical sense) you're making assume that your RGB primaries are perfect when in fact they may not be so...

:confused:

Sort of, the key point is that RGB(CMY) have no physical meaning, they are just labels for your display's primaries/secondaries. You can use them in a relative sense to get the correct xy location of the white point because colors are additive. To calibrate any of the colors directly simply means matching their physical measures x and y (or X,Y,Z) to the specification.

There exists other sets of units for describing that same point, and a set of mathematical transformations between the units. There is no loss of information in those transforms. I can freely convert among all of them. Using different transforms results in different numbers, but they all describe the same point in color space. If you know the units, and you know the point, you can convert back to any other set of units.

This is not correct because each display has a different set of primaries. RGB does not describe the same point in colorspace for each device. That's the whole point of calibration.

This is not correct because each display has a different set of primaries. RGB does not describe the same point in colorspace for each device. That's the whole point of calibration.

Do you mean RGB = 16-235, 16-235, 16-235? If so, I agree with you. That's never been the interpretation of RGB I've been using.

If you mean that I can take two different displays, with different sets of phosphors, and adjust them so that they measure the same RGB value as computed from a colorimeter, and yet the two displays will not be displaying very close to the same color, then I agree I'm not correct. But if that's the case then I don't know why. If the colorimeter measures the same point in color space, would it not have to appear as the same color? Else what's the point of calibration?
Roy

Do you mean RGB = 16-235, 16-235, 16-235? If so, I agree with you. That's never been the interpretation of RGB I've been using.

If you mean that I can take two different displays, with different sets of phosphors, and adjust them so that they measure the same RGB value as computed from a colorimeter, and yet the two displays will not be displaying very close to the same color, then I agree I'm not correct. But if that's the case then I don't know why. If the colorimeter measures the same point in color space, would it not have to appear as the same color? Else what's the point of calibration?
Roy

That's not how I interpreted your comment, it sounded to me like you were confusing a display's RGB primaries with a mathematical definition. I think you get the picture though, there is a mathematically defined colorspace, Rec709 which has defined coordinates of R709(x,y,Y),G709(x,y,Y),B709(x,y,Y) and additionally D65 defines W(x,y,Y). You measure what the display produces R(x,y,Y),G(x,y,Y),B(x,y,Y),W(x,y,Y) and adjust what you can to make the two agree, correct? I think what might be confusing you is the interpretation of what HCFR displays when you click the RGB button. This displays only the Y component of whatever you are measuring and tells you nothing about x,y.

Yes, I measure the display and try to match the defined colorspace. I think we agree.

The only thing I'm confused about is that krasmuzik says I'm confused but I don't know what he thinks I'm confused about.

He keeps telling me to read Poynton and learn how it's supposed to work, but I don't think my confusion lies in how it's supposed to work. Poynton describes the theory of how to do "perfect" color implementation as we currently understand it, but I am not designing a reproduction system. I have an imperfectly designed television set that I'm trying to calibrate. My primary colors are not perfect and my calibration controls are not complete. So I'm trying to understand what is the definition of "best case" taking into consideration what I can and cannot manipulate on my television set.

Since my color gamut does not match the rec 709 gamut, and I cannot make it do so, how should I compromise? In a perfect gamut, all my primary and secondary points would lie on the CIE triangle; the primaries forming the vertices of the triangle and the secondaries lying on the edge of the triangle at a point on a line passing from the opposite vertex through D65. Because my primaries cannot be made to agree with the rec 709 standard, they create a triangle that lies partially inside and partially outside the CIE triangle. So in my reality, what is the "correct" place for a secondary? Is it the rec 709 location on the rec 709 triangle even though that's outside my real triangle? Or is it the point on my triangle that lies on the line that passes from my particular opposite vertex through D65? The truth is I can't achieve either of those, but which one should I try to get close to?

The only thing I'm confused about is that krasmuzik says I'm confused but I don't know what he thinks I'm confused about.


My guess is he interpreted your comments the same way I did, it sounded like you were confusing the Y values of the RGB measures that HCFR reports as full colorspace coordinates.


Since my color gamut does not match the rec 709 gamut, and I cannot make it do so, how should I compromise? In a perfect gamut, all my primary and secondary points would lie on the CIE triangle; the primaries forming the vertices of the triangle and the secondaries lying on the edge of the triangle at a point on a line passing from the opposite vertex through D65. Because my primaries cannot be made to agree with the rec 709 standard, they create a triangle that lies partially inside and partially outside the CIE triangle. So in my reality, what is the "correct" place for a secondary? Is it the rec 709 location on the rec 709 triangle even though that's outside my real triangle? Or is it the point on my triangle that lies on the line that passes from my particular opposite vertex through D65? The truth is I can't achieve either of those, but which one should I try to get close to?

The problem here is that there are many answers to the question "What is the best I can get if I can't be perfect?" when you are optimizing on 6 variables. The approach I have been using that works well for my display and was recommended by krasmuzik is to get the secondary gamut as close to spec as possible since most real world images lie within this gamut. But it's really a trade study that can only be performed by you on your display until you get the results you prefer, there is no right answer.

I've always meant "RGB" as defined on Bruce Lindbloom's page - computed from xyY from the colorimeter and the (I presume) SMPTE-C transform matrix.

So by "secondary gamut as close to spec as possible" you mean as close to the secondary point on the true rec 709 triangle? So that means each individual point, both primary and secondary, should be as close as possible to the rec 709 point. That is more important than that the secondaries be consistent with the actual triangle formed by the primaries. Cool. That's exactly what I've been worrying about.

So now I've got a complete theory. All I have to do is figure out how badly I have to mangle it to put it into practice! Thanks.
Roy

So by "secondary gamut as close to spec as possible" you mean as close to the secondary point on the true rec 709 triangle?

yes, but as you mention, if it torques the primaries away from spec by too much it may not be any better than what you have. That is what I meant by my original suggestions to "weight" the secondaries deltaE more heavily than the primaries, the hard part is determining how much weight to apply and that depends on your specific display and how it implements it's color control. I give priority to yellow, magenta and red since my display pushs red and I want accurate fleshtones.

Not if you think of it as

cyan = not red = white - red.

Which is what it is in reality on a gamut diagram - the primaries align with the secondaries thru the white point - secondaries are NOT the mid point of the edge. If Red+Blue+Green is balanced for a 10000K white - then Red+Blue would be more Purple than Magenta. This assumes color/tint are correct - what color/tint does is rotate colors around or push away (towards) the white point - or it may just make colors brighter. You really need gear if you want to fine tune these individually beyond what you can do with AVIA filter checks.

While it will not help with individual color controls which would require calibration gear that you know works well out at the primaries - when using color bars visually I would look at yellow first - much easier to tell if it is lime or orange - or even lemon more than banana yellow. I have heard some criticism of calibrated PJs having a mustard yellow - that is what it is supposed to be - people get used to the limy yellows which is very wrong!I am "R-G colorblind" (a misnomer, actually, as you probably know). But the LAST thing -- and I do mean L-A-S-T -- the last thing I would EVER do is mess with the color settings on my TV.

:D

My wife is happy and if she's happy, I'm happy. I'm in charge of brightness and contrast and she has left the color alone on our new 57831. I wish one of you calibrators lived closer to me; I wouldn't mind having this done one day. Surely there are such around here, and I'll check into it later. Right now I'm having my hands full figuring out how to change channels, get picture aspect ratios correct for various program sources and just having a lot of fun in general with this new machine. And I do love it!!

I understand a fair amount of the color theory, so I know what you're talking about; but I'm one of those people who can't tell that someone is blushing. :-)

Central Alabama, BTW

Sounds like Tom Huffman's sticky post about Color Calibration with a CMS is perfect for you, then. It sets color and tint with the colorimeter. No messy judging color balance with a filter, just read numbers off your colorimeter.

My wife is happy and if she's happy, I'm happy. I'm in charge of brightness and contrast and she has left the color alone on our new 57831.

Interesting comment. While I’m not color blind, I too appreciate my wifes’ perspective when it comes to color discrimination and I pay heed to her comments. I’ve seen reports that generally the female of the species has more and denser cones/rods than the male does especially for the reds. She certainly has a much lager vocabulary when it comes to colors. I recently tried to color match some paint by memory and failed miserably. She walked into the paint store and nailed it by just looking at the sample swatches. Its Puse, not Mauve Taupe dummy. Any woman would know that!

Dave

Its Puse, not Mauve Taupe dummy. Any woman would know that!

DaveLOL

I think you're right. The common color perception deficiency which I have is "sex-linked" and affects about 5% of male Caucasians. It usually skips generations and is "carried" by (some of) the daughters of affected males. I have three sons and no daughters, so (presumably) the problem stops with me.

I can see differences in the colors of things that you would say are the same color. For example, take ten pairs of identical black socks. Wash them; dry them; ask me to sort them into pairs. I may end up with eight pairs and four socks which I claim do not match --- even though I will agree they ARE black.

My wife sorts the same group of socks into ten pairs and when I pull a pair, it looks to me as though the socks are slightly mismatched.

In the spring, the fresh young leaves of a red Azalea are indistinguishable from the blooms. I can see the pink ones, the purple ones, and, of course, the white ones. If I get close enough to the red ones, I see the flowers but probably more because of the textural differences than the colors.

Yet, I readily distinguish between red and green traffic lights, and I've never thought a fire engine was green, etc.

So, the bottom line is, I don't mess with color settings other than perhaps saturation.

One day, I may very well hire a certified tech with a good reputation to set up my TV -- although I wonder if I'd really appreciate it enough to warrant the effort.