- This topic has 36 replies, 8 voices, and was last updated 4 months, 1 week ago by
.
-
Posts
-
Hi there,
I’m hoping that I can get some help to finally figure out the absolute correct way to calibrate the Eizo CG2420 monitor for use with Davinci Resolve. It seems to be a monitor that a few people have and that sparks a few questions so I thought it would be good to try and get one comprehensive thread put together so that people can be pointed in it’s direction in the future.
Personally, I’ve found the whole thing really confusing and each time I think I’ve found out the ‘correct’ way to do it, I then discover something that I didn’t know before that I should have been doing which leads to a big visible shift in the calibration. It’s meant that I don’t know whether or not I can trust my calibration which isn’t great when you’re trying to colour grade.
What’s really annoying is that each time I think I figure it out I see a big change but have no way of knowing if that change is right as I’ve got nothing to compare it to.
I’m going to post my current settings here but hopefully as people suggest changes I’ll combine them in the post bellow so that anyone coming to this post in the future can more easily workout what to do.
I’ve really done my best to go through other threads and have literally spent cumulative days looking through things.
I would really appreciate it if someone could just walk me through the process step by step. Hopefully it will also save some time for people in the future.
If you see anything bellow that should be changed then please say!
Setup
– PC (Windows 10)
– Windows program monitor – (Eizo CG2420 – display port input via GPU)
– Resolve grading monitor – (Eizo CG2420 – HDMI input via Blackmagic Mini Decklink 4k)
Step #1 – Monitor Settings
– Signal
- Input Range: Full
- Input Color Format: RGB
Step #2 – ColorNavigator7
– Calibration Target
- Brightness: Standard Input – 120 cd/m2
- Black Level: Minimum
- White Point: Color Temperature – 6500K
- Gamma (EOTF): Standard Input – 2.20 (2.2 for online delivery/ 2.4 for broadcast)
- Priority: Standard
- Gamut: Native
– Calibrate using the above target
Step #3 – DisplayCal
– Display & Instrument
- Display: Resolve
- Instrument: i1DisplayPro
- Mode: Refresh (generic)
- White level drift compensation: Unchecked
- Black level drift compensation: Unchecked
- Correction: LCD PFS Phosphor WLED IPS, 98% Adobe RGB/96% P3 (HP DreamColor Z24x G2)
– Calibration
- Interactive display adjustment: Checked
- Whitepoint: Chromaticity coordinates – 0.3127x, 0.3290y
- White level: As measured
- Tone curve: As measured
- Calibration speed: High
– Profiling
- Profile quality: High
- Test chart: Auto-optimized
- Amount of patches: 1553
– 3D LUT
- Create 3D LUT after profiling: Checked
- Source colorspace: Rec709 ITU-R BT.709
- Tone Curve: Custom – Gamma 2.2 (Absolute)
- Black output offset: 100%
- Rendering intent: Absolute colorimetric with white point scaling
- 3D LUT file format: IRIDAS (.cube)
- Input encoding: Full range RGB 0-255
- Output encoding: Full range RGB 0-255
- 3D LUT resolution: 65x65x65
– Verification
- Testchart or reference: Extended verification testchart
- Simulation profile: Rec709 ITU-R BT.709
- Use simulation profile as display profile: Checked
- Tone curve: Custom – Gamma 2.2 (Absolute)
- Black output offset: 100%
- Device link profile: Checked
Step #4 – Resolve
- Switch monitor over to Blackmagic Decklink output (make sure you’re using the profile generated in ColorNavigator earlier)
- Disable any video monitoring LUTs currently applied and proceed to calibrate through DisplayCal
- As you calibrated in ColorNavigator earlier the interactive display adjustments can’t be changed at this point
- Finish the calibration and load LUT into Resolve
Calibrite Display Pro HL on Amazon
Disclosure: As an Amazon Associate I earn from qualifying purchases.Step #2 – ColorNavigator7
– Calibration Target
- Gamma (EOTF): Standard Input – 2.20 (2.2 for online delivery/ 2.4 for broadcast)
– 3D LUT
- Tone Curve: Custom – Gamma 2.2 (Absolute)
You should set LUT3D to emulate gamma 2.4 too if you want that your display simulates that. Otherwise even if you set CN to g2.4 resulting LUT3D will make display to behave like a gamma 2.2 display.
Actually… you can even choose whatever gamma you want in CN, but you have to choose g2.2/g2.4 in LUT3D.
I mean you can use the same HW calibration (let’s say 2.2) for your 2 Resolve workflows (online/broadcast), just change Resolve LUT3D from one to another (2.2/2.4).Your guide seems fine & will be useful
Step #2 – ColorNavigator7
– Calibration Target
- Gamma (EOTF): Standard Input – 2.20 (2.2 for online delivery/ 2.4 for broadcast)
– 3D LUT
- Tone Curve: Custom – Gamma 2.2 (Absolute)
You should set LUT3D to emulate gamma 2.4 too if you want that your display simulates that. Otherwise even if you set CN to g2.4 resulting LUT3D will make display to behave like a gamma 2.2 display.
Actually… you can even choose whatever gamma you want in CN, but you have to choose g2.2/g2.4 in LUT3D.
I mean you can use the same HW calibration (let’s say 2.2) for your 2 Resolve workflows (online/broadcast), just change Resolve LUT3D from one to another (2.2/2.4).Your guide seems fine & will be useful
That’s good to know, thanks. Do I need to have the monitor set to 2.4 or can I have it set to 2.2 and it won’t matter either way?
Also when it comes to interactive display adjustment. Does it matter if the bars don’t perfectly line up (picture attached)? As I’ve calibrated in ColorNavigator the screen adjustments are locked so I can’t adjust.
Finally, is there a way that I can check if a calibration is ‘correct’? I’ve done calibrations before which I’ve assumed are right and which a report seems to show are good, but when I started using the HP DreamColor calibration on there was a massive magenta shift in my perception but the report still shows it as being fine.
Step #2 – ColorNavigator7
– Calibration Target
- Gamma (EOTF): Standard Input – 2.20 (2.2 for online delivery/ 2.4 for broadcast)
– 3D LUT
- Tone Curve: Custom – Gamma 2.2 (Absolute)
You should set LUT3D to emulate gamma 2.4 too if you want that your display simulates that. Otherwise even if you set CN to g2.4 resulting LUT3D will make display to behave like a gamma 2.2 display.
Actually… you can even choose whatever gamma you want in CN, but you have to choose g2.2/g2.4 in LUT3D.
I mean you can use the same HW calibration (let’s say 2.2) for your 2 Resolve workflows (online/broadcast), just change Resolve LUT3D from one to another (2.2/2.4).Your guide seems fine & will be useful
That’s good to know, thanks. Do I need to have the monitor set to 2.4 or can I have it set to 2.2 and it won’t matter either way?
As explained before, resolve gamma is fixed through LUT3D configuration. CN with 2.2 seems a reasonable aproach for both and for using a photo edition app since it is close to common colospaces TRC (color managed appd like Photoshop will color manage “less” if TRC match)
Also when it comes to interactive display adjustment. Does it matter if the bars don’t perfectly line up (picture attached)? As I’ve calibrated in ColorNavigator the screen adjustments are locked so I can’t adjust.
Profile will take account of it, so LUT3D will compensate with abs. col. intent. If you wish to trust CN’s white do not use a white targe in DisplayCAL and use some relative intent in LUT3D.
Finally, is there a way that I can check if a calibration is ‘correct’? I’ve done calibrations before which I’ve assumed are right and which a report seems to show are good, but when I started using the HP DreamColor calibration on there was a massive magenta shift in my perception but the report still shows it as being fine.
Device link profile is equivalent to a LUT3D. DisplayCAL wil generate one when computing LUT3D. Use it in additin to simulate a colorspace while verifying.
Thanks as always for your advice. What I’m still not understanding is how to confirm that a result is correct. I’ve attached two measurement reports for calibrations carried out with and without a correction profile. As far as I can tell, both reports say that everything is looking good. But in reality there is a big visual difference between the two. The profile without correction has a big magenta shift relative to the profile with correction.
How come the measurement reports are the same? And would you be able to confirm (based on the report) that my new process using a spectral correction is giving me the correct result?
Thanks!
If you do not apply a correction to colorimeter it can measure a not accurate value, it depends on colorimeter “observer”. Usually an i1d3 has a good but not perfect matching to CIE 1931 2 degree observer, but it stores in firmware its own sensivity curves = you may know where those 2 observer are different. If you feed a spectral power distribution sample (like that HP Z24x) you can compensate where these observers drift.
Hence a “true d65 white” (by CIE 1931 2 degree observer) can be seen as not OK by this uncorreted observer so calibration will compensate this “non d65 white” seen by uncorrecetd colorimeter and you may end with a green or pink cast (but “white” by uncorrecetd observer). Yellow or blue cast are easier to go unnoticed because it looks “white”, see wikipedia or displaycal plots for daylight/blackbody curves.
If you validate an uncorrected observer calibration with the same uncorrected observer it sees all OK. No measurement from this workflow its accurate at all but since you are measuring the same wrong coordinates all time it goes unnoticed numerically.Thanks Vincent. That all makes sense. So out of interest does the attached measurement report look good or is there anything that looks off to you? Sorry, I’m still just trying to learn what everything means.
a*b* range in greys with device link applied, “tint coloration of greys” is a bit unusually high since typical CG/CS series have a superb ultra low grey range (extremely neutral grey).
Choose “RGB+gray balance”, look for “RGB gray balance (>= 1% luminance) combined Δa*00 and Δb*00 range”.
That range value is trying to sumarize color oscillations you see in greys if you compare a*b* columns in the right with perfect neutral grey on the left. For example a* <0 is a tiny green and a*> is a tiny pink tint.
It goes to pink and the to green across grey ramp, like a wave.IDNK what causes this in your setup. It may be a device link /LUT3D interpolation limitation or it may be caused by user trying to correct display white. Ask Florian or ArgyllCMS maillist.
If the second is the cause, then if you do a new Resolve LUT3D from the same HW calibration (do no modify it) but you choose native white in DisplayCAL and/or relative white point instead of absolute when making LUT3D, then range value should go down. Also assumed vs measured whitepoint error will go higher since you are using white calibrated by Colornavigation with its unknown correction for colorimeters.
IDNK if that reported not so good range is visible, I can spot visually color tint in non color managed grey gradient with 1.5 range in sme displays, you “see it”.
Try to import to a sample Resolve project a PNG black to white smooth (no steps) 8 bit gradient like lagom’s PNG gradient and check if tint is visible. If you see tint and it bothers you then try to keep Color Navigator white as I said, without correcting it.************
Also as a question for @fhoech, IDNK if it is possible to make LUT3D/device link profile with some “preserve neutral grey priority” even if it lowers accuracy of most neighbor nodes, like trying to print a proof in B&W with RGB driver like most common printers or if people want to grade a partcular scene in B&W (or a color movie) without unwanted tints in screen acused by high grey a*b* range.
-Is it possible now?
-Is it an interesting upgrade for future versions?
-Can improve current range going to 17x17x17 measurements when making the profile for reslve LUT3D with current ArgyllCMS/DisplayCAL?@macleod92 BTW, Color Navigator has some spectrophotometer to colorimeter matching functionality, typical matrix correction I suppose (CN is a black box).
Spectral corrections in the end compute on the fly a matrix correction for YOUR colorimeter based on standard observer to match, colorimeter observer (firmware sensivity curves) and sample backlight spectral power distribution.
i1d3 raw data read from USB are not CIE XYZ coordinates but raw RGB data. It is needed a RGB -> XYZ matrix conversion.
“No correction” uses AFAIK colorimeter own sensivity curves as a display backlight sample to compute RGB->XYZ
“XXXX.ccss” computes its own particular RGB->XYZ
ArgyllCMS command line plot those matrices if you enable a very verbose oupput, -d 3 or somethig like that witha simple spotread command, no actual readings are needed, this is computed on i1d3 initialization.Hence it is possible (very basic matrix manipulation) to built a 3×3 matrix from XYZ “no correction” readings to “CCSS corrected” readings (from ArgyllCMS console output) and it will be equivalent to apply a CCSS to your colorimeter.
So in principle it is be possible with this workarround to use CCSS equivalent correction in software that only allows custom 3×3 matrices like Calman & such but no custom CCSS, even some of them do the maths for you.
If you are worried about CN readings in whitepoint not matching DisplayCAL’s so LUT3D will have some absolute white point scaling and all that stuff we talked about, then you may ask Eizo as a customer of their premium CG line about how to feed a custom 3×3 matrix to CN.
I would even say that manual 3×3 correction should be mandatory for all these apps: Eizo, NEC, Dell, Benq, LG… etcHi Vincent – thanks for sending the above through. I think I understand what you’re saying although before I clarify the point about adding 3×3 LUTs I actually wanted to run the latest issue by you.
Based on your message above I decided to try using ColorNavigator7’s white point and set DisplayCAL’s Calibration settings to the following:
– Whitepoint: As measured
– White level: As measured
– Tone curve: As measured
Then within the 3D LUT tab I made sure to change Rendering intent to ‘Relative colorimetric’.
Otherwise I left all CN7 and DisplayCal settings the same as above.
When comparing the measurement report (both attached) for the new calibration I see that, as you predicted above, DisplayCal sees the white point as being way off. There is however an improvement in RGB gray balance (not perfect but an improvement).
My problem lies in the visual difference between the two calibrations (CN7’s white point vs. DisplayCal corrected).
The CN7 calibration looks like it has a gamma shift/ is brighter where as the DisplayCal corrected version is noticeably darker.
In terms of the tint there is also a noticeable difference between the two with one looking more magenta whereas the other looks more green.
In regards to tint my problem is not knowing which one is correct.
For the CN7 side of the calibration I’m using the CG2420’s built in calibration prob. As it’s built in I would assume that it has some form of spectral calibration applied by CN7 (although as CN7 is a black box I’m aware I can’t confirm this).
I’ve been comparing the two DisplayCal calibrations by looking at an 18% gray as well as a pure white screen in Resolve and while there’s definitely a color shift between the two I really can’t tell which one is correct. The longer I stare at each the more it ends up looking like the ‘right’ white.
It’s not overly clear but I’ve attached a couple of photos that I took on my phone of the two different calibrations.
So basically:
– Are there any settings I’ve got wrong in DisplayCal which would explain the gamma difference in either of the calibrations?
– Can you think of any way for me to confirm which ‘white’ is the correct tint?
P.S. I should say thank you for all your help, I really appreciate it. It really feels like monitor calibration shouldn’t be this difficult!
P.S. – for what it’s worth, I have two calibration profiles set up on my CG2420. A Resolve specific one using the settings outlined in the original post, and a Windows one which is set to sRGB, 2.2, 6500K, 120 cd/m2.
When I toggle between Windows and Resolve (using the CN7’s white point calibration LUT) using the image reference I attached above they seem to match fairly well whereas as the DisplayCal White calibration exhibits the same issue outlined above.
Not sure if this makes any difference but I thought I would mention it.
IDNK “lesser non-LUT3D CG’s” colorimeter accuracy. CN can use i1DisplayPro and will apply “some” correction, try it. Also Eizo’s CN or NEC Spectraview II have “visual whitepoint editor” feature like DisplayCAL, look in its manual.
Unless you have some reference measurement device is difficult to compare. I would say that i1d3 + 1nm AdobeRGB WLED PFS correction from Z24x should be close to actual color coordinates.
So you can use CN + i1d3, fix that little whitepoint tint in white (2dE is not very much, and it’s mostly in a* axis ) using CN visual white point editor until it looks white (if needed), then verify CN profile before all that LUT3D stuff with i1d3 and Z24x correction and see if white now falls close to daylight white.
Or just ask Eizo if it is possible to feed a manual 3×3 correction to CN (a 3×3 that you can cook from CCSS spectral data and your i1d3 firmw data)Once all looks ok, do LUT3D. If whitepoint is on spot you can even use abs col
IDNK which may cause your gamma issue, maye Florian can help you better than me regading rel. col. limitations in LUT3D maker.
P.S. – for what it’s worth, I have two calibration profiles set up on my CG2420. A Resolve specific one using the settings outlined in the original post, and a Windows one which is set to sRGB, 2.2, 6500K, 120 cd/m2.
When I toggle between Windows and Resolve (using the CN7’s white point calibration LUT) using the image reference I attached above they seem to match fairly well whereas as the DisplayCal White calibration exhibits the same issue outlined above.
Not sure if this makes any difference but I thought I would mention it.
They use the same reference, hence their D65 white must match.
Hi again,
I decided to try bypassing ColorNavigator entirely to see what sort of result I got. I therefore did all of the calibration within DisplayCal.
I manually set the monitor settings to the same thing that I would if using ColorNavigator.
I used the Resolve preset with the HP DreamColor Z24x G2 Correction. I pretty much left everything to default with the exception of changing the following in the 3D LUT Settings:
- Tone curve – Custom: Gamma 2.2 (Absolute)
- Black output offset – 0%
Otherwise everything was just left as I would normally have it. The only difference this time was that when the interactive display adjustment options came up I was able to adjust with the monitor controls whereas normally this would be based on ColorNavigator’s calibration from the built in probe.
What I’ve ended up with is an even worse profile than before which doesn’t make sense to me. The RGB gray balance combined range is 2.67 which is obviously not great.
I assumed that doing everything through DisplayCal would take a lot of the complication out of things. I’ve attached the measurement report. If anyone has any insight as to why my RGB gray balance is so bad that would be amazing.
Thanks!
