Home › Forums › General Discussion › How accurate are the spectral corrections from i1d3 really?
- This topic has 7 replies, 3 voices, and was last updated 5 years, 4 months ago by Henk Poley.
-
AuthorPosts
-
2018-11-10 at 1:02 #14416
Seems like they have not been updated in a few years, and i believe that newer w-led monitors have different spectral characteristics as my old tn cheap tn monitor from 2012 actually has a more correct white color when calibrated.
2018-11-12 at 9:31 #14428X-rite updated its bundle to include W-LED PFS phosphor for multimedia monitors with 9x% of P3 colorspace. It’s called “Panasonic VX[…]”. DisplayCAL bundle for i1d3 include this correction.
W-LED monitors with just sRGB coverage should keep using ol’ WLED CCSS from Xrite or community made for your specific model.
New models with W-LED PFS phosphor backlight for graphic arts (9x% AdobeRGB + 9x% P3) have a slighty different spectral power distribution than that “panasonic vx…”. You can find it in HP Z32x calibration software or Lightspace.
There are several threads with more information and spectral power distribution plots in this forum.
- This reply was modified 5 years, 5 months ago by Vincent.
Calibrite Display Pro HL on Amazon
Disclosure: As an Amazon Associate I earn from qualifying purchases.2018-11-20 at 20:32 #14554You have to understand that none of the cheap colorimeters give “accurate readings” out of the box. Since they don’t look at colours the same way that human eyes do. The colorimeters are sensitive to a few small chunks of the visible light spectrum. Surely the readings are reproducible, but that’s not very useful perse if you want to hit a certain color that we humans agree on.
What needs to be done to make a colorimeter actually accurate for color reproduction is to take a spectrometer and the colorimeter, apply them to your particular screen, and then tweak the interpretation of the numbers coming out of he colorimeter to match the numbers coming out of the spectrometer (DisplayCal > Tools > Correction > Create Colorimeter Correction). From the spectrometer data we can predict what humans see. And since only 3 filtered spectra of color (RGB) will come out of your screen we can mix and match the spectrometer data for each color filter to the colorimeter output.
If you are lucky somebody has already done this, and you might get pretty passable results if you use their numbers. But if say your screen’s backlight color output changes with age, it won’t be correct. There’s a database with screen corrections for DisplayCal here: https://colorimetercorrections.displaycal.net
If your screen is about the same age as your colorimeter and uses a popular backlight and popular color filters, then maybe the settings & corrections that it came with out of the box will be acceptable. But the correct way would be to get a spectrometer too, and create a colorimeter correction.
2018-11-20 at 21:28 #14555You have to understand that none of the cheap colorimeters give “accurate readings” out of the box.
True
Since they don’t look at colours the same way that human eyes do.
Almost true, true for some colorimeters, others see very very close to what humans do. See below.
The colorimeters are sensitive to a few small chunks of the visible light spectrum.
False, not all of them. I think that you have not see what is an i1DisplayPro (or i1d3 clones) or its sensivity curves, otherwise you won’t say that.
Since you can se its guts in youtube and dump sensivity curves if you have one at your disposal, try it. It is very didactical.What needs to be done to make a colorimeter actually accurate for color reproduction is to take a spectrometer and the colorimeter, apply them to your particular screen, and then tweak the interpretation of the numbers coming out of he colorimeter to match the numbers coming out of the spectrometer (DisplayCal > Tools > Correction > Create Colorimeter Correction).
True IF screen backlight SPD could be “read” with that particular spectrophotomer. Displays with P3 colorspace (or newer graphic arts displays with W-LED PFS backlight) are not good friends of 10nm spectros but… we have argyllcms hi-res driver and that is good.
I mean you can overcome at least partially this issue because ArgyllCMS 3nm driver. Other calibration tools trying to mach colorimeters to 10nm spectros like xrite’s do not have that advantage.From the spectrometer data we can predict what humans see.
Same as above. Too optimistic for exmaple with new P3 backlights with default xrite spectros driver.
If you are lucky somebody has already done this, and you might get pretty passable results if you use their numbers.
Good for CCSS spectral corrections, not so good with matrices.
If your screen is about the same age as your colorimeter and uses a popular backlight and popular color filters, then maybe the settings & corrections that it came with out of the box will be acceptable. But the correct way would be to get a spectrometer too, and create a colorimeter correction.
True if your spectrophotometer is “better” reading your particular screen than “popular” backlight spectral sample (CCSS) that you get from vendor or community.
For example take a look at Benq EW3270U (W-LED pfs phosphor P3 display) in community database. JETI 1511 spectral sample (CCSS) is almost equal to Xrite default CCSS for those backlights. User made CCSS with Xrite’s spectros are not close to that.
So there is some uncertain even using a spectrophotometer.
-Some displays are known to be avery very close match to generic CCSS from vendor EDRs: almost all graphic arts LED/CCFL displays and some P3 displays. I would avoid using 10nm spectros here “as reference”, no gain doing that.
-Some displays would have and advantage if you can measure it with an spectro (hi res mode if you use Xrite’s): the others- This reply was modified 5 years, 5 months ago by Vincent.
2018-11-21 at 12:06 #14562@Vincent, All very good remarks that I wasn’t particularly aware of.
I’m personally coming from the point that an i1 Display 2 didn’t really give any good results over the years on decidedly non-pro screens. And I was mainly just wondering if somewhere in DisplayCAL a warning could given to nudge people into understanding the problem. It is not at all clear how important the calibration to your particular type of screen is, for somebody just buying a tool to make an ICC profile.
X-Rite certainly won’t be adding this information clearly in their colorimeter prospects and apps, since they are probably making a good chunk of money from fools like me.
btw, I have this iSPEX doohickey laying around that were given away in 2013 for next to nothing. It has a film inside that splits light into a spectrum, that can then be read with a camera CMOS. It was good enough for actual science. Why don’t we have “super high resolution” screen spectrometers by now, instead of these apparently lacking (700nm-380nm)/3nm = 110-sensor arrays?
- This reply was modified 5 years, 4 months ago by Henk Poley.
- This reply was modified 5 years, 4 months ago by Henk Poley.
2018-11-21 at 13:30 #14567It was not a critic, just a set of clarifications of the uncertains that we need to play with, at least for measuring equipment under 2000 euro (+/-). That’s why I put examples that can help to visualize what is going on.
i1d3 devices have 3 RGB-like filters that mimic (in some way with some innacuracies) the response of CIE 2º observer.
Manufacturer stored (let’s say with some QC of what they are doing) actual filter’s spectral response in firmware, so we can dump it (thanks to free ArgyllCMS code & i1d3 unlock codes) and compare to CIE 1931 2º observer. Not equal to it but close.
That spectral sensivity gives you the chance to do not rely just in other people’s 3×3 matrix correction (like you did with your i1d2) but to compute “on the fly” an specific matrix for your i1d3 if you use a “spectral sample” of the display your are going to measure (WRGB samples).
This is a huge advance since this allow you use “good” corrections knowing just the backlight type of the display. Something like… “buy once”, and with community powered spectral samples, “you’ll have support forever”, at least for popular display types.
Of course devices will age but the nature of i1d3 filters seems that those are going to age reasonably well: there was a small test (Czornyj, LuLa) of 7 i1d3 units 2011-2017 that kept aging and low interinstrument error -spectrally corrected measurements as I explained- under 1.5dE.So now I hope to have explained what kind of “accuracy” you can expect from this kind of inexpensive devices let’s take a look on current backlights, for example with that EW3270U multimedia/movies 32″ UHD display.
This is what a very high end spectro reads:
This is what a Xrite 10nm spectro reads:
This is what Xrite default spectral sample for P3 displays looks like:
The first two SPD give close whitepoints but… had the display that was measured at 10nm that whitepoint coordinates now that we know the missing data? Same applies to other SPDs with spikes.
Even for smooth ol’ backlights like GB-LEDs the particular shape of z-bar in standard observer seems to systematically move a little towards yellow the whitepoint coordinates of those Xrite’s 10nm spectros evaluated against hi-end spectros or spectrally correted i1d3. Not too much, just a drift towards b*+.Of course this kind of “well known” displays is just a small amount of all models.
Even you can use Xrite’s generic WLED sRGB (“smooth” & “slow” spikes) correction for office displays & laptops & non P3 LED TVs (90% of LED displays out there?), it seems wise to try to find a 3.3nm correction for your particular model made by a kind xrite spectro owner as your first option.2018-11-22 at 0:19 #14578Hmm i found a Matrix, and CCSS correction for my panel, and both looks the same to me, after calibrating the panel with the correction in place i used that monitor as reference visually to manually tweak my other display that is also w-led via screen mirroring to adjust the white temp to look very close to other one, seems like i only needed to reduce the green a bit.
2018-11-22 at 10:01 #14583@Vincent, any annoyance you might have read in my reply was merely because of the opaqueness how to appropriately use these sensors. You are making good points at clarifying some areas I was not aware of.
I just wish software that drives it all would be more clear how wrong things are without calibrating the sensors. It’s not just some small improvements.
How to resolve the spikiness of modern backlight LEDs I don’t know either. If only a few LED types are used maybe you can still figure out which is which from 10/3.3nm spectrograms.
Fairly off-topic, but if you care I can upload some photo’s through the iSpex spectrograph. If I’m looking at it correctly I should get close to 950 pixels in red to blue resolution on my iPhone SE, of course filtered through the camera pigments. But I think only a few thousand iSpex were made, so I don’t know how useful that is, per se.
-
AuthorPosts