LG 32EP950 (JOLED panel) and i1 Display Pro issues with Red

Home Forums General Discussion LG 32EP950 (JOLED panel) and i1 Display Pro issues with Red

Viewing 15 posts - 16 through 30 (of 117 total)
  • Author
    Posts
  • #34068

    Raymond L
    Participant
    • Offline

    Hello! I was thinking about buying this monitor! Also own a i1displaypro so this post made me a bit hesitant..

    Did you work out the calibration issues? And how is the monitor behaving so far?

    The monitor itself is great. In my opinion, the last time I got this excited about display technology was years ago when IPS first became available in > 27 inches ( I didn’t find wide gamut pfs that amazing by comparison as that came between these two events).

    One thing I did learn is that targeting D65 white point with 1931 CIE 2 degree CMF simply does not work well – a bit yellowish/greenish. This is similar to the experience that Sony had with RGB OLED (e.g. https://pro.sony/s3/2021/01/22153606/ColourMatching_Between_OLED_and_CRT_E.pdf). I tried the Sony Judd hack and it works OK, but even better and cleaner is to target D65 with proposed CIE 2012 2 degree.  (Though in fact, the R/G/B OSD values only differ from the Judd hack by one click in Red.)

    This does mean you end up not taking full advantage of doing the calibrations in hardware unless you do something along the lines of the Judd hack in LG Calibration Studio (that does work).

    The good news is — I have done profiles with both LG and DisplayCAL using the i1 Display Pro, and despite the glitch, subsequent measurements using either a i1 Studio or SpyderX do not show a problem. Probably because the number of patches implicated in the profiles with problematic colors is too small to impact the matrix. (I did not experiment enough with 3D LUTs in hardware to express an opinion there.)  But as I prefer CIE 2012, I’ve settled on using DisplayCAL with i1 Studio as my production solution. In fact, the profiles with shaper+matrix in all my various whitepoint experiments differ only by small numbers (<=3) in the 3rd decimal place.

    The current theory from argyllcms mailing list is that it’s a i3 Display Pro/Plus v2 firmware incompat — but that’s a work in process.

    So I guess I am saying, I love it. Yea, it’s a bit inconvenient to use the slower i1 Studio path (it does not present problems with darker colors in my experience — the usual reason for preferring colorimetry)… and if you want to use LG Calibration Studio (to set the presets in hardware) with i1 Display Pro(or Plus), it’s OK but know that you will need to tweak your white point a bit.

    The studio professionals probably all use Calman, but at USD $2K for a Calman Studio license, I have not considered that path. Calman added support for this monitor in its recent update.

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

    #34518

    Wisechoice
    Participant
    • Offline

    Hey there. I can strongly  recommend using LightIllusion’s ColourSpace with this display and a Blackmagic Micro Converter Bi-directional 12G as an external LUT box (using HDMI to HDMI). Especially since you have a spectro that you can use for probe matching. I started out using DisplayCal for calibration. Then I  bought CalMan Home for LG to use with my LG CX 48″ OLED. It was a nightmare trying to get a decent calibration. They refunded me, no questions asked, and I went with ColourSpace instead. I haven’t looked back since.

    Like you, I’ve also put my SypderX on the shelf and have been happy with an OEM version of the i1D3.

    Ted at displaycalibrations.com  offers fantastic support if you buy ColourSpace through him. He also made the discovery of that converter’s better-than-advertised LUT capability and has developed a way to detect and compensate for very subtle panel variance that can result in clipping, via analysis of RGB separation graphs in ColourSpace.

    I don’t have a spectro yet myself, but just last night, I was able to get average dE 2000 readings of 0.1376 (yes, you read that correctly) on a 1K point volumetric verification with the i1D3 (Wacom version).

    This solution allows you to affordably bypass the bare bones LG hardware calibration and problems like the one you had. You then get an extremely high caliber reference quality display for both SDR and HDR (although I haven’t calibrated for HDR yet, I’m told the performance is just as good). It will also allow you to harness both of your probes for fast and accurate calibration.

    Before profiling, I did a perceptual match with a calibrated Dell U2713HM (also calibrated using the BMD converter as LUT box) and measured that new white point on the JOLED. Then I created a custom colour space based on Rec709 with that measured white point, as a source for the creation of the 33-point LUT. The old LED backlights don’t suffer from the metamerism that newer OLED displays do, and can serve as a reliable reference if you don’t have a spectro. ColourSpace gives you the tools to do all this, and to check the results in a way that gives you full confidence in your display.

    The only warning I would give to anyone else reading this, who might be considering a purchase  of the 32EP950, is that LG displays are known to sometimes  suffer reliability issues. My own unit arrived virtually DOA from B&H. It worked for a few hours and then quit. But it was so beautiful while it lasted, that I knew I would never be able to go back to what I was using before lol. So I’ll likely purchase an extended warranty from them, although I don’t usually do that.

    Attachments:
    You must be logged in to view attached files.
    #34520

    Raymond L
    Participant
    • Offline

    Thanks for the tip.  If I find that I need something more, I will give a standalone LUT box a try. Honestly, as it stands now, calibration to get the RGB curves aligned (for good gray balance) and the desired white point dialed in is really all it takes.  I find that if I took the vcgt and grafted it over to a linear profile, the de00 barely budges rising from 0.22 to somewhere in the 0.5 range avg (the max for both were around the same 1.7).

    I did get the warranty from B&H 🙂 bec I’ve been burned for an unusually large number of premature display failures.

    I should re-iterate that when sticking with a curve+matrix profile, the failures seen with display pro don’t really change the overall characterization (it ends up being at most 2 patches out of several 100), but my initial alarm was due to not being sure that was the RCA instead of a faulty display.

    I actually have an update on the display pro issue bec I’ve managed to reproduce it with an iPhone 13 pro (!). It turns out to be orientation dependent. If the phone is sitting flat on the table, it measures a dark R=19,G=0,B=0 / 80 nits whitepoint patch ok, but if the phone is upright so colorimeter is in horizontal position, it fails.  This does not happen with an iPad Pro 12.9 (which is a pfs phosphor LED instead of an RGB OLED display). And I’ve even moved the phone way far away from my usual desk w/o any difference in behavior (bec one of the suspicions during the investigation with ArgyllCMS was electrical interference) and am attaching it.  It is 2 cycles each of measuring the 19-0-0 patch with phone flat then standing upright. Valid values when flat, zeroes (spotread timeout, cannot find edge) when standing upright.

    I shared it with Calibrite about 10 days ago, but I did mention that I’ve switched to a spectro so am not really awaiting a reply — they have not replied beyond saying it’s been passed onto x-rite. I cannot be sure if it’s just my unit or a more general issue and was actually hoping that maybe Calibrite can tell me (and swap the unit if it’s just mine, as it’s still under warranty).

    -Ray

    Attachments:
    You must be logged in to view attached files.
    #34522

    Wisechoice
    Participant
    • Offline

    Holy moly. Keep us posted as to whether they give you any response!

    #34976

    Blowi
    Participant
    • Offline

    My thought is that, the spectral sensitivity of i1diaplay pro is so weak around the wavelengths of the OLED red stimulus, so that when the brightness of “RED” is too low,  within  the integration time the colorimeter cannot gathers enough signal.

    That may also mean the sensitivity of  i1Display pro is not suitable to measure this type of display, or at least the R LED.  The XYZ tristimulus may have large errors if the spectrum of the R LED is too different from what the colorimeter is intended for (LCD with not-too-red red).

    #34978

    Raymond L
    Participant
    • Offline

    As to whether it’s the technology or the unit, without a 2nd unit to compare, the only people who can say are Calibrite or x-rite.  Unfortunately, they have not provided any updates, but to be fair, I did say that I’ve switched to using a spectro, which works fine and also supports using CIE 2012 2deg, which gives me what looks to be the correct whitepoint.

    What is fair to say is, while Calibrite has been friendly, if I were them, I would seek to resolve the issue and maybe offer to send a replacement unit. The unit is still technically under warranty. It is also evident from the interactions that all the actual R&D is done by x-rite, without any expertise at Calibrite. So, overall, I would say that my impressions of Calibrite have been lowered.

    Hopefully, this won’t apply to x-rite for its own products as I ended up upgrading to an x-rite i1Pro3.

    Strangely, while many say SpyderX is inferior, it works for me and the final numbers when measured with the i1Pro3 are only slightly worse than using the i1Pro3  (using 1931 model) — that could as easily be instrument agreement. The numbers were still well within acceptable ranges.  But rather than “hack” the whitepoint, I feel better getting it calculated with 2012 model, so felt the i1Pro3 was a worthwhile investment — esp as I negotiated a deal to receive the Publish upgrade included for an extra $80 — in case I want to do print profiling in the future.

    #36775

    Mark Walter
    Participant
    • Offline

    Hello everybody,

    I encountered the same problems with the ColorChecker Display Plus. When I profiled my LG 31EP950 with this device, I had exactly the same wrong response in the dark red patterns (see report attached). When I profiled the display with the Colormunki Photo, the result was very good. So what do you think. Should I return the ColorChecker Plus, as I was specifically ordering it for the LG 31EP950, and keep the Colormunki Photo even though it’s not a very good spectro?

    Maybe I could rent a i1Pro2 or 3 and redo the calibration entirely with this device, just to be sure it’s accurate.

    Also, I want to connect there is a thread on LiftGammaGain, where I brought this up, and Ted Aspiotis argues that this behaviour is due to the metering method of ArgyllCMS. But I don’t think this is correct, because with other devices DisplayCal seems to work with this panel. So I also guess it’s a problem with the probe. BTW: I was wondering if you heard anything from Calibrite or X-Rite, Raymond?

    All the best,
    Mark

    • This reply was modified 1 year, 6 months ago by Mark Walter.
    Attachments:
    You must be logged in to view attached files.
    #36780

    Raymond L
    Participant
    • Offline

    Unfortunately, I have heard nothing — but in all fairness, I told Calibrite that I’ve switched to using a spectro and no longer need a resolution.

    It is great to hear that this problem has been reproduced on another device — now it is clear I did not have a defective unit.

    #36786

    Vincent
    Participant
    • Offline

    Hello everybody,

    I encountered the same problems with the ColorChecker Display Plus. When I profiled my LG 31EP950 with this device, I had exactly the same wrong response in the dark red patterns (see report attached). When I profiled the display with the Colormunki Photo, the result was very good. So what do you think. Should I return the ColorChecker Plus, as I was specifically ordering it for the LG 31EP950, and keep the Colormunki Photo even though it’s not a very good spectro?

    Make an hybrid ti3…. although munki overreading in low light may cause inconsinstences like Red 64 L*9 (i1d3) vs  L*11(munki) , but Red 51 reading with munki is L*7 so that hybrid ti3 looks that will be monotonic in red gamma ramp.

    Also, I want to connect there is a thread on LiftGammaGain, where I brought this up, and Ted Aspiotis argues that this behaviour is due to the metering method of ArgyllCMS.

    Very likely to be bullshit.

    Easy to test, change refresh on CCSS or even measure without correction. If that changes RGB [0-51, 0, 0] it’s due to the slight differences in counter correction of ouputed RGB counters from USB HID device in low light conditions (Referesh vs no refresh).
    There is no magic using an i1d3. It’s a HID command requesting a counter value.
    Also you can discuss this issue with ARgyllCMS creator in the argyllcms maillist for implementation details.

    But I don’t think this is correct, because with other devices DisplayCal seems to work with this panel. So I also guess it’s a problem with the probe. BTW: I was wondering if you heard anything from Calibrite or X-Rite, Raymond?

    All the best,
    Mark

    #36789

    Vincent
    Participant
    • Offline

    Strangely, while many say SpyderX is inferior, it works for me and the final numbers when measured with the i1Pro3 are only slightly worse than using the i1Pro3  (using 1931 model) — that could as easily be instrument agreement. The numbers were still well within acceptable ranges.

    It is inferior because it cannot be corrected in a distributed way and i1d3 can be corrected in such way.
    Of course owning an i1ProX or munkiphoto to do a custom correction for each of your screens solves this issue.

    #36825

    Mark Walter
    Participant
    • Offline

    Thank you Vincent,

    Hello everybody,

    I encountered the same problems with the ColorChecker Display Plus. When I profiled my LG 31EP950 with this device, I had exactly the same wrong response in the dark red patterns (see report attached). When I profiled the display with the Colormunki Photo, the result was very good. So what do you think. Should I return the ColorChecker Plus, as I was specifically ordering it for the LG 31EP950, and keep the Colormunki Photo even though it’s not a very good spectro?

    Make an hybrid ti3…. although munki overreading in low light may cause inconsinstences like Red 64 L*9 (i1d3) vs  L*11(munki) , but Red 51 reading with munki is L*7 so that hybrid ti3 looks that will be monotonic in red gamma ramp.

    Also, I want to connect there is a thread on LiftGammaGain, where I brought this up, and Ted Aspiotis argues that this behaviour is due to the metering method of ArgyllCMS.

    Very likely to be bullshit.

    Easy to test, change refresh on CCSS or even measure without correction. If that changes RGB [0-51, 0, 0] it’s due to the slight differences in counter correction of ouputed RGB counters from USB HID device in low light conditions (Referesh vs no refresh).
    There is no magic using an i1d3. It’s a HID command requesting a counter value.
    Also you can discuss this issue with ARgyllCMS creator in the argyllcms maillist for implementation details.

    But I don’t think this is correct, because with other devices DisplayCal seems to work with this panel. So I also guess it’s a problem with the probe. BTW: I was wondering if you heard anything from Calibrite or X-Rite, Raymond?

    All the best,
    Mark

    I will try that when I got some more time. I need to read more into it.

    In the meantime I did a verification on my Rec709 LUT (based on the ICC-Profil I did with my Colormunki Photo). Looks pretty OK, to me, no?

    But just to be sure that I understand it correctly. If I did the profiling and testing with the same probe (the Colormunki Photo, which isn’t supposed to be a very good spectro), that doesn’t say much about the quality of the probe, right? In other words: That doesn’t mean that I have a good result, just because it shows an Average ΔE*00 of 0.26? But if I do the exact procedure with an i1Pro2, I can trust the result, right? Because then I would just rent one and send back the iDisplay Plus…

    • This reply was modified 1 year, 6 months ago by Mark Walter.
    • This reply was modified 1 year, 6 months ago by Mark Walter.
    Attachments:
    You must be logged in to view attached files.
    #36831

    Vincent
    Participant
    • Offline

    Measuremengs match LUT3D prediction generated with the same device. That’s what it means. If device is accurate LUT3D is working as expected.
    But in grey scale that spectro (and maybe an i1Pro2) in darker shades is giving Red 64 L*9 (i1d3) vs L*11(munki) so gamma for non color managed apps or apps running through a LUT3D may be not so accurate. I explained it before. Maybe you changed something in monitor because now (last report) it is behaving in a different way (closer to i1d3 gamam tracking), or maybe it’s just display or munki photo that are not very consistent over time.

    • This reply was modified 1 year, 6 months ago by Vincent.
    • This reply was modified 1 year, 6 months ago by Vincent.
    • This reply was modified 1 year, 6 months ago by Vincent.
    #36835

    Mark Walter
    Participant
    • Offline

    Make an hybrid ti3

    Sorry, but I don’t know what that means. Could you explain it to me?

    Attached is a verification with the iDisplay Plus WITHOUT the Spectral Correction / verify_xxxl.ti1

    Attachments:
    You must be logged in to view attached files.
    #36837

    Mark Walter
    Participant
    • Offline

    This time with the Colorchecker Display Plus without a spectral correction. Same result in the dark reds. I am just wondering how I can correctly profile the LG 31EP950 with the Colorchecker Display Plus.

    Attachments:
    You must be logged in to view attached files.
    #36839

    Raymond L
    Participant
    • Offline

    Honestly, my belief is that you cannot if you are seeing the same phenomenon I saw — except maybe laying it flat on a table?.  See my video with the iPhone…..  Or you can stay with a simple matrix and trust the linearity of the display — the few samples in dark reds are too few to throw it off, in my opinion.

Viewing 15 posts - 16 through 30 (of 117 total)

You must be logged in to reply to this topic.

Log in or Register

Display Calibration and Characterization powered by ArgyllCMS