A new page has been added to the website discussing accuracy from different calibration approaches, and discusses the use of Delta-E as proof of calibration.

See: http://www.lightillusion.com/delta-e.html

Hope it's of interest.

Steve

Steve

Have read this page with interest despite not yet having bought the software.
I have a faulty LCD panel that will not reach correct Rec 709 blue at maximum saturation value.

It falls short in both saturation and hue erring on the Cyan side when I use Calman or Chromapure software. Otherwise it is surprisingly linear (as measured by Colorchecker that is).
I notice that starting with over saturated colours is preffered which in my case is impossible for blue.

Obviously your software would not correct my panel but would it interfere with the rest of its 'Calibration' or just 'bend' those points around its 'no go' areas?

LightSpace will always calibrate any display to the best of its ability, and not let those areas that can't be calibrated interfere with those that can.

The very limited colour set in Colorchecker is not at all an indication of linearity...

Many thanks for reply Steve.

I agree it is very limited but is it not still the industry standard for both Video Cameras and Reference Monitors?
Probably the Tec papers should be updated.

Would any of your 17x17x17 points coincide with the colour patterns within Colorchecker or will it still be representative of interpolation of the calibrated points?

Nope, it's not an industry 'standard'. It's just a useful guide for the image after shooting.

A 17^3 cube will cover all the Macbeth colour points very accurately.

Sorry Steve, I obviously used the wrong word.

Presumably my use of it is still a 'Guide' to the linearity of my LCD as well then?
Not talking accuracy of course because I can see the value of using the eeColorBox and your software.

However I take your point that perhaps it is possible that my LCD accidentally gives me results that are not reproduced throughout the full 3D Gamut but that's the same when used as a 'guide' for Camaras and Reference Monitors is it not?

It's a very informative page!


Think, you've got many inquires, why LightSpace don't offer any deltaE indicator. Guess, they coming from people, working with other solutions.

It's clear: LightSpace has a different concept than the "guesswork" solutions. No doubts - the profiling system has a big benefit regarding speed calibration!

But such a system also has it's restrictions. LightSpace is addressed to users having a LUT-supported equipment. I think, many users who got attention of LightSpace first do not really know this. Think, they do calibrate their devices using the built-in features of their displays. In most cases this will be kind of a gamma/grayscale equalizer, + in modern devices a 2D RGBCMYO CMS with hue/sat/luminance control. Therefore e.g. Calman is well-known to have many deltaE indicators for targeting D65/Rec709. And so I can imagen, this kind of user group is "spoilt" with such functions ...

At the moment I use Calman for pre-cal my JVC projector, to have a better "start"-condition than with stock settings (X30/70/90's are far from Gamma 2.2 and good 6500K RGB balance). Therefor Calman's DDC support is not the bad (Calman's autocal "Guesswork" could be better in lower stimulus levels, but the development seems not to be really interested in improvement suggestions like a smaller adjustment stepping size to get the results closer to the targets).

LightSpace achieve very perfect calibrations. But regardless, what calibration SW we are working with - a 17x17x17 cube calibration in every situation needs much too many hours - and that's always a problem for devices with (high voltage) "bulbs". They slightly fluctuate between minutes in brightness and spectrum/CCT, so we will never get a perfect calibration, even by use of a very expensive probe ... no way!

We just can hope to have low fluctuations during the entire profiling time ...

However: I like to work with LightSpace. Last visual results were very impressive on my X70 (even with just doing a 10-side cube profiling). After some experiments with the enhanced colour space capability of the X70, my best LUT converting result was 98%!

Ok, if I post-check e.g. the grayscale calibration using Calman there are differences in the RGB balance histogram, but I think that's because of calculating a LUT will always have some differences to "Guesswork".

No, it's not the same - it's just a visual guide. Nothing really to do with calibration.
You need a lot more info for that (hence the need for 17^3 profiling).

Steve

Harry, the Manual Mode within LightSpace is ideal for pre-calibration set-up.
Just get the cross in the target circle throughout the grey scale
Then use a Quick Profile to check the underlying gamut.

No need, and no benefit in using any other calibration system.

The do the full profile for final accurate calibration.

And no, the guess-work approach is not better in the shadows - that is simple a function of the probe accuracy, and display limitations (back-light bleed).

As for speed, we have no problems with any display when using a K10-A (possibly matched to a spectro if needed).
Any display fluctuations are just not an issue as the calibration process (the LUT generation) uses all the profile data to calibrate each individual point, so also managed any fluctuations.

But, as we say, Delat-E values really do not tell all the story - the final images do that

(We may add Delta-E reports in a future release, but would like them to be better than the standard offerings, so are researching ideas at the moment...)

Steve

hi steve one question about WB .

would be possible have the kelvin temperature in addition to the XY readings ?
also what is considered "ok" ? meaning there is a circle for WB and is everything inside the circle considered ok ?

for example if the center of the circle is 6500K for 709 , in am wondering what is the values range all around the edge of the circle ..
thanks

g

I currently am referring to the Colorchecker used in Calman that actually measures the dE values of the individual patterns. (I know you see dE as a no go measurement but it is a bit more than just a visual guide).

Calman also use a Color Comparator that is a visual comparison.

Sorry, not yet got LS but am obviously interested.

Yorkie, it is the limited number of colour used that I am referring too.
You need many more than is used within Calman to have any idea of a display's linearity.

Steve

Gabriele, Kelvin values do not define a colour temp in the same way as xy values, and as a result are not accurate for calibration.
The Kelvin value has many points on the same intersecting locus, so many xy values can be the same Kelvin value.

See: http://en.wikipedia.org/wiki/Color_temperature



And the circle within LightSpace CIE diagram defines the level of acceptance - the more central within the circle the better the accuracy.

Steve

Hi Steve,

sure, I could do that by using the remote control to align all crosses into the middle of the circle.
But CM do this automatically for me via RS232 DirectDataControl Calman <-> JVC, up to 21 pts in grayscale. By hand (using the remote control) I could adjust much less than 21 points in the JVC's OSD user gamma equalizer.



Oh, I think you got me wrong. I meant, Calman's autocal functionality via DDC is not as accurate as it could be regarding how "fine" one grayscale point (IRE level) could be adjusted JVC internally. CM use a 0,3 stepping between 2 values for approaching the target, but the JVC supports 0,1 as lowest step size. Therewith a grayscale autocal result would be much more accurate. I do this always by hand after the Calman grayscale autocal, because especially the lower levels (5 ... 15 IRE) are not really "dark gray", they sometimes are slightly green or red or blue. And it's not an issue of my probe - the accuracy of the i1D3 (profiled with the i1D2) is very good at all levels!


If this could be true some day, I think, many of your customers would appreciate it! That's a point I miss in LS, because if people talk together about their display calibrations, they commonly talking about deltaE results (or just only to compare calibrations from time to time using data with "numbers"). So it's always a bit hard to say "my LightSpace 3D Cube view looks great, but sorry, can't say anything about grayscale deltaE ... " ... or so. But as said - not really important; the visual result makes it all.


Hmmm ... but what, if the luminance will fluctuate during measuring all the patches - or is saturation + hue of a patch all, LightSpace needs for LUT calculation? But remember, these bad fluctuations are sporadically (caused by slightly varying heat inside the lamp, I guess) and not always on the same "patch". So a calculated LUT will match to the situation as the measurement was done, but at runtime the LUT may not match anymore because of new fluctuations. I think, that's a problem, no calibration system can solve. So I hope, in a few years we will have affordable home cinema projectors with bright LEDs / much more stable light sources than in actual PJ's with UHP bulbs.

That's also my experience ... I have such observations using Calmans RGB absolute histogram + CCT indicator. The same CCT can be reached with different R/G/B luminances!