Originally Posted by lovingdvd
Thanks. I'm a bit confused... I have a Radiance (XS) with a 5^3 cube and I've been using CalMAN 5's 3D LUT calibration for a couple years now. The last time I used it was about 6 months ago, before all these 3D LUT improvements and changes were made. So I can't decipher what has changed as it applies to 5^3. What is the "interactive autocube (legacy)"? Since it says legacy I assume that is what I have been using and am used to. And if that's right, then what has replaced it and how does it work differently? Under what case/conditions might someone want to go back to the interactive autocube legacy?
In some instances, like with my old JVC X30 which was getting tricky to fully correct with all its gamma droop, wonky gamut etc, the old interactive autocube (the one you know) could give slightly better results with a large LUT (1400 points) than with the default profile (2000 reads) of the new engine, which might be one of the reasons why it has been added back as an option. For various reasons I was running the MadVR LUT over the 5x5x5 Radiance LUT, itself run after a JVC greyscale autocal as I needed a correction for my non-MadVR sources, so that might have added to the problem.
Since I've upgraded to the X500, all my issues with the new engine (using the larger MadVR LUT, not the Radiance LUT or the JVC Autocal) are gone.
As far as I know, with a 5x5x5 Radiance the interactive mode is still the only viable option because you don't have enough points to be able to interpolate anything. That's what I was still using with the Mini 3D.
So for the 5x5x5 Radiance you'll get one choice for the LUT, which is the same old interactive method, and all the 125 points are corrected. You still have to run a 1D 21 points greyscale autocal before as there are not enough greyscale points in the 125 points 3D LUT to get good results otherwise.
To clarify, the interactive method (legacy for larger radiances and still current for 5x5x5 ones) corrects each of the points reading and correcting them. For example, the old standard interactive calibration used around 1419 points and needed around 3000 reads total.
The new method (for larger LUTs only) reads a subset of points (you can choose how many, from 65 to 6000 points and more), concentrating on the most problematic areas of the display, and interpolates a 3D LUT from that profile of the display. So a new 2000 points profile only takes 2000 reads (in about 110mn with my Discus), as no point is corrected. A 3D LUT is then generated according to the gamut/gamma targets, and you can quickly generate as many LUTs as you want from the same profile. It's a much better method as it can be much quicker without any significant compromise re accuracy compared to say a full 17x17x17 LUT which would require more than 4900 reads (and take about 4 hours with my Discus) for no significant benefit.
I'm sure Joel will comment on that if I'm wrong.