or Connect
AVS › AVS Forum › Display Devices › Display Calibration › The Official ChromaPure thread
New Posts  All Forums:Forum Nav:

The Official ChromaPure thread - Page 153

post #4561 of 5345
Quote:
Originally Posted by TomHuffman View Post

With a captured 10 sec video of a full field test pattern, you would take a screen capture, and then import into Photoshop or Paint Shop Pro. Either will tell you the exact RGB levels of the captured image. Anyway, that's how I would do it.

Yes, but the screen capture would always be 8bit RGB because that's what PCs are working in - even if the capture is 10bit. And we don't know if the video playback application is using dithering or any other sort of processing. Especially if the source happens to be YCbCr, a screen capture won't do. It might be better to look at the raw uncompressed video stream with a hex editor. When using a full field test pattern, 99.9% of the raw stream should have the exact same byte sequence. So it should be easy enough to figure out the exact YCbCr or RGB data this way.
post #4562 of 5345
Quote:
Originally Posted by PE06MCG View Post

I recently changed my TV from a CCFL LCD Toshiba TV to a Sharp LED model.
Setup is Input > Duo > TV.

One of the big differences I find is if I change Colorspace output from Duo.

On the Toshiba there was a massive difference in brightness level changing from one to the other. Grayscale was completely changed and gamma was all over the place. In other words a complete recalibration was necessary with any colorspace change.

What a difference with the Sharp. The TV takes any change in Colorspace in its stride with ease.
Not checked all measurements but PQ up to now looks identical irrespective of choice.

I am wondering which of these 2 TV's is typical.
I never worked on a Toshiba display that I didn't think was a disaster. Their ColorMaster CMS introduced terrible artifacts and was basically unusable. The Sharp is working correctly. The display should handle RGB and YCbCr is such a way that they look identical to the viewer. No recalibration should be required.
post #4563 of 5345
Quote:
Originally Posted by TomHuffman View Post

I never worked on a Toshiba display that I didn't think was a disaster. Their ColorMaster CMS introduced terrible artifacts and was basically unusable. The Sharp is working correctly. The display should handle RGB and YCbCr is such a way that they look identical to the viewer. No recalibration should be required.

Thanks Tom.
I must admit the Sharp is a massive improvement despite getting a great PQ out of the Toshiba (as long as I did not use its CMS).

I did 'switch off' the Toshiba CMS allowing all changes to be made by my Duo but probably it still changed any RGB it received to YCbCr in anticipation of CMS conversion then changed it back again.
post #4564 of 5345
Quote:
Originally Posted by TomHuffman View Post

I don't know. There are still 0.002-0.005 discrepancies at the very low end of the grayscale that I am at a loss to explain.

You could be working near the repeatability limit of the sensor/display measurement. I did some repeatability tests on the D3 measuring large sets of color patches and found the precision to be 0.4 dE94 (1-sigma). This included both low and high luminance patches so it could be even worse considering just low stimulus patches. This was on a plasma so I don't know how much of that is sensor vs. display.
post #4565 of 5345
Quote:
Originally Posted by TomHuffman View Post

Video Generator Re-do

OK, here are the results from the exact same test using the Sony LED. I also switched to the Klein K-10 for an added bit of power at very low light levels.

The result profile is essentially the same--discrepancies, where they exist, are most prominent at the very low end and the Blu-ray player tracked the AccuPel the best. The difference is that the absolute error levels are lower. In fact, regarding color the Quantum Data tracked the AccuPel almost exactly. The Blu-ray player's consistency with the AccuPel in grayscale and gamma is almost scary.

Summary Results
Using the AccuPel as the reference, I found the following (all dE values are in CIE94 units):

Lumagen Mini
Ave. grayscale difference: 0.5
Ave. gamma difference: 0.03
Ave. color difference: 0.5

Quantum Data 804a
Ave. grayscale difference: 0.6
Ave. gamma difference: 0.02
Ave. color difference: 0.1

Samsung Blu-ray
Ave. grayscale difference: 0.2
Ave. gamma difference: 0.00
Ave. color difference: n/a

SignalGeneratorTest.pdf 16k .pdf file

This makes it appear as though the most accurate source is to use a Blu-Ray disc & player and not a signal generator (if I read the results properly). Since I don't have a signal generator, that's great news!
post #4566 of 5345
Quote:
Originally Posted by TomHuffman View Post

I never worked on a Toshiba display that I didn't think was a disaster. Their ColorMaster CMS introduced terrible artifacts and was basically unusable. The Sharp is working correctly. The display should handle RGB and YCbCr is such a way that they look identical to the viewer. No recalibration should be required.

I would throw in a caution based upon my experience with a Sharp Elite (70X5). I used to have an original release Darbee placed after a Lumagen Radiance. The original release Darbee's had a bug in that they only output RGB. I didn't bother to sent it in for a firmware update that fixed this. Bought another Darbee and placed both of them on input side of Radiance with Radiance sending 422 (12 bit) to Elite. Immediately noticed a slight reduction in the "pinkness" of skin color on very familiar faces (e.g. Today show hosts). Like hosts were using less rouge in makeup. Conclusion is Elite does not like RGB as input. Glad I calibrated with the original Darbee removed from the chain.
post #4567 of 5345
Quote:
Originally Posted by dsinger View Post

I would throw in a caution based upon my experience with a Sharp Elite (70X5). I used to have an original release Darbee placed after a Lumagen Radiance. The original release Darbee's had a bug in that they only output RGB. I didn't bother to sent it in for a firmware update that fixed this. Bought another Darbee and placed both of them on input side of Radiance with Radiance sending 422 (12 bit) to Elite. Immediately noticed a slight reduction in the "pinkness" of skin color on very familiar faces (e.g. Today show hosts). Like hosts were using less rouge in makeup. Conclusion is Elite does not like RGB as input. Glad I calibrated with the original Darbee removed from the chain.

Luckily my Sharp is an Aqeous Model which is effectively an Edge Lit LED.Model LC-60 LE651K which responds very favourably to calibration via my Duo whether I use RGB or YCbCr 444 as output to the TV, but thanks for the warning.
Skin tones are approx the same when Colorchecker is used as a measure.
post #4568 of 5345
Quote:
Originally Posted by TomHuffman View Post

Video Generator Re-do

OK, here are the results from the exact same test using the Sony LED. I also switched to the Klein K-10 for an added bit of power at very low light levels.

The result profile is essentially the same--discrepancies, where they exist, are most prominent at the very low end and the Blu-ray player tracked the AccuPel the best. The difference is that the absolute error levels are lower. In fact, regarding color the Quantum Data tracked the AccuPel almost exactly. The Blu-ray player's consistency with the AccuPel in grayscale and gamma is almost scary.

Summary Results
Using the AccuPel as the reference, I found the following (all dE values are in CIE94 units):

Lumagen Mini
Ave. grayscale difference: 0.5
Ave. gamma difference: 0.03
Ave. color difference: 0.5

Quantum Data 804a
Ave. grayscale difference: 0.6
Ave. gamma difference: 0.02
Ave. color difference: 0.1

Samsung Blu-ray
Ave. grayscale difference: 0.2
Ave. gamma difference: 0.00
Ave. color difference: n/a

SignalGeneratorTest.pdf 16k .pdf file

Hi Tom,
Just wondering if you have thought to supply this info to Lumagen for comment? If they can identify the source of the differences, they will probably release new firmware, if appropriate of course.
Regards, Mike.
post #4569 of 5345
Quote:
Originally Posted by zoyd View Post

You could be working near the repeatability limit of the sensor/display measurement. I did some repeatability tests on the D3 measuring large sets of color patches and found the precision to be 0.4 dE94 (1-sigma). This included both low and high luminance patches so it could be even worse considering just low stimulus patches. This was on a plasma so I don't know how much of that is sensor vs. display.
I thought of that, which is why for the latest test I used the K-10. The results were also averages of three readings, which is another step I took (that I didn't mention) to ensure repeatability.
post #4570 of 5345
The error between round down and round nearest is largest at 10% stimulus, mascior's disk uses round nearest.
post #4571 of 5345
Quote:
Originally Posted by AV_mike View Post

Just wondering if you have thought to supply this info to Lumagen for comment? If they can identify the source of the differences, they will probably release new firmware, if appropriate of course.

The problem is that there's no way to know for sure which of the TPGs is really the "best". Tom has selected the AccuPel as the reference and then tested how much the others differ from the AccuPel. But I think selecting the AccuPel as the reference was an arbitrary choice (or was it not?). Maybe the Lumagen is actually correct and the AccuPel slightly off? The only thing we could report to Lumagen is that the Lumagen measures differently to the AccuPel. But which of those two is correct (if one at all) we do not know. At least that is my understanding from what I've read so far...
post #4572 of 5345
it sounds like you have not researched this (??).. it sounds like you have not looked into Greg Roger's background, the Accupel or the Professional Industry in this regard (??)..

I would suggest you do so.

The Accupel 5000 has Reference Output!

One:
Quote:
Originally Posted by Smackrabbit View Post

I put the AccuPel 5000 on the QuantumData 882 this month and the values from the AccuPel 5000 are spot on accurate, so don't worry about it being incorrect.
Chris Heinonen
Senior Editor, Secrets of Home Theater and High Fidelity, www.hometheaterhifi.com
Displays Editor, AnandTech.com
ISF Level II Certified Calibrator, ReferenceHomeTheater.com


.
Edited by turbe - 7/26/13 at 11:24pm
post #4573 of 5345
How did he check/confirm that the AccuPel is "spot on accurate"? Sorry for playing the devil's advocate here. Just wondering if anybody ever really checked the exact digital output of all those test pattern generators with appropriate tools (like HDMI capture cards), to confirm that they're really doing what they're supposed to be doing down to the very last digital bit? If AccuPel was tested that way, I'll happily shut up and give all hail to Greg Roger. However, we're on a "scientific" forum here, and I'm not a fan of simply trusting in anyone's reputation if there are multiple devices on the market and they all measure differently somehow. This shows that creating TPG hardware is not as easy as it might seem, and although Greg Roger's creation might be the most probable candidate for being the correct/perfect one, a scientifically thinking person is not happy until he's got absolute prove.
Edited by madshi - 7/27/13 at 1:04am
post #4574 of 5345
Quote:
Originally Posted by madshi View Post

How did he check/confirm that the AccuPel is "spot on accurate"? Sorry for playing the devil's advocate here. Just wondering if anybody ever really checked the exact digital output of all those test pattern generators with appropriate tools (like HDMI capture cards), to confirm that they're really doing what they're supposed to be doing down to the very last digital bit? If AccuPel was tested that way, I'll happily shut up and give all hail to Greg Roger. However, we're on a "scientific" forum here, and I'm not a fan of simply trusting in anyone's reputation if there are multiple devices on the market and they all measure differently somehow. This shows that creating TPG hardware is not as easy as it might seem, and although Greg Roger's creation might be the most probable candidate for being the correct/perfect one, a scientifically thinking person is not happy until he's got absolute prove.

''Using a Quantum Data HDMI Analyzer, we can read these values directly and compare them to those that were encoded on the disc.'' Using the same way they can test pattern generators also...

Quantum Data's cost is around 5-6.000$.
post #4575 of 5345
Lumagen Radiance XE/XE+/XE3D/XE/XE+/XE3D/Mini-3D/XD/XD3D Series New 071713 Firmware Update

Release Notes

Fix for A bug only in previous release (071613) caused incorrect video colors after exiting test patterns.

Download Link
Edited by ConnecTEDDD - 7/27/13 at 4:25am
post #4576 of 5345
Quote:
Originally Posted by ConnecTEDDD View Post

''Using a Quantum Data HDMI Analyzer, we can read these values directly and compare them to those that were encoded on the disc.'' Using the same way they can test pattern generators also...

Quantum Data's cost is around 5-6.000$.

So the AccuPel was tested with a Quantum Data HDMI Analyzer, I suppose? That should be prove enough for me - thanks!

However, that makes me wonder: If Quantum builds a Data HDMI Analyzer, didn't they test their own "Quantum Data 804a" TPG with it? That doesn't make much sense to me. Since Tom measured a difference between the AccuPel and the Quantum Data 804a, only one of them can produce correct output, right?
post #4577 of 5345
Quote:
Originally Posted by madshi View Post

Quote:
Originally Posted by ConnecTEDDD View Post

''Using a Quantum Data HDMI Analyzer, we can read these values directly and compare them to those that were encoded on the disc.'' Using the same way they can test pattern generators also...

Quantum Data's cost is around 5-6.000$.

So the AccuPel was tested with a Quantum Data HDMI Analyzer, I suppose? That should be prove enough for me - thanks!

However, that makes me wonder: If Quantum builds a Data HDMI Analyzer, didn't they test their own "Quantum Data 804a" TPG with it? That doesn't make much sense to me. Since Tom measured a difference between the AccuPel and the Quantum Data 804a, only one of them can produce correct output, right?

Good Point!

Like Blu-Ray Players that each model (or sometimes a different firmware version) of any brand has different output, each brand Pattern Generators have different output also, and Tom has examined 3 different pattern generators that is showing that this different exists.

So this thing makes clearer the need to use only your current Movie Playback Blu-Ray Player a pattern source for HC, or an internal patttern generator (and not external) for HTPC calibrations.
post #4578 of 5345
Quote:
Originally Posted by ConnecTEDDD View Post

Like Blu-Ray Players that each model (or sometimes a different firmware version) of any brand has different output, each brand Pattern Generators have different output also, and Tom has examined 3 different pattern generators that is showing that this different exists.

Which, to be honest, is not something I understand. All of this is strictly digital processing, and the specs clearly say what to do, so every Blu-Ray player should have the same output *) and every test pattern generator should have the same output. They do not, and that makes me wonder whether all those companies really know what they're doing.

*) Chroma resolution upsampling is necessary because HDMI doesn't support 4:2:0 transport. And chroma upsampling algorithms can vary between different Blu-Ray players. But that doesn't really explain measured color differences in test patterns.
post #4579 of 5345
Hi Tom,
With regards to the Gamma module target line not following the selected target gamma value - the follow procedure appears to be repeatable:-

Started Chromapure (version - 2.4.2.19491)
Selected meter - i1D3 - initialised.
Selected Gamma module - selected target of 2.40 (just so that its different from standard of 2.22)
Reset form - target line moves to 2.40 (correct)
Took full 10% sweep of readings, okay
Selected White Balance module - took three different stimulus readings, okay.
Returned to Gamma module - reset form ready to take new full sweep, target line moves back to 2.22 (wrong). Target box was still reading 2.40.
To correct this - the target has to be changed, form reset, target changed again, form reset.

I think this may be what others have seen and posted.
Repeatable on my desktop & laptop (running Win XP & Win 7 respectively)

Regards, Mike.
post #4580 of 5345
Quote:
Originally Posted by madshi View Post

Quote:
Originally Posted by ConnecTEDDD View Post

Like Blu-Ray Players that each model (or sometimes a different firmware version) of any brand has different output, each brand Pattern Generators have different output also, and Tom has examined 3 different pattern generators that is showing that this different exists.

Which, to be honest, is not something I understand. All of this is strictly digital processing, and the specs clearly say what to do, so every Blu-Ray player should have the same output *) and every test pattern generator should have the same output. They do not, and that makes me wonder whether all those companies really know what they're doing.

*) Chroma resolution upsampling is necessary because HDMI doesn't support 4:2:0 transport. And chroma upsampling algorithms can vary between different Blu-Ray players. But that doesn't really explain measured color differences in test patterns.

Different Brand/Chipset/Model PC Video Cards have different output also. Nothing is perfect in this digital world... frown.gif
post #4581 of 5345
Quote:
Originally Posted by ConnecTEDDD View Post

Different Brand/Chipset/Model PC Video Cards have different output also. Nothing is perfect in this digital world... frown.gif

After proper configuration, they should have identical output. But maybe they don't in real life. I can't claim that I've properly tested that.
post #4582 of 5345
Tom did you use windows or fields for the test?
post #4583 of 5345
We also have to keep in mind the generator's control implementation in the application. Both of the current commercial calibration applications have had errors in regards to this in the past and application updates released. I wouldn't be surprised if there are still (or after future minor/major version updates) errors in control.. in both apps.

Greg has developed the 5000's own control application which can be helpful if needed.

.
Edited by turbe - 7/27/13 at 7:59am
post #4584 of 5345
Seem like GregR and Jim Peterson could specify what RGB levels their devices use for 21pt greyscale stimulus to help narrow down the discrepancies....
post #4585 of 5345
Well, better to verify each command for the RGB mix from ChromaPure for the Accupel and for the QD (assuming the 8bit / 10bit triplet commands are being used in that CP module).
post #4586 of 5345
Tom!
Why do I have a mistake in choosing ColorChecker ?
See the end of this message for details on invoking.doc 24k .doc file
post #4587 of 5345
Quote:
Originally Posted by ConnecTEDDD View Post

''Using a Quantum Data HDMI Analyzer, we can read these values directly and compare them to those that were encoded on the disc.'' Using the same way they can test pattern generators also...

Quantum Data's cost is around 5-6.000$.
I wish. The Quantum Data model in question is the 882. It is $25,000.
post #4588 of 5345
Everyone, I appreciate the spirited discussion and debate over this can of worms I opened. I spent quite a bit of time last night on this going back and forth between the AccuPel and the Quantum Data and I have come to the conclusion that these discrepancies--which again were small and only seemed to appear at the very low end of the gray scale--are not in any obvious way that I can discern caused by the signal generators themselves.

I could not even get consistent readings at 10% between different measuring sessions when using the SAME signal generator. This led right back to the issue that I thought I had solved by moving to the Sony. Mass market commercial displays are not manufactured to extremely high tolerances and expecting a precise level of precision at very low input levels is probably unrealistic. Short of a direct comparison of the actual digital output of the various devices--which I don't have the equipment for--I am going to leave this until a future time when I can draw some more definitive conclusions.
Edited by TomHuffman - 7/27/13 at 1:41pm
post #4589 of 5345
Quote:
Originally Posted by AV_mike View Post

Hi Tom,
With regards to the Gamma module target line not following the selected target gamma value - the follow procedure appears to be repeatable:-

Started Chromapure (version - 2.4.2.19491)
Selected meter - i1D3 - initialised.
Selected Gamma module - selected target of 2.40 (just so that its different from standard of 2.22)
Reset form - target line moves to 2.40 (correct)
Took full 10% sweep of readings, okay
Selected White Balance module - took three different stimulus readings, okay.
Returned to Gamma module - reset form ready to take new full sweep, target line moves back to 2.22 (wrong). Target box was still reading 2.40.
To correct this - the target has to be changed, form reset, target changed again, form reset.

I think this may be what others have seen and posted.
Repeatable on my desktop & laptop (running Win XP & Win 7 respectively)
OK, now that makes sense. The module resets to the default 2.22 when reentering it, even if the drop-down reports some other value. This is an easy fix.
post #4590 of 5345
Quote:
Originally Posted by Chad B View Post

Tom did you use windows or fields for the test?
Fields with no labels.
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Display Calibration
AVS › AVS Forum › Display Devices › Display Calibration › The Official ChromaPure thread