BenQ PD2720U Not reaching advertised colour gamut coverage?

Home Forums Help and Support BenQ PD2720U Not reaching advertised colour gamut coverage?

Viewing 15 posts - 31 through 45 (of 78 total)
  • Author
    Posts
  • #28339

    Wire
    Participant
    • Offline

    Marcel, UP2516D gamut coverage numbers are attached…

    Attachments:
    You must be logged in to view attached files.
    #28344

    Marcel
    Participant
    • Offline

    The gamut viewer is a companion program with DCal called ICC Profile Info. Prolly in DCal program group in Start

    Profile Info just brings up the page I attached to my first post?

    Yes

    The UP2516D meets spec according to DCal. See the 2D plot I attached to my first post.

    Didn’t see any percents on that image just diagrams?

    % are calculated by another command line application. When you profile a display, after ICC is made, a set ot other commandl ine apps are triggered to provide feedback, also you can visually preview profile like in your 1st post by clicking on a checkbox. But that displayprofile info is different from volume intersection.

    If you want to see volume intersection % of an arbitrary profile with your display profile (or another kind of profile) you need to execute a few commandline orders:
    iccgamut to compute .gam volume for each profile
    https://www.argyllcms.com/doc/iccgamut.html
    viewgam -i to compute intersection
    https://www.argyllcms.com/doc/viewgam.html

    Also fell free to request a separate app for this feature in future DisplayCAL releases, if this is not supported now.

    If you want to avoid using command line:
    Reinstall profile without redoing it, just installing it with DisplayCAL will trigger them
    Maybe running a 3D view on browser from display profile info triggers them too (depending on -i param presence), so check DisplayCAL logs when running a 3D plot.

    Isn’t this just the same as selecting the profile in DisplayCal and hitting the profile information “I” next to it?  It shows the gamut in 2d or 3d and in the list of info, the coverage is listed at the very bottom of the list. I’m confused what you guys want me to look up haha. Probably not important though because its not going to get me better results. 😛

    Marcel, UP2516D gamut coverage numbers are attached…

     
    Ah so yours is a couple percent lower then advertised too. Glad it’s not just benq overestimating their specs.  >_<
    What calibrator do you use?

    • This reply was modified 6 months, 1 week ago by Marcel.
    #28346

    Vincent
    Participant
    • Offline

    The gamut viewer is a companion program with DCal called ICC Profile Info. Prolly in DCal program group in Start

    Profile Info just brings up the page I attached to my first post?

    Yes

    The UP2516D meets spec according to DCal. See the 2D plot I attached to my first post.

    Didn’t see any percents on that image just diagrams?

    % are calculated by another command line application. When you profile a display, after ICC is made, a set ot other commandl ine apps are triggered to provide feedback, also you can visually preview profile like in your 1st post by clicking on a checkbox. But that displayprofile info is different from volume intersection.

    If you want to see volume intersection % of an arbitrary profile with your display profile (or another kind of profile) you need to execute a few commandline orders:
    iccgamut to compute .gam volume for each profile
    https://www.argyllcms.com/doc/iccgamut.html
    viewgam -i to compute intersection
    https://www.argyllcms.com/doc/viewgam.html

    Also fell free to request a separate app for this feature in future DisplayCAL releases, if this is not supported now.

    If you want to avoid using command line:
    Reinstall profile without redoing it, just installing it with DisplayCAL will trigger them
    Maybe running a 3D view on browser from display profile info triggers them too (depending on -i param presence), so check DisplayCAL logs when running a 3D plot.

    Isn’t this just the same as selecting the profile and hitting the profile information “I” next to it?  It shows the gamut in 2d or 3d and in the list of info, the coverage is listed at the very bottom of the list. I’m confused what you guys want me to look up haha.

    Don’t look up this then.
    I just said how % are computed between some colorspace and your display colorspace. When running 3D view it need to run those commands behind, so it is possible to check gamut coverage having .gam files: typical sRGB, AdobeRGB but also eciRGBv2, P3, some offset printing standard, …
    So if you want to check % coverage with whatever colorspace it is done this way. Don’t you need it? don’t use it.

    #28347

    Wire
    Participant
    • Offline

    Vincent, thanks for that how-to on gamut maps, super helpful

    Marcel, so is 96% xy vs 118% xyY equal to 98% Dell spec?

    The primaries are  determined by physics and material economies, etc. As to how ITU specs get written, that’s its own fascinating bailiwick.

    My pet peeve about specs is everyone thinks BT.1886 is about a new standard for video, when the reason it was created was so the content industry doesn’t forget CRT display dynamics when working with legacy HD.

    #28348

    Marcel
    Participant
    • Offline

    Marcel, so is 96% xy vs 118% xyY equal to 98% Dell spec?

     

    IDK, I don’t know my way around this stuff. They only advertise the xy don’t they? Isn’t that 98% supposed to be just gamut coverage? This stuff is making my head spin. Too many variables haha. >_<

    #28349

    Wire
    Participant
    • Offline

    Marcel,

    I forgot to mention, I am using i1 Display Pro Plus (i1d3) and a spectral correction provided by a DCal user for this specific model.

    Just changing the colorimeter correction file can gain or lose a couple percent of coverage, so buyer beware.

    As to the importance of the delta:

    I have found that a small gamut delta can be visible for key colors. For example, Macbeth ColorChecker cyan (patch 18) lives just a tiny distance outside of sRGB but within almost equal gamut ColorMatch RGB. There’s just a sliver of difference on an xy plot. But while in side-by-side test chart you can see the diff, in a photograph, no one would ever notice anything missing.

    In photography and video, there’s a lot to be said for smoothness of response over absolute range, and thoughtful rendering is paramount.

    My personal perception of gamut is variable, not unlike white adaptation. I will never confuse two different domains of saturation, but for a single view, proportional, smooth response beats range. The designers of sRGB did their homework. And yes, something was left on the table which new tech can help us claim. It’s interesting to explore for its own sake.

    In my opinion: amplifying the above comment, increased gamut over sRGB (709)  for creative content has vanishing small merit for most renderings. You have to know what to look for, while increasing gamut introduces large hazards.

    I also find that a well-aligned lower-gamut display really makes content come to life in a wonderful way, when a poorly aligned high-gamut display can destroy it.

    Rendering intent — not just in the sense of ICC transformation parameter, but in the broad sense of adapting displays to environments and creative vision — is the undiscovered country of color management. And an area with a lot of promise and pitfalls for UHD tech.

    Don’t fret about your swimming head with this topic, it’s part of the fun! It’s a bit like noticing a picture is hanging crooked, so you go to adjust it, then it looks crooked the other way. Then you find the walls are not perpendicular.

    Rule of thumb: If you go looking for a difference, you will find it!

    i1Display Pro Plus on Amazon   i1Display Pro on Amazon  
    Disclosure: As an Amazon Associate I earn from qualifying purchases.

    #28371

    Marcel
    Participant
    • Offline

    I forgot to mention, I am using i1 Display Pro Plus (i1d3) and a spectral correction provided by a DCal user for this specific model.

    Just changing the colorimeter correction file can gain or lose a couple percent of coverage, so buyer beware.

    Ah you have the newer more expensive version. Maybe that’s why you can get closer to advertised specs.

    Yeah I have noticed changing correction can change results which is a little concerning. Comparing the spectral correction vs the matrix correction that DisplayCal pulled from the internet for my monitor, they don’t even read white or luminance the same. They differ by 250K for colour temperature detection and by about 8 cd/m². The matrix one also lowers my % of gamut coverage slightly. I read on another thread that matrix corrections are supposed to be more accurate so I decided to use that one but how can you truly know if your results are accurate if they differ by so much depending on which correction you choose and if corrections are just user submitted?

    In photography and video, there’s a lot to be said for smoothness of response over absolute range, and thoughtful rendering is paramount.

    How do you check this?

    In my opinion: amplifying the above comment, increased gamut over sRGB (709) for creative content has vanishing small merit for most renderings. You have to know what to look for, while increasing gamut introduces large hazards.

    Yeah I figure colour grading on a wide gamut display could make things look off on a srgb or less display, but I have a 3 monitor setup, 2 of which only cover srgb so I can see what my edit is looking like on both types of monitors at the same time, and adjust accordingly if something looks off on the lesser displays.

    It’s a bit like noticing a picture is hanging crooked, so you go to adjust it, then it looks crooked the other way. Then you find the walls are not perpendicular.

    hahaha, yeah that’s a good comparison.

    #28372

    Vincent
    Participant
    • Offline

    I forgot to mention, I am using i1 Display Pro Plus (i1d3) and a spectral correction provided by a DCal user for this specific model.

    Just changing the colorimeter correction file can gain or lose a couple percent of coverage, so buyer beware.

    Ah you have the newer more expensive version. Maybe that’s why you can get closer to advertised specs.

    No, HW uses the same filter tech. He can reach greater coverage because of backlight. User made CCSS for your model shows that yours cannot go further.

    Yeah I have noticed changing correction can change results which is a little concerning. Comparing the spectral correction vs the matrix correction that DisplayCal pulled from the internet for my monitor, they don’t even read white or luminance the same. They differ by 250K for colour temperature detection and by about 8 cd/m². The matrix one also lowers my % of gamut coverage slightly. I read on another thread that matrix corrections are supposed to be more accurate so I decided to use that one but how can you truly know if your results are accurate if they differ by so much depending on which correction you choose and if corrections are just user submitted?

    You cannot use matrix unless you did it by yourself with another device you call “reference”. You should not “port” them.
    Otherwise choose a CCSS at native gamut of the closest backlight SPD to your model you can get (unless it is suspected to have very narrow spikes where you have to prioritize nm). These can be ported between i1d3 (and displays with same backlight) because they store no information relative to a certain i1d3.

    • This reply was modified 6 months, 1 week ago by Vincent.
    #28380

    Wire
    Participant
    • Offline

    Vincent, Marcel’s point was about comparing discrepancy between claimed and measured coverage, not comparing absolute coverage.

    I would expect an instrument that can take spectral corrections to outperform matrix correction because the shape of the standard observer response for the display primaries can be better accommodated by a spectral map, whereas a matrix correction can only accomodare the position of the spectral peaks — unless I  fundamentally misunderstand the nature of the corrections?

    Marcel,

    Based on e everything I’ve read, the I1 Display Pro Plus is not a new design. It simply is adjusted and qualified for an additional 1000 nits to support HDR10.  The design is many years old and well proven.

    As I recall, old DTP-94 can correctly map the WGC display primaries, but this seems only part of story on accuracy. I should double check this.

    #28381

    Vincent
    Participant
    • Offline

    I would expect an instrument that can take spectral corrections to outperform matrix correction because the shape of the standard observer response for the display primaries can be better accommodated by a spectral map, whereas a matrix correction can only accomodare the position of the spectral peaks — unless I  fundamentally misunderstand the nature of the corrections?

    No,

    A matrix is to match some device (your i1d3, an spyder, an i1d2, a Klein…) to other device you call “reference” (maybe better or not, like an i1pro2 measuring a WLED PFS at 10nm).
    This way you can match an innacurate spyder4 to a JETI, they will measure same WP & primaries (black is another tale..). It’s for matching numbers of two measurement devices, that’s all. And it works for every colorimeter, it’s an universal solution.
    This need TWO measurement devices and ONE display, on the same place (that’s the ugly part). That’s why it is not “portable”.

    CCSS are to correct one device measurement based on its manufacturer spectral sensivities (stored in firmware, supposed to be accurate) for a given sample SPD.
    This needs ONE device (your i1d3, which firmware data  we “trust”) and ONE display that we also trust that has very close spectralpower distribution in primaries.
    This is the reason we can share CCSS between us for X display model (as long as CCSS data is accurate, like not using 10nm reading fro a WLED PFS, or a WGCCFL)

    • This reply was modified 6 months, 1 week ago by Vincent.
    • This reply was modified 6 months, 1 week ago by Vincent.
    • This reply was modified 6 months, 1 week ago by Vincent.
    #28414

    Marcel
    Participant
    • Offline

    You cannot use matrix unless you did it by yourself with another device you call “reference”. You should not “port” them.
    Otherwise choose a CCSS at native gamut of the closest backlight SPD to your model you can get (unless it is suspected to have very narrow spikes where you have to prioritize nm). These can be ported between i1d3 (and displays with same backlight) because they store no information relative to a certain i1d3.

    Oh? When I read this thread https://hub.displaycal.net/forums/topic/creating-matrix-correction-vs-spectral-correction/ it seems like Florian was saying the matrix one is better to use because it’s more accurate. Maybe I miss understood the question. To my eyes the matrix does look better though. There seems to be too much red in the whites with the spectral one, especially when comparing to my other 2 monitors.

    Based on everything I’ve read, the I1 Display Pro Plus is not a new design. It simply is adjusted and qualified for an additional 1000 nits to support HDR10.  The design is many years old and well proven.

    Hmm thought it would of had to have some upgrade to read the extra 1000 nits. Guess dell just doesn’t stretch the truth as much as benQ does then. 😛

    • This reply was modified 6 months, 1 week ago by Marcel.
    #28421

    Vincent
    Participant
    • Offline

    You cannot use matrix unless you did it by yourself with another device you call “reference”. You should not “port” them.
    Otherwise choose a CCSS at native gamut of the closest backlight SPD to your model you can get (unless it is suspected to have very narrow spikes where you have to prioritize nm). These can be ported between i1d3 (and displays with same backlight) because they store no information relative to a certain i1d3.

    Oh? When I read this thread https://hub.displaycal.net/forums/topic/creating-matrix-correction-vs-spectral-correction/ it seems like Florian was saying the matrix one is better to use because it’s more accurate. Maybe I miss understood the question. To my eyes the matrix does look better though. There seems to be too much red in the whites with the spectral one, especially when comparing to my other 2 monitors.

    If YOU made it for YOUR display, using a device YOU call  reference (on certain configuration). And it is more accurate because numbers match and YOU call the other device “reference”, hence all measurements that approaches better to YOUR reference is more accurate. That is what it means, it is accuracy relative to reference. That chain does not exists if you download a CCMX made by others on other display an other colorimeter.
    Also if you measure a WLED PFS @10nm with an i1Pro2 that reference measurement is fake, but you call it “reference” for a CCMX (so it will be wrong against actual color coordinates)

    You can use std GB-LED CCSS (U2413) rather than user made CCSS for PD2720U. It had bumps in red, maybe its true, maybe it’s other users device malfunction / wrong operation / dirt.

    #28428

    Wire
    Participant
    • Offline

    Also if you measure a WLED PFS @10nm with an i1Pro2 that reference measurement is fake, but you call it “reference” for a CCMX (so it will be wrong against actual color coordinates)

    VIncent,

    I couldn’t quite follow the antecedent of “fake”.

    Do you mean that we have to keep in  mind the chain of references at all times, with the understanding that some references are purer than others? To amplify my question with a statement: So “fake” means that while YOU have constructed a nice WLED PFS local chain which gives good local results, don’t confuse this with a truth for that entire class of devices because you can’t generalize from your specific devices due to local variables, binning / sort issues, etc. (classical issues in standardization of anything). Fake works up to a point, but don’t lose sight of its provisionality

    Or are you making another point… ?

    So what is a user to think about sharing of CCMX? It’s sometimes better than nothing? For those who can’t afford to build out own chain of trust to a level of standards, which still can experiment and maybe get closer to ideals?

    It seems to me (speaking Socratically so I’m expecting to be wrong) that the basic problem with a colorimeter as opposed to a spectro is that a tristimulus sensor, rather than a spectral sensor, is meeting a tristimulus emitter (the RGB display)  where the spectral passbands (S,M,L) don’t agree, so the sensor function has to be compensated to account for the mismatch. A spectro has very wide sensitivity so doesn’t suffer this mismatch.

    #28429

    Vincent
    Participant
    • Offline

    Sorry, wrong word. Change “fake” for “innacurate”.

    PFS is a good example:
    Measure a imac P3 with an i1pro2 at 20nm, measure RGB. 3 CIE XYZ coordinates. Measure it with an SpyderX and no correction, 3 CIE XYZ coordinates. A CCMX is a matrix that allows you to transform Spyder coordinates to i1Pro2 coordinates. That’s all.
    Are i1Pro2 coordinates accurate? NO, at 10nm you loose information in red channel (and blue too for all LED backlights).
    Is corrected SPyder by CCMX accurate if you believe that i1pro2 10nm readinsg is YOUR reference? YES.

    CCMX is the most accurate way for matching a device to a reference. Just happens that sometimes that is not a true reference (10nm PFS). Also happens that MY spyderx/i1d3 IS NOT YOURS (the one you corrected by reference), hence that matching is not valid for me because it will carry an error due to variantions between device. It is not portable, do not use (if possible) CCMX made by other people.

    CCSS works in a sligthly different way, manufacturer is supossed to store on each device it sensivities, so you get rid of intrument variantion, you have YOURS.
    Then if we believe taht the display we are going to measure has some SPD (RGB samples):
    -measure that SPD for each primary and std observer
    -measure each SPD with “colorimeter observer”
    -compute a matrix.
    You end with a matrix too, but made FOR YOU, not for other people’s colorimeters.
    The data we share among us is SPD, it is not colorimeter dependent, hence “it is portable”.
    If reference SPD is not like your display, an error arises. if you measure a p3 mac with 10nm i1pro2, you’ll end up with same error.

    *****

    The mistake from Marcel is that he believed that accuracy was “for all people”, but it is just for the one that makes a matrix for HIS device and HIS monitor, “trusting” as “reference” some device. If you did not make CCMX, then comparison & accuracy it is out of scope.

    If you own that reference (so you can compare), “global” accuracy depends on the accuracy of reference. If I measure and make a matrix for an i1d3 with an i1pro2 at 10nm in this sample, overall accuracy will be worse than making a CCSS at 3nm fro teh same i1d3. Why? beacuse “reference” at 10nm is not accurate and it seems that 1nm spectral data stored in i1d3 is accurate, that such spectarl data characterizes that individual i1d3 (or its batch at worse). As you go to better resolutions in reference, and as you go up in QC for spectral sensivity data from an hypotetical i1display4, both corrections should converge on the corrected CIE XYZ measurement.
    AGAIN, this only applies if YOU HAVE BOTH DEVICES, reference and the one to be corrected. OTHERWISE the only corrections to be trusted are the ones that do not carry information from “colorimeter to be corrected”, because that device is not yours

    • This reply was modified 6 months ago by Vincent.
    • This reply was modified 6 months ago by Vincent.
    #28435

    Wire
    Participant
    • Offline

    The reason you focus on 10nm PFS is because the relatively wide, as compared to 1nm, spectral sampling internal is not tight enough to capture the peakiness of the primaries for that red/blue LED tech, so the effective response is attenuated and leads to an under-measure, and attendant bias — color cast, lumpy correction, etc?

    If I follow your writeup, you are strongly implying that per-device variance within a class is a problem, which brings us to a distinction between consumer-class devices, which we can accommodate locally to make workable alignments, and reference-class devices, which support a chain of trust and allow us to compare devices.  IOW, the average consumer, me, Marcel, may think we are buying a sort of reference (the colorimeter) that lets us compare our displays, but we are just buying a link in a chain of trust. The ordinary user faces uncertainty about how well his instrument (probe) matches a class of displays based on illumination and the correction is a way to try to handle this uncertainty, but without a solid chain of reference, in the case you describe, the weak link is i1Pro2 low sampling interval for the peakiness of PFD, you can end up with a lacking overall alignment (in comparison to…?)  in spite of the ordinary user thinking he is appealing to accuracy by referring to a spectrophotometer. Basically, the informed user has to be aware of the chain of trust in references and avoid clinging to certain over-simple assumptions about how devices play together.

    In the post Marcel referenced from Florian, I think Florian responds by saying that internally you end up with a 3×3 correction matrix whether you get there via a CCMX or the internal processing of the colorimeter firmware and CCSS.

    The key being to get the combined device response into the domain of the standard observer well enough to make the whole rest of the model work.

    This makes me wonder what the eye does with extremely narrow band stimulus and how the tristimulus model copes with extremely narrow-band tech like lasers?

Viewing 15 posts - 31 through 45 (of 78 total)

You must be logged in to reply to this topic.

Log in or Register

Display Calibration and Characterization powered by ArgyllCMS