Originally Posted by ConnecTEDDD
***SNIPPED FOR QUOTING***
Mr. Aspiotis, I'm curious. How are you taking measurements? Your process, I mean. Like sweeps? Anything shown between measurements? What do you leave on the screen before or after measuring? Single readings or multiple? If multiple, averaging them? Do you re-measure previous patterns (none, randomly, switching back and forth, specific points only, etc)?
I know my lowly ColorMunki Display has nothing on your meters, but I've seen a few posts from people with high quality meters that seem to have much more stable measurements (or perhaps never checked stability-time-repeatability of their panel/measurements). So automated pattern measurement sequences, like grayscale/saturation sweeps, the previous pattern would always "taint" or "skew" the current one. Buzzard, on an older LG OLED, claims to be unable to reproduce what another user reported with a lower quality meter (as good or better than mine, and basically had the same behavior I'm seeing).
Since you don't have one of these panels yourself, did you by chance happen to check for something like this?
Perhaps this might be related to tecnezio's results?
Watching content between patches would also do the same thing (taint/skew). I assume that when profiling a display it would take stability into account but not so much this particular kind of stability or lack thereof. Basically content shown from the last few minutes are tainting/skewing my measurements.
Example: Put a 20% grey full screen field up, put the LG magic wand pointer on screen for ~5 seconds then move it. You can see the retention after image and it stays for about 1 minute.
This is not my meter as far as I can tell as it behaves fine on various LG's LCD displays my friends and family have. I don't get repeatable measurements that are rock solid stable, but it's definitely not the case when measuring my E6. This is something I highly doubt a correction matrix or spectral sample of my panel would help (my assumption, could be wrong).
Contact and non-contact modes have similar results, so OLED heat is not an issue for the meter itself (i.e. OLED heat in contact mode is not an issue.)
No random jumps/dips of 10 or more nits. This is very obvious on my E6, where I just measure twice from HCFR and see a large enough difference to throw off white balance and gamma to a small perceptional degree (ala warm 1 vs warm 2 kind of white)
Constant measurements over 1 minute match results of back to back measurements or slightly delayed measurements for the same length of time. This is not the case on my E6, where the constant measurements will be LOWER, which makes no sense unless this means the display is unstable and needs time to stabilize which will greatly impact profiling results (!) for creating a 3D LUT.
And this is why I'm curious, as I'm finalizing my manual measurements (my process is described below), and I'm wanting to feel comfortable doing automatic measurements for profiling and running the hour plus long patch sets. My manual process is already long and tedious (which I'm fine with) but it's not applicable for these processes.
I've tried using 0%, 5%, 10% 15%, 25%, 30% 50% 75% and 100% patterns (fields and different window sizes larger and smaller than 10%) as my content between measurements (basically same as process below but with different "content").
I would suspect if wide gamut displays were a problem with the meter's hardware (aka i1 Display Pro variants) then others would have more discussion on it in the various forums.
I only have experience with my display, so I'm trying to go through everything I can think of before simply blaming the technology (LG's WRGB OLEds), which I don't have a problem with either. I just want to be satisfied with my conclusion on my display and not go "but what if..." the next morning.
And by all means, this question goes to everyone, so feel free to share how you're taking measurements in detail.
Also, I'm really curious if anyone has any thoughts to share regarding this interesting behavior
using jScreenFix (in that post of mine), or anything that changes pixels extremely fast.
Using that as my content between measurements compared to using the Colors of Journey demo, the closer to black the pattern is the higher Y compares (5% demo's 0.19 nits vs .jar's .62 nits), and closer to white the lower Y compares (100% demo's 170 nits vs .jar's 130 nits).
IMO, a significant difference. RGB balance is just as bad (non-linear).
This does lower repeatability differences as the rapidly changing pixels prevent voltage buildup (read: image retention), but real content will not behave like this so I've been trying to recreate real world conditions by using the process outlined below.
I have my computer connected to the screen, so I can switch between measurements with no time spent flipping through menus or switching inputs (which causes a shift in white balance control alignment, initially).
I've been using MadVR* to convert LG's HDR demo (Colors of Journey
) to SDR. This clip covers nearly everything. Bright colors, quick motion, slow motion, panning of a single picture, brights, darks, slow and fast pixel transitions, some quick scenes some slow.
I take one measurement every two or more play throughs, as taking one per playthrough the pattern will still slightly skew measurements, two is much less but usually longer as I do other things and forget about measuring.
I continue for 30 measurements, each taken at a different point in the video, and used for all sets. I randomize the order (bright scene, dark scene, color-heavy scene, etc) then average them.
I'm going by RGB values, by the way.
Once averaged, I make adjustments for white balance controls.
Repeating until satisfied.
This is very
time consuming, but has been working wonderfully for me to help combat the measurement behavior I see on my panel.
But this is not something I can do for profiling via DisplayCAL or anything that needs to be done in an automated process.
I've also tried quick measurements (low integration time; 300ms-1000ms), delayed measurements (dynamic iris value; 200ms-3000ms), which works just fine but does nothing to effect erratic measurements.
*MadVR shows a 20% black window a few seconds before each measurement., which seems to not actually effect measurements at all.