AVS Forum banner
1 - 14 of 14 Posts

·
Registered
Joined
·
7,998 Posts
Discussion Starter · #1 ·
I just read a review in HTM of the Epson 9500 LCD projector.

http://www.hometheatermag.com/frontp...or/index3.html


At the end of the measurements page, the reviewer, Tom Norton, says the following.

The CIE chart above shows the Epson's color gamut in the THX mode out of the box with the white triangle. It's nearly an exact overlay of the (Rec.709) HDTV color standard (the black triangle). The brightness of the three primary colors (red, green, and blue) was slightly off; brightness or intensity is the third dimension of color and is not visible in a CIE chart. This could not be corrected (using the RGBCMY controls) without altering the color decoding, which was excellent out of the box. This was easy to spot with the projector's color isolation controls. I've found that minor differences in the brightness of the individual colors are one of the least significant details to get spot on (within reason) for visibly good color reproduction. But this result does suggest that the Epson's RGBCMY color management system functions by altering the parameters of the color decodera very common color management implementation, but not a desirable one.
I have read this several times and for the life of me I don't quite understand what he is saying, so I thought that I would throw it open to the rest of you to see if this can be clarified.


He seems to say that the CMS should not be used to correct errors that he measured in the brightness of the primary colors BECAUSE doing so screws up the color decoding as evidenced by the color isolation patterns with which he presumably used a SMPTE color bar.


What bewilders me about this is the following: the only purpose of a color decoder is to get accurate color. It ensures that the YCbCr-to-RGB conversion is without errors. That's it.


However, there are two ways to determine if there are errors.

1) Measure the xyY values of the color directly using standard test patterns and compare the measurement with the specification for the target gamut.

2) Look at SMPTE color bars using isolation color controls (or worse filters). This will give you visual cues about the brightness of the primary colors and the hues of the secondary colors.


The reviewer seems to be saying that he sees a discrepancy between these two--when he adjusts primary brightness to meet the spec using the CMS, then the SMPTE color bars look wrong.


Here's what I don't get: How is this possible? Assuming that his instrumentation is reasonably accurate, shouldn't the measurements of the gamut match the visual evidence from the SMPTE color bars? Secondly, if for whatever reason they don't match, wouldn't the direct measurement of the color be a more reliable guide than SMPTE color bars?


He seems to imply that there is value to how the SMPTE color bars appear beyond how the color actually measures. I just don't see what value that would be.


Finally, compare this statement from Tom's Norton's review of the Epson

"I've found that minor differences in the brightness of the individual colors are one of the least significant details to get spot on (within reason) for visibly good color reproduction."
with the following statement made in Greg Rogers recent WSR review of the JVC RS35 projector.

"The errors from having the wrong RGBYCM luminance values can be more visually significant than having the wrong CIE x,y values."
Maybe I am being picky, but it seems to me that these two statements cannot both be true. I am inclined to go with Greg on this, but Norton's review has me scratching my head.
 

·
Registered
Joined
·
8,864 Posts
Yeah ... doesn't make sense to me either. Unless he was using an improperly encoded SMPTE chart for the decoder mode ... 601 vs 709 for example?


As someone with a 709 decoder challenged display, I gotta agree Mr. Rogers on this one.


I think the addition of CMS has created no small amount of confusion about what is in the color decoding domain (big Y) and what is in the gamut domain (x,y).
 

·
Registered
Joined
·
7,998 Posts
Discussion Starter · #3 ·

Quote:
Originally Posted by HDTVChallenged /forum/post/18285824


Yeah ... doesn't make sense to me either. Unless he was using an improperly encoded SMPTE chart for the decoder mode ... 601 vs 709 for example?

That's the only thing that makes sense to me, but you would think that this would have been obvious to the reviewer as well.
 

·
Banned
Joined
·
3,550 Posts
Are they talking about digital representation? Since these displays offer higher input than 8-bit and PC level, 236 through 254 involves more than checking SMPTC color bars because lower significant bits are combined BEFORE color decoding even starts. For luminance of Y, this involves checking levels that are outside of SMPTC color bars since these are encoded using video. If you don't account for this, you get overly saturated primaries as you may in-inadvertently attempt to increase chroma/saturation using only values for x and y between 16-235? Maybe I misunderstood what they are saying.

http://en.wikipedia.org/wiki/Rec._709

Quote:
Digital representation


Rec. 709 coding uses studio-swing levels where reference black is defined as 8-bit interface code 16 and reference white is defined as 8-bit interface code 235. Interface codes 0 and 255 are used for synchronization, and are prohibited from video data. Eight-bit codes between 1 and 15 provide footroom, and can be used to accommodate transient signal content such as filter undershoots. Eight-bit interface codes 236 through 254 provide headroom, and can be used to accommodate transient signal content such as filter overshoots and specular highlights. Bit-depths deeper than 8 bits are obtained by appending least-significant bits. Ten-bit systems are commonplace in studios. (Desktop computer graphic systems ordinarily use full-swing encoding that places reference black at code 0 and reference white at code 255, and provide no footroom or headroom.) The 16..235 limits (for luma; 16..240 for chroma) originated with ITU Rec. 601.[1]
 

·
Registered
Joined
·
661 Posts

Quote:
Originally Posted by TomHuffman /forum/post/18284954


"I’ve found that minor differences in the brightness of the individual colors are one of the least significant details to get spot on (within reason) for visibly good color reproduction."

I also read this review last month and wasn't sure how he backed up this statement. As for his comment that using the Epson's CMS messes up the color decoding, I interpreted it to mean that when he attempted to adjust the luminance of the primaries that it messed up other aspects of the color accuracy - perhaps the secondaries were then thrown off? As for how he determined they were off, I would agree that if he had instrumentation to measure, I'd assume he'd use that to see that the color accuracy was thrown off by using the Epson CMS. Why he'd look at color bar patterns is not clear. I like HTM's recent reviews since they now at least show meaningful graphs for color and grayscale performance. He even mentions dE with regard to grayscale accuracy. I don't know why he doesn't start talking about dE with regard to color accuracy. An article in the magazine introducing people to color error measures and color accuracy in general would probably be good for their readership - especially given how more and more displays are coming with CMS controls.



--tom
 

·
Registered
Joined
·
7,998 Posts
Discussion Starter · #6 ·

Quote:
Originally Posted by thomasl /forum/post/18288897


As for his comment that using the Epson's CMS messes up the color decoding, I interpreted it to mean that when he attempted to adjust the luminance of the primaries that it messed up other aspects of the color accuracy - perhaps the secondaries were then thrown off? As for how he determined they were off, I would agree that if he had instrumentation to measure, I'd assume he'd use that to see that the color accuracy was thrown off by using the Epson CMS. Why he'd look at color bar patterns is not clear.

I don't think so. He specifically refers to testing the color decoding using the color isolation controls. Some displays offer a way to show red or blue or green only. Then you use these modes with SMPTE color bars to check color decoding. The color isolation controls are a more accurate stand-in for gel filters.
 

·
Registered
Joined
·
661 Posts

Quote:
Originally Posted by TomHuffman /forum/post/18289270


I don't think so. He specifically refers to testing the color decoding using the color isolation controls. Some displays offer a way to show red or blue or green only. Then you use these modes with SMPTE color bars to check color decoding. The color isolation controls are a more accurate stand-in for gel filters.

Right - one of my Samsung's has a blue only mode but I've never actually done any comparisons between using it to visually set color/tint and using the meter. In thinking about it a bit more, he says that the color decoding was quite accurate to begin with. If this is true, then the primary luminance values would have been spot on given their locations and an assumed D65 white point. So, when he says the primary luminance values are slightly off, he obviously means that they don't match the HD709 targets. Of course, they shouldn't if the color decoding is correct since his primaries aren't an exact match for HD709. At that point, I'd have to ask him what was he adjusting to when he tried to adjust the luminance of the primaries. At that point, you don't want to simply target the HD709 luminance targets. You want to target luminance values that minimize your color error when referenced against the HD709 targets. Assuming he did adjust the CMS to match the HD709 luminance targets exactly, would this not then show up as being "off" with the color isolation controls and SMPTE color bars pattern? The only way the color bars patterns will show up as perfect is if the color decoding is correct for his given primaries and white point. The minute he alters the luminance of the primaries to something other than their optimal values given his primaries' locations, the color decoding will appear off visually I would suspect unless I'm missing something.



--tom
 

·
Registered
Joined
·
7,998 Posts
Discussion Starter · #8 ·

Quote:
Originally Posted by thomasl /forum/post/18289745


Right - one of my Samsung's has a blue only mode but I've never actually done any comparisons between using it to visually set color/tint and using the meter. In thinking about it a bit more, he says that the color decoding was quite accurate to begin with. If this is true, then the primary luminance values would have been spot on given their locations and an assumed D65 white point. So, when he says the primary luminance values are slightly off, he obviously means that they don't match the HD709 targets. Of course, they shouldn't if the color decoding is correct since his primaries aren't an exact match for HD709. At that point, I'd have to ask him what was he adjusting to when he tried to adjust the luminance of the primaries. At that point, you don't want to simply target the HD709 luminance targets. You want to target luminance values that minimize your color error when referenced against the HD709 targets. Assuming he did adjust the CMS to match the HD709 luminance targets exactly, would this not then show up as being "off" with the color isolation controls and SMPTE color bars pattern? The only way the color bars patterns will show up as perfect is if the color decoding is correct for his given primaries and white point. The minute he alters the luminance of the primaries to something other than their optimal values given his primaries' locations, the color decoding will appear off visually I would suspect unless I'm missing something.

I have had some experience with the Epson projectors. The way the CMS works is that reducing the saturation to meet the gamut target also considerably lowers the brightness of the colors. You then have to raise the brightness to get them to their proper targets. When completed, both brightness and hue/saturation are very close to the intended target.


What I understand him to say is that at this point the color isolation method shows that color decoding performance (primary brightnesses, secondary hues) is off. This is what seems impossible. On the other hand, he claims that in the THX mode, the brightness targets measure incorrectly--even while the saturation and hues are nearly perfect--and using the CMS to correct this results in incorrect values according to the color isolation method. Again, he sees correct values using direct measurements but incorrect values using color isolation.
 

·
Registered
Joined
·
661 Posts

Quote:
Originally Posted by TomHuffman /forum/post/18290193


What I understand him to say is that at this point the color isolation method shows that color decoding performance (primary brightnesses, secondary hues) is off. This is what seems impossible. On the other hand, he claims that in the THX mode, the brightness targets measure incorrectly--even while the saturation and hues are nearly perfect--and using the CMS to correct this results in incorrect values according to the color isolation method. Again, he sees correct values using direct measurements but incorrect values using color isolation.

Right - there are only a few possibilities to explain that - one is that he doesn't trust his measuring instrument which would seem odd since that would throw all other measures (contrast, grayscale, etc.) into question; second is that there was somehow some kind of color decoding matrix mismatch issue when he was viewing the color bars pattern with the color isolation method. Once again, that seems odd since I'd assume that these patterns are on the same disc/pattern generator using the same connections/video chain that he is using to display the windowed color test patterns;


I guess I was thinking since he mentions that he only adjusted brightness of the primaries in the CMS and from the graph, it at least looks like green is over-saturated then that might somehow explain things. If he lowered green's brightness to match HD709 but didn't de-saturate the point to match HD709 then green might appear off when looking at the color bars with green isolated. Of course, red and blue look pretty darn spot on to HD709 - so....what doesn't make sense to me then is if the primary points are so close and the color decoding is claimed to be "accurate" out of the box then their brightness values should be pretty close to "accurate" as well. But he says, before touching the CMS, color decoding is accurate and the primary brightness values are slightly off. Both of those statements can't really be true. Unless he used the color isolation method to determine color decoding accuracy and then used his instrument measures to determine that the brightness values of the primaries are slightly off. So, what this would say to me is that, assuming I trust the measuring device, is that my color isolation method was less accurate and only an approximation. I then alter the brightness values of the primaries and the color isolation method shows things to be off - which makes sense since it's only an approximation and I trust the measuring device more than that method. In the end, it seems to me what he is saying is that he trusts the color isolation method more than whatever method he's using with his measuring device.



--tom
 

·
Registered
Joined
·
1,011 Posts
Is there the possibility that he's using 100% windows with instrumentation, and 75% color bars with the isolation controls, and that there are non-linearities?
 

·
Banned
Joined
·
3,550 Posts

Quote:
Originally Posted by glaufman /forum/post/18291704


Is there the possibility that he's using 100% windows with instrumentation, and 75% color bars with the isolation controls, and that there are non-linearities?

Possible they did that but I doubt it. One way to check this as linear instead is to check greyscale at different saturation points (assuming your saturation control is linear) to determine where your best location for illuminant D65 will be at before you do CIE gamut. If you can't get 2.2 or lower gamma this becomes difficult. There is no use for using 100% windows or even checking these unless your checking your meter for tolerances, or checking saturation...
 

·
Banned
Joined
·
20,735 Posts
I think that all hinges on what he means by the statement: "This could not be corrected (using the RGBCMY controls) without altering the color decoding, which was excellent out of the box."


I don't understand what he means by that, and it appears not to make sense to me.


If he means that by using the CMS as it is implemented he can get the correct Y value but then the xy location moves away from the intended target and you can't achieve both correct at the same time then I could understand I suppose, but is that the case with this display? It isn't clear and I do not know.


Second, I'm not sure how he is judging "color decoding" in this context. How is he ascertaining the correctness of color bars? Is he using a blue filter, which in this case with a CMS will not work post-CMS because of the impact of mixing the filtered-out primaries, let alone with the inherent potential weaknesses of the filter. Is he using a blue-only mode on the display (again I am totally not familiar with the display so not sure of its options), and how is that blue-only mode implemented. If it comes after the CMS, then it also won't give you the correct view of the colorbars because of the mixed-in primaries, so it would appear to be off even though it is actually correct.


In other words: Tom your confusion makes sense to me, the way you interpreted it would be two statements seemingly in conflict. But it's vague what he means and it isn't clear what exactly he's looking at to judge the color decoding error on bars and he may just be making a judgement on bars that look wrong because the method of judging them is not sufficient. Without some clarification we're just kind of guessing at intent...
 

·
Registered
Joined
·
1,011 Posts

Quote:
Originally Posted by SteveMo /forum/post/18292174


Possible they did that but I doubt it. One way to check this as linear instead is to check greyscale at different saturation points (assuming your saturation control is linear) to determine where your best location for illuminant D65 will be at before you do CIE gamut. If you can't get 2.2 or lower gamma this becomes difficult. There is no use for using 100% windows or even checking these unless your checking your meter for tolerances, or checking saturation...

What? How on earth do you check grayscale at different saturations? Gray should have no sautration by definition.
 

·
Banned
Joined
·
3,550 Posts

Quote:
Originally Posted by glaufman /forum/post/18297029


What? How on earth do you check grayscale at different saturations? Gray should have no sautration by definition.

Moving the saturation control up, or down. On my projector (I'm not sure why this would be different) saturation effects luma, luma effects greyscale.
 
1 - 14 of 14 Posts
Top