Originally Posted by Light Illusion
More likely display drift, rather than probe issues or user error.
WOLEDS are inherently unstable.
Hopefully I have now understood the situation correctly. Real question at the bottom, wall of text to just make sure the basics are correct.
The reason for all these questions is just to learn more and to make sure I can get my meter (i1D3) to get the most stable and repeatable results from my OLED TV (LG C8). This is also due to having learned that getting my meter profiled by a professional would cost 400€ (about 200% of the price of the meter) + travel charges from the neighboring city. If my meter (individual) isn't good for profiling accurately or my TV (individual) is too diffucult to tame - no point in paying 400€+. Once again I have to say: The calibration results that I already get with LighSpace look amazing. This is all just about taking a next step and getting even better results with a profiling against a reference meter - or not. And if I'm not on a wild goose chase, someone else might benefit from me hitting my head against a wall. If nothing more, a knowledge that it doesn't work or it has been tested already and the results weren't good.
The fast integration time is less accurate with dark patches, but that can be defeated with intelligent integration.
The OLED will always be instable at high nits patches, slower integration time just shows the problem better.
OLED is not only instable at high nits, having produced variable delays on patches below 5 nits using intelligent integration @5 (with integration time of 0.35), the drift increased. Measurements with integration time of 0.8 and intelligent integration @1 thus showed more stable results. And as I'm causing bigger drift the problem can be found anywhere as the next patch reading can be anywhere and will have bigger drift to handle. Long dark patch cooling the panel (up to 20s below 5 nits) is clearly not a good idea.
So managing drift is the key to accurate profiling on OLED.
This brings me back to the idea of "frankenprofile". If the display profiling would be run without intelligent integration, wouldn't that produce the most stable drift handling on OLEDs? Getting only the very low nits readings retaken separately with intelligent integration enabled (or just very long integration time?) and manually replacing those values to exported .bcs -> import -> convert colorspace. That would only require testing what integration time and stabilization time combination (with extra delay of 0.5) would produce the best drift handling results. (Two good candidates to start with: [email protected]
, [email protected]
and [email protected]
, [email protected]
I'm eager to test that myself, just don't know when I'll have the opportunity. I know that I can filter the results from latest profiling (L <= 1) to get low end readings, but is there a way to copy or export those to create a .csv? With increasing luminance filter (=lower integration time), the number of patches would increase and the possibility of errors when manually typing those would increase...
PS. Before posting, I got an afterthought:
Am I just overcomplicating things? Is there a minimum for intelligent integration? How big of an error can you see in low nits? Setting intelligent integration to ~0.1 would leave just a handful of patches for slow readings. Here's the variation between 4 consecutive readings per patch with [email protected]
R=G=B, Y(min) - Y, x, y difference between measurements
100, 12.7330 - Y 0.0082, x 0.0003, y 0.0 (the Y and x changed at the same time, 2 measurements of each - panel heating up?)
95, 11.1243 - Y 0.0085, x 0.0001, y 0.0005 (same as above + y change at the same time)
90, 9.6265 - Y 0.0, x 0.0, y 0.0
85, 8.5436 - Y 0.0, x 0.0, y 0.0
80, 7.2657 - Y 0.0, x 0.0, y 0.0
75, 6.1165 - Y 0.0, x 0.0, y 0.0
70, 5.2976 - Y 0.0, x 0.0, y 0.0
65, 4.3959 - Y 0.0, x 0.0, y 0.0
60, 3.6045 - Y 0.0002, x 0.0007, y 0.0016
55, 3.0408 - Y 0.0, x 0.0, y 0.0
50, 2.3705 - Y 0.0, x 0.0, y 0.0
45, 1.8069 - Y 0.0001, x 0.0, y 0.0001
40, 1.3369 - Y 0.0, x 0.0, y 0.0002
35, 1.0206 - Y 0.0002, x 0.0, y 0.0
30, 0.6664 - Y 0.0007, x 0.0005, y 0.0003
25, 0.4043 - Y 0.0006, x 0.0005, y 0.0005
20, 0.2432 - Y 0.0002, x 0.0003, y 0.0003
15, 0.1063 - Y 0.0003, x 0.0004, y 0.0009
10, 0.0226 - Y 0.0001, x 0.0001, y 0.0009
The 10,10,10 actually took 2781ms to read, so have to set intelligent integration above it.