Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion. Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules. |
25th October 2015, 00:56 | #1 | Link |
Registered User
Join Date: Jul 2014
Location: Las Vegas
Posts: 177
|
Display Calibration
Continuing from the other thread
As already mentioned, the Spyder 4 can't correctly read black level, or at least reads it in a fashion that is not a standard. The i1 display pro reads it correctly. Calibrating with dispcalgui, I set a brightness of 120 cd/m^2, white point of 6500k. End result is somewhere around 6456k, okay, not that big of a deal. Adjusting RGB sliders around, (most annoying task in the world), I get an average delta E of 0.2. Perfect. After applying the profile, I calibrate again just to see the level. Obviously starting a new calibration temporarily removes and previous profile. With the same monitor settings, now my delta is 0.6... Measuring the delta E with HCFR, my delta E of red for example is 3.2, and average of 2.12. Seriously, wtf? I don't understand how there is zero consistency between measurements. The i1 software is fairly bad, and the spyder's is worse. I'm not paying for commercial trash either. So the only thing left is the argyII stuff. Too bad there are basically no guides on anything about it. The dispcalgui wiki doesn't actually tell you what to set, just tells you what each option does. Not very useful if you don't have a Ph.D in the physics of color. All other guides are about something else completely or outdated by 5 years. Same goes HCFR. What kind of piss poor operation is being run here? The entire color calibration business at this point seems like complete BS to me. Luckily I can still return everything I bought, and I'm fairly close to doing so. tl;dr: Not impressed by high end displays or calibration techniques. Same result, or worse. Waste of $ and time. |
25th October 2015, 05:09 | #2 | Link |
Registered User
Join Date: Feb 2002
Location: San Jose, California
Posts: 4,407
|
Do not target a brightness with the calibration at all, do you have it at ~450cd/m^2 native but are targeting 120 cd/m^2 with the calibration? That would be very bad.
I agree the i1Profiler isn't that great. It is quite important to set the right settings or a calibration can be a lot worse than nothing. I suggest using i1Profiler or dispcal to set your brightness and RGB sliders before calibration using only your monitor's controls. I use this, it targets the native white point and brightness. Native is optimal for anything but professional color work as long as they are already close to correct. Code:
"C:\Argyllx64\bin\dispcal" -v -dmadvr -qh -ye -G2.4 -f0 -k0 -O "[NAME] matrix" -o "[NAME]_1.icm" "[NAME]" "C:\Argyllx64\bin\targen" -v -d3 -G -e8 -B8 -s40 -g128 -f1800 -c "[NAME]_1.icm" "[NAME]" "C:\Argyllx64\bin\dispread" -v -dmadvr -ye -K "[NAME].cal" "[NAME]" "C:\Argyllx64\bin\colprof" -v -M "[NAME]" -D "[NAME] BT.1886" -qh -aX "[NAME]" "C:\Argyllx64\bin\collink" -v -M "[NAME]" -D "[NAME] BT.709" -3m -qh -et -Et -IB -G -ims -a "[NAME].cal" Rec709.icm "[NAME].icm" "[NAME] BT.709.icm" The Argyll documentation tells you what everything means but please ask if you want to know why I picked any particular setting. You need to set "-ye" in both the dispcal and the dispread line to your display's backlight type. "e" is white LED on my system. edit: run "dispcal" without any options to get the list of letters for the backlight types on your system, this is selecting the spectral correction file for your display type. Also make sure to enlarge the test pattern in the madVR TPG window, it starts very small, both for the dispcal and dispread steps. The file "Rec709.icm" also needs to be copied into the same directory where you are running these, it is in Argyll's "ref" folder. In the last line, collink, you can use other reference files like "EBU2312_PAL.icm" for EBU/PAL or "SMPTE_RP145_NTSC.icm" for SMPTE C to generate a 3DLUT for other gamuts. I would also change the BT.709 in "[NAME] BT.709.icm" to match. BT.2020 and P3 do not work very well on a nominally sRGB display so I don't think those are useful in your case. This generates several argyllcms files as well as: "[NAME]_1.icm", a fast matrix profile - simple windows calibration only. "[NAME.icm", a profile along with the windows calibration, photoshop and other color managed applications can use this for gamut conversations. "[NAME] BT.709.icm", a profile that includes the 3DLUT to convert from BT.709 to the monitor. Not useful as far as I know. "[NAME] BT.709.3dlut, the BT.709 3DLUT for madVR.
__________________
madVR options explained Last edited by Asmodian; 25th October 2015 at 06:02. |
25th October 2015, 11:08 | #3 | Link | |
Registered User
Join Date: Mar 2002
Posts: 2,323
|
Quote:
(I'm still hesitating whether to buy the 200 pounds value iDisplay Pro or not.)
__________________
Ryzen 5 2600,Asus Prime b450-Plus,16GB,MSI GTX 1060 Gaming X 6GB(v398.18),Win10 LTSC 1809,MPC-BEx64+LAV+MadVR,Yamaha RX-A870,LG OLED77G2(2160p@23/24/25/29/30/50/59/60Hz) | madvr config |
|
25th October 2015, 11:36 | #4 | Link |
Registered User
Join Date: Oct 2012
Posts: 7,923
|
a meter needs time to heat up and the display too.
so before you calibrated the meter should be an the screen for at least 30 mins and the screen should run at this time too. i the correct correction loaded? and most important windows icm profiles are low quality no matter what you do. |
25th October 2015, 17:26 | #5 | Link |
Guest
Posts: n/a
|
For sure - meter does need to be heated up for at least 20-30 minutes and so does the monitor (20-30 minutes). It is also advised to use USB Hub for colorimeters and spectrometers because they need more power than normal USB 2.0/3.0 can provide to work @ their highest potential.
I also recommend dispcalGUI + ArgyllCMS 1.8.2 over using raw ArgyllCMS. DispcalGUI, at this point, is very stable and almost bug-less. UNINSTALL i1Profiler and delete ALL ICC profiles for now!!! Not only that, but go through your Services and make sure X-Rite services are disabled, along with XRD service (gamma-related). Make sure there is nothing X-Rite related in StartUp either. i1Profiler tries to load ICC profiles upon your PC starting up and that can seriously mess up your calibration. Use dispcalGUI Install (not stand-alone archive) and once you install and open it up, a menu window should up, asking you if you want dispcalGUI manage ICC profiles. Select "Yes, let dispcalGUI manage profiles". Are you planning on GAMING on your PC a lot? In that case, there is no need for desktop calibration or ICC profiles. Instead, you should create a 3DLUT for ReShade Framework, which allows the use of such 3DLUT. You can now FULLY calibrate your display for gaming, including both - grayscale AND colorspace / color gamut calibration! |
26th October 2015, 00:02 | #6 | Link | |||||||
Registered User
Join Date: Jul 2014
Location: Las Vegas
Posts: 177
|
Quote:
Quote:
I have white level drift compensation on, but have tried it off, and I think on looks better. Black level drift compensation is off. Should I enable this? i1 is really fast so I don't think it slows things down by much. I have spectral: LCD White LED IPS (WLED AC LG Samsung) correction on, which is what my panel is. I target a color temperature of 6500k. The native is 6300k which is a bit off. Isn't one of the main points of calibration to get a correct white point? Tone curve is gamma 2.2 and speed is low. I use auto optimized 2681 patches and high profile quality. Verification is extended chart (51 patches). When I try it with advanced options I have the black level as measured. Gamma 2.2 relative. Black output offset to 100% and black point correction to 0% and a rate of 4. See anything wrong here? Quote:
Quote:
When I lower RGB levels, luminance goes down as expected. So it's kind of annoying to set it to 120cd/m^2, then after adjusting it goes down to 114... or if you have it to 124 and after adjusting it goes to 120. But then I get a delta E of .4 instead of .2. Quote:
Quote:
Quote:
When I did make a 3D lut, my movies looked lifeless, like the color was drained. It looked like my 10 year old piece of crap TN panel. It was not at all enjoyable. Contrast ratio affects colors other than black but it shouldn't look this bad at 1050:1. Something is definitely wrong. I'll try again with the 3DLUT because it is easier to tell if the calibration succeeded. Testing windows profile is harder. Last edited by Arm3nian; 26th October 2015 at 00:06. |
|||||||
26th October 2015, 03:45 | #7 | Link |
Registered User
Join Date: Jul 2014
Location: Las Vegas
Posts: 177
|
I profiled and made a 3D lut following the AVS guide. I see no difference. Still looks like crap. Games look fine to me, without any calibration.
The low bitrate stream from cable going into the living room TV looks better than what I'm watching. This is really depressing. |
26th October 2015, 05:01 | #8 | Link |
Registered User
Join Date: Feb 2002
Location: San Jose, California
Posts: 4,407
|
Did you try my settings? The batch file I posted is for doing a normal calibration, only the last line is for creating the 3DLUT for madVR. Feel free to increase the patch count to 2681 if you want to (change the -f1800 to -f2681 in the targen line). I don't like the newer dispcalgui versions because I don't know what it is doing.
On my system HCFR matches perfectly with Argyllcms, they also use the same meter reading code. Are you using madTPG and the same spectral correction everywhere? Changing the RGB sliders is to change the white point, which also changes dE measurements. Are you not able to get close to 6500K by only adjusting the RGB sliders? You do have to adjust brightness and the RGB sliders together, that is simply how it is. Any dE below 1 is quite good, simply moving the meter on the screen will often change a 0.2 dE to 0.4 or the other way around. Most displays aren't very uniform. Don't use white or black point correction, that is mostly for Plasma, or at least not a warmed up LCD.
__________________
madVR options explained |
26th October 2015, 19:29 | #9 | Link |
Registered User
Join Date: Jul 2014
Location: Las Vegas
Posts: 177
|
Those settings are what I used when I tried again. It looks the same as before.
I'm going to return my display. I don't see a point in 4k anymore. Nothing scales correctly when windows is at 150% and if it isn't then you can't see anything on a 27inch. Originally I thought non integer scaling looks bad, but with madVR, it looks good. 1440->4k looks great with doubling. 1080->1440 should also look good. 10bit also isn't very useful since most of the time I'm doing things other than watching movies. I don't notice much banding on a true 8bit display. Maybe I'll get the new ROG swift (pg279q). I'm convinced at this point that almost all IPS look the same. No point in paying $700 for my current display when I can get the same thing but gsync/ulmb and 165Hz refresh rate for almost the same price. I'll also probably return the i1. It looks the same as the spyder even though it can read black. Maybe I'll callibrate my new display before sending it back. |
29th October 2015, 00:35 | #10 | Link |
Registered User
Join Date: Jul 2014
Location: Las Vegas
Posts: 177
|
Why didn't I think of this before... too busy arguing with smart asses on oc.net. Here are screenshots of no calibration vs 3D LUT.
Pics from the hobbit (imgur compression kind of killed quality but the color differences are still visible) Group 1: normal vs 3D LUT Group 2: normal vs 3D LUT The ones without the LUT look darker, I guess you could say saturated. Calibrated results look more natural? Seems a bit dull though. I tried some other sources and the same results. |
29th October 2015, 06:39 | #11 | Link |
Registered User
Join Date: Feb 2002
Location: San Jose, California
Posts: 4,407
|
I don't think you could have used my command lines to generate those images. Especially if you want an image more like un-calibrated you really should try them.
However, calibration can only remove total saturation. For example, the way you fix the white point with a calibration is by reducing the saturation of the two colors that are too saturated. This is why you don't target a white point unless it is off enough to bother you. Human vision adapts white anyway so minor errors are not a big deal, the loss of contrast looks worse. In the mid-range it is possible to add saturation and a good calibration can fix a bad gamma which can make the image seem higher contrast. If you want a bit deeper/darker shadows try this: Code:
"C:\Argyllx64\bin\dispcal" -v -dmadvr -qh -ye -G2.4 -f0.6 -k0 -O "PG278Q matrix" -o PG278Q_1.icm PG278Q "C:\Argyllx64\bin\targen" -v -d3 -G -e8 -B8 -s40 -g128 -f2259 -c PG278Q_1.icm PG278Q "C:\Argyllx64\bin\dispread" -v -dmadvr -ye -K PG278Q.cal PG278Q "C:\Argyllx64\bin\colprof" -v -M "ROG Swift PG278Q" -D "PG278Q BT.1886" -qh -aX PG278Q "C:\Argyllx64\bin\collink" -v -M "ROG Swift PG278Q" -D "PG278Q BT.709" -3m -qh -et -Et -IB:0.6:2.4 -G -ims -a PG278Q.cal Rec709.icm PG278Q.icm "PG278Q BT.709.icm"
__________________
madVR options explained Last edited by Asmodian; 29th October 2015 at 06:45. |
30th October 2015, 05:49 | #13 | Link | |
Registered User
Join Date: Feb 2002
Location: San Jose, California
Posts: 4,407
|
Quote:
I, and others who's screens I have calibrated, much prefer the calibration when not targeting a white point or (shudder) a brightness. Targeting a brightness with the calibration for viewing video is simply a bad idea. There is no way you can notice a small change in brightness but you reduce saturation on all colors in order to target a brightness, which reduces contrast too. The only way to reduce brightness with a calibration is to have the video card output 254 or less when the source pixel's value is 255, this requires lowering all values to avoid banding, reducing the dynamic range for no visible benefit. White point calibration is more debatable. My view is that you should only target a white point if the current white has a color tint you do not like, and then only maybe. For almost all my calibrations on displays with a reasonable white point (~6200 to 6800K), targeting a white point does not improve my perception of white but it does make the image look more faded or washed out, like in your screen shots. Tuning with the RGB sliders on the display, with feedback from a meter, has always allowed me to get a good enough white point so I do not need to target one with the calibration. The sad part is that on displays with a really bad white point (e.g. >9000K) targeting a white point reduces contrast and saturation enough that most people have preferred not targeting a white point even in extreme cases. If blue can only use 0-217 and green 0-188 it reduces the saturation of all colors by a lot. I have enough experience trying to notice display color accuracy that I do get bothered by white points too far from 6500K and I do prefer them calibrated to 6500K, even with the loss of contrast and saturation. I constantly notice the bad white instead of watching the video. I have handled it by giving away or returning all my displays with bad white points, rather than calibrating them into a faded mess. This may give the impression that I think calibrations are somewhat useless but the calibrated gamma curve is generally much better, as long as you do not get banding, and tuning the white point with the meter and the monitor's controls often helps a lot.
__________________
madVR options explained |
|
31st October 2015, 20:53 | #14 | Link |
Registered User
Join Date: Jul 2014
Location: Las Vegas
Posts: 177
|
Now I understand. So "as measured" simply uses what you set the white point and white level to be via monitor controls. And targeting a value means adjustments are made in software to get you where you couldn't with the monitor controls.
I thought that "as measured" meant it used the initial reading without adjustments and targeting meant it just brought up a value and told you if you were close or not... fail I can get very close to 6500k and 120cd/m^2 with a delta E of 0.2-0.3 using the sliders. I'm guessing as measured is recommended for system wide calibration as well? What should I use for the tone curve for a windows profile, gamma 2.2 or as measured (I set the settings to default (gamma 2.2) in the top)? I am not doing pro image work or printing, just want nice colors. I also don't use the advanced options as they seem to make things worse if you don't need it. Is auto-optimized 2681 patches good for windows? |
2nd November 2015, 05:24 | #15 | Link | |
Registered User
Join Date: Feb 2002
Location: San Jose, California
Posts: 4,407
|
Quote:
Also definitely do the same for the normal system calibration, I would argue it is even more important to not limit saturation globally. For gamma I think 2.4 actual (-G2.4) or 2.2 effective (-g2.2) are both reasonable targets for a Windows calibration. I prefer actual but effective is better for matching displays with different black levels. 2681 patches is probably overkill to be honest, but the only reason to use fewer is if you get banding with too many (speed too, of course).
__________________
madVR options explained Last edited by Asmodian; 2nd November 2015 at 05:34. |
|
7th November 2015, 11:49 | #16 | Link |
Registered User
Join Date: Apr 2006
Posts: 299
|
I curious if anyone here uses f.lux together with a display profile and/or MadVR 3DLUT. Sometimes I turn it off for color critical work but most of the time it's on and does a excellent job and reducing eye strain. So I've been concerned with what kind of implication this might have on image quality. The f.lux filtering seems to sit on top of both the 3dlut and display profile so I guess it's an overlay, which should allow a luma corrected filter. But how precise can it be when it's not aware of the specific display characteristics?
Last edited by markanini; 7th November 2015 at 14:51. |
8th November 2015, 03:46 | #17 | Link | |
Registered User
Join Date: Feb 2002
Location: San Jose, California
Posts: 4,407
|
Quote:
edit: Actually f.lux on top of a calibration is probably better than without the calibration; f.lux does try to do a good job visually, even if it cannot be accurate, so it is still likely to look better after the calibration. However, it is especially important to not target a white point with the software calibration when using f.lux. This is expected based on how it works; it is doing its own calibration without any understanding of the display, as you mentioned, and it is targeting reduced eye strain instead of color accuracy. It wants color accuracy to be bad, you cannot arbitrarily remove blue light and get accurate colors. There have also been reports of f.lux interacting badly with madVR, causing dropped frames and/or choppiness. I would do a normal calibration but targeting 5003K (illuminant D50, x0.34567, y0.35850) or maybe 5503K (D55, x0.33242, y0.34743) for the white point if I wanted a low blue light calibration. This might not be as low eye-strain as using f.lux on top of a calibration targeting 6503K (D65, x0.31271, y0.32902) but it would give accurate colors relative to white.
__________________
madVR options explained Last edited by Asmodian; 8th November 2015 at 08:47. |
|
8th November 2015, 08:58 | #18 | Link |
Registered User
Join Date: Jul 2014
Location: Las Vegas
Posts: 177
|
I calibrated again using "as measured", but using targeted values and then cancelling before to get my desired levels. After setting levels, my color temperature was 6484K, white level was 119cd/m^2, and delta E was .1 to .3. After, measuring with HCFR, my color temperature went lower to 6450K, white level went up to 121cd/m^2, and delta E of white is 1. So not bad, but why is it worse than before calibrating? Contrast ratio is around 1126:1.
The individual delta E's are 3.4, 3.8 and 1.1 for RGB. Why are they so high? Do you get the same results? Turning the profile on or off makes no changes in HCFR readings... To compare subjectively, the calibration looks the same as before except a bit dimmer. It looks good, but the objective measurements are fishy. Speaking about eye care, my display has an internal blue light filter that is set to 100 by default. How does this affect color accuracy? Last edited by Arm3nian; 8th November 2015 at 09:03. |
9th November 2015, 03:31 | #20 | Link | |||
Registered User
Join Date: Feb 2002
Location: San Jose, California
Posts: 4,407
|
Quote:
Quote:
Quote:
I don't know, assuming it was on when you calibrated it sounds like it is fine, a max dE of 3.8 isn't bad. What is the dE for the cyan and magenta saturation measurements? What happens to the appearance and HCFR measurements with it off?
__________________
madVR options explained |
|||
Thread Tools | Search this Thread |
Display Modes | |
|
|