Originally Posted by neo_2009
I'm considering buying a Spyder5 to perform a gamma calibration with JVC Autocal V10 (for a JVC X7900), and later an I1 Display PRO to be able to do a 3D LUT calibration on madVR.
When searching for autocal information, i saw this post
I think i also saw another post (i've tried to find it, but i couldn't), were Manni also states than when using a 3DLUT, there may be no need for an Autocal calibration.
, please forgive me if i'm incorrectly interpreting what you posted ... Would you mind clarifying the above? Thank you!
When you run a JVC Autocal with a spyder for gamma and an i1pro2 (or a known good or corrected Spyder) for color, you end up with something like a dE max of around 5 dE and an average dE of around 1.5 if all went well (measuring a large colorchecker set). That means that for many, it will be good enough and a 3D LUT isn't really necessary. If you want to get closer, you can run over this a small 3D LUT in 10 minutes (around 100 points, I use a Calman Lightning LUT) that will bring you close to reference because the baseline after the Autocal is very linear. That will bring you roughly to a dE max of 1.5 and a dE average under 1, which is near reference. Of course, you'll get a few bigger errors if you run a larger verification set, but for most, that's more than good enough, especially given that you have to do this every 200-500 hours. Total time: 20-30 minutes. Running an Autocal and no 3D LUT will most of the times give you a much better result if your target gamut is way over the native gamut (for example if you want BT2020 but the display is only able to reach P3).
The alternative is to not run a JVC Autocal first, and to run a large 3D LUT (because in that case the baseline for the LUT is not as linear). You then need to use a LUT that is at least 1,500points, preferably 5,000 points, even 10,000 if you're a maniac or your starting point isn't great. That takes hours to run (here, with my Discus trained to the ipro2, it takes anything between 2.5 and 5 hours to run a 5,000 points 3D LUT, depending on the software used). Visually, you get a similar result (or reference if you used a reference meter to profile the display), but it takes 3-5.5 hours (unless you have a K10A or similar). From a dE point of view, you will probably get less "spikes", but you could also get more artifacts overall due to the fact that you have more external corrections. Better numbers doesn't necessarily mean better picture overall, it only means theoretically more accurate, but you can still have banding, posterization, etc). For example if you run a large LUT with a BT2020 target but your display only cover P3, you'll most likely get a lot of posterization, so Autocal + no LUT in that case is much better.
That's why, for the JVCs, with a good Spyder (especially when using an i1pro2 for color), I prefer the first method (Autocal then small LUT if target gamut isn't far above native gamut) or no LUT (if target gamut is far above native gamut). That saves dozens of lamp hours vs no Auotcal and large LUT, and the result is more or less the same if you interest is in watching movies and not in watching patterns.
Hope this clarifies.