Is there any recommended alternate whitepoints for PFS phospor monitors?

Home Forums General Discussion Is there any recommended alternate whitepoints for PFS phospor monitors?

Viewing 15 posts - 61 through 75 (of 93 total)
  • Author
    Posts
  • #25045

    A.ces
    Participant
    • Offline

    Maybe, but for compatibility reasons and since all image editing color managed apps(*) use relative whitepoint intents for “image file to screen”, I’ll use CIE 1931 2 degree + whatever visual white point adjustment you need (usually none).

    (*) excluding SW or HW LUT3D with abs col intents, for visual wp you may want rel col intent in such LUT3D

    Hmm i tried out the CIE2012 observer D65 whitepoint on my standard gamut monitor to rule out any mismatch with WCG monitor, and the D65 of CIE 2012 looks kinda blueish compared to CIE1931 D65 when switching between them using the visual whitepoint editor and spotread, maybe D65 of CIE2012 is technically more accurate but i’m just used to how it looks on the older observer?

    it does look very neutral though after a minute, or two when i adapt instead of when instantly switching between those two whitepoints.

    • This reply was modified 3 years, 9 months ago by A.ces.
    #25047

    Vincent
    Participant
    • Offline

    Maybe, but for compatibility reasons and since all image editing color managed apps(*) use relative whitepoint intents for “image file to screen”, I’ll use CIE 1931 2 degree + whatever visual white point adjustment you need (usually none).

    (*) excluding SW or HW LUT3D with abs col intents, for visual wp you may want rel col intent in such LUT3D

    Hmm i tried out the CIE2012 observer D65 whitepoint on my standard gamut monitor to rule out any mismatch with WCG monitor, and the D65 of CIE 2012 looks kinda blueish compared to CIE1931 D65 when switching between them using the visual whitepoint editor and spotread, maybe D65 of CIE2012 is technically more accurate but i’m just used to how it looks on the older observer?

    it does look very neutral though after a minute, or two when i adapt instead of when instantly switching between those two whitepoints.

    Here, writen before:

    AFAIK a WP in CIE 2012 should look very close to CIE 1931 in a* axis. In b* axis D65 2012 is seen as cooler in 1931, but white.

    As I wrote before, it’s cooler but “white”

    #25048

    AstralStorm
    Participant
    • Offline

    Personally I can recommend CIE 1964 10″ for correcting monitors.
    10″ observer is the correct choice because you’re sitting really close to the monitor and using some of the peripheral vision. I found the 2012 ones to actually underestimate blue. Actual base Stockman&Sharpe 2006 10″ (ArgyllCMS 2.1.1) did not have this issue, it’s likely caused by linearization.

    The i1 Display Pro correction saga continues. With the help i1Studio CCSS, matrix corrections and also HCFR (rebuilt with new Argyll) to support i1Studio), I found out what’s the problem with my i1d3. It just cannot handle the primaries the monitor outputs, causing a correction nonlinearity that does not truly exist – due to spectral shape, compounded by instrument nonlinearity at the edges of its gamut. I verified with a manual prismatic spectroscope that i1Studio reports correct primaries.

    DisplayCAL/ArgyllCMS probably should skip correction out of gamut of the colorimeter, but someone has to characterize (profile) its gamut first. I don’t have hardware that can do this, as you need a tunable, high quality monochromizer. All I can say is that it does detect part of the primaries but it’s not enough.

    I’ve uploaded the CCSS for the weird Acer XV273K monitor in case someone wants a peek. It’s GB-r but with different, DCI target green ~550 nm and deep ~440 nm blue. Red phosphor is almost identical to other GB-r screens, fired off blue led. This together results in something of a magenta whitepoint out of the box with about 6650K CCT and 7250K VCT, a bit further down magenta line than Illuminant C.
    The IPS panel has a flaw where it lets through blue (and minimally green) light in near black more than red or green, resulting in deepest blacks starting at 10000K and being quite off color. Of course this is impossible to correct besides raising the black point. (To the tune of losing 1/3 of the contrast.)

    Maybe your monitor has this problem too. It makes i1d3 worthless for calibrating it in full gamut for now. You can calibrate smaller gamut with it fine.

    • This reply was modified 3 years, 9 months ago by AstralStorm.

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

    #25050

    AstralStorm
    Participant
    • Offline

    Short version: even i1d3 filters or detectors are too narrowband to capture wide gamut displays and it cheats to do so, being inaccurate at it. I wonder if the Plus model is better at it, and how K10 fares.

    Of course for a well behaved color mixing display (which it is), you could just calibrate in gamut of the colorimeter and assume out of gamut is fine afterwards, or use mostly white or lower saturations. Profiling a LUT will be invalid at high saturations, but a profile using white to calculate curves or gamma should be fine.

    • This reply was modified 3 years, 9 months ago by AstralStorm.
    • This reply was modified 3 years, 9 months ago by AstralStorm.
    • This reply was modified 3 years, 9 months ago by AstralStorm.
    #25054

    A.ces
    Participant
    • Offline

    I do believe around 2-5% gray balance  it’s  measuring around 9000-10000K then immediately  comes down to around 6500K when calibrated, displaycal even skips  measuring the last 3 patches at high 96pt calibration  on this monitor, though my i1 2 pro does measure very close to the i1d3 at 3nm being only 50K higher in VDT, and around 1-2dE off,.

    The CIE2012 observer works pretty well on my other standard gamut display no underreported blue there but of course on WCG displays its  likely different.

    I did try 10° once but it looks less neutral for me than 2°.

    • This reply was modified 3 years, 9 months ago by A.ces.
    #25056

    Vincent
    Participant
    • Offline

    Personally I can recommend CIE 1964 10″ for correcting monitors.
    10″ observer is the correct choice because you’re sitting really close to the monitor and using some of the peripheral vision. I found the 2012 ones to actually underestimate blue. Actual base Stockman&Sharpe 2006 10″ (ArgyllCMS 2.1.1) did not have this issue, it’s likely caused by linearization.

    It’s meant for walls, maybe this is one of the many things that you didn’t know

    The i1 Display Pro correction saga continues. With the help i1Studio CCSS, matrix corrections and also HCFR (rebuilt with new Argyll) to support i1Studio), I found out what’s the problem with my i1d3.

    A broken i1d3 is a possiblity

    It just cannot handle the primaries the monitor outputs, causing a correction nonlinearity that does not truly exist – due to spectral shape, compounded by instrument nonlinearity at the edges of its gamut. I verified with a manual prismatic spectroscope that i1Studio reports correct primaries.

    Actually it looks like you do not know even you are talking about… which is “surprising”.
    If spectral sensivities are accurate (firmw) some non linearities happen when actual SPD for one channel output is not just that SPD multiplied by a factor. An spectro is handly there. And all of these things ares totally unrelated to the nonsense you wrote.

    DisplayCAL/ArgyllCMS probably should skip correction out of gamut of the colorimeter, but someone has to characterize (profile) its gamut first. I don’t have hardware that can do this, as you need a tunable, high quality monochromizer. All I can say is that it does detect part of the primaries but it’s not enough.

    Again, you do not even know what you are talking about, just the same nonsense every time you write in this thread. Colorimeter has no gamut, it is not an output device. It has spectral sensivities and when fed by any SPD reports tristimulus value. You can fed any laser like SPD from hard UV to microwave go throgh its filters (weighted against it) and report tristumulus color values.
    Your 1st step is to learn this basic principles.

    Also its funny to say that a ***10nm*** spectro can “delect” some “special signature” in some display SPD but an i1d3 cannot… which points to the same cause: that you do not know even where those XYZ numbers come from.

    The IPS panel has a flaw where it lets through blue (and minimally green) light in near black more than red or green, resulting in deepest blacks starting at 10000K and being quite off color. Of course this is impossible to correct besides raising the black point. (To the tune of losing 1/3 of the contrast.)

    This a common issue not limited to that display.

    It makes i1d3 worthless for calibrating it in full gamut for now. You can calibrate smaller gamut with it fine.

    No just makes your lack of knowledge more visible to forum audience.

    Short version: even i1d3 filters or detectors are too narrowband to capture wide gamut displays and it cheats to do so, being inaccurate at it. I wonder if the Plus model is better at it, and how K10 fares.

    Short version, you lack of the minimum knowledge about what CIE XYZ coordinates are, so you write nonsense caused by that ignorance.

    i1d3, or Klein K10 or whatever colorimeter you can imagine, even you own eyes DO NOT NEED TO BE NARROWBAND to capture a widegamut light. They are the weighting functions for whatever SPD from a lightsource. If they were narrowband…. you, your own eyes, won’t see much of the visible spectrum.

    Read and learn the basics so you can cure that ignorance
    https://en.wikipedia.org/wiki/CIE_1931_color_space

    Of course for a well behaved color mixing display (which it is),

    That rules out “non linearities”.

    you could just calibrate in gamut of the colorimeter and assume out of gamut is fine afterwards, or use mostly white or lower saturations. Profiling a LUT will be invalid at high saturations, but a profile using white to calculate curves or gamma should be fine.

    No because colorimeter has no gamut. It will output color coordinates for whetever SPD light you aim at it.
    Your writings are pure ignorance… please solve it.

    I do believe around 2-5% gray balance  it’s  measuring around 9000-10000K then immediately  comes down to around 6500K when calibrated, displaycal even skips  measuring the last 3 patches at high 96pt calibration  on this monitor, though my i1 2 pro does measure very close to the i1d3 at 3nm being only 50K higher in VDT, and around 1-2dE off,.

    Yes, it is typical behavior: back point and i1d3 vs 3nm i1Pro2

    The CIE2012 observer works pretty well on my other standard gamut display no underreported blue there but of course on WCG displays its  likely different.

    I did try 10° once but it looks less neutral for me than 2°.

    Spectral corrections works better if colorimeter observer (the one stored in firmware) is close to reference observer (std) .
    10 degree is not meant for that, just happens that AstralStroms knows nothing. That does not rule out the possibility of a broken/defective i1d3 of it it happens often, just means that at June 13th 2020 he still does not know what color coordinates are or where they come from, mostly for the nonsene present in his messages.

    • This reply was modified 3 years, 9 months ago by Vincent.
    #25058

    AstralStorm
    Participant
    • Offline

    Will you stop up with this condescending crap. I’m sick of it. If you want to feel smarter go elsewhere. You do not even read the post. And you’re goddamn wrong.

    What do you mean by “meant for walls”? The 2 degree observer is actually the one made for with intent of use for projections or far tv screens or for observing some tiny details on a monitor, which you indeed do with just the fovea.

    However, it falls flat if you use a big (or close enough) screen or try to evaluate general look of it at a glance rather than a detail. So it’s not particularly good for general monitor usage, where big areas of same color or tone prevail. It works as intended if you keep the screen mostly black with some highlights on it.

    According to most ergonomic styles, the monitor should be at arm’s length. At that range, the circle of fovea area on the monitor is about 5 cm diameter. That’s not a lot. However it is enough to cover most of even a big TV screen at a few meters distance.

    The error *in measurement* is introduced by i1d3. What it specifically does is  They’re not part of the display – as measured by i1Studio. It happens that you don’t even read what I did. I literally did a raw measurement off i1d3 (uncorrected) and compared that to i1Studio result. And no correction helps enough as the color blue of this screen falls into a deep hole in i1d3 sensitivity.

    10k black artifact is not caused by i1d3 or related to i1d3. It’s a feature of the IPS screen technology. It’s exactly the same when using a spectroradiometer. *You* do not know what you’re talking about here. OLED screens do not have this kind of an artifact, nor do CRTs. Perhaps some finely zoned HDR screens can avoid it as well by turning the pixels on a bit more while modulating lowest dark with backlight, but these won’t calibrate well when dynamic backlight is left on.

    CCSS are *not related at all* to observer functions. They work with all observer functions the same. The correction is applied *before* observer is applied. The default observer embodied in i1d3 is used (with its built-in calibration and filters mimicking CIE 1934 2 degree allegedly) then CCSS weighting is applied to it in XYZ space after conversion from CIE 1934 2 deg LMS. Which is why CCSS allows you to pick any observer function as it’s applied later.

    On the other hand, CCMX is computed *using* observer matrix on both devices, so if you, say, enter correction matrix in HCFR, you need to pick the same observer as you did when computing it or it will be inaccurate. (CCMX files do contain the observer used for measurement and observer allegedly used by the probe.) CCMX is essentially a perceptual match of a few colors in a matrix form, not a colorimetric one, so it does not make sense for another observer.

    #25062

    Vincent
    Participant
    • Offline

    Will you stop up with this condescending crap. I’m sick of it. If you want to feel smarter go elsewhere. You do not even read the post. And you’re goddamn wrong.

    Just happens that you are confusing people with your nonsense.
    Colorimeters have no gamut. i1d3 (it’s design, we can have doubts about its QC) does not need narrow filters. Did you take a look of what is inside it? It is no 7-filter plastic design like old spyders. An i1d3 by design just need that its filters sensivities are measured.
    Learn what is an observer. I’ve provided you a free link that can solve at least partially your ignorance.

    What do you mean by “meant for walls”? The 2 degree observer is actually the one made for with intent of use for projections or far tv screens or for observing some tiny details on a monitor, which you indeed do with just the fovea.

    However, it falls flat if you use a big (or close enough) screen or try to evaluate general look of it at a glance rather than a detail. So it’s not particularly good for general monitor usage, where big areas of same color or tone prevail. It works as intended if you keep the screen mostly black with some highlights on it.

    According to most ergonomic styles, the monitor should be at arm’s length. At that range, the circle of fovea area on the monitor is about 5 cm diameter. That’s not a lot. However it is enough to cover most of even a big TV screen at a few meters distance.

    I meant literally that.

    The error *in measurement* is introduced by i1d3. What it specifically does is  They’re not part of the display – as measured by i1Studio. It happens that you don’t even read what I did. I literally did a raw measurement off i1d3 (uncorrected) and compared that to i1Studio result. And no correction helps enough as the color blue of this screen falls into a deep hole in i1d3 sensitivity.

    No, you still do know nothing although I’ve explained it several times. Uncorrected i1d3 readings (CIE XYZ) use a spectral correction, their own sensivities curves. Measuring an unknown display SPD with uncorrected i1d3 anc compare it to whatever reference device you have(and an i1studio is not a reference device) means… NOTHING. It is an useless test. Yo are providing wrong data to i1d3, an error in measurement is expected… because of you.

    If you trully want to check if your i1d3 firmware spectral sensivity curves and its actual sensivity curves, so you can check if they are off and you cannot use CCSS with that device with accuracy… I wrote a detailed test some post ago.
    But to do it you need a reference device, like a JETI, not an i1Studio because SPD measured by an i1studio and actual SPD entering i1d3 lens MAY BE NOT EQUAL BY FAR. Since they are not equal CCSS correcetd i1d3 and i1studio readings may not match and you cannot know which one is closer to “actual” color coordinates.

    10k black artifact is not caused by i1d3 or related to i1d3. It’s a feature of the IPS screen technology. It’s exactly the same when using a spectroradiometer. *You* do not know what you’re talking about here. OLED screens do not have this kind of an artifact, nor do CRTs. Perhaps some finely zoned HDR screens can avoid it as well by turning the pixels on a bit more while modulating lowest dark with backlight, but these won’t calibrate well when dynamic backlight is left on.

    I’ve never said that was caused by measurement device. Again is not my fault if what I wrote and what you imagined are not the same.

    CCSS are *not related at all* to observer functions. They work with all observer functions the same.

    It seems that you learn a basic concept.. at last.

    The correction is applied *before* observer is applied.

    The default observer embodied in i1d3 is used (with its built-in calibration and filters mimicking CIE 1934 2 degree allegedly) then CCSS weighting is applied to it in XYZ space after conversion from CIE 1934 2 deg LMS. Which is why CCSS allows you to pick any observer function as it’s applied later.

    Well… it looks that you don’t. Same ignorance as before. A way to solve your ignorance is to look into argyllCMS code.

    Let’s summarize it, so you stop writing nonsense and learn a little about this:

    -i1d3, its HW, reports RGB values. Raw data from USB port and HID functions.
    -Before making meaurements firmware is read: serial no, etc (1 firmw slot) sensivity curves (the other slot)
    -Since i1d3 reports RGB values, not CIE color coordinates,  a RGB to XYZ matrix is needed. Here comes CCSS
    -Taking CIE XYZ readings from an  i1d3 RGB reading uses 1) a ref observer , 2) sensivity curves, 3) a refrence CCSS for incoming light. A REFERNCE OBSERVER IS NEEDED to get CIE XYZ coordinates from an i1d3. Do you get it now?
    Think of it as a black box: ref CCSS, colorimeter observer and ref observer IN, 3×3 RGB to XYZ matrix OUT. It does not matter that user provides a CCSS or does not. A default CCSS is used, same for observer

    -CCSS UNCORRECTED i1d3 takes as “reference CCSS” its own spectral sensivities = It’s measuring a display like their own filters. Now you have XYZ coordinates and you can use an 3×3 matrix XYZ->XYZ like a CCMX to make your i1d3 measure like other device.

    -CCSS CORRECTED i1d3 use the user provided SPD with colorimeter observer and refrence obsever to compute RGB to XYZ matrix. Like with uncorrectd i1d3 you apply that matrix to raw RGB readings from i1d3 hardware and get CIE XYZ coordinates.

    It is in ArgyllCMS code, and you are totally wrong.

    For example:

    i1d3_take_XYZ_measurement function.
    -INPUTS are RGB values from i1d3 HW (labeled there as XYZ as an IO parameter)
    -when doing a display measurement 1st applied matrix is the one computed from CCSS, ref observer and colorimeter observer (“RGB” to XYZ)
    -2nd applied matrix is XYZ to XYZ optional matrix from a CCMX. No user CCMX means identity matrix.

    Where does this “RGB” from XYZ come from?
    i1d3_comp_calmat funtion.
    If you provide a CCSS that CCSS is used as reference (i1d3_set_cal function), if you DO NOT provide a CCSS, device’s sensivity curves are used as sample display SPD (i1d3_set_cal function, again)
    Calculations are the same, just varies input data AND A REFERNCE OBSERVER IS NEEDED. Otherwise there will be no XYZ readings because i1d3 does not provide them

    For other people reading this, you can get RGB to XYZ matrices both CCSS corrected and CCSS uncorrected (Corrected by a CCSS equal to col obsever) runing Aryll with enough verbose.

    On the other hand, CCMX is computed *using* observer matrix on both devices, so if you, say, enter correction matrix in HCFR, you need to pick the same observer as you did when computing it or it will be inaccurate. (CCMX files do contain the observer used for measurement and observer allegedly used by the probe.) CCMX is essentially a perceptual match of a few colors in a matrix form, not a colorimetric one, so it does not make sense for another observer.

    Not that’s not 100% accurate. i1d3 readings are “RGB” not color coordinates. You are applying a RGB to XYZ matrix using a reference observer and a reference CCSS (device’s own spectral sensivities) as inputs. With that matrix you get uncorrected CIE XYZ readings from an i1d3 (wich are actually corrected to an observer but with a not accurate CCSS). There is an implicit reference observer in CIE XYZ values reported by an uncorrected i1d3.
    This matrix is not close to an identity matrix by far.

    A 3×3 XYZ to XYZ is just a correction where your reference device says what readings are OK and a a matrix is computed to make the uncorrected device measure the same for the sample measurements used as reference.
    That new matrix is usually extremely close to identity matrix.

    The problem is that SPD measured by “a not so reference” device, an i1studio @10nm or 3nm, may be not the same as actual SPD going through colorimeter lens.
    That is the very reason an i1d3 with accurate spectral sensivities in its firmware may report different CIE readings CCSS correcetd vs CCMX corrected.
    That does NOT mean i1d3 spectral sensivities are innacurate UNLESS you use a very accurate reference device which porvides enough spectral resolution and accuracy (so actual SPD and CCSS are almost equal) . As I’ve explained before then you can claim i1d3 frmw is off from its actual behavior.

    As said before, do you get it now? Then if you learnt it now how an i1d3 works, move on and do not write more nonsense

    • This reply was modified 3 years, 9 months ago by Vincent.
    #25063

    AstralStorm
    Participant
    • Offline

    i1d3 correction code uses some rather old least squares optimizer, but it’s still synthesizing a linear matrix correction when using CCSS. The observer type is folded into the correction just like with CCMX. So they’re both actually applied as a matrix and not anything better, and anyway it cannot fix this issue of nonlinearity. (By nonlinearity I mean the sensor is reading  I guess someone will have to implement applying it as a weighting to the RGB sensitivity matrix. The extra spectral data is mostly unused, only making for overdetermined optimization when constructing the matrix. It’s not a particularly good way of correcting i1d3 and won’t do anything to the nonlinearities.

    In the meantime of your ranting about how I don’t know shit, I’ve added proper PQ support to argyll (made the original patch by Florian actually useful for any blackpoint and whitepoint, allowing you to pick any chunk of that absolute gamma your monitor uses) and fixed and found a bunch of annoyances in DisplayCAL. Have to send some emails to mailing lists and contact Florian. I have not managed to improve the loader yet to support over unity gamma though.

    Next I’ll try changing the ICC generation or calibration/profiling code to try to create a “broken” ICC profile designed for Windows that really tries to make a everything happen in vcgt. As is vcgt seems to overcorrect if you try to set gamma. Could be just a bug though.

    Then comes the better correction attempt, using various luminances to correct linearity on top of actually using the whole CCSS as it should be – for weighting and not as matrix source.

    Sorry, I wrote this before the post above. I’ll see what’s off further. RGB to XYZ matrix used is the one decoded from device, yes. However the weighting used to synthesize the matrix is based on the observer function. It is not quite close to unity, being more than 10% off. And vastly off from the matrix someone else has provided. This alone suggests a problem my i1d3 has with reading this screen.

    • This reply was modified 3 years, 9 months ago by AstralStorm.
    • This reply was modified 3 years, 9 months ago by AstralStorm.
    #25067

    AstralStorm
    Participant
    • Offline

    Correction to above: By nonlinearity I mean the sensor is reading wrong luminance values. It’s somewhat similar to a flat gamma.

    Just measuring the dichroic filter in i1d3 is an incomplete correction, the CCDs/CMOS elements used to measure light level are not exactly linear and have their own sensitivity. I thought for a second to use a white light at angle reflected off the dichroic filter measured with the spectro, but it’s likely not enough. Albeit useful to a point. The lens somewhat in the way of doing that, but it’s actually good as this would also measure chromatic aberration of that collimator.

    #25070

    Vincent
    Participant
    • Offline

    i1d3 correction code uses some rather old least squares optimizer, but it’s still synthesizing a linear matrix correction when using CCSS. The observer type is folded into the correction just like with CCMX. So they’re both actually applied as a matrix and not anything better,

    Not exactly, read again what I wrote.
    You can compute a CCMX that is exactly the same as a CCSS, just get CCSS uncorrecetd RGB_to_XYZ and CCSS corrected RGB_to_XYZ . Find a XYZ to XYZ that transform 1st matrix to the 2nd one.

    But that does not mean that CCSS corection (from an i1studio) and CCMX correction (from an i1Studio) will report the same values.
    When computing a RGB to XYZ matrix a “synthetic” measurement is made: Which color coordinates will come from my observer and that SPD? Which color coordinates will come from  ref observer and that SPD? Then you compute a RGB_to_XYZ matrices.

    When you make a CCMX with ARgyll related software such synthetic measurement is not made (to be accurate there is one, from a display with SPD equal to filters). ACTUAL SPD pass through colorimeter lens and you get raw RGB readings. “Crippled” SPD from i1studio is weighted against a ref observer. If actual SPD and “crippled” SPD (from i1Studio) are different you are making a comparision with apples and oranges.

    Do you get it now?

    -You can make a CCMX that measures the same as a CCSS. No readings are needed, just firmware data. That will be a “synthetic” CCMX
    -You can make a CCMX from actual menasuremenst but UNLESS you use a reference device that provides spectral readings almost equal to actual SPD…you cannot find if CCMX vs CCSS measurement difference come from i1d3 firmw innacuracies or from SPD innacuracies from that “not so reference” spcetrophotometer.

    In the meantime of your ranting about how I don’t know shit, I’ve added proper PQ support to argyll (made the original patch by Florian actually useful for any blackpoint and whitepoint, allowing you to pick any chunk of that absolute gamma your monitor uses) and fixed and found a bunch of annoyances in DisplayCAL. Have to send some emails to mailing lists and contact Florian. I have not managed to improve the loader yet to support over unity gamma though.

    And yet you did not know where those matrices come from. DID, past, from your last meassage.

    Sorry, I wrote this before the post above. I’ll see what’s off further. RGB to XYZ matrix used is the one decoded from device, yes. However the weighting used to synthesize the matrix is based on the observer function.

    Actually 3 params IN (there are more, just to simplify):
    -ref display SPD (a synthetic display equal to filter response if no ref SPD is provided)
    -device obsever , 3 curves, provided by manufacturer (supposed to be accurate)
    -ref observer
    OUT:
    RGB to XYZ

    So it seems that now we agree.
    If you do want to test how off are firmw data in your i1d3 from actual i1d3 behavior you need a JETI-like device so actual SPD an “crippled” SPD reported by spectro are the same. Take a few i1d3 and you can make an statistic. That will be interesting to see.
    You cannot do that with an i1Pro2 or an i1Studio, or to be accurate, you cannot do it for some backlights like most LEDs or some WG CCFLs

    • This reply was modified 3 years, 9 months ago by Vincent.
    #25072

    Vincent
    Participant
    • Offline

    Correction to above: By nonlinearity I mean the sensor is reading wrong luminance values. It’s somewhat similar to a flat gamma.

    Just measuring the dichroic filter in i1d3 is an incomplete correction, the CCDs/CMOS elements used to measure light level are not exactly linear and have their own sensitivity. I thought for a second to use a white light at angle reflected off the dichroic filter measured with the spectro, but it’s likely not enough. Albeit useful to a point. The lens somewhat in the way of doing that, but it’s actually good as this would also measure chromatic aberration of that collimator.

    But you do not know what Xrite does. As I wrote above im my previous message you can take actual RGB readings from HW (filter+sensor) and a monochromator as light source, I mean to measure assembled behavior, then write it to EEPROM. It seems a task that can be automated and once it is built it has little to no operational cost.

    As I said before, a proper test of i1d3 firmw accuracy requires a JETI or equivalent device. All people who dit it like those guys from Poland point to accurate firmw data in their samples… at least for GB-LEDs

    • This reply was modified 3 years, 9 months ago by Vincent.
    • This reply was modified 3 years, 9 months ago by Vincent.
    #25141

    A.ces
    Participant
    • Offline

    Just curious but can i get a very accurate D65 whitepoint for all monitors by using a colorchecker/white balance card, and a around 95% CRI D65 light, if so that would really be a cheap solution for finding the correct whitepoint.

    #25143

    AstralStorm
    Participant
    • Offline

    For my monitor, a matrix correction with CIE 2012 2degree match is very accurate. CIE 1931 2degree with matrix results in magenta output and banding, with a much “flatter” correction matrix – it underestimates how violet the maximum blue looks on this display and Argyll tries to use it for maximum saturation (as it looks blue to this observer), causing all sorts of issues.

    I think the cause is that the spectral peaks and also eye sensitivity get wider with higher brightness.

    Perhaps using the original nonlinear basis for CIE 2012 matrices would work best, including avoiding the “cold” flaw in darker settings, which to my eyes is minor compared to what using other observers cause.

    The problem is dependent on RGB bias/gain, but not brightness, as  the white itself cannot go bright enough to exhibit it – the display has another flaw that exceeding green brightness of ~125 nits causes the green diode to clip. It can go about 10% brighter (= more saturated) in blue and by consequence of design, also red.

    This is the root of problem i1d3 has with my display, and potentially other high brightness HDR RG displays. The spectrometer partly avoids it because it measures actual spectrum for the color which gets wider, removing half of the problem, but with the observer that does not attempt to model high brightness behavior of the eye the results are still inferior.

    If I bias and gain it with monitor controls so that excess blue and red is not available (cutting gamut), every observer works smoothly, as does i1d3. It’s just rather hard to pull off while keeping a correct white level due to monitor control granularity.

    As for the difference between 10 degrees and 2 degrees, it’s not very big and seems mostly relevant for low brightness, where the screens are inaccurate anyway.

    • This reply was modified 3 years, 9 months ago by AstralStorm.
    #25147

    Vincent
    Participant
    • Offline

    For my monitor, a matrix correction with CIE 2012 2degree match is very accurate. CIE 1931 2degree with matrix results in magenta output and banding, with a much “flatter” correction matrix – it underestimates how violet the maximum blue looks on this display and Argyll tries to use it for maximum saturation (as it looks blue to this observer), causing all sorts of issues.

    Not related to observer at all but to GPU HW or driver limitations or poor display electronics. Chances to be GPU are very high.

    I think the cause is that the spectral peaks and also eye sensitivity get wider with higher brightness.

    Measure it, plot it, share it. TI3.

    The problem is dependent on RGB bias/gain, but not brightness, as  the white itself cannot go bright enough to exhibit it – the display has another flaw that exceeding green brightness of ~125 nits causes the green diode to clip. It can go about 10% brighter (= more saturated) in blue and by consequence of design, also red.

    All this seems related to poor electronics controlling the panel and or GPU. Not to the chosen observer or measurement device. Within certain radius you should be able to set with RGB gain/offset or GPU any white, numerically or visually.

    This is the root of problem i1d3 has with my display, and potentially other high brightness HDR RG displays. The spectrometer partly avoids it because it measures actual spectrum for the color which gets wider, removing half of the problem, but with the observer that does not attempt to model high brightness behavior of the eye the results are still inferior.

    If I bias and gain it with monitor controls so that excess blue and red is not available (cutting gamut), every observer works smoothly, as does i1d3. It’s just rather hard to pull off while keeping a correct white level due to monitor control granularity.

    As for the difference between 10 degrees and 2 degrees, it’s not very big and seems mostly relevant for low brightness, where the screens are inaccurate anyway.

    As said before none of your issues seem to be related to measurement device at all.

Viewing 15 posts - 61 through 75 (of 93 total)

You must be logged in to reply to this topic.

Log in or Register

Display Calibration and Characterization powered by ArgyllCMS