- This topic has 18 replies, 6 voices, and was last updated 8 months, 4 weeks ago by
.
-
Posts
-
Display: LG OLED55CX (PC mode)
Graphics Card: GTX 1650 @ rgb 12bit (0-255)
Madvr: 0.92.17 beta 140 (16-235)Windows 10: 19043.1165 , 21H1
Nvidia drivers: 471.68 WHQLI’ve been using a 3dlut for BT.709 material for a while now and the results are
amazing.But trying to get a working 3dlut for HDR using Madvr just does´nt work.
When bringing up the MadTPG window and turning on HDR mode (BT.2020 / 0.000 / 10,000)
the actual window looks grey and washed out and this is before even starting Displaycal.Running Displaycal with default settings for madVR HDR (D65, REC.2020/SMPTE2084/BT.2390)
and everything set to “As measured” using 425 patches and Target peak luminance 700 nits
with the 3dlut file format set to “Process HDR content”.The 3dlut is successfully loaded into the HDR section of Madvr but the results are terrible.
First of all Madvr does not change into HDR mode but the picture is so dark that you can
barely see what you are looking at.Did I miss something? Any suggestions?
The latest beta version of Madvr (v140)uses a SDR 3DLUT in HDR mode.
If you have already generated an ICC profile in RGB or Gamma 2.2 using Displaycal, you can convert that to a rec709 and rec2020 Madvr 3DLUT using the the 3DLUT maker.
Then in the Calibration section of Madvr you place the rec709 3DLUT in the rec709 part and the rec2020 3DLUT in the BT2020 part.
In the HDR section use “tone map using pixel shaders” set your displays peak luminance and it will apply that rec2020 SDR 3DLUT in HDR mode thus you get accurate colours.
The attachment is the example I use when creating the rec2020 3Dlut.
EDIT: I have calibrated my display twice to create two ICC profiles in RGB mode. The first one is at 100 nits which is the base for the rec709 3DLUT and the second is at my displays highest brightness mode which was measured at 900nits which became the basis for the rec2020 3DLUT which is what gets applied in HDR mode.
Jaylumx, great advice!
One of the first things that I tried was to place the 3dlut created from the Madvr HDR preset in Displaycal
in the Madvr calibration settings only to get an error that the 3dlut was in the wrong format.Using the 3dlut maker this finaly works, thank you! You are a lifesaver.
Question:
When making a 3dlut for BT.709 I do the measurements in none HDR mode and when doing the measurements
for BT.2020 I should be in the HDR mode. Correct?When making a 3dlut for BT.709 I do the measurements in none HDR mode and when doing the measurements
for BT.2020 I should be in the HDR mode. Correct?I did the bt2020 measurements in SDR mode but at my displays highest brightness setting (900 nits as measured). That way I was able to get 94% DCI-P3 coverage as opposed to 91%
Ok , I have done some more testing.
Using madTPG and everything set to “as measured” gamut coverage was terrible, 68%.
But when trying out the 3dlut in madvr the picture is really dark.
Movies are unwatchable. Is this due to the gamut coverage?Also, 3dlut only works when selecting “tone map HDR using pixel shaders” with the “output video in HDR format”.
Using Displaycal to generate the window with the same settings, gamut coverage went upp to 74%.
What am I doing wrong?
There is little you can do on an HDR display using its HDR mode. You cannot disable Rec2020->panel internal translation and thanks to ABL it will be difficult to map, to capture overalcolorspace behavior in an ICC so you can use it as source for a LUT3D. So do not use madVR for that. Try to fix some minor issues in HCFR with TV CMS on OSD, there is little more you can do on HDR.
For SDR using TV as a PC, check that such 12bit setting is not forcing 4:2:0 chroma subsamplig os something like that. 4:4:4 is what you want and bitdepth is the least important thing. Then keeping patch window small so ABL dows not kick in, profile as always for making a LUT3D.
Ok , I have done some more testing.
Using madTPG and everything set to “as measured” gamut coverage was terrible, 68%.
But when trying out the 3dlut in madvr the picture is really dark.
Movies are unwatchable. Is this due to the gamut coverage?Also, 3dlut only works when selecting “tone map HDR using pixel shaders” with the “output video in HDR format”.
Using Displaycal to generate the window with the same settings, gamut coverage went upp to 74%.
What am I doing wrong?
I did not use MadTPG for the calibration. I set Whitepoint to 6500K and tone curve to Gamma 2.2 with the testchart on Auto-optimized. I basically calibrated for SDR in my displays highest brightness.
I remember someone mentioning in these forums not to calibrate display in HDR mode in order to avoid some displays tone mapping which could mess up the measurements.
Update:
I Finaly figured out the problem I was having with madTpg.
Due to a Nvidia driver issue the madtpg window was shifting into HDR mode
properly even when clicking on the HDR button in madtpg.I rolled back the driver to the last one and everything works.
I am still getting poor gamut coverage though, DCI-P3 coverage 84%.
I hav’nt tried changing anything under the calibration tab, any
suggestions?One way I improved my coverage was to use a colormeter correction for my specific display and colormeter. I got them from the following link.
Jaylumx, I was wondering about the actual setup that you are running.
You said that you are using the SDR mode instead of the HDR mode and
that you did the BT2020 measurement in SDR mode but with the display
highest brightness setting.Did you change the gamut for the SDR mode when doing this or did you
just use the SDR mode as it is?Also, when later viewing HDR content are you then still using the SDR
mode and doing HDR to SDR?In madvr are you then using the “tone map using pixel shader” but without
the “output in HDR” option?Jaylumx, I was wondering about the actual setup that you are running.
You said that you are using the SDR mode instead of the HDR mode and
that you did the BT2020 measurement in SDR mode but with the display
highest brightness setting.Did you change the gamut for the SDR mode when doing this or did you
just use the SDR mode as it is?Also, when later viewing HDR content are you then still using the SDR
mode and doing HDR to SDR?In madvr are you then using the “tone map using pixel shader” but without
the “output in HDR” option?I calibrated my display using sRGB as the tone curve, used D65 as whitepoint and set white level as measured to my displays hightest SDR brightness level. I also set the testcart as default auto-optimized. The generated ICC colour profile is what displaycal loader uses as the IDLUT.
I copy that ICC profile into another directory and use that as the destination profile in 3DLUT maker while using BT2020 as the source (see earlier screen cap for my settings).
When 3DLUT maker generates the 3DLUT I set that to the BT2020 part of the calibration section in Madvr and in the HDR section i use “tone map using pixel shader” with the “output in HDR” option enabled? My display clips above 1000 nits anyway.
From what i read in the Avforums, madshi mentioned that this was a new feature (can’t remember from what beta version) which converts gamma to PQ and the process is nearly lossless.
The content I watch on my display is in HDR and it tone maps nicely to 1000 nits better that my displays native tone mapping with the added benefit that Madvr measures the peak luminance per pixel (GPU hog) similar to Dolby Vision.
The trick is to calibrate in SDR mode and get the best coverage possible and Madvr will convert the result into PQ.
This is interesting stuff.
To summarize, you adjust peak brightness of the SDR mode to its highest output.
You, calibrate the display with Displaycal using sRGB, D65 whitepoint and white level as measured.
Then using the 3dlut maker, you use the settings posted earlier.I assume that you then turn the settings in the SDR mode back to what they where AND that since
you are outputting HDR with the tonemap HDR using pixelshaders that the display in the end is in
HDR mode.Follow up question:
When you adjust the peak brightness of the SDR mode, I assume that its the OLED lights that you adjust.
How about contrast? Any adjustment there?This is interesting stuff.
To summarize, you adjust peak brightness of the SDR mode to its highest output.
You, calibrate the display with Displaycal using sRGB, D65 whitepoint and white level as measured.
Then using the 3dlut maker, you use the settings posted earlier.I assume that you then turn the settings in the SDR mode back to what they where AND that since
you are outputting HDR with the tonemap HDR using pixelshaders that the display in the end is in
HDR mode.Follow up question:
When you adjust the peak brightness of the SDR mode, I assume that its the OLED lights that you adjust.
How about contrast? Any adjustment there?Correct for your summary.
To answer your follow up question, I adjust the brightness. I have an Acer CG437KP which uses an LED backlight and will not allow me to adjust the contrast when I set colour space to HDR.
*Update on HDR 3dlut progress**
So, I have been doing alot of testing and I the results are “better” but not perfect.
The verification measurement report shows a huge differance before and after 3dlut.
Grayscale and colours are all in the “green”.But, what I am concerned is the “measured luminance” that I am getting and the gamma curve.
Measured luminance: 350 cd/m2
Is the gamma curve supposed to look like that?
I have measured display peak luminance with hdr test patterns using HCFR and I am hitting
just over 700 cd/m2 using a HDR code 940, 100% white pattern.Any thoughts?
