| Forums | Register | Polls | Search | Statistics |
 (?)  
You must be logged in to post content on this forum.
Display Calibration Light Illusion Forums / Display Calibration /  
 

Newbie calibrating a standard display monitor (Dell s2415h)

 
Author andreachi
ZRO
#1 | Posted: 2 Jan 2021 09:22 
Hello and happy New Year!
First off Colourspace is an amazing piece of software, very powerful.
So far my only experience is with a i1Display PRO for basic calibration with HCFR and Ligthspace on a Pioneer Kuro tv and a Panasonic AE5000 projector.
Results were very good mostly thanks to the relentless inputs from Master Ted.

As a source I use HTPC's be it in the form of a recent i7 Dell laptop where Colourspace is installed and which I use for checking 3dluts and testing all madvr settings (not really for playing movies) and other 2 htpcs with Nvidia cards respectively connected to the projector and a LG C9. Player is MPC-BE + 3dlut\madvr.

Before taking the task to calibrate the Oled and the vpr again (since 1500 hours have passed and a new screen) I want to play a bit and improve my knowledge of CS as much as possible while playing with a couple of spare monitors I have around..a Dell S2415h and a Benq E277HDR.
They're not reference monitors at all but that's the point.

My workflow is somewhere on those lines:
1) connect hdmi directly to the display or run madTPG remotely on the HTPC's display

2) set the display in the best uncalibrated mode leaving most settings at default with the least correction of RGB high values while avoiding extreme brightness and contrast values
> aiming for 100 nits

The s2415h monitor is a basic fullHD w-led display with only few color presets including a custom one which I set against a 100% white patch as: r -99 g -100 b -100 brightness 32 contrast 64

Native display seems pretty 2.4 gamma and around rec709 coverage altough not quite there looking at B and R ..
I believe this is the reason why no values I tried could better the rgb separation graph ? altough switching hdmi output from RGB to YCbCr did seems to provide a (very tiny) lesser separation

3) Center the i1Display PRO probe and check CS patterns making sure brightness and contrast are all displayed except the extreme ones

4) Profile at least 10^3 or better 17^3 + drift (more necessary on plasma and oled though?) using madTPG as pattern generator
>> madTPG pc levels 0-255 + CS patch scale 0-255 (settings in attachment)

5) Confirm measured contrast values with data specs of the monitor
>> in this case data spec of 1000:1 vs a measured one around 989 ..pretty close !

6) Creates different 3dlut with\without disable gamut and in all flavours: peak chroma\luma, fit\map space and hybrid
>> in this case: source rec709 native 2.4 gamma > destination either 10^3 or 17'3 profile

7) Check the created 3dlut with the LUT viewer looking for obvious graphical glitches (attached), banding, color deviation etc.

8) Run primary only\edge or 1000 point to verify the different 3dlut and pick the "best" one

9) Add the best videoscale out of the best 3dlut >> last round of verification mostly with edge patterns
I picked passblack in this case being the most "neutral" one

Finally grab some well deserved coffee and play known movie parts to see what the new 3dlut brings out
- are the skin colours natural ?
- is the perceived contrast "better" ?
- are all whites now pink lol ? etc.

Is this workflow formaly correct ?

Couple of random quirks:
- I found that 3dlut loaded in madvr and displayed thru madtpg Vs the same 3dlut loaded in CS and displayed thru CS patch generator will return up to a 0.5 value or even more difference
Please note that without any 3dlut loaded on either will return the same value on same patch (for instance 70-80% red where it is more obvious then other colours\values)
I always use madTPG since this will repilcate the source\display chain during movie watching

- Generally speaking can the rgb separation be bettered by the 3dlut correction or will it at best be the same if not more spread out after a 3d lut?

- Would it be possibile to add somewhere in the graphs option (or anywhere else) a count of green, orange and red ?
something like green (578) orange (223) red (199)... this would help in the evaluation !

- Comparing the 10^3 and the 17^3 profile with marginally different settings between the 2 the latter provided slightly better rgb separation but then 3 out of 5 3dlut presents graphical glitch in the 3dlut viewer (in attachment)..is it because the monitor is just a cheapo one and not worth more then 1000 points or I messed up somewhere in between?

To cut short after dozens of measurements I believe mapspace + passblack is the right 3dlut that squeeze the most out of display considering:
- no graphical glitch or banding in the 3dlut viewer
- among the lowest dE value of the 100% white (0,39) ?
- the best gamut coverage adherent to the native rec709 gamut ?

In any case picture is perfect to my eyes and looks totally spot on while without 3dlut seems lifeless (no contrast, exaggerated skin tones etc).
Sorry for the long post please share any idea\correction\input..much appreciated !
Greetings from Italy

pre and post profiling.zip Attached file:
pre and post calibration

 
edge verification.zip Attached file:
edge verification

 
3dlut viewer.zip Attached file:
3dlut viewer

 
settings.zip Attached file:
settings

 

Author Steve

INF
Male
#2 | Posted: 3 Jan 2021 12:17 
Thanks for the post.

A couple of replies.
The LUT Processing in madVR has been reported to have some issues.
The issues have been flagged to Mathius by one of our users, who has tested this rather fully.
(You can test this yourself by using the LUT Image to 'rip' the effect of the LUT from nay system that applies LUTs.)

RGB Separation can be 'worse' after calibration, if the display has a poor gamut, and/or poor linearity.
This is expected for such poor displays.

When 'Reports' are released there will be more information added to those regarding all aspects of the calibration.

With any display that is poor, there will likely be artefacts on the generated LUT if the profile data is not able to accurately 'map' the issues in the displays pre-calibration. Using more or less points can bot show less artefacts - using less can reduce artefacts as the fewer points men less 'correction' is being applied, so such issues are 'masked' (but the correct will be less accurate overall). using more points can reduce artefacts as the increased data can help improve the calibration results.
(The seeming artefact you have circled is just due to the low gamut, and is not an artefact - it is excepted.)

As for the correct processes to use - the application of VideoScale, or not, depends on your signal path.
You either need it, or not - there is no 'what looks best'.
Using Pass Black, or not, also depends on your signal path having the ability so show blacker than black or not.
(With home video there is never any valid data blacker than black, or whiter than white, regardless what anyone else says.)

It is also rare that Peak Chroma is not the best LUT process to use, unless the display is really poor.
This displays seem borderline.

Hope that helps.

Steve
Steve Shaw
Mob Boss at Light Illusion

Author andreachi
ZRO
#3 | Posted: 3 Jan 2021 20:36 
Thank you Steve !
glad to see im not far away from a coherent workflow , obviously with this display in essence being a limiting factor on what is possible to achieve.
Your insights (and Ted's) are highly appreciated
Andrea

You must be logged in to post content on this forum.
Display Calibration Light Illusion Forums / Display Calibration /
 Newbie calibrating a standard display monitor (Dell s2415h)

 

 
Online now: Guests - 1
Members - 0
Max. ever online: 192 [11 Jan 2023 08:39]
Guests - 192 / Members - 0