LightSpace 3D LUT Home Cinema Calibration Software - Page 2

Hello, to perform professional calibration services (paid) you need the business version of the calibration software which is much more expensive than the home user versions...

I have verified the results of profiling using LightSpace/eeColor with LightSpace Quick Profiling (141 Colors = 21-Step Grayscale + 20-Step Luminance of RGBCMY).

I have verified also using 21-Step Grayscale + 5-Step Saturation (Total 46 Colors) with CalMAN 5.

CalMAN 5 CIE Charts/dE Reports are available here: http://www.avsforum.com/t/1454804/teds-lightspace-cms-calibration-disk

Interesting. I already own calman 5 along with an I1pro meter. I was considering a lumagen for auto cal and 3d lut purposes, but after reading this the lightspace software along with eeColor box is intriguing. I dont need the scaling options the lumagen provides so i would be purchasing strictly for auto-cal and 3d lut capability. I own a mits wd73835 dlp that has horrible cms tracking so the higher level of correction with lightspace could certainly be beneficial. I will be watching this thread closely.

Thanks. I wanted to ask about this again since this info wasn't included in your previous post but didn't want to give the impression of badgering you.


As long as you have a BD player that outputs bit perfect or the BD player is the only source you use I can see LS being perfect for you. I think your usage highlights a flaw in the pricing comparisons of the various products that are being posted. Unlike you most people use a VP for more than just the calibration capabilities. For those that fall in this group a VP will still be required when using LS so at a minimum a Mini 3D or Duo needs to be added to the price comparison.

Actaully, I think something like the Edge could provide most of the scaling/deinterlacing and switching duties.

Yes possibly, Still adds $400.00 or so to the total cost.

I have an integra avr model dtr 80.3 that handles all my scaling and switching duties. Im fully aware it does not possess the same scaling qualities as a lumagen but it is fine for me. I use a ps3 as my BD player or i could use ConnecTEDDD new disc with my vaio laptop for calibration.

Probably a bunch of them out there available used for nearly nothing. I have a DUO and an Edge sitting unused as I graduated to the Lumagen XS.. I have ChromaPure so I don't expect any imiminent support for eecolor in any case. Just trying to say that there are ways.

I thought that LS could already use the Radiance for pattern and LUT holders/processors.

Or are you saying that LS/Radiance will be able to use a 1000 point cube and generate the patterns?

ss

LightSpace will have support for Lumagen Radiance Series as a pattern generator to profile a display/projector up to 17-Point Cube (4913 Colors).

If you have eeColor, LightSpace is calculating the correction LUT of 65-Point Cube (274.625 InterpollaTED Colors) or you can export to any other supporTED hardware like DAVIO, LUTher, Pluto. Tcube etc...

I have a Radiance mini 3d so will LS write to the mini the 4913 or just the 125 like Calman does. I am guessing that LS will write to the mini 21 point grayscale/gamma.

Thanks.

ss

It's Lumagen's hardware limitation, 5-Point Cube + 21-Step Grayscale Points can be stored to Lumagen only due to it's design.

To experience the difference of 125 to 4913 CalibraTED Colors, buy the eeColor (for 600$), for it's price now, it's the best value for money hardware i have ever seen. A must have!

Lumagen integration for patch generation is now done

Steve

Latest Build of LightSpace has a feature for iD3 users that is not available at any other software this time.

You can now manually set the integration time to any value....1.56....2.32...etc with editing the integration time value box, rather than using the LightSpace Intergration Time Slider.

This, according to X-Rite, will help with plasma profiling.

So, if you know the PWM Rate of the source you can set the integration time exactly... to do this:

1. Set the integration time to be an integer multiple of the PWM frequency of the corresponding display.

2. So, as an example, in PWM-based Display (Pioneer KURO with 24p signal), the frequency of the panel is 71.93Hz (PureCinema In Advance Mode) so the integration time should be an integer multiple, i.e., n*1/71.93Hz so if n were 100 then we obtain the following:

integration time to be = 100*1/71.93 = 1.39 seconds.

The more accurate you set, the better will be the results, gaining more stable repeatability

If you need to practice with different integer multiplier, there is a Measure & Log feature of LightSpace that it exports to an Excel file all your readings, xyY + the actual meter measuring time in ms. of each reading (3800ms, 5430ms.. etc).
Edited by ConnecTEDDD - 2/11/13 at 10:44am

Light Illuision has add a new interesting website page about Calibration Accuracy.

Chasing Delta-E, Grey Scale and Primary Colours for Perfect Calibration
http://www.lightillusion.com/delta-e.html

That is interesting considering the debate Steve had with Joel and Derek.
Anyway I have a eecolor box ordered from Buzz, so first I will see how well it does with Calman 5.1E LUT 17^3 and then go from there.

ss
Edited by sillysally - 3/11/13 at 2:58am

When I read stuff like this I know it is desperation to provide misinformation. Fact a 17x9 on a Pluto and a K-10a takes less than 35 minutes with around 2000 reads, about 2.5 per point. Fact CalMAN Studio also supports a full 17^3 if you wish and that will take about the same time as other products on the market in a similar setup.

"The guess-work based approach provides for only 780 individual colour points to be measured in 4.5 hours, with an average of 5 reads per patch, making a total of 3800 individual reads for just 780 actual calibration points."

... and that one was measured with CalMAN ... Sorry, but I feel like I am in a kindergarden here

Derek, have you read what you are writing? You are telling that the CalMAN supports 17-Point Cube and it takes the same time as the other products on the market.....the same time you are telling that it takes about 2.5 reads per color....

2.5 reads per color point (as you telling that it takes only) is still the slowest performance versus any other software solution that supports large 3D Cube in the market.

It will require 2.5 times more total time versus the 1 read per color point of THX CineCube/TruLight/LightSpace approach.

BTW The ColorBox/CalMAN Studio the next month is celebrating the full one year after it's released but i haven't seen any successful calibration report at any forum from any user, i have found only topics with problems.

Noone has post calibration reports without problems and the most important is the actual image verify report that is visible to the eyes and not shown to the possible perfect charts. I haven't seen one user to write that is happy and has great image.

Calibration report can't tell you if the image has distortion or clipping or other problems...

If you have seen any ColorBox/CalMAN Studio report from anywhere in the web please provide me the link to check it out.

And again, I'm not selling software, i'm the end user.

At the moment I'm agnostic on the subject of LUT sampling granularity but the thrust of the argument in that article is that one should ignore dE metrics in deciding how well a particular 3dLUT configuration works and to just trust LI when they claim 17^3 is the best solution. It is a very self-serving point-of-view. It also makes the claim that dE metrics at the gamut edge are deceptive (which is true) and because of that dE in general is untrustworthy (which is false). There will be diminishing returns on the granularity one uses in a look up table that depends on the inherent non-linearity you are trying to fix. dE is a powerful tool if used properly to get an objective measure of when "enough is enough".

Software: CalMAN 5.1.0.1130
Meter: Klein K-10A
Pattern Generator: Lumagen Mini (Pattern Delay 0.5 sec)



This shows that for a generally medium 3D Cube of about 771 Color Points, guesswork method took 3158 reads with total time of 3:04:50 using the fastest colormeter available, the Klein K10-A.

The average Read per Color Point is 4.09, so it's 4 times slower that any other 3D Cube solution that is using 1 read per color.

The following picture it shows the result of 1 Read per Color Point (LightSpace) 17-Point Cube Profiling.
As we know LightSpace or THX CineCube or TruLight don't have any type of dE Report so i used CalMAN for that dE report.

These results shows that 1 Read per Color Point works.

If CalMAN in the future will use more advance color science tech, to minimize the average of 4 Reads per Color Point to 1, then CalMAN true 17-Point Cube Calibration will take same-equal speed versus the current professional solutions or THX,LightSpace,TruLight.


CalMAN Report of Pioneer KURO 17-Point Cube using eeColor with LightSpace (iD3 Meter)

Does Calman 5.1 for the LUT cube calibration use lerp (Linear interpolation) method ?
What method does LS use?

ss

Tedd-

thank you for showing us this. Can you show us the Calman report for the LUT that Calman created? I only see the report of the LUT that LS created. For users that could care less who wins the time race and are more concerned with the final result, it would be nice to do a stare and compare between the 2 LUT's generated. Thank you.

CalMAN 5 uses a custom blend of algorithms to build the LUT. It is effectively using the data it gets from all the readings it takes, even if it's not obvious.

Ted these numbers just don't add up. We all know the K10 is one of the fastest meters even at low light and on avg takes less than 0.5 per a 5 sample reading and with the mini set to 0.5 delay you should be seeing at least 3000 reads per hours. So that should have only taken you around an hour. We need to find out why your system is taking 3 seconds per reading when it should be around 1.

My point is that even with our read more than once method and advanced algorithms we can match or better the speed of the other systems. On most displays we see an avg of 2.5 reads per points so for 4913 that would be just over 12000 reads. But we can also do around 3000 reads per hour so about 4 hours for a full 17^3 which is what everyone is saying the others systems take, about 4 hours.

We can show results for a K-10, Radiance as a generator and a ColorBox. So the same setup Ted has. We don’t have a Kuro but do have a VT60 so very close to the Kuro. I will have Joel post some very detailed charts and numbers tonight. We have some very advanced LUT diagnostics tools we use to determine if any issues happen during LUT create and if so what type.

Also a point on how we test and validate our cube flattening method. We first calibrate a display using a full 17^3 and record all the data. Then we do the same but now using a 17x9 and compare the same 17^3 points to what the 17x9 produced. I will let Joel explain more but we do have some very advanced math in how we do this.

Ted I did a quick check yesterday and a 17x9 with around 3200 reads took about an hour and ten minutes with a similar setup K-10, Radiance, VT60. So we do need to look into why your system is running 1/3 of what it should be.

Joel and I ran out of time yesterday to setup this test but we will by the end of this week. BTW we were working on an updated method of AutoCal for what we are calling unaligned grayscale/gamma, our test subject was the Sharp Elite and by the end we have a AutoCal that works on the SE with dE's less than 0.4 more on that later in the CalMAN thread.