Official JVC RS20 / HD750 Calibration and CMS thread (NEW FIRMWARE V1.1) - Page 8

Hi,

I have some more questions, maybe they are not all related to the DLA-RS20, but at least it is my intention to put the JVC at its best performance...

- Since I currently don't own (yet) a Blu-Ray player I want to perform a temporary calibration for SDTV, being the 601 standard. I know that the primary colors are almost equal (only x for green is 0.29 instead of 0.30), but what luminance values I have to use for the primaries and the secondaries? Keeping in mind that today my DVD player is performing the YUV to RGB conversion and sending it over through DVI to the JVC.

- I'm using the "standard" range on the DVI (RGB colorspace); being 16-235. However I noticed that the JVC is clipping on specular highlights on the demo footage on the DVE DVD, as well as on the graybars (the white only). Since I don't want to use DVI in "enhanced" mode (0-246), because of contouring artifacts, I put the JVC in enhanced mode and adjusted the brightness to -7 to get black (16) to real black (not seing blacker-then-black) and the contrast to +11 to get "highlight" white (a bit higher then 235 I assume) at white of the JVC. Why? Because it looked to me that the JVC throws away everything higher then 235, even if you lower the contrast. Did other people notice similar behaviour?

- Are the full screen primary and secondary color test patterns on the DVE DVD 100% saturated?

- What are the most recommended DVD and eventually Blu-Ray disks for performing calibrations? Where can they be ordered?

- Is the gamma calibrations performed using the standard menu system, or did some people use service (factory) menu's?

- Is there an inherent difference between profiles Cinema2, Natural and User1-3? If yes, what is the preferred profile for performing a calibration on?

- What gamma setting is recommended for wathing movies; gamma "B" or gamma "Normal" or another one? Or is it recommended to create a user one? I read somewhere that B is very close to gamma 2.2?

Hope to get some reactions on these...

Thanks in advance!

Ignace.

LovingDVD,
I loaded in what Manni calls his third attempt settings when I reported how they looked on my projector.

You mention that DVE can be used for this Filter method.
I used the the DVE color filters approach with DVE color bars to look at the color decoder behaviors (for example all color bars containing blue should exactly match the white background when viewed through a blue filter--etc.) . Am I correct in assuming the filter method you are describing is essentially also looking at color decoder errors as you adjust by eye for the correct Y values.

If so, In trying out Manni's settings, I made a mistake in trimming the Main Color and Tint controls away from 0,0 while looking through the filters in an attempt to compensate for the color decoder errors. ( I have no meter and understand the color decoder is not part of the CMS, but I am having some success using these make-do methods).
What I apparently should have done is to correct the Y values of Blue, Red and Green with the CMS Y values as you describe instead of trimming the Main Color and Tint controls. Is that anywhere near correct?

Thanks
KT

There are several (way too many) different CIE color space models that produce different dE results (very different models, very different equations, very different scales i.e. dE=1 is the JND [just noticeable difference] in all systems, but dE=2 is barely noticeable in some systems and a large error in others). I use the CIE LUV 1976 color space/dE calculations in my reviews and for the calculator. It is (I think) the favorite of us "old video engineers" for some technical reasons and because we have developed a good feel for what a 3 dE or 6 dE error means over the years. Roughly speaking, 1 dE difference can barely be detected under careful comparison conditions, a 3 dE difference is barely noticeable in real video, a 6-7 dE difference is quite noticeable in real video, and 10 dE differences are really big (unacceptable) errors to a discerning viewer. These would be my estimates for highly saturated primary and complementary colors. When measuring grayscale we drop the luminance contribution to the error (and pick that up as a gamma error) so that the given dE variances (without luminance) are much more obvious. i.e. a 1 dE deviation is the JND, but a 2 dE deviation is obvious on a grayscale test pattern, a 3 dE deviation becomes quite noticeable on a B&W film, and a 4 dE deviation is too much for me.

Anyway, I don't know what color space HCFR is using for dE calculations. Most of the programs are probably using newer color spaces (that were developed at the insistence of the paint and textile industries rather than video engineers), because newer is always better, right? That is as much as you will get me to say or argue about dE. There are lots of things worth spending my time on, and that isn't one of them. But there is lots of discussion about this topic in other threads.

Greg,

HCFR uses the Luv 1976 dE's as well. The HCFR dE's are the same as the dE's I get calculating them with the CIE Luv method.

The CIE94 and CIELAB dE's don't match HCFR in my experience.

Dan

Then I don't know why Manni01 got different results. My calculator has been up for months and compared against other CIELUV programs so I don't think there are any errors, other than possible negligible differences that can occur from using different precision arithmetic.

Well I'm not so sure since Greg was saying it is incorrect to use this method for Brightness adjustment within a CMS, and I think I understand why based on his explanation. However I believe it is valid for checking/correcting the color decoder. I'll leave it to someone else to answer your questions since I am not sure.

Thanks Greg. I think I understand. Essentially you are saying there is really no correct Y value for a custom gamut since there is no reference and one is basically doing whatever they want anyway, so at that point who is to say what Y should be. Is that the gist of it?

I was thinking that a given brightness (Y) of a color, however, would have a technically correct, derivable value. Not because of the shade/coloring of a value, but because it could be too bright or too dark relative to the other colors. No?

Ric. Aren`t the L values artificial in that they are specified by a formula in the standard for the primary values specified in the standard ts only right with respect to meeting the standard. If you reduce the value without making any changes to the sauturation or hue, the color will appear darker. But the only right is the artificial right set in the standard.

Greg, I'll try to look into this tomorrow. I highly doubt your calculations are wrong. We'll get it sorted out.

Dan

The discrepancies are very small.

The source is the fact that HCFR rounds its numbers to 3 decimals for display, but it uses the unseen extra digits in the dE calculation. When I used four digits I got exactly the same results. If you typed the 4-digit values into Greg's app, I'll bet you would get the same values as well.

It was some silly travel program from cable and I don't have it any more. I guess it was shot with video equipment instead of film. Nevertheless, I guess I could try to check out my Blurays for this.

Yep, that's a lot more concise than I said it.


I'm sure you would find values that you didn't like because they seem too bright or too dark to you, but there is no "technically right" value since you went on your own with a subjective x,y value. Remember that it's the total combination of the x,y point and its brightness that combine to produce a perceived color. So if you could imagine a particular (non-standard) color that you wanted you could get equally close to it (same dE) by choosing a variety of x,y and Y combinations. But there is no calculation that can know what xyY color you really want, and hence no way to calculate a Y value for your x,y point.

Yes, color filters can be used for adjusting the color decoder, IF you use the projector's native primaries since in that case the luminance (Y) of the primary, complementary, and reference white colors are all strictly related, i.e. Y(White) = Y(R) + Y(G) + Y(B), Y(Yellow) = Y(R) + Y(G), Y(Magenta) = Y(R) + Y(B), and Y(Cyan) = Y(G) + Y(B), when the magnitude of the RGB signals is R = G = B. The color filters visually verify that R = G = B if you input RGB color bar signals. So when you instead input YCbCr encoded color bar signals, the filters allow you to verify that the YCbCr signals are correctly decoded back to RGB signals where R = G = B.

Thanks for figuring that out Tom, and for your input.

Dan

Well, the way that you are looking at the error (up to 5%) makes it sound worse than what your eyes are able to notice.

Yes, there is a 5% error from one dE calculation to the other, but the overall difference in dE is only 0.2 at worst case from one calculation to the next. A dE of 1 is barely noticeable under the most scutinous viewing, so a dE of 0.2 will not be noticeable in the least. Hope that helps.

Dan

Here's an open spreadsheet that calculates CIELUV and makes the math explicit.

 

dE-LUV.zip 30.697265625k . file

Feedback on THX Trick:
Guys, I know this report is late --but last night I also found that the THX Color temperature trick does need to be redone (set to my Jason's calibrated Custom 1 setting).

However, with my bulb at 470 hours, I now find the Color Temperature for Jason's original low power mode has visibly drifted toward green, and at high power it has drifted toward red. I also have measured a 1 stop loss, (50% loss in brightness) due to bulb aging to this point, so I now need to run at high power.

From previous comments it looks like this is all consistent/normal with what others are finding. Does this sound correct?

KT

I am a proud owner and really love my Sony G90 CRT projector used in a 1080p mode. It looks great on a 100" Stewart screen even with 10k+ hrs on the tubes. However, my wife and I are getting tired of looking at a Volkswagon hanging in the rafters. I've read a lot on the forums and everyone raves about the RS20/HD750, but my question is will it be a step up, down, or lateral from the G90.

The LCD sets we have purchased, including high end Sony Bravia to Vizio all look "digital" at some point and handle motion poorly.

I have all the set-up toys to do the job, I just don't want to feel as I have taken a step down.

Comments, please.

John

John -

You might consider posting this in the general RS20 thread since this isn't a calibration issue.
You might also get more "bites" there as well.

Good luck.

Mike

Even though this is the wrong thread to have posted that in, the G90 is still imo the king of projectors when all things are considered and we are just talking performance. I love my HD750 but it isn't as good as my buddies G90.

Guys,
For what it is worth---I have loaded in Manni's third attempt and Jeffy's new settings and compared them as objectively as I can by eye, to the THX CMS setup (I have no metering capability). I want to report my findings and have a question on the logic and validity of my conclusion that some of these settings look like they may be transferable to my projector with excellent results.

What I find is that that low saturation/lighter looking blues on Jeffy's settings match THX almost exactly, but Manni's are very clearly darker/or more saturated in appearance. Using THX as a reference, this is the only clear subjective difference I can see viewing many real world reference Video and Film segments I use for such judgments .

I also have been using 2 other objective ways to judge the color accuracy which may -or may not be valid:
1. First it seems to me that if the Color settings in the CMS are calibrated properly that the old standby filter method of setting the Color and Tint controls using the VDE BD disk should ideally yield Tint and Color control settings of exactly zero. Deviations here would indicate that the color CMS performance is not spot on (I am making the assumption that the Color Decoder in the RS20 is accurate). (I should note here that I found the THX mode unexpectedly required a color setting of 5 to satisfy this criteria. All comparisons below were made with this THX setting).

2. Secondly, after the Color and tint controls have been properly set, if Color filters are used with the VDE basics BD test disk to view the color bar pattern for testing color decoder errors.. then the R,G and B color decoders should look perfect (example... all of the primary and secondary colors and white viewed through the red filter should have the same red intensity).

Now I realize that many other criteria for excellent performance of the CMS setting exist ... but it seems to me that these 2 criteria above represent necessary but not sufficient conditions for a properly functioning CMS.
That said I found all of the CMS work around settings developed on the original Calibration site did not meet the 2 criteria above (even though they produced very reasonable -good looking images and helped us get through the "broken CMS era.) Since the CMS was defective this could be expected.

But now with the new version of the CMS firmware - and with simply copying in settings arrived at by Calibrations performed on projectors other than mine -I find that the CMS performance now unexpectedly comes extremely close to satisfying the above two "necessary" criteria.

Specifically, I found that Jeffy's new CMS settings produce almost perfect Color decoding while simultaneously satisfying the Color /Tint controls setting of 0. In other words Jeffy's settings seem to satisfy both of my Necessary conditions (and as discussed above - the less saturated blues match the THX blues perfectly).

I found Manni's 3rd attemt produces near perfect Color decoder behavior- but only if I set the Color control at +10 (Tint is fine at 0). And as mentioned above less saturated blues do not match the THX blues) So, at least on my projector, this case comes close to satisfying the 2 necessary conditions.

So my question is this.. do these results indicate a high probability that Jeffy's settings installed on my projector, results in a reasonable CMS setup?
If checked with, lets say, Jeffy's Calibration meter and software, would he end up with a pretty good Calibration report..or is it simply not the case that the 2 criteria I have set out are sufficient to allow drawing such a conclusion?

Thoughts on this would be appreciated.
And many thanks to Manni and Jeffy for contribution of their new settings.

KT

See Greg R's comments below about using filters with pseudo primaries.

http://www.avsforum.com/avs-vb/showp...&postcount=206

In such cases, filters can yield grossly inaccurate recommendations. Not a good idea.