Hello,

I have read a few things about MPVM and watched the Light Illusion video tutorials. I am working with an LG WOLED tv which is not particularly linear and thought that MPVM might help me with my characterization. I am using an I1Pro3 and an i1D3.

In my reading about MPVM, there are two suggestions that I have come accross that I wanted to ask about:

- Using a sequential patch sequence to build the profile for the probe data. What size of profile would be a good starting point? The video demo uses a ^5 cube.

- Running the patch sequence through an inverted LUT so that the patches are mostly bright. Would I make a custom colourspace based on something wide like P3 or Rec2020 and change the EOTF to something like .5, and build a lut from that? Would I then use this as the "active LUT" in the settings? Does passing the patches through the "active LUT" modify the patch values stored in the profile along with the probe readings of the patches? Not sure if this is even the right question - there is lots I don't know here.

Inquiring minds would be grateful for any input on this. Thanks.

There are some very fundamental issues here.

First, LG TVs are not so much non-linear when used for SDR, but are unstable due to heat related drift, etc..
That means a Sequential patch set will likely cause issues related to head build-up.
And the concept of inverting a LUT is virtually mathematically impossible... You would never get a perfect invert.
(Yes, we have an Invert function in the LUT Tools, but it does state in the User Manual: Inverting 3D LUTs will rarely generate perfect results, due to the associated complexities of the original LUT.)

Steve

Thanks, Steve.

In doing MPVM for an LG WOLED, would a better approach be to do a preroll followed by an anisometric cube patch sequence to ensure that the readings are more stable? Or would it be better to work with the four colour volumetric matching? If I do use a cube, what would be a good approach for the cube to move the patches toward brighter values?

The options are really down to you.
But you can make a patch set with whatever values you want - there are no restrictions.

And you can use Preroll, Stabilisation, and Drift.

As to whether MPVM will be better than FCVM, testing ot the only way to verify.

Steve

I just completed an array of testing MPVM with my i1D3, i1Pro3 and LG G3. I worked with small 4^3 cubes. I tried sequential, anisometric and sort by ascending luma sequences. The anisometric sequence worked best in these tests. I found, though, that I was unable to get a consistently good match using MPVM. Using FCVM, I am able to get the probematch to +/- .001 x/y and <1.5% on the Luminance.

I think you may be confusing what a 'probe match' may be telling you.

Steve

Perhaps. I am testing different ways of doing probe matching: FCVM, and various MPVM strategies. Once I have created a probe match I am testing it by comparing probe matched colourimeter readings of the small "matching" patch set to the spectro reading of the same set using a spreadsheet created by a colourspace user. When testing the matched colourimeter, I am using a preset patch sequence in the measurement window, advancing through it manually and extracting the readings from the measurement log.

The spreadsheet calculates the difference between the probe matched colourimeter readings and the spectro readings. My understanding is that the difference should be within .001 x/y and 1.5% W.

Is this a useful methodology to verify the probe match?

There are no set value for probe match accuracy.
I know some quote values, stating they are NIST values, but there is no such thing.

Further, when checking a probe match, the different modes will act very differently volumetrically.
One is a simple matrix, the other true volumetric.
A small volumetric match may indeed show variations, especially if the display is non-linear volumetrically.

And with TVs that are unstable, as with LG WOLEDs, unless you are using Stabilisation patches, along with Drift patches, and a good Preroll, you will always get variations over time.
Both while doing the match profiling, as well as when doing any verification.
Screen stabilisation is a large issue with such TVs.
And as the screen do tend to be rather non-linear volumetrically, large MPVM profiles will likely be needed.

So what the probe match verification is telling you is there are likely stability issues withe the display, being picked-up during the whole process.

Steve

I understand. I had hoped for more with the LG G3, but its unstable nature really makes it quite labour intensive to calibrate. The learning has been fun but rather than spend more studio time fighting it, I think I will settle for seeing it as a "for-now-good-enough" secondary pixel-checking/client monitor. My next set of tests will be to try out the limited "+" alternate-colour-engine patch sets to see if these easily get me to "good enough".