- This topic has 16 replies, 2 voices, and was last updated 8 years, 8 months ago by
.
-
Posts
-
Display: Late 2013 Macbook Pro Retina display
Colorimeter: NEC SpectraView MDSVSensor3 (custom i1 Display Pro, I believe?)I’m trying to create a 3D LUT for Resolve so that what I see in the UI matches what I see onscreen in Quicktime or other programs which use the ColorSync monitor profiles.
However, every single time I try to create a Resolve 3d LUT, pure video black shows as an elevated green cast in the Resolve viewer, and all other tones contain some element of excess green and foggy shadows continue into the mids.
Since the monitor seems to have profiled OK with the display calibration routine (although I’m not sure it’s perfect, it’s at least clean with decent blacks) I’m assuming the sensor is reasonably suited to this display despite being customized for the NEC PA272W I have.
Am I doing something wrong in my 3D LUT settings, or is this truly just the wrong sensor for this display, and is there a way I can correct the sensor to become suitable?
Calibrite Display Pro HL on Amazon
Disclosure: As an Amazon Associate I earn from qualifying purchases.Hi,
the black point shouldn’t change. Please create an archive of the related files (there’s a button that looks like a package next to the “Settings” dropdown in DCG that allows you to do that) and attach it, or send it to me. Thanks.
You used black point correction for the calibration, which lifted the black point. My recommendation would be to disable it, or not do 1D LUT calibration. The profile itself is also unusable due to five misreads (which explains the high self check error of around 18 dE peak), which in itself would be fixable by removing the misread patches, but you’ll have to re-profile anyway.
Ok, I’ll try that. Other attempts without black point compensation seemed to do the same thing. I’ll re-run and report back. Thanks for the quick diagnosis!
Note that black point compensation and black point correction are two entirely different settings that do entirely different things (black point correction alters the black point hue to match the whitepoint, black point compensation maps the measured black to zero by scaling in XYZ). You’ll find the black point correction setting on the “Calibration” tab when you enable “Show advanced options” in the “Options” menu, and the tone curve is not set to “As measured” (otherwise it won’t have any effect).
As a follow-up, due to the five misreads you may want to enable “Automatically check measurements” in the “Tools” menu.
- This reply was modified on 2015-07-26 16:51:54 by fhoech.
Florian, first off, thank you so much for taking your time to help us on this forum. I’ll definitely be contributing to the project.
So, I re-ran the 3D Lut profile, but I’m not familiar with how to remove the misread patches. Attached is my latest run. It also had several misread patches, with an even higher dE this time, and when I tried it in Resolve it had a lower but still very visible green cast at video black.
Florian, first off, thank you so much for taking your time to help us on this forum. I’ll definitely be contributing to the project.
Appreciated 🙂
So, I re-ran the 3D Lut profile, but I’m not familiar with how to remove the misread patches.
This time I can’t make sense of most of the misreads either unfortunately. I think the way forward is figuring out why Resolve has hiccups, it seems to fail to update the patterns in time on several occasions despite the rather generous additional delay. Is it running on the same machine as DCG? How does the machine load (both memory/CPU) look like during the reads? I would recommend upping the display update delay (on the “Display & instrument” tab) to around 1000ms and try again with a lower patch count (e.g. around 500).
Well, so I ran it again, with “Automatically check measurements” selected, 596 patches. File attached.
System was rebooted from scratch, nothing else running but Resolve and DispcalGUI. I put display update to 1000ms, observed the entire run, and noted that every single time the camera shutter sound played the patch changed and the numbers sent to Resolve updated promptly. The extra delay did seem unnecessary.
There were still several misreads indicated in the review window that popped up, all in the very-near-black region. Ignoring them left me with a LUT that had a very good delta-E and gave no black balance issues but, when I displayed a grey gradient, the lower 5% or so of the gradient was clipped straight to black, with a hard cutoff compared to no LUT. The overall gradient visually appeared integral, though, no colour banding or strange issues there.
So, we’re getting closer, but it’s still unusable. I’m half tempted to just go borrow an i1 Pro spectroradiometer for this particular display.
Side note: is it “normal” for resolve to show a contrasting colour shade in the portion of the frame not containing the patch, instead of just black? It looks so strange…
There were still several misreads indicated in the review window that popped up, all in the very-near-black region.
Those could also have been false positives. The check assumes an sRGB-like display.
when I displayed a grey gradient, the lower 5% or so of the gradient was clipped straight to black, with a hard cutoff compared to no LUT
I noticed that the different profiles all indicated different peak luminances (192cd/m2, 241cd/m2 and 117cd/m2). Is the display auto-dimming? If so, this should be disabled during profiling. Auto-dimming could also explain the clipping near black (unless the display itself is clipping, in which case you can circumvent the problem by adjusting it to not clip or by re-enabling 1D calibration).
I’m half tempted to just go borrow an i1 Pro spectroradiometer for this particular display.
A spectrometer won’t help, the instrument isn’t the issue.
Side note: is it “normal” for resolve to show a contrasting colour shade in the portion of the frame not containing the patch, instead of just black? It looks so strange…
Yes, this ensures an APL (average picture level) of roughly 22%. This helps with profiling displays like Plasmas, and doesn’t hurt otherwise.
The extra delay did seem unnecessary.
Without a look at the unaltered measurements, it’s hard to tell. Best to disable the automatic measurement check again.
Well, we’re getting somewhere! Turning off ambient light correction (whoops! I forgot that was even enabled!) gave me a raw delta-E of less than 1, which is superb. However, applying the LUT in Resolve still gives me a clipped black in a gradient. I’ll attach two screengrabs so you can see what I mean.
The bluer of the two is without the LUT, the warmer is with it, you can clearly see the hard line in the low luminance values where the gradient has a sharp transient.
Any ideas on why that’s happening? The whole kit is attached, as usual.
Can you attach the original (source) gradient image as well? From the above, it looks like a levels mismatch.
