Comprehensive Data on Lumagen/ChromaPure LUT calibration of JVC Projector - AVS Forum
Forum Jump: 
 5Likes
  • 10 Post By TomHuffman
  • 4 Post By TomHuffman
  • 1 Post By TomHuffman
 
Thread Tools
post #1 of 17 Old 05-26-2014, 11:50 PM - Thread Starter
AVS Special Member
 
TomHuffman's Avatar
 
Join Date: Jan 2003
Location: Springfield, MO
Posts: 6,419
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 52 Post(s)
Liked: 217
Goal
To characterize as accurately as possible the results of various calibration strategies for a JVC projector using the Lumagen video processor and ChromaPure software.

Methodology
Using Excel's random number generation function, I created a random sequence of 1000 test patterns in the video range. After each calibration session I evaluated the results using the set of 1000 test patterns and then compared each. I selected four different calibration strategies:
  • Standard grayscale, gamma, and 6-point CMS
  • Grayscale, gamma, and 6-point CMS using the 75% of Rec. 709 gamut
  • A full LUT calibration using the Lumagen's 125 point matrix using ChromaPure's Advanced Auto-calibration feature.
  • A full LUT calibration using the Lumagen's 729 point matrix using ChromaPure's Advanced Auto-calibration feature.

I calibrated to the Rec. 709 High Definition gamut using a 2.22 gamma. Finally I plotted the results on a chart using the CIE94 dE method.

Equipment Used
  • JVC RS-45 projector
  • Klein K-10 Colorimeter
  • ChromaPure Professional
  • Lumagen 2021 video processor

The Test
First, I zeroed out all calibration adjustments in the Lumagen and JVC, except brightness, contrast, and sharpness. I tested the JVC's performance in an uncalibrated state using the best preset (Standard Color Space, Cinema Mode). As shown below, the results were quite poor. These results look worse than they actually are. This is because the bulk of the recorded dE error is due to problems with the JVC's gamma response, which is less visibly troublesome than colorimetric errors. The RS-45, when set to the Standard gamma, tracked at about 1.6. Since the gamma target was 2.22, this resulted in relatively large dE errors.

Second, I ran a standard grayscale, gamma, and CMS calibration using the Lumagen controls only. Although the results looked excellent on what used to pass as a standard calibration report, 1000 random measurements told a different story. Although the dE errors were MUCH lower than what was recorded in the uncalibrated test (largely due to the drastic reduction of all the gamma errors), the results were still considerably worse than what would be considered acceptable for a competent calibration.

Third, I repeated the same calibration, except this time using the 75% of Rec. 709 gamut. This choice places the color targets well inside the high definition color space, instead of directly on the gamut boundary. In theory this should yield better results because the targets correspond to colors more representative of the entire color space. I was surprised how much this choice improved the calibration, reducing the measured errors by about half. Finally, we now had results that would be considered professionally acceptable, though there was still a considerable amount of error left on the table.

Fourth, I reset the Lumagen to its factory default and then ran a 2-pass 125 point LUT calibration using ChromaPure's Advanced Auto-calibrate feature. The results were, not surprisingly, excellent. In fact, the errors were now so small (ave. dE error below 1.0) that I question whether any additional improvement would offer any visible improvement. Another interesting result was the the average dE error for the measured points and the average dE error for the 1000 random test patterns were almost exactly the same, indicating that the Lumagen's interpolation performed very well.

Last, again I reset the Lumagen to its factory default and then ran a 2-pass 729 point LUT calibration using ChromaPure's Advanced Auto-calibrate feature. The results showed a small, but measurable improvement over the 125-point LUT. Since the 729 point calibration requires 6X the amount of time to run and achieves only a very small improvement, each individual will have to decide for themselves whether the minor increase in accuracy justifies the considerable increase in time required to run the process. The measured performance was now essentially perfect, at least insofar as gamma, grayscale, and color were concerned.

Summary Results



Conclusions

  • At least for displays with non-linear color performance, such as the JVC projectors, a standard gamma, grayscale, and CMS calibration should no longer be used. In absence of an LUT solution, users should calibrate to targets inside the color space, rather than at the gamut boundary.
  • An LUT calibration can yield results that approach measurable perfection. However, a very large number of control points does not seem to be necessary. Using only 125 points resulted in 87% of the measured colors below 1.5 dE, which I consider to be a performance threshold. Using 729 control points improved this further to 95% of all colors below 1.5 dE. With an average dE of 0.51 and a maximum dE of 2.84, 729 control points yield almost exactly the same results that Zoyd observed when testing eeColor LUT calibrations using 4913 points.
ScottJ, JimP, Chad B and 7 others like this.

Tom Huffman
ChromaPure Software/AccuPel Video Signal Generators
ISF/THX Calibrations
Springfield, MO


Last edited by TomHuffman; 07-31-2014 at 12:08 AM.
TomHuffman is offline  
Sponsored Links
Advertisement
 
post #2 of 17 Old 05-27-2014, 10:12 AM
AVS Special Member
 
Kris Deering's Avatar
 
Join Date: Oct 2002
Location: The Pacific Northwet
Posts: 7,107
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 218 Post(s)
Liked: 354
Outstanding data Tom!! Thanks for the awesome effort and data! How many hours were on the RS-45? I think the later models have been a bit better in terms of linearity over their lamp life so I wonder if the results would have been different.

Senior Video Editor
Secrets of Home Theater and High Fidelity


To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.
Kris Deering is offline  
post #3 of 17 Old 05-27-2014, 01:35 PM
AVS Special Member
 
zoyd's Avatar
 
Join Date: Sep 2006
Location: Planet Dog
Posts: 4,619
Mentioned: 29 Post(s)
Tagged: 0 Thread(s)
Quoted: 194 Post(s)
Liked: 341
Quote:
Originally Posted by TomHuffman View Post


[*] An LUT calibration can yield results that approach measurable perfection. However, a very large number of control points does not seem to be necessary. Using only 125 points resulted in 87% of the measured colors below 1.5 dE, which I consider to be a performance threshold. Using 729 control points improved this further to 95% of all colors below 1.5 dE. With an average dE of 0.51 and a maximum dE of 2.84, 729 control points yield almost exactly the same results that Zoyd observed when testing eeColor LUT calibrations using 4913 points.

The eeColor uses 64^3 = 262144 control points but I've performed statistical tests on the residual errors present using a variety of measured points ranging from 2500 to 9261 and using various algorithms to populate the control points. In all these tests the best achievable in your "excellent catagory [< 1dE]" was 96.3%. I've also tested the Lumagen 729 autocal w/CalMAN and obtained results similar to those posted above. All of these calibrations are qualitatively the same for color reproduction on my (moderately non-linear) display.

-----

To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.


To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.


To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.


To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.
zoyd is online now  
post #4 of 17 Old 05-27-2014, 08:17 PM - Thread Starter
AVS Special Member
 
TomHuffman's Avatar
 
Join Date: Jan 2003
Location: Springfield, MO
Posts: 6,419
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 52 Post(s)
Liked: 217
Quote:
Originally Posted by Kris Deering View Post

Outstanding data Tom!! Thanks for the awesome effort and data! How many hours were on the RS-45? I think the later models have been a bit better in terms of linearity over their lamp life so I wonder if the results would have been different.
About 500 hours.

Tom Huffman
ChromaPure Software/AccuPel Video Signal Generators
ISF/THX Calibrations
Springfield, MO

TomHuffman is offline  
post #5 of 17 Old 05-27-2014, 09:46 PM
AVS Special Member
 
Kris Deering's Avatar
 
Join Date: Oct 2002
Location: The Pacific Northwet
Posts: 7,107
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 218 Post(s)
Liked: 354
With that many hours and the older light engine it was very hard to keep the grayscale/gamma linear. Gamma was the hardest. The newer light engine and bulb are drastically better in this regard and calibrations stay stable for far longer than before.

Senior Video Editor
Secrets of Home Theater and High Fidelity


To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.
Kris Deering is offline  
post #6 of 17 Old 05-28-2014, 12:47 AM
Senior Member
 
<^..^>Smokey Joe's Avatar
 
Join Date: Aug 2006
Location: Christchurch, New Zealand
Posts: 384
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 15 Post(s)
Liked: 48
I did some similar tests on a 75 series at one point trying to overcome the gamut collapsing after JVC interning WB/gamma adjustments.
But found the lumagen ran out of range and created a horrid mess.

Ending up with replacing the projector with a 95 series.

Curious though that you did not do a JVC WB and gamma tweak(leave CMS alone), then tidy up the result with the lumagen. Did you try this?

Masterpiece Calibration Ltd
Christchurch NZ
<^..^>Smokey Joe is online now  
post #7 of 17 Old 05-28-2014, 01:20 AM - Thread Starter
AVS Special Member
 
TomHuffman's Avatar
 
Join Date: Jan 2003
Location: Springfield, MO
Posts: 6,419
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 52 Post(s)
Liked: 217
Quote:
Originally Posted by <^..^>Smokey Joe View Post

I did some similar tests on a 75 series at one point trying to overcome the gamut collapsing after JVC interning WB/gamma adjustments.
But found the lumagen ran out of range and created a horrid mess.

Ending up with replacing the projector with a 95 series.

Curious though that you did not do a JVC WB and gamma tweak(leave CMS alone), then tidy up the result with the Lumagen. Did you try this?
There was certainly no problem with the Lumagen running out of adjustment range. The gamut and grayscale adjustments were not that large. The gamma adjustments were the largest, but the Lumagen handled them fine with no issues that I saw.

This was intended primarily as a Lumagen test. The choice of a JVC projector was made simply out of convenience. I could have just as easily chosen another display for the test.

Tom Huffman
ChromaPure Software/AccuPel Video Signal Generators
ISF/THX Calibrations
Springfield, MO

TomHuffman is offline  
post #8 of 17 Old 05-28-2014, 01:31 AM - Thread Starter
AVS Special Member
 
TomHuffman's Avatar
 
Join Date: Jan 2003
Location: Springfield, MO
Posts: 6,419
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 52 Post(s)
Liked: 217
BTW, there are different ways to display the data on a graph. For example, here is another way to display exactly the same data comparing the recommended calibration techniques that shows the differences in more dramatic relief.


Tom Huffman
ChromaPure Software/AccuPel Video Signal Generators
ISF/THX Calibrations
Springfield, MO

TomHuffman is offline  
post #9 of 17 Old 06-26-2014, 04:01 PM
jrp
AVS Special Member
 
jrp's Avatar
 
Join Date: Jan 2003
Location: Portland, Oregon, USA
Posts: 1,040
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 18 Post(s)
Liked: 35
Quote:
Originally Posted by <^..^>Smokey Joe View Post
I did some similar tests on a 75 series at one point trying to overcome the gamut collapsing after JVC interning WB/gamma adjustments.
But found the lumagen ran out of range and created a horrid mess.

Ending up with replacing the projector with a 95 series.

Curious though that you did not do a JVC WB and gamma tweak(leave CMS alone), then tidy up the result with the lumagen. Did you try this?
Just checking that you used the "Gamma Factor" first to get the global Gamma close and then used the 21 point for fine corrections. The Gamma Factor is a true power curve and will give the best results for large Gamma corrections. The 21-point gray-scale/Gamma calibration is "piece-wise-linear" and intended only for moderate corrections.

It is of course certainly possible the Gamma was so far out that the precision lost in calibration for the error was too much.

Jim Peterson
Lumagen
jrp is offline  
post #10 of 17 Old 07-02-2014, 08:22 PM - Thread Starter
AVS Special Member
 
TomHuffman's Avatar
 
Join Date: Jan 2003
Location: Springfield, MO
Posts: 6,419
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 52 Post(s)
Liked: 217
Update

I just completed the same test on a 9th generation Pioneer Kuro plasma, the PDP-5020FD.



There are two aspects of these results that really interest me.

First, a standard 6-point CMS, grayscale, and gamma calibration actually made the image worse, despite the fact that it measured perfectly along the gamut boundary. It resulted in a slightly higher number of colors in the > 1.0 dE range, but the average dE was higher than if you had left the display alone in the Movie mode. Calibrating at 75% of the gamut helped, though not as dramatically as it did with the JVC projector. This plasma display--unlike the recent Panasonics--is an ideal candidate for LUT calibration.

Second, the results I got from a 729 point Lumagen LUT calibration were incredibly good, better even than the JVC. I am not sure how to explain this. Perhaps it is because of the much higher light output of the display; or, perhaps it is explained by the fact that the display was not that far off to begin with. In particular, there were no extremely large initial gamma errors. The pre-calibration gamma was about 2.12.
venkatesh_m likes this.

Tom Huffman
ChromaPure Software/AccuPel Video Signal Generators
ISF/THX Calibrations
Springfield, MO


Last edited by TomHuffman; 07-02-2014 at 10:34 PM.
TomHuffman is offline  
post #11 of 17 Old 07-02-2014, 08:54 PM
zel
Member
 
zel's Avatar
 
Join Date: Mar 2004
Location: California
Posts: 49
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 2 Post(s)
Liked: 18
Quote:
Originally Posted by TomHuffman View Post
I just completed the same test on a 9th generation Pioneer Kuro plasma, the PDP-5020FD.
I'm very interested in trying a LUT calibration on my Kuro. What kind of a pattern (window size, background) did you use for profiling and measuring the results?
zel is online now  
post #12 of 17 Old 07-02-2014, 09:06 PM - Thread Starter
AVS Special Member
 
TomHuffman's Avatar
 
Join Date: Jan 2003
Location: Springfield, MO
Posts: 6,419
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 52 Post(s)
Liked: 217
Quote:
Originally Posted by zel View Post
I'm very interested in trying a LUT calibration on my Kuro. What kind of a pattern (window size, background) did you use for profiling and measuring the results?
5% window, standard background.

Tom Huffman
ChromaPure Software/AccuPel Video Signal Generators
ISF/THX Calibrations
Springfield, MO

TomHuffman is offline  
post #13 of 17 Old 07-02-2014, 09:07 PM
zel
Member
 
zel's Avatar
 
Join Date: Mar 2004
Location: California
Posts: 49
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 2 Post(s)
Liked: 18
Quote:
Originally Posted by TomHuffman View Post
5% window, standard background.
Thank you!

Very impressive result.
zel is online now  
post #14 of 17 Old 07-02-2014, 09:31 PM
Senior Member
 
<^..^>Smokey Joe's Avatar
 
Join Date: Aug 2006
Location: Christchurch, New Zealand
Posts: 384
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 15 Post(s)
Liked: 48
Quote:
Originally Posted by jrp View Post
Just checking that you used the "Gamma Factor" first to get the global Gamma close and then used the 21 point for fine corrections. The Gamma Factor is a true power curve and will give the best results for large Gamma corrections. The 21-point gray-scale/Gamma calibration is "piece-wise-linear" and intended only for moderate corrections.

It is of course certainly possible the Gamma was so far out that the precision lost in calibration for the error was too much.
Hi JRP,
I made a slight error, I was referinging to the 70 series then switched to the 95 series.
The 70 series, it has been confirm to me by JVC Japan engineers a design feature effected Oceanic models where by the white balance controls effected the primary colours(CMS), the gamma controls are dodgy aswell.
So when I tried to compensate with the Lumagen, the software and lumagen threw the toys out of the pram.

Tom, what you see from precal them simple cal and larger error is something I have seen alot of, especially since having alot more internal colour checker routines. I've been using 75% patterns at 75% gamut for a while. In some cases I've found the CMS lies and does nothing internally.

Masterpiece Calibration Ltd
Christchurch NZ
<^..^>Smokey Joe is online now  
post #15 of 17 Old 07-04-2014, 02:33 PM - Thread Starter
AVS Special Member
 
TomHuffman's Avatar
 
Join Date: Jan 2003
Location: Springfield, MO
Posts: 6,419
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 52 Post(s)
Liked: 217
Quote:
Originally Posted by zel View Post
Thank you!

Very impressive result.
To add some additional information, Of the 1,000 colors I sampled, only 4 had a dE of 2.0 or higher.

224, 225, 220 dE 3.03
235, 235, 235 dE 2.86
48, 56, 43 dE 2.34
55, 64, 51 dE 2.37

This particular display has no built-in white balance controls, which explains the first two values. 100% is left uncalibrated. If I could have pre-calibrated 100%, the first two colors would have been less than dE 1.0.

The second two are at the very low end of the video range, about 15% and 19% respectively.

Tom Huffman
ChromaPure Software/AccuPel Video Signal Generators
ISF/THX Calibrations
Springfield, MO


Last edited by TomHuffman; 07-04-2014 at 02:36 PM.
TomHuffman is offline  
post #16 of 17 Old 07-31-2014, 12:14 AM - Thread Starter
AVS Special Member
 
TomHuffman's Avatar
 
Join Date: Jan 2003
Location: Springfield, MO
Posts: 6,419
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 52 Post(s)
Liked: 217
Panasonic Results

Having sampled two displays that definitely benefit from an LUT calibration, I thought it would be useful to collect data on a display that preliminary data (ColorChecker and saturation sweep) suggested does not. Here is that data.



This display doesn't even have a basic CMS. All I did was put it in the best presets and calibrate white balance. With an average dE of only 0.78 out of the 1,000 random colors sampled, any benefit from an LUT calibration would be negligible.

It is a shame that Panasonic discontinued their line of plasmas. The last couple of model years were very good displays.

Tom Huffman
ChromaPure Software/AccuPel Video Signal Generators
ISF/THX Calibrations
Springfield, MO

TomHuffman is offline  
post #17 of 17 Old 07-31-2014, 02:00 AM
Senior Member
 
<^..^>Smokey Joe's Avatar
 
Join Date: Aug 2006
Location: Christchurch, New Zealand
Posts: 384
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 15 Post(s)
Liked: 48
Tom, this makes a very good point for the general consumer that not all displays warrant or justify the cost of 3DLUT boxes and the calibration time it would take.

Masterpiece Calibration Ltd
Christchurch NZ
<^..^>Smokey Joe is online now  
Reply Display Calibration

User Tag List

Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off