Originally Posted by bobof
In the interest of balance, I think DisplayCAL LUTs are "often rougher" than LS and not as performant where they are working equally, which is why I persist - for me this looks just like a math / algorithm issue when displays have a certain characteristic. It is heartening to see that other folk get the same characteristic "random areas on an otherwise OK looking LUT" - this to me says if addressed it could be really high performing.
I don't pay as much attention to the low end GS anymore as I tend not to correct the GS before profiling and loading a LUT. For me the GS got much better and more predictable since doing this. I don't even adjust the WP any more, in case the extra level of colour temp adjustment adds something the LUT has to work around (I dislike the pivoting nature of the gain / bias colour temp controls - the unadjusted colour temp balance graphs look nicer - flat parallel lines). For what it is worth I think there are 2 inherent dangers in correcting the GS before profiling:
1) you can be targeting using the dE or zoom widgets, but end up with all your greyscale points in slightly different rotation around the greyscale neutral axis, which being pragmatic about things is quite likely to trip up an algorithm. The closer you can get to the actual values just being sent to the panels the better. The one exception I make is I think there is value in adjusting the panel gamma of the projector using autocal first.
2) even the Discus off screen has clear issues measuring colour accurately between JVC black and close to 10%, so there is a question of how valid the adjustments actually are for the GS that can be come up with. When I've played with fixing up the GS after the fact in the Radiance like
is doing (you get 21pt GS you can adjust easily from the UI) what you realise is at the low values the adjustments that make reasonably large differences to the GS measurement are pretty tiny, so I think you have to temper your expectations a little about how good this can really be.
I do think a lot of the above that I wrote is just my personal preference and prejudice from having spent time fiddling with these units, it probably doesn't merit much in-depth discourse!
Remember that I don't have a Radiance, but use MadVR LUTs. This means that my LUTs can only correct content from MadVR.
So what usually do (and have done successfully for a long time) is get a baseline as close as possible with the JVC Autocal for my non-madVR source, and create a 3D LUT on top to get near reference results. Not perfect, I don't care about perfect, my projector drifts so perfect won't be perfect for more than a session. Just very good for 300-500 hours, before the next calibration.
I don't have the issue near black you report with the Discus. I use dark to bright sequence patterns though because the 2 hours dark cal window of the Discus is a bit short for a 5000point LUT (even with LS that handle the Discus faster than CM), and that works fine for me. I always wonder why 10% and below is worse once the LUT is applied, but it's not the biggest issue I have.
With this undersaturation issue, I've tried the other way around (that doesn't give me significantly better results with a non-undersaturated baseline) and used color profile off (no processing, no gamma calibration, no white point calibration). The RGB separation before the 3D LUT wasn't better, and the posterization issue on cyan was still present. Plus the gamut was smaller as the filter can't be enabled with profile off, as you know, so made no sense to use in this situation. In fact my SDR DCI-P3 calibration is only used by MadVR, so I could do whatever I want with it, but no processing makes no positive difference here.
I agree with you though that DisplayCAL LUT's are not overall necessarily better, just that they don't exhibit this posterization issue that Lioghtspace LUT have without using the LUT concatenation procedure here.
I have always been careful to say "here" or "in my experience". I'm not trying to bash Lighspace. I'm only expressing frustration at the way feedback is received (it's always incredibly painful, frustrating and alienating, no credit is ever given to my experience in calibration or my track record) and processed (nothing happens unless it's either minor, such as a missing gamut target, or a blatant error that would affect 100% of the users, such as the BT1886 implementation errors, and even that took me weeks to get nowhere, the issue was only corrected when the entire world told Ligh Illusion it was wrong). If it's only a problem for MadVR or JVC users, the answer is always the same: it's the user's fault or the display's fault, or the meter's fault, or whatever justifies doing nothing. Lightspace is working as intended. Why don't you apply five different convoluted manual steps and stop complaining?
I have never and will never make blanket statements about Lightspace. It is a very good LUT generator when it works. It even works for me in some situations (rec-709 when not undersaturated), and I agree that in that case the results tend to be excellent. I'm hoping it will work significantly better with my rs2000, as I have said, as my gamut won't be undersaturated in rec-709 or in P3 (touch wood). Still, I know that I won't have the choice to calibrate to P3 without the filter to save brightness, because my gamut will only cover 95% of P3, or BT2020.
I mainly regret the non-existant support/maintenance for MadVR and or JVC users. I don't think that Steve uses MadVR or or that anyone in house knows the software. For example he never simply generated a P3 LUT to confirm that the header was invalid once uploaded, instead he found suspicious that I was the only one to report the issue, as if that was suspicious...
I would also like to know if any of the issues reported (clearly not only by myself) are going to get improved in Lighspace or in Colourspace, at least by the time we are asked to make the decision to trade up or not.
If there is no improvement, I am unlikely to trade-up, because I don't see the point in purchasing a software that offers some support but no maintenance for JVC/MadVR users.
If Light Illusion doesn't want our business, tell the user that these are not supported. At least it will be clearer, and no one will complain (or buy) the software and feel like they become a nuisance rather than a customer.
This is especially frustrating when friends, who are using more mainstream displays or VPs, keep lauding how Steve's support is amazing, how reactive he is, and so on.
I don't doubt that it's the case for some, and they are lucky. It's just not my experience as a JVC/MadVR user.