Measuring a wrong whitepoint

Home Forums Help and Support Measuring a wrong whitepoint

Viewing 10 posts - 1 through 10 (of 10 total)
  • Author
    Posts
  • #17000

    Armin
    Participant
    • Offline

    Hello everyone! 🙂

    I am here looking for help, after I tried (and googled) everything I could think of.

    I have a mac mini 2018 and a BenQ PD2700U display. The PD2700U is factory calibrated to sRGB. When I set the monitor to sRGB and run a report on uncalibrated display device with DisplayCal, it measures a whitepoint (CCT) of around 7000 K. Shouldn’t it be 6500 K in sRGB?

    I also tried to verify the sRGB mode of the BenQ with an sRGB simulation profile as display profile. The CCT showed an deltaE of over 6, because it was again at around 7000K instead of 6500K.

    I am using an i1 Display Pro and I set the correction to White LED, which should be the right setting for the display.

    I have tried it with another model of the i1 Display Pro and a Macbook Pro instead of the Mac Mini with the same results. I also tried a different display with a similar result.

    Does anyone have an idea where this deviation to a cooler temperature could be coming from?

    Best regards,

    Armin

    Calibrite Display Pro HL on Amazon  
    Disclosure: As an Amazon Associate I earn from qualifying purchases.

    #17002

    Vincent
    Participant
    • Offline

    1- Factory calibration usually is not accurate, that’s why you bought measurement devices (also you bought them because display response difts over time)

    2- Color Correlated Temperature IS NOT a valid number for white point. It only gives you info about blue-yellow axis. You can get exactly 6500K CCT/CDT and a disgusting green(or pink) tint. So… you get >6dE over assumed dayligt white because of this, not because of having ~7000K CCT.
    Lots of explanations of this fact in internet.

    P.S. The photographer Andrew Rodney (Digital Dog) should have a free PDF explaining all this stuff for unexperienced users. Take a look:
    http://digitaldog.net/files/22Thecolorofwhite.pdf

    • This reply was modified 4 years, 11 months ago by Vincent.
    #17013

    Armin
    Participant
    • Offline

    Thank you Vincent for your reply. I read the article, which was quite interesting, and I think I understand what’s written in it.

    Of course I bought the measurement device for the reasons you mentioned, I am just trying out different stuff to see if I have fully understood the whole matter, which I still haven’t it seems. 😉

    The monitor (BenQ PD2700U) is brandnew and factory calibrated with an average delta E of 0,3563 (it says so on the calibration sheet). It’s white chromaticities coordinates are 0,3127/0,3287, which is very close to sRGB (0,3127 / 0,3290). When the monitor is in sRGB mode and I run a report on uncalibrated display device in display cal I get white chromaticities coordinates of 0,3056/0,3262 which is quite off with an delta E of over 6.

    In order to see if I have a faulty device, I tried another monitor with similar results.

    Could some wrong settings in display cal lead to such results?

    Could it be that Mac OS with no installed calibrated display profil, by default outputs a more bluish white (I haven’t found anything on the internet about this theory yet)?

    #17014

    Vincent
    Participant
    • Offline

    a) PD2700U should be a sRGB-like LED display so you need to measure with WLED spectral correction applied.
    Same for sRGB like macbooks. P3 macbooks should use Panasonic VXX**** or some user made “Apple P3” CCSS.

    b) factory calibration is not to be trusted. Your monitor is not faulty because “factory calibrated white is ~7000K CCT

    c) use RGB gains to tweak whote point if you need to but as explained in that article 7000K could be at 0-1dE from daylight locus… hence be “white” (a bit cooler but white) so for everyday use you may need no change. DisplayCAL verifies that proximity to “natural” whites in its reports.

    • This reply was modified 4 years, 11 months ago by Vincent.
    #17020

    Armin
    Participant
    • Offline

    Thanks again Vincent!

    #17284

    Armin
    Participant
    • Offline

    I might have found a solution to my problem.

    The reason why I measured a wrong white point is, that the color correction I used (White LED ccss) is not really suitable for my displays, although it is the recommended one.

    I found a correction matrix (ccmx file) for my particular display. With this correction I measure around 6600K, which is much closer to the desired 6500K.

    One thing just still makes me uncertain. I tried all general ccss correction files for the i1 Display Pro, and also no correction at all, and everytime I got a slightly different result, but still around 7000 K. Therefore I thought that the correction file doesn’t have such a big impact, and is only there for slight corrections.

    Only when I use the ccmx correction matrix for my particular display I get a completely different result, but which is actually much more accurate.

    Is this really the case, that the result depends so much on the correction matrix?

    I read that a ccmx matrix should only be used with the display and the colorimeter with which the measurement had been done. So, I am wondering if it is just a coincident that I get such good results with the ccmx matrix, when it is really completely off with my display and my colorimeter?

    Is the only way to get a reliable result, to get a spectrometer and make my own matrix?

    #17285

    Vincent
    Participant
    • Offline

    I might have found a solution to my problem.

    The reason why I measured a wrong white point is, that the color correction I used (White LED ccss) is not really suitable for my displays, although it is the recommended one.

    I found a correction matrix (ccmx file) for my particular display. With this correction I measure around 6600K, which is much closer to the desired 6500K.

    No it is not. It is wishful thinking from your side based on the false belief that your factory calibration “has to be OK“, and that is a false premise.

    CCMX are not portable, they are matrix made for a display and a colorimeter. Unless you own/rent a spectrophotometer and make that CCMS yourselft… it is not actually for you.
    It is better than nothing, but you have something: a sRGB LED spectral distribution sample (WLED sRGB CCSS), so what you are using (CCMX made by other) is worse.

    CCSS uses internal colorimeter sensivity curves weighted against a spectral power distribution used as “reference”. Of ourse there are slight differences between WLED sRGB displays spectral power distribution but if you think that generic ones are “not good for you” you should better try another WLED sRGB CCSS suitable sample to be used as reference (like those sRGB99% as U2415 with a little gamut extension beyond sRGB), not a CCMX made by someone else.

    One thing just still makes me uncertain. I tried all general ccss correction files for the i1 Display Pro, and also no correction at all, and everytime I got a slightly different result, but still around 7000 K. Therefore I thought that the correction file doesn’t have such a big impact, and is only there for slight corrections.

    Explanied above. It uses colorimeter sensivity curves, the close they are to CIE 1931 2Âș, the less impact a CCSS correction has.

    Second, Color Correlated Temperature IS NOT A VALID MEAN TO EVALUATE WHITE. It is explanied in this forum, in books… even in wikipedia. So do not use it. Use deltaE to target (to D65), deltaE to closest daylight curve of similar approach (to some white no matter if it is cool or warm as long as it is “white”).

    Third, since it’s a LED sRGB display is… useless to test the other ones like WG CCFL or a P3 screen backlight. It could be useful to test is something is wrong but since you keep reading CCT values instead of a propre values fro white, you are doing t wrong form start.

    Only when I use the ccmx correction matrix for my particular display I get a completely different result, but which is actually much more accurate.

    No, it is not. Your statement is wrong and it is based in a false belief about factory calibration. Since initial statement is not true, all your deductions fail.

    Is this really the case, that the result depends so much on the correction matrix?

    I read that a ccmx matrix should only be used with the display and the colorimeter with which the measurement had been done. So, I am wondering if it is just a coincident that I get such good results with the ccmx matrix, when it is really completely off with my display and my colorimeter?

    It is mostly useless for you. Use a suitable CCSS (WLED sRGB type, like bundled correctuion named that way) or get/rent an spectrophotometer and make one for your display and colorimeter.

    Is the only way to get a reliable result, to get a spectrometer and make my own matrix?

    It depends on spectrophotometer. For WLED sRGB displays like yours, spectros form Xrite should be accurate.

    So keep using WLEDFamily_07Feb11.ccss or try another CCSS from similar WLED sRGB displays.
    And of course, use proper values to evaluate white point. From your last message (these quotes) it looks like you did not understood what you read about CCT in Andrew Rodney’s paper.

    #17301

    Armin
    Participant
    • Offline

    No it is not. It is wishful thinking from your side based on the false belief that your factory calibration “has to be OK“, and that is a false premise.

    Well, I don’t think it is normal to measure a deltaE of over 6, when the factory sheet says that the delta E should be 0,4. I neither think that the factory calibration should be really as precise as the calibration sheet says it should be, but in a normal working device the deviation shouldn’t be that high.

    To rule out a faulty device, I tried 2 other screens, where I was able to manually set the color temperature to 6500 K. On both screens, I also measured a very high delta E of over 6 (with chromaticity coordinates that correspond to around 7000K) on the uncalibrated devices. And I tested the Macbook Pro, which I assume has a CCT of around 6500K as well, but again I measured very high deltaE (with chromaticity coordinates that correspond to around 7000K).

    In theory all these devices could be very inaccurate, but I think this is very unlikely. Therefore I am trying to find the error somewhere else.

    Second, Color Correlated Temperature IS NOT A VALID MEAN TO EVALUATE WHITE. It is explanied in this forum, in books
 even in wikipedia. So do not use it. Use deltaE to target (to D65), deltaE to closest daylight curve of similar approach (to some white no matter if it is cool or warm as long as it is “white”).

    Well, as I understood, chromaticity values correspond to a certain color temperature (CCT). This can be calculated for example here on waveforumlightning.com:

    https://www.waveformlighting.com/tech/calculate-color-temperature-cct-from-cie-1931-xy-coordinates

    Several coordinates can correspond to the same temperature, which is why you shouldn’t use it to evaluate a white point, because it could still have a shift on the green/mangenta axis. Is this correct?

    However, if I measure a coordinates which correspond to 7000K, when the coordinates should correspond to 6500K, then I have a high deltaE for sure. Since I am at this point only concerned about the high deltaE, I shorten things and sometimes just us color temperature. When I am concerned about a correct white point I use the coordinates.

    Third, since it’s a LED sRGB display is
 useless to test the other ones like WG CCFL or a P3 screen backlight. It could be useful to test is something is wrong but since you keep reading CCT values instead of a propre values fro white, you are doing t wrong form start.

    I read the coordinates and the CCT.

    So keep using WLEDFamily_07Feb11.ccss or try another CCSS from similar WLED sRGB displays.

    The problem is that I can actually see with my eyes that something isn’t right when I use any CCSS. The screens turn quite yellow.

    or get/rent an spectrophotometer and make one for your display and colorimeter.

    If I don’t find any other reason for my problem, I will probably soon try this.

    I will post it, if I find out more.

    Thanks again Vincent for your time and detailed answers.

    #17303

    Vincent
    Participant
    • Offline

    No it is not. It is wishful thinking from your side based on the false belief that your factory calibration “has to be OK“, and that is a false premise.

    Well, I don’t think it is normal to measure a deltaE of over 6, when the factory sheet says that the delta E should be 0,4. I neither think that the factory calibration should be really as precise as the calibration sheet says it should be, but in a normal working device the deviation shouldn’t be that high.

    Factory calibration is not to be trusted, no matter what they advertise. It is as simple as this.
    Hence when you validate a factory calibration you do not have to choose the correction that gives you the closest result to “what it should be”, but the one that matches better your display backlight technology, like generic WLED sRGB.

    And again… deltaE00 =6 from what?
    -From D65 when your white point is placed in CCT curve 7000K? likely to  happen. That does not mean than factory white is not white.
    -dE ~6 from daylight locus? That wound mean “not white” but pink or green.
    I’m not sure if yo understood that from Rodney’s paper.
    In 1st case you “can” have  factory white pleasing to the eye but a bit cool.
    2nd one means disgusting factory calibration white point with noticeable color tint.

    To rule out a faulty device, I tried 2 other screens, where I was able to manually set the color temperature to 6500 K. On both screens, I also measured a very high delta E of over 6 (with chromaticity coordinates that correspond to around 7000K) on the uncalibrated devices. And I tested the Macbook Pro, which I assume has a CCT of around 6500K as well, but again I measured very high deltaE (with chromaticity coordinates that correspond to around 7000K).

    I think that you missed the point in all your tests. Read what I wrote previously as stop using CCT as a mean to evaluate white. It does not work.

    In theory all these devices could be very inaccurate, but I think this is very unlikely. Therefore I am trying to find the error somewhere else.

    Second, Color Correlated Temperature IS NOT A VALID MEAN TO EVALUATE WHITE. It is explanied in this forum, in books
 even in wikipedia. So do not use it. Use deltaE to target (to D65), deltaE to closest daylight curve of similar approach (to some white no matter if it is cool or warm as long as it is “white”).

    Well, as I understood, chromaticity values correspond to a certain color temperature (CCT). This can be calculated for example here on waveforumlightning.com:

    https://www.waveformlighting.com/tech/calculate-color-temperature-cct-from-cie-1931-xy-coordinates

    Several coordinates can correspond to the same temperature, which is why you shouldn’t use it to evaluate a white point, because it could still have a shift on the green/mangenta axis. Is this correct?

    Color temperature is “CT” points in a certain model curve (daylight or blackbody), when you read kelvin temperature in this kind of calibration software you use CORRELATED color temperature “CCT”.
    The new “C” stands for those light sources that are not the same kind of light sources as reference curves (like a WLED form a display), hence their white could be moved along magenta-green axis from your chosen color temperature curve.

     

    However, if I measure a coordinates which correspond to 7000K, when the coordinates should correspond to 6500K, then I have a high deltaE for sure. Since I am at this point only concerned about the high deltaE, I shorten things and sometimes just us color temperature. When I am concerned about a correct white point I use the coordinates.

    As explained previously, first focus on check actual deviation from daylight locus, displacement along magenta-green. DIsplayCAL does that in console reports or in HTML reports.
    7000K with <2dE from daykight locus looks white, even pleasing, but a bit cool (blue).
    Most macbooks factory white are arround there, “cool” but “white”, with no tint. The closest you are to panel uncalibrated native white, the higher the contrast is (the closest it is to norminal “advertised” value).
    They are calibrated to ensure grey neutrality and aproximate gamma of 2.2. I’m afraid that matching D65 on the cost of contrast is not a comercial target for Apple. Neutral grey, 2.2 and higher contrast for multimedia seems a more natural commercial target for them

    #17306

    Vincent
    Participant
    • Offline

    It is as simple as this:

    -choose no profile or vendor/driver/EDID profile in you OS for that display as default display. That means no calibration in GPU is applied.
    -Run DisplayCAL, choose proper correction (like bundled WLED sRGB)
    -Run tools > report > calibrated/uncalibrated screen report

    I should write an xyY coordinate and 4 color correlated temperatures in a log window.
    Take a look on Correlated Daylight Temperature (CDT, 2nd one maybe, I do not remember). Kelvin value says position along daylight whites curve. dE says how far it is in green magenta axis.

    7000K  CDT, <2dE from daylight locus is “white”, usable… but not D65 although most users won’t care because it is asn looks white.

    CCT, is measured against blackbody curve. A perfect D65 white won’t have low dE from blackbody curve. Same goes for a perfect 7000KCDT dE=0 from daylight curve. CCT dE from blackbody won’t be 0.

    We usually choose daylight curve as reference for whites, not blackbody. Maybe this is what keeps confusing you.

Viewing 10 posts - 1 through 10 (of 10 total)

You must be logged in to reply to this topic.

Log in or Register

Display Calibration and Characterization powered by ArgyllCMS