i1 Pro and i1D3 with CCSS don’t match

[Myrick]

Sorry if this question has been answered before.

I have an i1 Pro and an i1 Display Pro.  The monitor in question is a DreamColor z24x G2 (WLED).  I’ve made both matrix and spectral corrections with the two instruments and I’ve tried corrections downloaded from the community database, but when taking whitepoint measurements I get different readings no matter what correction I use.  Shouldn’t whitepoint readings from either instrument be pretty much the same when using a good correction?

Here’s an example where the two disagree by at least 0.0011 and at most 0.0111.

i1Pro x=0.3090 y=0.3356

i1D3 (with homemade spectral correction) x=0.3101 y=0.3288

i1D3 (with homemade matrix correction) x=0.3045 y=0.3245  (I also found that using the matrix for calibration/profiling produces a higher delta E than with the spectral option).

I tried measuring a Sony OLED production monitor (but used corrections from the web for the i1D3) and the measurements were much closer (off by ~0.0005 in each coordinate) between the i1Pro and i1D3.  Does this rule out a faulty instrument?

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

[Vincent]

i1Pro/i1Pro2/i1pro3 cannot measure WLED PFS backlights properly because of their poor resolution.
Plot bundled HP Z24x CCSS (1nm) with Displaycal and you’ll see why your spectro won’t be very accurate.

3nm readings should be close to actual color coordinates than 10nm readings, and i1d3 + DisplayCAL HP Z24x CCSS (1nm) is likely to be morte accurate than your 3nm readings. i1d3 + 1nm CCSS > i1Pro 3nm (DisplayCAL/Argyll/HCFR) > i1Pro 10nm (Calman, LS, i1profiler, basiccolor… asldo ARgyllCMS too if you select 10nm readings)
If you wish to check this by yourself make a 10nm CCSS, then plot it with DisplayCAL. Then do the same at 3nm. Then plot bundled 1nm CCSS.

[Myrick]

Ah, that makes sense.  I didn’t realize the WLED was especially hard to read.

I put the two (the bundled 1mm and my 3mm) side by side and it was clear that the red spikes were averaged out quite a bit by my i1 Pro.  The white point and primary coordinates were different between the two as well, which I thought was due to the bundled one being done in AdobeRGB or due to variances from screen to screen.

Thanks for your help.  Time to sell my i1 Pro.

[Vincent]

WLED is easy to read. WLED PFS phosphor is difficult to read for a 10nm spectrophotometer

[Vincent]

Time to sell my i1 Pro.

It is useful, I would not sell it. You’ll need $8000 more or less to buy something able to measure a WLED PFS with the accuracy of that 1nm CCSS sample.

For well known widegamut backlights use 1nm bundled samples. For unknown backlights use your 3nm readings as a reference (matrix or CCSS)

• This reply was modified 4 years, 1 month ago by Vincent.

[A.ces]

Time to sell my i1 Pro.

It is useful, I would not sell it. You’ll need $8000 more or less to buy something able to measure a WLED PFS with the accuracy of that 1nm CCSS sample.

For well known widegamut backlights use 1nm bundled samples. For unknown backlights use your 3nm readings as a reference (matrix or CCSS)

I don’t think the 1nm spectral samples are so accurate either when taking inter-instrument variability of the i1d3 into account, i have 2 of them, and they both measure around 150-200k from each other.

i think one of the most accurate options would be to perceptually match against a CCFL monitor, or a standard srgb laptop screen, The Panasonic 94% correction in my case was around 300k off (too much blue)  from the cheap CCFL display i bought for matching my OLED TV.

• This reply was modified 4 years, 1 month ago by A.ces.

[Vincent]

Time to sell my i1 Pro.

It is useful, I would not sell it. You’ll need $8000 more or less to buy something able to measure a WLED PFS with the accuracy of that 1nm CCSS sample.

For well known widegamut backlights use 1nm bundled samples. For unknown backlights use your 3nm readings as a reference (matrix or CCSS)

I don’t think the 1nm spectral samples are so accurate either when taking inter-instrument variability of the i1d3 into account, i have 2 of them, and they both measure around 150-200k from each other.

i think one of the most accurate options would be to perceptually match against a CCFL monitor, or a standard srgb laptop screen, The Panasonic 94% correction in my case was around 300k off (too much blue)  from the cheap CCFL display i bought for matching my OLED TV.

That is not right, inter instrument  variarion “should” be recorded in firmware (spectral sensivity curves). CCSS (generic) + firmw curves (per device, or batch)  => individual 3×3 matrix RGB to XYZ correction for each i1d3.

You approach (CCFL matching) is valid IF you want to get a visual match for yourself but “you” can be off from CIE 1931 2 degree observer. It is not the same problem as the one in this thread.

• This reply was modified 4 years, 1 month ago by Vincent.

[A.ces]

Time to sell my i1 Pro.

It is useful, I would not sell it. You’ll need $8000 more or less to buy something able to measure a WLED PFS with the accuracy of that 1nm CCSS sample.

For well known widegamut backlights use 1nm bundled samples. For unknown backlights use your 3nm readings as a reference (matrix or CCSS)

I don’t think the 1nm spectral samples are so accurate either when taking inter-instrument variability of the i1d3 into account, i have 2 of them, and they both measure around 150-200k from each other.

i think one of the most accurate options would be to perceptually match against a CCFL monitor, or a standard srgb laptop screen, The Panasonic 94% correction in my case was around 300k off (too much blue)  from the cheap CCFL display i bought for matching my OLED TV.

That is not right, inter instrument  variarion “should” be recorded in firmware (spectral sensivity curves). CCSS (generic) + firmw curves (per device, or batch)  => individual 3×3 matrix RGB to XYZ correction for each i1d3.

You approach (CCFL matching) is valid IF you want to get a visual match for yourself but “you” can be off from CIE 1931 2 degree observer. It is not the same problem as the one in this thread.

Hmm true but wouldn’t a Perceptual match with backlights that are guaranteed to have a accuracte spectral correction + no metameric failure be better, as displays with WCG are prone to metamerism failure so even if the instrument measures correct, it would still look off to our eyes

[Vincent]

I would say no. Perceptual match is always oriented to an human observer, for that human observer (not aiming for CIE XXXX 2 degree)
sRGB CCFL backlight grants nothing over other displays to use them as reference. Plot that sRGB CCFL SPD then take a look of several samples of human observers from statistical studies like the ones in RIT.

[A.ces]

I would say no. Perceptual match is always oriented to an human observer, for that human observer (not aiming for CIE XXXX 2 degree)
sRGB CCFL backlight grants nothing over other displays to use them as reference. Plot that sRGB CCFL SPD then take a look of several samples of human observers from statistical studies like the ones in RIT.

I see do you know if observers like CIE 170-2:2015 if argyllcms/displaycal supports them would be better against this problem with observer differences/metamerism failure,and is the CIE2012 observer in displaycal CIE 170-2:2015, or something else?

[Vincent]

AFAIK HTML verification report is run against CIE 1931 2degree, but you may choose other for calibartion or ArgyllCMS command line tools. Even if you change oberver to be it a better mean of humans, like proprosed CIE 2102 2 degree… it’s a mean, so there will be people futher from it than another.

MY best guess will be to use 1nm if it a well known backlight, custom 3nm if it’s unknown adn using it as a starting point if “white does not look white” use visual whitepoint editor for white (but readings for allother colors will use the CCSS or CCMX you made)

• This reply was modified 4 years, 1 month ago by Vincent.

[A.ces]

AFAIK HTML verification report is run against CIE 1931 2degree, but you may choose other for calibartion or ArgyllCMS command line tools. Even if you change oberver to be it a better mean of humans, like proprosed CIE 2102 2 degree… it’s a mean, so there will be people futher from it than another.

MY best guess will be to use 1nm if it a well known backlight, custom 3nm if it’s unknown adn using it as a starting point if “white does not look white” use visual whitepoint editor for white (but readings for allother colors will use the CCSS or CCMX you made)

We are gonna need a solution soon though with all these wide color gamuts displays, and the CIE1931 2 observer breaking especially with the release of 92%+ BT2020 color gamut displays later in 2020/2021.

[Vincent]

Then don’t use those 92% BT2020 displays based on R+G+B primaries and keep using current AdobeRGB+P3 backlights.
CIE 1931 2 degree is not breaking at all (and it is not going to break)… it’s a mean (closest one to actual statistical mean or not). Any observer model – any other mean- will have the same problem… it’s a mean asociated with some statistical distribution that may have relatively large standard deviation.

Do you want to minimize it? => more primaries. RIT had a 7 LED prototype. You can work with RGB colors and RGB colospaces as we have worked before. Just need a LUT3D with RGB address (input) and a n-primary vector as LUT3D node contents in display (output).
Google it and you’ll see how it performs.

• This reply was modified 4 years, 1 month ago by Vincent.

[Author]

Posts