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 - 16 through 30 (of 93 total)
  • Author
    Posts
  • #24869

    Vincent
    Participant
    • Offline

    Interesting. How far are the primaries from the target color space after calibration?

    That would be a hint of backlight type change

    Color can only look blue (and still report 6.5k VCT) if x coordinate is off by negative offset. According to certain report (linked on AVSforum), i1d3 can have problems sometimes in x plane with accuracy out of the box, reporting false smaller x values. Recalibrating it fixes that. (I1 Studio having different offset accuracy issue, a cone of confusion mostly due to observer positioning, producing warmer results rather than offset.)

    Versus an i1pro/i1studio? No, the problem is in the other side, those cheap spectross.

    There are typical situations to show this. A GBLED. 10nm reading will show +A*dx +A*dy vs actual coordinates (3de, but in daylight curve, hence “white” => lower Z in 10nm) and 3nm will lower that A by a huge margin (~1.XdE to CCSS corrected i1d3).

    Versus high end spectros? may be

    • This reply was modified 3 years, 10 months ago by Vincent.

    Calibrite ColorChecker Studio on Amazon  
    Disclosure: As an Amazon Associate I earn from qualifying purchases.

    #24871

    AstralStorm
    Participant
    • Offline

    The spectros were more accurate actually *in x only but not in absolute*. The said report used an SMPTE reference probe and a few samples of the devices. (You can dig it up, I don’t have the URL on hand right now.)

    The matrix correction might be good enough for i1d3 indeed, though I’d recommend evaluating a nonlinear stretch of the current curve too just in case.

    • This reply was modified 3 years, 10 months ago by AstralStorm.
    #24872

    Vincent
    Participant
    • Offline

    An i1pro2 or i1studio?

    NO. They measure XYZ, x is dependent on Z by definition.

    10nm readings in a not too narrow SPD like a GB-LED report:
    -hugely averaged blue SPD (look what is z-bar)
    -slightly averaged green SPD
    -reasonably accurate red SPD

    As you move to 3nm mode blue SPD gets more accurate => more accurate Z more accurate xy.

    ***

    Also out of the box i1d3 is… not very accurate by definition. You may claim if it is accurate or it is not once CCSS corrected. ONLY THEN you can make claims about if firmware curves and actual i1d3 oberver curves match.

    Hence claims about if XYZ readings from an uncorrected i1d3 are useless and a nonsense. They are SPD dependent.

    • This reply was modified 3 years, 10 months ago by Vincent.
    • This reply was modified 3 years, 10 months ago by Vincent.
    #24876

    AstralStorm
    Participant
    • Offline

    Claims about uncorrected i1d3 accuracy and especially manufacturing variance are useful for normal people who do not own a reference spectro. (I’d like to, but it costs some $30k.)

    In fact useful to know if you need to calibrate the probe and how, and whether it stays stable, and for how long. And the absolute reproducibility is also important.

    Specifically it means x-rite i1d3 is doing a poor job out of factory sometimes, vs say ChromaPure calibrated i1d3.

    As is, out of factory i1d3 tends to be very precise but sometimes inaccurate, with a scale inaccuracy. (Mostly primary offset but brightness too.)

    • This reply was modified 3 years, 10 months ago by AstralStorm.
    #24878

    AstralStorm
    Participant
    • Offline

    In comparison, i1 Studio is less precise but more accurate. And i1 Pro further slightly better than that. As in it returns “smoothed” kind of error rather than offsets and scales.

    Wrong z affects calibration accuracy most of the time but not primary value accuracy, as the middle of tristimulus tends to stay very close to true value if the curve is mostly symmetric.
    Which means you could see if sharpening or scaling the detected spectral peaks helps, using a reference spectral light bulb for $100. (Or a few.)

    Can’t do that with a tristimulus colorimeter, you need a reference probe to calibrate one of these.

    • This reply was modified 3 years, 10 months ago by AstralStorm.
    #24880

    AstralStorm
    Participant
    • Offline

    By correcting I mean deconvolving the spectrometer response from reference light. Similar to audio equalisation.

    #24882

    Vincent
    Participant
    • Offline

    Claims about uncorrected i1d3 accuracy and especially manufacturing variance are useful for normal people who do not own a reference spectro. (I’d like to, but it costs some $30k.)

    False. They’re meaningless. Use closest CCSS of if you have some hint of some unknown backligt SPD (“nano” from LG) made it from scratch (from an image to Excel to CCSS, not easy… but that’s all you can do without renting an spectro)
    As long as that manufacturing variance is stored in sensivity curves in firmware it means NOTHING. A CCSS corrected i1d3 will take this into account, hence there will be no significative variance between i1d3 measuring the same display while CCSS corrected.

    If you do not know how CCSS correction  works take a look on ArgyllCMS code. This will solve many of your misunderstanding.

    An uncorrected i1d3 is equal to CCSS corrected by firmware sensivity curves used as SPD = an ideal display with the same SPD as colorimeter RGB sensivity curves. It is SPD dependent!

    Specifically it means x-rite i1d3 is doing a poor job out of factory sometimes, vs say ChromaPure calibrated i1d3.

    Chroma pure are matrix calibrated to certain displays… so it’s mostly useless for the target you named before (people with just an i1d3 no spectro)= non portable.
    As some Spectracal representative said to me there, Chromapure dev does not even know how to deal with SPD+firmware correction. It was a joke… but pointed in the right direction.
    Are you commercially related to Chromapure? Chromapure approach has been ridiculized several times in avsforum because this.
    It will be better to avoid the commercial fight Spectracal vs Chromapure that contaminates AVSforum out of here… IMHO.

    As is, out of factory i1d3 tends to be very precise but sometimes inaccurate, with a scale inaccuracy. (Mostly primary offset but brightness too.)

    No… read what I wrote. It is accurate to a display backlight with the same SPD as Firmw sensivity curves. It’s SPD dependent. Your whole point is wrong from start.

    #24883

    Vincent
    Participant
    • Offline

    Wrong z affects calibration accuracy most of the time but not primary value accuracy, as the middle of tristimulus tends to stay very close to true value if the curve is mostly symmetric.
    Which means you could see if sharpening or scaling the detected spectral peaks helps, using a reference spectral light bulb for $100. (Or a few.)

    Z, not z. I meant what I wrote.
    X measured with some tolerance, Y measured with some tolerance, Z measured with lower accuracy… x & y gets a “delta” (x = X/(X+Y+Z)) , almost the same a 1.x multiplication factor (Z undermeasured), the same, small but pushes WP up & right… and near D65, dailight curve derivative may match where this delta error pushes xy (xy ,not XY) = 10nm measurement on a GBLED white looks white, just warmer than actual D65. It can go unnoticed unless a visual comparision

    3nm lowers delta (more accurate Z) hence 3nm reading with an i1pro and a GB-LED CCSS corrected i1d3 match withing reasonable tolerance, 10nm readings do not (~3de)

    And GB-LED is “easy” to measure. WLED PFS at 10nm can go wild.

    • This reply was modified 3 years, 10 months ago by Vincent.
    • This reply was modified 3 years, 10 months ago by Vincent.
    #24886

    A.ces
    Participant
    • Offline

    There are outliers though for example my older i1d3 with spectral correction CCSS consistently measures 200K over what it should be, while my newer i1d3 + i1 2 pro CCSS measures VDT at 0.3126 0.3290, and I know the newer one is more accurate due to it measuring very
    closely to the TV that has been calibrated by 1nm spectro + colorimeter.

    So therefore there still seems to be some unit to unit variability depending on how accurate the firmware data is… Or my first probe is just somehow broken.

    Also I believe only xrite has the capability to upload new firmware data/recalibrate their probes, all these other services can only tell you if you are within NIST limits.

    • This reply was modified 3 years, 10 months ago by A.ces.
    #24888

    Vincent
    Participant
    • Offline

    There are outliers though for example my older i1d3 with spectral correction CCSS consistently measures 200K over what it should be, while my newer i1d3 + i1 2 pro CCSS measures VDT at 0.3126 0.3290, and I know the newer one is more accurate due to it measuring very
    closely to the TV that has been calibrated by 1nm spectro + colorimeter.

    No. You do not get it.

    a) CCT is meaningless. Use a proper distance.

    b) out of spec device claims  = spectral sensivity curves  in firmware do not match actual colorimeter curves = that device cannot use a CCSS properly (because Firmw data is wrong), should be done with A device (i1d3), A reference device and A CCSS made with that reference device. None of your claims supports that you have done it… not even the same CCSS.
    If you had/rent a JETI, measure display with a JETI, make a CCSS for that display with that JETI, measure that display with an i1d3 and that CCSS you can make claims about having an “outlier i1d3”, or out of spec i1d3.

    If you do the same with an i1pro2 you cannnot make such claims , or at least you can’t for several backlight types including WLED PFS.
    If you use a generic JETI/CS 1nm CCSS in one i1d3 and a CCSS from an i1pro2 on another i1d3… it is not even a TEST at all.

    None of you cliams about outlier i1d3 holds. It does not mean that there are no out of spec i1d3 in stores, just that your claims do not hold.

    So therefore there still seems to be some unit to unit variability depending on how accurate the firmware data is… Or my first probe is just somehow broken.

    No, it does not imply that, not from your data, read above ^

    Also I believe only xrite has the capability to upload new firmware data/recalibrate their probes, all these other services can only tell you if you are within NIST limits.

    AFAIK it is not possible for i1d3 for you as consumer. If you were a reseller like NEC or Spectracal maybe there are other services pior to sell the rebranded devices, IDNK.
    Note that this is different than to request a matrix correction from a NIST traceable device.

    • This reply was modified 3 years, 10 months ago by Vincent.
    #24891

    A.ces
    Participant
    • Offline

    There are outliers though for example my older i1d3 with spectral correction CCSS consistently measures 200K over what it should be, while my newer i1d3 + i1 2 pro CCSS measures VDT at 0.3126 0.3290, and I know the newer one is more accurate due to it measuring very<br>
    closely to the TV that has been calibrated by 1nm spectro + colorimeter.

    No. You do not get it.
    a) CCT is meaningless. Use a proper distance.
    b) out of spec device claims  = spectral sensivity curves  in firmware do not match actual colorimeter curves = that device cannot use a CCSS properly (because Firmw data is wrong), should be done with A device (i1d3), A reference device and A CCSS made with that reference device. None of your claims supports that you have done it… not even the same CCSS.<br>
    If you had/rent a JETI, measure display with a JETI, make a CCSS for that display with that JETI, measure that display with an i1d3 and that CCSS you can make claims about having an “outlier i1d3”, or out of spec i1d3.
    If you do the same with an i1pro2 you cannnot make such claims , or at least you can’t for several backlight types including WLED PFS.<br>
    If you use a generic JETI/CS 1nm CCSS in one i1d3 and a CCSS from an i1pro2 on another i1d3… it is not even a TEST at all.
    None of you cliams about outlier i1d3 holds. It does not mean that there are no out of spec i1d3 in stores, just that your claims do not hold.

    So therefore there still seems to be some unit to unit variability depending on how accurate the firmware data is… Or my first probe is just somehow broken.

    No, it does not imply that, not from your data, read above ^

    Also I believe only xrite has the capability to upload new firmware data/recalibrate their probes, all these other services can only tell you if you are within NIST limits.

    AFAIK it is not possible for i1d3 for you as consumer. If you were a reseller like NEC or Spectracal maybe there are other services pior to sell the rebranded devices, IDNK.<br>
    Note that this is different than to request a matrix correction from a NIST traceable device.

    3nm CCSS should be very accurate when done for a standard gamut display aka W-LED IPS right? If I remove data from other samples from the xrite EDR then the spectral curves are spot on, I meant that between those two i1d3 using my 3nm CCSS, or a pruned xrite EDR(same thing really as they measure exactly the same then), the VDT and the resulting XYZ coordinates themself are shifted 200K.

    • This reply was modified 3 years, 10 months ago by A.ces.
    #24893

    Vincent
    Participant
    • Offline

    There are outliers though for example my older i1d3 with spectral correction CCSS consistently measures 200K over what it should be, while my newer i1d3 + i1 2 pro CCSS measures VDT at 0.3126 0.3290, and I know the newer one is more accurate due to it measuring very<br>
    closely to the TV that has been calibrated by 1nm spectro + colorimeter.

    No. You do not get it.
    a) CCT is meaningless. Use a proper distance.
    b) out of spec device claims  = spectral sensivity curves  in firmware do not match actual colorimeter curves = that device cannot use a CCSS properly (because Firmw data is wrong), should be done with A device (i1d3), A reference device and A CCSS made with that reference device. None of your claims supports that you have done it… not even the same CCSS.<br>
    If you had/rent a JETI, measure display with a JETI, make a CCSS for that display with that JETI, measure that display with an i1d3 and that CCSS you can make claims about having an “outlier i1d3”, or out of spec i1d3.
    If you do the same with an i1pro2 you cannnot make such claims , or at least you can’t for several backlight types including WLED PFS.<br>
    If you use a generic JETI/CS 1nm CCSS in one i1d3 and a CCSS from an i1pro2 on another i1d3… it is not even a TEST at all.
    None of you cliams about outlier i1d3 holds. It does not mean that there are no out of spec i1d3 in stores, just that your claims do not hold.

    So therefore there still seems to be some unit to unit variability depending on how accurate the firmware data is… Or my first probe is just somehow broken.

    No, it does not imply that, not from your data, read above ^

    Also I believe only xrite has the capability to upload new firmware data/recalibrate their probes, all these other services can only tell you if you are within NIST limits.

    AFAIK it is not possible for i1d3 for you as consumer. If you were a reseller like NEC or Spectracal maybe there are other services pior to sell the rebranded devices, IDNK.<br>
    Note that this is different than to request a matrix correction from a NIST traceable device.

    3nm CCSS should be very accurate when done for a standard gamut display aka W-LED IPS right? If I remove data from other samples from the xrite EDR then the spectral curves are spot on, I meant that between those two i1d3 using my 3nm CCSS, or a pruned xrite EDR(same thing really as they measure exactly the same then), the VDT and the resulting XYZ coordinates themself are shifted 200K.

    Did you make the 3 integrals numerically? 3nm * x-bar, 3nm * y-bar, 3nm * z-bar, then the same with 1nm? They MAY look equal although they are not (XZY coordinates and distance between them).
    Or try to make them match numerically (XYZ coordinates) by a simple RGB gain to each channel (1×3, a vector, not a matrix) in one CCSS, then compare resulting “RGB-gained” SPD. Maybe they do not match after that. (1)

    SPD is weighted by/against a standard observer. Huge differences in SPD in some wavelegth result in little to none coordinate variation. Small SPD diferences in some wavelengths result in significative coordinate variation: my GB-LED example.
    Argyll did them for you in older versions with specplot. INDK right now.

    Same for i1d3 observer and for “std to i1d3” observer difference: I can feed an RGB LED CCSS to my i1d3 measure a GB-LED and error be small (1.x dE) because inter-observer diferences are small where those SPD (GB-LED vs RGBLED) drift from each other.

    And as I said your 200K claim means nothing. Get actual distance, actual error. 200K is nothing.

    If you want to make claims you 1st need to understand underlying maths and apply them.

    (1) It’s a very good example if you do it by yourself.
    -Get RGB spectral samples from a GB-LED
    -Get RGB spectral samples from a RGB LED
    -Calculate 3 gains R1,G1 and B1 that makes GB-LED sample to have the same XYZcoodinates as 2nd one or some white (d65, and do the same to 2ns one)
    -Plot those 2 SPD with gains applied. Are they equal? Huge Red peak in RGB LED is noticeable.
    That’s an easy example.

    Now try with some sRGB WLEDs or WLED PFS CCSS: Apply an scalar gain (multiply) to each channel spectral power distribution to make PSD match some XYZ coordinates. Then compare SPD with those gains applied.

    • This reply was modified 3 years, 10 months ago by Vincent.
    • This reply was modified 3 years, 10 months ago by Vincent.
    #24900

    AstralStorm
    Participant
    • Offline

    ChromaPure does not just matrix calibrate their i1d3. They do a spectral calibration with a reference spectrometer over a few screens and bake a composite spectral correction in firmware. You do get the CCSS of the measured screens for reference *after* firmware correction. (Because the internal spectral calibration curve is changed.) They do *not* just sell you a matrix and bunch of CCSS and be done with it.

    Have you even read their site, much less bought one? This is why it costs a hundo over plain i1d3.

    See: https://www.chromapure.com/newgear_i1%20Display%20Pro%20Accuracy.asp

    Why do you insist CCT is worthless? It’s literally xy coordinate line in specified observer.

    So it delta CCT is delta xy (unspecified Y) in a specified observer, typically sad CIE 1931 2″ one, so not ideally a linear error, but definitely comparable and *mostly* linear. You can draw a numeric “fat curved plane” isotherm of true potential XYZ values with that.

    VCT is better, that makes it even more accurate as it’s now small chunk of that CCT area, making it a “fat curved line” over Yxy curve.
    Mind you, you can also get Y error separately, and with those 3 numbers you get a real XYZ error in given observer.
    XYZ for given observer is after all convertible to Yxy.

    What do you think dE is? (Barring the various functions and CIE 1931 being mediocre for comparison purposes in colors, it’s ok for white.)

    #24901

    Vincent
    Participant
    • Offline

    ChromaPure does not just matrix calibrate their i1d3. They do a spectral calibration with a reference spectrometer over a few screens and bake a composite spectral correction in firmware. You do get the CCSS of the measured screens for reference *after* firmware correction. (Because the internal spectral calibration curve is changed.) They do *not* just sell you a matrix and bunch of CCSS and be done with it.

    Have you even read their site, much less bought one? This is why it costs a hundo over plain i1d3.

    See: https://www.chromapure.com/newgear_i1%20Display%20Pro%20Accuracy.asp

    It is not worthly. They are matrix corrections. They are NOT PORTABLE. It’s almost scam that’s why he gest ridiculized every time he tries to sell that snake oil, Tyler gets in and flame starts after it was made clear what is his business model. But that’s offtopic.

    They guy at Xrite that designed i1d3 was very clever, it solved inter instrument agreement as long as units (or batches) store their actual sensivity curves… that made PORTABLE corrections possible skiping all  offset matrices market nonsense. The whole “study” you link is a nonsense.

    Now even newer Kleins copy that behavior: store spectral sensivity so you can use portable corrections for **well known backlights**

    The “BAD” part is that outside DisplayCAL community or “EDR rips translated to EDRs” you replace offset matrices market by SDKs with limited EDR support, unlock codes or encrypted EDR (Spectracal business)… and that is not good to consumer… but at least they are portable = what they sell or pack is able to provide accurate correction for that well known backlight. Matrices are not.

    Why do you insist CCT is worthless? It’s literally xy coordinate line in specified observer.

    Because it is not. It is WP PROJECTION into some whites curve. 20 CCT delta could be very very noticeable, 200K almost unticeable.
    It is a 3D space with a non euclidiean metric (on XYZ). If you exclude Y, it is a 2D space… and CCT is mostly meaningless.
    It’s maths… Relying oin CCT you lack of the more important information: pink-green deviation from curve. THAT is noticeable.

    What do you think dE is? (Barring the various functions and CIE 1931 being mediocre for comparison purposes in colors, it’s ok for white.)

    It’s a distance using several metrics in a 3D space (put whertaver surname you choose to dExx). It holds more information than your beloved CCT.

    • This reply was modified 3 years, 10 months ago by Vincent.
    #24903

    Vincent
    Participant
    • Offline

    I belive that you have missed the point all the time.

    Actual diferences between an i1d3 sensivity curves and another i1d3 may happen. But that does not correlate to actual inter instrument error as long as you use CCSS.
    Why? Because they (SHOULD) have stored in firmware their actual (& maybe different) sensivity curves. Provide actual display SPD, read those curves from firmware and inter instrument error is gone (between tolerances of actual sensivity curves and stored data in firmware). That xrite guy was a genius! And that is a VERY VERY VERY good reason to avoid the new Spyder X (builtin matrices, no spectral sensivity data AFAIK).

    That makes Chromapure “built in” matrix correction business a NONSENSE. Also it is a sensible hint to DO NOT USE DisplayCAL’s community CCMXs, just the CCMX you made for YOUR devices and YOUR displays with YOUR (rented) reference device to correct your colorimeter.

    PS: I said “built in”, not custom ON SITE colorimeter correction based on 3×3 matrices. I chose the words carefully.

    • This reply was modified 3 years, 10 months ago by Vincent.
Viewing 15 posts - 16 through 30 (of 93 total)

You must be logged in to reply to this topic.

Log in or Register

Display Calibration and Characterization powered by ArgyllCMS