Possible issue in HP LP2475w review

  • If you compare Prad's HP LP2475w review with Tftcentral review you would notice that for exactly the same starting point of calibration (compare deltaE for uncalibrated 6500K mode, User Mode and sRGB mode from the two sources) after calibration to sRGB Tftcentral achieved excellent deltaE of 0.3 (max. 0.5) whilst Prad's sRGB calibration resulted in average deltaE of 3.7 and huge errors of 10 for green and cyan.


    There is clearly something wrong with the part of the Prad's review. Could someone explain it sensibly?


    I think it could be a good idea to update the part of the review.

  • We have updated the whole test procedure beginning from 2009. Older reviews can't be directly compared.


    Regarding tftcentral I should add that the colorimetric workflow isn't correct. While we are using our own test software they (and other pages too) are relying on a simple profile validation which doesn't show what they really want to show (although I have the impression that it is often not quite clear what has to be shown at all) - such validations are just not suited for that kind of tasks and were never designed to be (apart from extensive external tools like the UDACT).


    Especially the direct comparison between the profile validation after profilation (against the actual display profile) and before (against whatever) makes no sense at all and is misleading.



    Extracts of our colorimetric test workflow today:


    - Comparison of factory setting and some picture modes against sRGB (primary, secondary and some tertiary colors) including chroma difference of grey axis, presentation of gradation and other important colorimetric parameters.


    - Comparison of the condition after calibration and profilation against working color spaces under consideration of the display profile (just like the UDACT measures the ugra/ fogra media wedge)


    - Area homogeneity (chroma difference (a simple comparison on basis of the CCT is not sufficient) and luminance difference)


    - UDACT


    Also very important is the right probe (link): Most review sites are measuring with equipment that is not suited for anything else than a non WCG-CCFL spectrum.


    => The LP2475w review was written before we have introduced the mentioned workflow and therefore has inaccuracies - which makes it somehow comparable to the review you have mentioned. But I think that is wouldn't make sense to test this quite old model again under the new conditions. So you have to rely more on the subjective impressions of the editor regarding these older tests.


    Best regards


    Denis

  • So Tftcentral in fact shows how close profilation target was achieved - this is the way one should interpret their "calibration reports"; deltaE in fact is only difference between profilation target colour chromacity and the colour actual chromacity, is that right?
    In other words they should compare against sRGB or AdobeRGB (defined colour space) after profilation, as they do compare before calibration and profilation, right?


    My main point of interest regarding the monitor is whether it is the best under 600e wide-gamut choice for photo edition. According to your old review (comparison of calibration to sRGB with sRGB standard) it cannot achieve low dE for colours of sRGB target (after calibration and profilation to sRGB). Do you sustain that or do you think that a test according to your new procedure would show it is in fact capable to be calibrated and profiled so to achieve dE<2 for colours?


    As to the right probe - I understand you mean that most of the used probes has spectral sensitivity characteristics chosen for XYZ coordinates of the most popular sRGB colour space and so they are not precise for colours outside sRGB colour space, is that right?
    What if one would prepare a correction matrix for such a colorimeter - would it become capable of accurate measurements in, for example, AdobeRGB?
    (I do not know German language so I only guess that is what you may describing in the article you gave a link to).


    I would be very interested in a review of colorimeters. How accurate they are in sRGB space, how do they perform in AdobeRGB, maybe you could derive correction matrices for AdobeRGB, etc.
    I think you have got a good background for this task - I think of your own software and reference class colorimeter - UDACT.
    What do you think of the idea?

  • In a meantime I have noted that in Dell U2410 review you have stated:
    Currently, no software we know of officially supports the integration of individual correction matrices.


    Currently Argyll CMS supports correction matrices, and there are some correction matrices available for some colorimeter and wide gamut LCDs. Take a look here for more detailed information.

  • Sorry for the late answer.


    Quote

    Currently Argyll CMS supports correction matrices, and there are some correction matrices available for some colorimeter and wide gamut LCDs


    Yes that is correct. We already mention this alternative (link). When the DELL review was created Argyll couldn't manage correction matrices yet. Anyway: Such matrices must be used with caution because a consumer spectrophotometer is no ideal reference and we often have quite big deviations between colorimeters of one type (but the DTP94 from X-Rite is quite stable ).


    Quote

    So Tftcentral in fact shows how close profilation target was achieved - this is the way one should interpret their "calibration reports"; deltaE in fact is only difference between profilation target colour chromacity and the colour actual chromacity, is that right?


    Yes. It gives some hints regarding neutrality and linearity (especially when a matrix profile was created) - but comparing it with their pre calibration measurements just makes no sense and shows that there is a general comprehension problem.


    Quote

    In other words they should compare against sRGB or AdobeRGB (defined colour space) after profilation, as they do compare before calibration and profilation, right?


    What is used as target in their pre calibration measurements is not quite clear. If I'm for example comparing results of color space emulations or fixed sRGB modes with our measurements there are high deviations. One problem is the wrong equipment for WCG-CCFL screens but this doesn't explain all.


    The comparison against a working color space after calibration/ profilation needs to account for both profiles (and gamut mapping if the working color space is bigger than the display color space) - just like a CMM does. A profile validation doesn't allow this.


    Quote

    Do you sustain that or do you think that a test according to your new procedure would show it is in fact capable to be calibrated and profiled so to achieve dE<2 for colours?


    Yes - as long as you are using color aware software. Our old test also used a simple profile validation without any transformations on basis of the ICC profiles. The only thing that was applied were the LUT corrections.


    Quote

    I understand you mean that most of the used probes has spectral sensitivity characteristics chosen for XYZ coordinates of the most popular sRGB colour space and so they are not precise for colours outside sRGB colour space, is that right?


    It's all about the emitted spectrum. No consumer colorimeter reaches the characteristic of the 2 degree CIE standard observer. These devices are internally corrected (via application of a 3x3 correction matrix) against a reference. Usually a CRT and a LCD with 72% NTSC CCFL backlight. Any spectral deviation will lead to rising absolute errors. You can of course apply another correction via software (Quato for example has embedded generic corrections for WCG-CCFL and White-LED spectra, Eizo also uses corrections in Color Navigator) or modify the correction in the colorimeter (NEC has a customized EyeOne Display2 just like HP for their RGB-LED Dreamcolor). A spectrophotometer works different but has limitations through the sample intervall (the 10nm of consumer devices lie over the recommendation in ISO3664 but we are achieving quite consistent results).


    Apart from that we must also consider observer metamerism but that is a different kettle of fish.


    Quote

    I would be very interested in a review of colorimeters. How accurate they are in sRGB space, how do they perform in AdobeRGB, maybe you could derive correction matrices for AdobeRGB, etc.


    There are two main problems:


    - Getting the right reference; the EyeOne Pro is not ideal for that task so you must define a tolerant deviation range for evaluating absolute measurement accuracy


    - You must include the inter instrument agreement which means that you have to test many colorimeters of one type (I'm thinking of at least 10 devices per colorimeter).


    That's why we haven't tested a colorimeter yet.


    edit: again - don't want to comment...


    Best regards


    Denis

  • Suppose I would compute correction matrices for some standard sRGB gamut sources like CRT and LCD and for a WCG-CCFL for my Huey Pro colorimeter (against good spectrometer like i1Photo Pro or ColorMunki Photo) and then I have used them accordingly during profilation of a monitor, esspecially WCG-CCFL one. Would it lead to consistent and good results?


    Are you going to check newer revisions of a Dell U2410 for the Custom Mode problem? I am wondering which monitor to choose - HP LP2475w or Dell U2410 for photo edition tasks.


    As far as I know U2410 has 8-bit+A-FRC panel LM240WU4-SLB1, 12-bit non-programmable LUT and accepts 10-bit input, and LP2475w has 8-bit panel without A-FRC - LM240WU4-SLA1, 10-bit LUT and 8-bit input (correct me if I'm wrong).
    In theory then U2410 should be able to produce finer gradations, and more precise colours in sRGB mode, but on the other hand malfunction of Custom Mode will force to use 10-bit (Radeon X1300) VCGT LUT. Which choice would you preffer?