- This topic has 278 replies, 20 voices, and was last updated 2 years, 10 months ago by
.
-
Posts
-
Happy to report I was able to calibrate my B8 in SDR, my 3DLUT (haven’t touched the 1D) yielded an average ΔE*00 of .23 (max 1.05), report is attached. DisplayCAL seriously puts CalMAN to shame, it’s embarrassing!
Is there any viable way for me to work on HDR now? I tried switching the TV to calibration mode, then calibrating to BT2020, D65, g2.2 (which is apparently what the TV switches to in that mode?). The resulting profile reported pretty bad coverage (75% sRGB vs. about 100% for the SDR mode), and I’m unable to verify my results properly (see attached report for an obvious example). All my attempts at producing a proper LUT have also ended in failures. What gamma should the LUT target, not 2.2 surely? Can I only change the 3D one here too? Any pointers would be appreciated, surely I must be missing a step here.
Happy to report I was able to calibrate my B8 in SDR, my 3DLUT (haven’t touched the 1D) yielded an average ΔE*00 of .23 (max 1.05), report is attached. DisplayCAL seriously puts CalMAN to shame, it’s embarrassing!
🙂 I’m happy for you, although I haven’t got similar results.
Would you mind:
– sharing what settings you used in Displaycal? (preferebly screenshots)
– which patch sets did you use for profiling and verification (can you upload them if it’s not available)And can you repeat your SDR verification with the built-in video_xxl (~750 patches) set? This one has way more grayscale points, so problems with gamma, color temp, etc will be more visible.
Thanks!
Yes, screenshots are attached. TV was set on expert2, g2.2, black level low (madvr set to 16-255), backlight at 27 (120cd/m² for me), contrast at 90. White balance is adjusted through the user settings from the warm2 profile. Sadly I blew everything up this morning while trying to work on HDR, still unsure why (my bash history confirmed I haven’t interacted with SDR profiles…), I’m running a new one right now. How can I make sure I get 100% repeatable results with aiopylgtv (i.e. how do I make sure I’m calibrating from a clean TV profile/LUT)?
You can use the ddc_reset function in aiopylgtv to reset to a consistent state. It has an additional boolean parameter reset_1d_lut to control whether the 1D LUT is reset or not. Since the 2018 sets at least behave differently once the 1D LUT has been set/reset you may want to keep the factory 1D LUT, ie call ddc_reset(picMode, False)
As soon as the 3d LUT is reset, the color gamut is locked to wide.
As soon as the 1D LUT is reset, gamma is locked to 2.2 and color temperature is locked to bypass (equivalent to 192/192/192/64/64/64 in the service menu)
When the 1D LUT is NOT reset, the closest you can get to replicating this is by setting gamma to 2.2, and then for example setting white balance gain/bias in the service menu to 192/192/192/64/64/64 for the cool color temperature, and then setting the white balance in the regular UI to cool. (for modes which instead have the slider, this is equivalent to setting the slider as cold as possible.). Using the cool mode for this has the advantage that the factory “warm” presets aren’t messed up and you can still use the factory “warm2” preset for accurate-ish white balance after a factory reset.
I’m still not sure what the best combination of black level/brightness/contrast settings on the TV and madvr output levels are (will investigate that further)
Yes, screenshots are attached. TV was set on expert2, g2.2, black level low (madvr set to 16-255), backlight at 27 (120cd/m² for me), contrast at 90.
Couple of thoughts:
– why did you raise the contrast from default 85? (I don’t think it ‘s a good idea unless you have some big issues with TV/firmware)
– full field pattern insertion: why didn’t you use 0 , 0.5 , 0 ? (why 10% level???)
– tone curve on 3dlut: why didn’t you use gamma 2.4 (for SDR dark room = expert2)?How can I make sure I get 100% repeatable results with aiopylgtv (i.e. how do I make sure I’m calibrating from a clean TV profile/LUT)?
After you created a facrtory reset in the TV (to get back all the factory 1dluts), you can use these commands to just remove a 3dlut (= upload unity 3dlut) from SDR modes:
aiopylgtvcommand 192.168.1.78 start_calibration expert2 aiopylgtvcommand 192.168.1.78 upload_3d_lut_bt709 expert2 aiopylgtvcommand 192.168.1.78 end_calibration expert2
Note, that when I do this I don’t see any difference in the image on my B8 (unlike when you upload unity 1dlut).
Didn’t know about that ddc_reset parameter, that’s neat, thanks! I’m pretty sure I’ll keep only using a 3DLUT though, at least for SDR.
My whitepoint isn’t quite perfect, which is frustrating, but I was able to get my SDR profile back on tracks, see attachments for the report. Maybe I’ll try running new measurements to fix it but I’m gonna leave everything alone for a few days, it probably wouldn’t make a huge difference to my eyes anyway.
chros: contrast 90 looked like a decent value for most people (both here and on avsforum), same thing for g2.2. BFI 1.0/0.5 does the same thing as 0/0.5 as far as I can tell. I picked 10% to make sure the ABL wouldn’t be an issue. Again, I’m not 100% satisfied with my recent results, if you have tweaks to suggest they’d be very much appreciated!
My whitepoint isn’t quite perfect
Yes, it’s too warm.
contrast 90 looked like a decent value for most people (both here and on avsforum)
🙂 I have never seen anybody doing like this: either we leave it at default (what ArgylCMS doc suggest) or set it to 100.
Try out with 85.same thing for g2.2
2.2 gamma is good for day usage, but for evening/night you need 2.4 gamma (= BT.1806 when black is 0)!
BFI 1.0/0.5 does the same thing as 0/0.5
1s vs 0s: 0 means inserting frame after every frame (while 1s is probably just enough with madvr).
Change it to 0.I’m not 100% satisfied with my recent results
Now the problem is visible with the video_xxl chart: the higher end gamma is not that beautiful, but the rest looks really good.
Maybe anybody can write instructions to aiopylgtv ?) Its looks like very hard to understand. Text or videos how to connect to tv and upload 1d, 3d luts ? it’s also embarrassing this “WARNING: Messing with the calibration data COULD brick your TV in some circumstances, requiring a mainboard replacement. All of the currently implemented functions SHOULD be safe, but no guarantees.” Its realy to broke TV?
Thanks chros, g2.4 seems indeed more comfortable (tracking still isn’t that great though). The new report is attached, I’m gonna go ahead and call it my final result, I’m absolutely done with chasing SDR perfection right now. 😀
Is HDR calibration still kind of hopeless? I’m supposed to part ways with my probe at the end of the month.Maybe anybody can write instructions to aiopylgtv ?) Its looks like very hard to understand. Text or videos how to connect to tv and upload 1d, 3d luts ? it’s also embarrassing this “WARNING: Messing with the calibration data COULD brick your TV in some circumstances, requiring a mainboard replacement. All of the currently implemented functions SHOULD be safe, but no guarantees.” Its realy to broke TV?
You’ll need Python, go grab it here. After installing it, run
pip install aiopylgtvfrom a command prompt, you should then be able to run stuff likeaiopylgtvcommand 192.168.x.x start_calibration expert2(192.168.x.x being whatever IP your TV has on your home network).Maybe anybody can write instructions to aiopylgtv ?) Its looks like very hard to understand. Text or videos how to connect to tv and upload 1d, 3d luts ?
I’d like to see something like this too. I know the library is still fairly new, but unless I’m missing something extremely obvious, there doesn’t seem to be any user documentation at all. There’s no list of parameters and what they do, for example, just a few scattered posts. Also, do we run DisplayCAL first, generate the 3dlut file, then use this tool to upload it, or do we need to use this tool to put the TV into a specific mode first, then run DisplayCAL and generate the 3dlut file? Is there a method to back up any factory LUT to restore later if needed/desired? A basic writeup of a start to finish upload of a 3dlut, and a brief explanation of what each parameter does, would be much appreciated.
What would be even better is if this could be folded into DisplayCAL and you could upload the generated 3dluts from the GUI after the profiling was done.
The documentation is intended to live in the Readme.md on github.
I was intending to write a full end to to end guide for using this with DisplayCal once I’m satisfied that I have a consistent procedure.
The documentation is intended to live in the Readme.md on github.
I read that, but it mostly looks like examples for using the library as part of another program, not using the aiopylgtvcommand program itself. I even skimmed through the source code, but couldn’t really seem to identify any section where I could isolate the different command line parameters and what they did.
I was intending to write a full end to to end guide for using this with DisplayCal once I’m satisfied that I have a consistent procedure.
It’s your code, so your call, but since you’ve been putting out a call for more testers on places like AVS Forum, how about at least documenting the different parameters and what they do? That should be enough to still weed out the general user class, but also open it up to more “power users” who at least have some idea of what they’re doing. Just a thought.
Ok I see your point, but for the moment I was intending to encourage users to call the library from a python script as in the documentation examples, rather than using the command line tool. (The use of named function parameters is less error prone, and aiopylgtvcommand is not really a proper client, but just a thin interface to call functions directly.)
I agree that the documentation can/should be improved in any case.
