Hello,

I have a question about what the optimal 100% IRE luminance should be for my LCD tv...

I have a Samsung LN46A630. The "Grayscale for Dummies" guide says the optimal range is 30-40 ftL. But I have seen elsewhere that the target range for newer LCDs (I got mine only a couple months ago) is more like 50-60 ftL.

So, in the very first calibration step when setting contrast on the 100% white window, how do I decide what is the correct target ftL reading? I don't quite understand why there would be a range (much less why I see different ranges in different guides). Shouldn't there be a specific number that is the standard?

Thanks in advance for any guidance.

Hello,

I have a question about what the optimal 100% IRE luminance should be for my LCD tv...

I have a Samsung LN46A630. The "Grayscale for Dummies" guide says the optimal range is 30-40 ftL. But I have seen elsewhere that the target range for newer LCDs (I got mine only a couple months ago) is more like 50-60 ftL.

So, in the very first calibration step when setting contrast on the 100% white window, how do I decide what is the correct target ftL reading? I don't quite understand why there would be a range (much less why I see different ranges in different guides). Shouldn't there be a specific number that is the standard?

Thanks in advance for any guidance.

There is a standard, for target illumination in a theater controlled environment. (I've seen a reference to it, but I didn't keep the link.)

I had the same question you did when I bought my LN46A650 in December. After some reading, it became clear to me that the different recommendations are driven by different assumptions about the environment. If you have an interior, darkened room with controlled lighting, that mimics a movie theater environment, you will find that the lower ftL recommendations, less than 40 ftL, work well. If your set is in a living room with exterior windows that allow in daytime lighting, you will need a higher level, perhaps 50 ftL or more.

Similar considerations arise with the question you did not ask, why is there a range of recommendations for gamma? If you have a controlled darkened environment, a gamma value higher in the range (2.2-2.35 or 2.2-2.5, depending on the author) will work. Your eyes will adjust to the dark environment and you will see shadow detail just fine. In an uncontrolled, brighter environment, your eyes will adjust to the higher ambient light, and a higher gamma will hide the shadow detail, so you will want to target the low end of the recommended range.

Tom Huffman, in his guide Basic Guide to Color Calibration using a CMS (http://www.avsforum.com/avs-vb/showthread.php?t=852536), used to offer the 30-40 ftL recommendation, but he received the question you raised so often that he revised his guide to now say 30-50ftL. So feel comfortable with any choice in this range. When I suffered the confusion you feel, I compromised between the 30-40 and 50-60 recommendations, and picked 45 ftL as a rough target for my somewhat dark living room with limited windows and thick curtains. If my wife watched TV only in the evening, I'd choose a lower target. What will work for you depends very much on your environment. The nice thing about doing the calibration yourself, is if you don't like it, you find it too bright or too dark, you can change it.

Put it as high as it will go without clipping. Then go as low a you like in terms of it feeling comfortable without introducing banding. Digital displays tend to only be transparent to about 6bits so beneffit from using as much available dynamic range as possible.

There is a standard, for target illumination in a theater controlled environment. (I've seen a reference to it, but I didn't keep the link.)

I had the same question you did when I bought my LN46A650 in December. After some reading, it became clear to me that the different recommendations are driven by different assumptions about the environment. If you have an interior, darkened room with controlled lighting, that mimics a movie theater environment, you will find that the lower ftL recommendations, less than 40 ftL, work well. If your set is in a living room with exterior windows that allow in daytime lighting, you will need a higher level, perhaps 50 ftL or more.

Similar considerations arise with the question you did not ask, why is there a range of recommendations for gamma? If you have a controlled darkened environment, a gamma value higher in the range (2.2-2.35 or 2.2-2.5, depending on the author) will work. Your eyes will adjust to the dark environment and you will see shadow detail just fine. In an uncontrolled, brighter environment, your eyes will adjust to the higher ambient light, and a higher gamma will hide the shadow detail, so you will want to target the low end of the recommended range.

Tom Huffman, in his guide Basic Guide to Color Calibration using a CMS (http://www.avsforum.com/avs-vb/showthread.php?t=852536), used to offer the 30-40 ftL recommendation, but he received the question you raised so often that he revised his guide to now say 30-50ftL. So feel comfortable with any choice in this range. When I suffered the confusion you feel, I compromised between the 30-40 and 50-60 recommendations, and picked 45 ftL as a rough target for my somewhat dark living room with limited windows and thick curtains. If my wife watched TV only in the evening, I'd choose a lower target. What will work for you depends very much on your environment. The nice thing about doing the calibration yourself, is if you don't like it, you find it too bright or too dark, you can change it.

Thanks very much for the explanation. Unfortunately I can't be a purist for all of my tv watching: only about 10-20% of my watching is during the day when there is sunlight coming into the room, but on the flipside only about 10-20% of my watching is with the lights completely off. The other 60-80% of the time it's evening but I'm doing other stuff so have one or two lamps on in the room while I'm watching.

I did my first calibration to about 46 ftL. That seems fine with a couple lights on - it's not blindingly bright but not dim either. When I turn the lights off though, sometimes it is a bit bright, especially in scenes with a lot of white in them. I guess that'll have to do though. I wish Samsung allowed me to store two movie-mode presets for each input, so I could have one calibrated for a light room and one for a dark room.

Sort of related question... on that very first step, if I'm targeting about 46 ftL, what should I set my backlight to? I currently have backlight = 4 and contrast = 92. Any reason to get to 46 ftL some other way (i.e., higher backlight and lower contrast)? Or does it not matter?

My first-pass calibration file is attached. I have only done grayscale and the primary levels so far. Have not gotten to the point of doing the xy calibrations for the primaries and secondaries.

Put it as high as it will go without clipping. Then go as low a you like in terms of it feeling comfortable without introducing banding. Digital displays tend to only be transparent to about 6bits so beneffit from using as much available dynamic range as possible.

Mr D touches on a point I did not address, as I wrote about the first half of your question, Luminance, and he wrote about the second part, the Contrast control. On the Samsung set you will find two basic controls that affect illumination, Contrast and Backlight. In most ranges, you will have two or three choices that give the same maximum illumination. As hypothetical values, it may be that a Contrast value of 84 and Backlight of 4 give about the same 45 ftL as a Contrast value of 65 and Backlight of 5. Obviously, lowering the Backlight and increasing the Contrast will give you the same maximum illumination while also giving you a greater dynamic range, and so it's a better choice.

Exactly how high you can set the Contrast varies from set to set. On my Samsung, I've found undesirable effects appearing in the grayscale whenever I try the Contrast above 90. The overall white never clips, but either my red component falls off abruptly, or just before that point the red/blue balance in the grayscale gets so far off that max white turns pinkish. So you may find, after you've chosen your initial Contrast value, that you will later want to fine tune it up or down a little bit, to give you a high overall contrast with as much RGB grayscale error as you are willing to accept. I've found the step and the continuous gray patterns on the AVSHD disk very helpful in looking for color shift, to let you determine how much error is acceptable. The general advice is that you can ignore color shift in the whiter-than-white region above IRE100.

...

Sort of related question... on that very first step, if I'm targeting about 46 ftL, what should I set my backlight to? I currently have backlight = 4 and contrast = 92. Any reason to get to 46 ftL some other way (i.e., higher backlight and lower contrast)? Or does it not matter?

My first-pass calibration file is attached. I have only done grayscale and the primary levels so far. Have not gotten to the point of doing the xy calibrations for the primaries and secondaries.

Obviously I was following up to Mr D's suggestion, while you were posting your file. I wrote above on Backlight. On my set, in this range, like you, I've picked a Backlight of 4.

If you look at the RGB lines in your grayscale, you can see the problem I've faced, with the high red component at IRE100. There seems to be a tradeoff, where you can back off the Contrast a little, and reduce this error which shows up as a pinkish tinge in your bright white. But you could defer facing this tradeoff and work on your color primaries first. I found when I adjusted the color primaries, looking at both the x-y and the Y values, the grayscale RGB errors were reduced and became easier to manage. When working on the colors, I followed Tom's advice and let the deltaE values drive me. From where you are now, I would certainly adjust the colors first and then come back to the grayscale/contrast. Calibration is very much an iterative process.

Maximize the contrast control, then set the backlight to the desired light output you want in your environment. Brighter room: set it higher, darker room: set it lower.

Sort of related question... on that very first step, if I'm targeting about 46 ftL, what should I set my backlight to? I currently have backlight = 4 and contrast = 92. Any reason to get to 46 ftL some other way (i.e., higher backlight and lower contrast)? Or does it not matter?
As others have pointed out, you want to minimize the backlight setting (saves power), and to maximize the contrast control up to the point where your grayscale isn't compromised.

Bill

thanks for the input everyone. so on these samsungs is the point at which higher contrast setting starts to throw off grayscale something around 90? mine is currently set at 92, with backlight at 4. so maybe if i up the backlight to 5 i can lower the contrast a bit and then the calibration will result in more accurate grayscale.

really appreciate the input.

thanks for the input everyone. so on these samsungs is the point at which higher contrast setting starts to throw off grayscale something around 90? mine is currently set at 92, with backlight at 4. so maybe if i up the backlight to 5 i can lower the contrast a bit and then the calibration will result in more accurate grayscale.

Dave, moving the contrast will affect the grayscale. If you raise contrast, you should see a slight shift from right to left of the grayscale response and vice versa if you lower contrast. On my Samsung, I have a pronounced bump in blue at around 50% stimulus. If I move my Contrast control, I can move the bump left or right (as well as change it's shape a bit - from pointy to more rounded - this is with a 21 point grayscale run) about 5%. As folks typically mention, it's best to experiment - you may have to sacrifice a bit of your dynamic range in order to get a better tracking grayscale. As for specific numbers, 92 sounds reasonable for Samsungs - it's currently where I have mine set (with a backlight of 2 which produces around 35 ftls for a 100% stimulus white window pattern). I found that I could put the contrast up to 100 without coloration in the high end of the grayscale - but it did cause a pinkish hue in the above white part of the grayscale ramp so I've backed it down a bit.

The main rules for setting contrast are:

(1) No clipping of high end/above white (use a white ramp of some kind to see this)
(2) No crushing/compressing of high end white (some displays - CRTs/plasmas - will have their gamma drop significantly at the high end when contrast is set too high)
(3) No coloration in the high end of the grayscale.
(4) No visible noise in the high end of the grayscale.
(5) Optimize for viewing environment/No eye fatigue (too bright/too dim for your viewing conditions).

On LCDs, backlight is where #5 comes in. You can set Contrast so the first 4 rules are satisfied and then use the backlight to control overall brightness. In theory, moving the backlight up or down should not alter your grayscale response. Of course, the backlight control may be too coarse (each tick of mine changes overall brightness by about 9-10 ftls at 100% white, while each tick of contrast alters it by about 1). So, you may need to tick Contrast up or down a few ticks if you're shooting for a specific luminance range/target (e.g. 30-35 ftls) and your backlight either undershoots or overshoots it.

hope this helps,


--tom

Dave, moving the contrast will affect the grayscale. If you raise contrast, you should see a slight shift from right to left of the grayscale response and vice versa if you lower contrast. On my Samsung, I have a pronounced bump in blue at around 50% stimulus. If I move my Contrast control, I can move the bump left or right (as well as change it's shape a bit - from pointy to more rounded - this is with a 21 point grayscale run) about 5%. As folks typically mention, it's best to experiment - you may have to sacrifice a bit of your dynamic range in order to get a better tracking grayscale. As for specific numbers, 92 sounds reasonable for Samsungs - it's currently where I have mine set (with a backlight of 2 which produces around 35 ftls for a 100% stimulus white window pattern). I found that I could put the contrast up to 100 without coloration in the high end of the grayscale - but it did cause a pinkish hue in the above white part of the grayscale ramp so I've backed it down a bit.

The main rules for setting contrast are:

(1) No clipping of high end/above white (use a white ramp of some kind to see this)
(2) No crushing/compressing of high end white (some displays - CRTs/plasmas - will have their gamma drop significantly at the high end when contrast is set too high)
(3) No coloration in the high end of the grayscale.
(4) No visible noise in the high end of the grayscale.
(5) Optimize for viewing environment/No eye fatigue (too bright/too dim for your viewing conditions).

On LCDs, backlight is where #5 comes in. You can set Contrast so the first 4 rules are satisfied and then use the backlight to control overall brightness. In theory, moving the backlight up or down should not alter your grayscale response. Of course, the backlight control may be too coarse (each tick of mine changes overall brightness by about 9-10 ftls at 100% white, while each tick of contrast alters it by about 1). So, you may need to tick Contrast up or down a few ticks if you're shooting for a specific luminance range/target (e.g. 30-35 ftls) and your backlight either undershoots or overshoots it.

hope this helps,


--tom

let's say i calibrate to 46 ftL (100% white) for watching most of the time in a bright room.

are you saying that in a dark room i can just lower the backlight from 4 to 3 and, while the 100% white luminance will be lower (maybe about 35 ftL), all of the rest of the readings will remain correct (meaning contrast, brightness, white balance, gamma, and color space settings will not need re-adjusting at all)?

are you saying that in a dark room i can just lower the backlight from 4 to 3 and, while the 100% white luminance will be lower (maybe about 35 ftL), all of the rest of the readings will remain correct (meaning contrast, brightness, white balance, gamma, and color space settings will not need re-adjusting at all)?

I would be skeptical about greyscale remaining fully correct, the exact color of the backlight can change depending on how bright it is, so you may need to tweak a different mode setting for a different backlight level. Obviously if you have measuring equipment you can test easily to see the effect of changing the backlight and if color temperature remains stable then it simplifies things. So yes, greyscale can be affected, but not necessarily so, measure to see.

Black level, white level should remain the same, though if you're watching with ambient light you may have your black level artificially high to fight that, so that may cause a change if you move to night viewing. Gamma shouldn't really be affected, however again you may choose a more accurate 2.5 gamma for reference night viewing, and a lower gamma for daytime viewing to fight ambient light a bit. In other words, these are settings that shouldn't be affected, but that you may want to change for the ambient lighting in the room.

Color decoder settings (saturation and color balance) should not change.

I would be skeptical about greyscale remaining fully correct, the exact color of the backlight can change depending on how bright it is, so you may need to tweak a different mode setting for a different backlight level. Obviously if you have measuring equipment you can test easily to see the effect of changing the backlight and if color temperature remains stable then it simplifies things. So yes, greyscale can be affected, but not necessarily so, measure to see.

Black level, white level should remain the same, though if you're watching with ambient light you may have your black level artificially high to fight that, so that may cause a change if you move to night viewing. Gamma shouldn't really be affected, however again you may choose a more accurate 2.5 gamma for reference night viewing, and a lower gamma for daytime viewing to fight ambient light a bit. In other words, these are settings that shouldn't be affected, but that you may want to change for the ambient lighting in the room.

Color decoder settings (saturation and color balance) should not change.

Thanks a lot.

The problem is, for each input (say, my cable box), I only have one "movie mode" preset that I can save. And movie mode is the only one that allows me to start with "Warm 1" color tone. There is a "standard mode" preset I could tweak and save for darker viewing, but it doesn't let me start with warm1 color tone (only "normal" "cool 1" or "cool 2"). And warm1 is the one that results in the best calibration.

Dave, in support of what Chris wrote (http://www.avsforum.com/avs-vb/showthread.php?p=16093727#post16093727) above, I did the experiment (http://www.avsforum.com/avs-vb/showthread.php?p=15881728#post15881728) on my Samsung LCD of varying the Backlight alone and measuring the results. As the Backlight was reduced, the grayscale color balance got worse, but generally maintained the same shape, and the measured gamma stayed pretty much the same. So on my set the experiment confirmed that I could change the Backlight by one or maybe two steps without worrying about the calibration getting radically out of whack.

I did find, when I have tried to calibrate to a lower Backlight setting, that the other controls each seemed to have a greater effect and it was a little harder to get a grayscale balance I liked.

Of course, your mileage may vary, so it's worth performing the measurements yourself if your worried.

Bill

are you saying that in a dark room i can just lower the backlight from 4 to 3 and, while the 100% white luminance will be lower (maybe about 35 ftL), all of the rest of the readings will remain correct (meaning contrast, brightness, white balance, gamma, and color space settings will not need re-adjusting at all)?

Dave,

In theory yes, but as Chris pointed out, it really depends on the particulars of the display and it's backlight - which is why experimenting and measurements are the best option to see what your set does. More than likely if you're just moving it from 4 to 3, you'll get minimal (if any) changes in your display's grayscale response. On my Samsung, going from 3 to 2/2 to 1 (or vice versa) has no real measurable grayscale effect (i.e. it's within the variance of my Eye-One Display meter on that display). I did a few other experiments early on with the backlight and don't recall bigger movements having much more of an affect either. Of course, how much you can move it without any grayscale affects may also depend on how fine-grained the control is. Mine changes light output by about 30 cd/m^2 at 100%. So, one tick in either direction is more than enough to increase/decrease light output if it was desired. Given that, I probably didn't do any experiments where I had the backlight at 8 for example, calibrated and then lowered it to 2 and re-measured. I think all of my experiments had the backlight in the 1 - 5 range.

If you can only save one set of optimal settings per input then I would calibrate and save your settings for nighttime viewing (assuming that this is when your movie watching would take place) and then simply tick the backlight up one notch during the day on days when it looks too dim. Or you could calibrate the Standard mode for daytime viewing even if it means a slightly worse grayscale. Or you could simply live with a slightly dimmer image during the day. We choose the latter since we don't do a lot of daytime tv watching. It really depends on when you watch the display the most and what you consider your "critical" viewing times.

hope this helps,


--tom

Dave,

In theory yes, but as Chris pointed out, it really depends on the particulars of the display and it's backlight - which is why experimenting and measurements are the best option to see what your set does. More than likely if you're just moving it from 4 to 3, you'll get minimal (if any) changes in your display's grayscale response. On my Samsung, going from 3 to 2/2 to 1 (or vice versa) has no real measurable grayscale effect (i.e. it's within the variance of my Eye-One Display meter on that display). I did a few other experiments early on with the backlight and don't recall bigger movements having much more of an affect either. Of course, how much you can move it without any grayscale affects may also depend on how fine-grained the control is. Mine changes light output by about 30 cd/m^2 at 100%. So, one tick in either direction is more than enough to increase/decrease light output if it was desired. Given that, I probably didn't do any experiments where I had the backlight at 8 for example, calibrated and then lowered it to 2 and re-measured. I think all of my experiments had the backlight in the 1 - 5 range.

If you can only save one set of optimal settings per input then I would calibrate and save your settings for nighttime viewing (assuming that this is when your movie watching would take place) and then simply tick the backlight up one notch during the day on days when it looks too dim. Or you could calibrate the Standard mode for daytime viewing even if it means a slightly worse grayscale. Or you could simply live with a slightly dimmer image during the day. We choose the latter since we don't do a lot of daytime tv watching. It really depends on when you watch the display the most and what you consider your "critical" viewing times.

hope this helps,


--tom

right, thanks a lot for your thoughts (and everyone). it's a toss up. obviously a more perfect calibration is more important when watching a movie in darkness with undivided attention. which is a reason to calibrate for those conditions. but that's really only 10-20% of my watching. i will do two calibrations and have a specific nighttime setting for blu ray watching, because that is on a different input (so can save it). but that's not the majority of my nighttime movie watching... most are movies from HD cable, so same input source as my daytime watching. i think in the end, i can live with a slightly dimmer picture during the day... doesn't bother me as much as having too bright a screen at night.

thanks again.

My recommendation would be to calibrate your cable box properly for night viewing, then during the day you can manually turn up the backlight significantly. The change in greyscale is unlikely to be very significant, and with a great deal of ambient lighting happening are you really going to care about that? It seems I would be more dissatisfied with a dim and washed out no-shadow detail picture with still perfect greyscale in the day, than a display with very good or excellent but not quite perfect greyscale (still certainly better than any out-of box setting) but that is bright enough to be satisfying for daytime and largely not super-critical viewing anyway.

We all must live with compromises. If you want perfection 100% of the time, live in a cave. Choosing to view in the daytime isn't ideal, but it's perfectly normal, and so crank away at the backlight I guarantee you'll probably be happier with the picture that way in a bright room. Turn it back down for the nighttime and you'll have your perfectly calibrated images for when it really counts anyway.

Sometimes you just gotta use a little bit of common sense, and be reasonable. ;)

And that's coming from someone who doesn't budge a whole lot when it comes to pursuit of accuracy wherever possible.

And it'll be easy to test my theory. Throw up a grey field. Do a perfect calibration at a dim viewing level at night. Then the next day, or even right then, put the meter away and crank the backlight up. Do you immediately say "UGGHHHHH it's HORRIBLE!" ? More likely you'll have trouble even telling a difference at all. Try the same thing while watching a TV channel. I guarantee you that you will be more angry at the crappy cable quality and macroblocking than you will be about any relatively minor change in greyscale. :)

I went through the rest of the calibration process. I was able to adjust the custom color space for the primaries to get closer to the xy targets. But for the secondaries I had no luck. As recommended I just tweaked the controls while taking readings to see the effect each had, but it didn't seem possible to get closer to the xy targets (without radically altering Y).

If anyone is willing to have a look at my final calibration file and give suggestions, I'd very much appreciate it. It's attached.

So far I haven't watched many programs with these settings, so I don't really know yet what I think of them.

I went through the rest of the calibration process. I was able to adjust the custom color space for the primaries to get closer to the xy targets. But for the secondaries I had no luck. As recommended I just tweaked the controls while taking readings to see the effect each had, but it didn't seem possible to get closer to the xy targets (without radically altering Y).

If anyone is willing to have a look at my final calibration file and give suggestions, I'd very much appreciate it. It's attached.

So far I haven't watched many programs with these settings, so I don't really know yet what I think of them.

I can see you're making progress.

So far, I've tended to follow the advice to minimize the measured error. As regards the primaries, the green Y seems a little low, and the blue Y in particular is high. I expect if you drop the blue luminance down a little, you'll give up some error in the x-y plane but the overall blue error will be reduced.

Each of the secondaries seems to have a different issue, which is kinda neat as you get to work on all three. Yellow is largely a luminance issue; by dropping its red and green components together you can bring its Y value in line. Cyan is outside the line between blue and green, so you can solve the saturation error here easily by adding enough red to make it line up. Magenta is largely a hue issue; if you add some blue and remove some red, you can make the green-magenta line in the CIE chart go through the intersection of the other two lines nicely at white. Its remaining error in saturation can be solved by adding a touch of green, to pull it in to the red-blue edge of the triangle.

Good luck,
Bill

I can see you're making progress.

Bill, thanks for your help... I have some follow-up questions, if you don't mind.

As regards the primaries, the green Y seems a little low,

how exactly do you know that? and how do i fix it?

and the blue Y in particular is high. I expect if you drop the blue luminance down a little, you'll give up some error in the x-y plane but the overall blue error will be reduced.

again, how exactly do you know that? and how do i drop the blue luminance down? and will that throw off other things i should watch for?


Each of the secondaries seems to have a different issue, which is kinda neat as you get to work on all three. Yellow is largely a luminance issue; by dropping its red and green components together you can bring its Y value in line.

how do you know yellow has a luminance issue (what value are you looking at)? i will try that fix.

Cyan is outside the line between blue and green, so you can solve the saturation error here easily by adding enough red to make it line up. Magenta is largely a hue issue; if you add some blue and remove some red, you can make the green-magenta line in the CIE chart go through the intersection of the other two lines nicely at white. Its remaining error in saturation can be solved by adding a touch of green, to pull it in to the red-blue edge of the triangle.

I will try these fixes, thanks!!!

Bill, thanks for your help... I have some follow-up questions, if you don't mind.

how exactly do you know that? and how do i fix it?

I know the green is a little low by looking at HCFR primary/secondary color grid. You may have the problem I had initially, that with my screen resolution I was not seeing the bottom line of the primary/secondary color grid, labeled delta luma. You may need to click the down arrow to the right of the "editable data" box at the top to expand the top part of the screen. When you do, you will see that HCFR reports green as 1.5% low relative to Rec709.

I also copied the numbers from your file into a spreadsheet (http://www.avsforum.com/avs-vb/showthread.php?p=15887531#post15887531) to see what the results would be when adjust for the current primary color x-y coordinates. Taking into account the actual green position, its Y value is 2.2% low. The Accupel Display Calibration calculator (http://www.accupel.com/HDG4000_manuals.html) can give you the same results, if you would rather not use a spreadsheet.


again, how exactly do you know that? and how do i drop the blue luminance down? and will that throw off other things i should watch for?

Same story, but more so, with the blue luminance. The delta luma line in HCFR reports it as 11% high relative to Rec709. Given the actual blue position, it calculates as 12.8% high.

For both green and blue, you should be able to adjust these by simply moving the corresponding slider in the Samsung CMS, after you have chosen the custom color space, i.e., the green slider for green and the blue slider for blue. Green may be close enough that one click down makes it low rather than high, and perhaps 1.5% is as close as you can get. But for the blue, I'm sure by increasing the blue slider you can reduce this error.

Fixing the Y values for the primary colors usually affects the high end of the grayscale so you will want to recheck the grayscale balance after getting the primaries as close as you can.


how do you know yellow has a luminance issue (what value are you looking at)? i will try that fix.

Similar story about the yellow luminance, but with a slight twist in the remedy. The bottom line of the HCFR color grid shows the yellow Y is 4.9% high. This is easily fixable again in the CMS, by tweaking down red and green in parallel. Take one or two clicks off each, remeasure and see how much closer you are. As yellow is the mixture of red and green, you want to move them both in tandem, otherwise you will unintentionally change the position of yellow.


I will try these fixes, thanks!!!

As regards cyan and magenta, one of the reasons I added the CIE94 measures to the color spreadsheet that angryht developed was to make it easy to determine where the problem is between luminance, saturation, and hue. HCFR itself gives you part of this answer for CIELAB76, by pulling out the percentage Y error in the delta luma line and reporting the rest of the color error in the delta xy line.

hey bill, if you wouldn't mind looking at this new calibration file, i'd appreciate it. i think everything is looking good except that the green primary is undersaturated and its hue is a bit off. however, i don't know what i can do about that. my custom color space settings for Green are currently: R=15, G=44, B=0. to increase green's saturation, i'd have to cut R and B together. but B is already 0. any thoughts?


thanks a lot for all your help.

hey bill, if you wouldn't mind looking at this new calibration file, i'd appreciate it. i think everything is looking good except that the green primary is undersaturated and its hue is a bit off. however, i don't know what i can do about that. my custom color space settings for Green are currently: R=15, G=44, B=0. to increase green's saturation, i'd have to cut R and B together. but B is already 0. any thoughts?


thanks a lot for all your help.

I think your colors are great already. It is a characteristic of the Samsung LCDs that the built-in gamut does not quite reach to the Rec709 green. So you cannot quite get the saturation where you want it, the best you can do is to minimize the error. And that you have already done by minimizing the
hue and luminance error.

Looking at the same results with dE 94, the green error is only 1.2, which should make you feel better about it. One of the characterisitics of CIE94 is that the error measure does a better job of weighting the regions, making red more important and green less important. That is the only error greater than 1; when compared against the actual primaries, the errors of all six RGBYCM are less than 1.

Looking at the grayscale, I suspect you can see a slight pink tinge at the high end of a Pluge pattern. As the rest of the grayscale looks so smooth, I expect you have already worked on this and found that anything you do to improve the balance at IRE100 makes the errors worse elsewhere. Having done so myself, I can assure you one can spend hours and hours trading off improvements in one part of the grayscale with errors elsewhere. So I think you should enjoy your set for a while.

bill- thanks. i really do appreciate your help. i do not plan on calibrating again for some time. i just like to have in mind what if anything there is to improve on. it's pretty cool once you get the hang of the samsung color space controls.

you're right about my grayscale - only controls are gain and offset, so i picked the best compromise i could find.

is there anything else in the calibration that looks to you like it could/should be improved on my next go around (whenever that is)?

the only things i can think of are:

1. there is a "color tone" control. right now i have it set on "warm 1". there is also a "warm 2" option, and i am wondering if starting a calibration from there would widen the color space and get the green primary xy to be more accurate (i.e., more saturated).

2. with my last calibration, the black detail was simply phenomenal (maybe dumb luck, but it was incredible - like 3D looking into any shadows). currently my gamma control is set at -1 (same as for last calibration), and the gamma curve is again pretty flat in this calibration, which is great. the average gamma is still about 2.1. but for some reason i'm seeing a little black crush now with this new calibration. i know the recommended gamma range is 2.2 to 2.5. i have found that lowering the gamma setting to -2 brings the average gamma up to about 2.4, but then the curve is not as flat. wondering if there is a way to keep a flat gamma curve but raise the average gamma a bit. or do something to get back the amazing black detail i had before? (i like my colors better now though.)

bill- thanks. i really do appreciate your help. i do not plan on calibrating again for some time. i just like to have in mind what if anything there is to improve on. it's pretty cool once you get the hang of the samsung color space controls.

you're right about my grayscale - only controls are gain and offset, so i picked the best compromise i could find.

is there anything else in the calibration that looks to you like it could/should be improved on my next go around (whenever that is)?

the only things i can think of are:

1. there is a "color tone" control. right now i have it set on "warm 1". there is also a "warm 2" option, and i am wondering if starting a calibration from there would widen the color space and get the green primary xy to be more accurate (i.e., more saturated).

2. currently my gamma control is set at -1, and the gamma curve is pretty flat in this calibration, which is great. the average gamma is about 2.1. i know the recommended range is 2.2 to 2.5 though. i have found that lowering the gamma setting to -2 brings the average gamma up to about 2.4, but then the curve is not as flat. wondering if there is a way to keep a flat gamma curve but raise the average gamma a bit.

Comparing Warm1 and Warm2, it changes the default color temperature, but has no effect on the color gamut. I measured and documented the results of changing color tone alone on my A650 here (http://www.avsforum.com/avs-vb/showthread.php?p=15827874#post15827874). Actually I started calibrating my A650 at Warm2, because it was closer to D65 to start with. I later found that there were some irregularities in the grayscale, particularly in blue, that were easier to avoid starting from Warm1. At that time, it seemed specific to my Panasonic blu-ray player, that my Pioneer DVD player gave a better calibration at Warm2. But I later worked on it some more and made Warm1 work there, too. Overall the differences were small; you could try on your set with your sources to see if the grayscale and gamma works out a little better at Warm2, I don't expect you would see much change.

Since we tend to watch blu-ray movies in the evening with lower ambient light, the gamma for my blu-ray player is calibrated a little higher, in the range 2.27-2.43. So I do have my control set to -2. If you target a higher gamma, the challenge is to keep the gamma low enough at the bottom so that near blacks stay visible. I ended up spending a lot of time tweaking up the RGB offsets together and dropping the brightness to push down the gamma at IRE5-10 to stay low. I have seen charts of Samsung plasmas with very flat gammas, but I have seen very few results like that for the Samsung LCD. If you do most of your viewing in a lower light environment, I expect you would be happy with the higher gamma, if you can manage the low end gamma satisfactorily. Flat would be nice, but I don't think you will notice any anomalies if all the gamma points are in a tight range, say between 2.25 and 2.40. Of course, you can check this with a continuous Pluge pattern after doing the adjustments.

Have fun,
Bill

bill, if you have thoughts on this (http://www.avsforum.com/avs-vb/showthread.php?p=16300545#post16300545) question, please let me know. and THANKS AGAIN!

hey bill, if you wouldn't mind looking at this new calibration file, i'd appreciate it. i think everything is looking good except that the green primary is undersaturated and its hue is a bit off. however, i don't know what i can do about that. my custom color space settings for Green are currently: R=15, G=44, B=0. to increase green's saturation, i'd have to cut R and B together. but B is already 0. any thoughts?

Dave, I'd agree with Bill - your color measures are pretty much spot on. As for green, the easiest way to think of things is that your display has it's native notion of what green is - there is no way to make it any 'greener' - thus there is no way to increase saturation. Still, the difference between your green and HD709 is so small that it is within the limits of visibility more than likely

The only thing I see that you might want to work on in the future is the low end grayscale. It seems to be trailing off on green in the low end with red increasing as well - there may not be much you can do with it since like with my Samsung, blue begins to dominate below 20% anyhow. You may be aided by doing a 21 point grayscale measure - it'll give you a bit more detail about how each color performs across the entire range. I find the 21 point measures allow one to easier see the bumps/dips in certain colors thus allowing one to trade off a bit of error in one spot for less error elsewhere.

As for black detail, re-check your black bars pattern and make sure that you are not crushing 1% into black. If so, tick up your brightness until you can see it. This will lower your gamma in the low end - I've found on my Samsung that it comes out of black fairly fast - i.e. my gamma at 5% is 2.05-2.1 with the 1% black bar barely visible while the rest of my luminance response from 15%-100% puts gamma at around 2.2-2.3

hope this helps,


--tom

that near blacks stay visible. I ended up spending a lot of time tweaking up the RGB offsets together and dropping the brightness to push down the gamma at IRE5-10 to stay low.

I have also used this technique (in the opposite direction though) to tweak gamma at the low end. The best thing to do is to do a set of controlled measures to see what your settings do to the luminance response when they change. What I did was to take free measures at 20/40/60/80/100% white. I then changed brightness/contrast/RGB Offsets/RGB Gains one at a time 5 ticks in one direction and re-measured so I could graph their affect on luminance. I was interested to see if I could lower luminance from 10-50 while still maintaining the proper brightness setting. My experiments showed that on my display moving Contrast or RGB Gains in unison resulted in the pretty much the same luminance affect (a flat drop across the entire range). But I found that Brightness and RGB Offsets were somewhat different with RGB offsets having a measurably bigger affect at 20/40/60% - thus allowing me to tweak my gamma by moving RGB Offsets down in unison and then re-checking/re-setting Brightness/Contrast accordingly to maintain black level and the same brightness level at 100%.

cheers,


--tom

Tom, sounds like one of those "Great minds think alike" experiences.
Bill

that's some tricky stuff to get my head around. in my new calibration, my gamma at 10%IRE is 2.22. in my previous calibration (which had better shadow detail), my gamma at 10%IRE was 2.17. so, the previous calibration came out of black slightly more quickly (showing more shadow detail), right?

if i want to tweak my new calibration to come out of black slightly more quickly and show more detail, i don't really want to touch the gamma control, right? because if i lower it to -2, that will bring up the average gamma, which will yield even less shadow detail. but if i raise it to zero, it will lower the average gamma well below 2.2, which will make everything washed out, and true black will only look dark gray. (couldn't i then just lower the brightness to fix that?)

so, you're saying i should somehow try and tweak only the low end of the gamma curve to try and lower the gamma at 10%IRE, while leaving the rest of the curve where it is. and i can do that by increasing my offsets in unison (which should add some shadow detail to dark scenes) and simultaneously lowering the brightness (which will keep true black looking black and not dark gray)?

if i have that right (a big if), what are the pitfalls?

that's some tricky stuff to get my head around. in my new calibration, my gamma at 10%IRE is 2.22. in my previous calibration (which had better shadow detail), my gamma at 10%IRE was 2.17. so, the previous calibration came out of black slightly more quickly (showing more shadow detail), right?

if i want to tweak my new calibration to come out of black slightly more quickly and show more detail, i don't really want to touch the gamma control, right? because if i lower it to -2, that will bring up the average gamma, which will yield even less shadow detail. but if i raise it to zero, it will lower the average gamma well below 2.2, which will make everything washed out, and true black will only look dark gray. (couldn't i then just lower the brightness to fix that?)

so, you're saying i should somehow try and tweak only the low end of the gamma curve to try and lower the gamma at 10%IRE, while leaving the rest of the curve where it is. and i can do that by increasing my offsets in unison (which should add some shadow detail to dark scenes) and simultaneously lowering the brightness (which will keep true black looking black and not dark gray)?

if i have that right (a big if), what are the pitfalls?

Yes, that was the suggestion, to try to keep the gamma from getting too high at the bottom. Tom uses the same technique, to keep the gamma from getting too low.

The question is what is happening between 0 and IRE10. As Tom suggested, you get some feel for this if you run the grayscale measures at 5% increments instead of 10%. There are patterns for this on both the GetGray disk and the AVSHD disk. You can configure this number in HCFR after opening a .chc file in the Measures->Parameters dialog. The default is 10 points; 20 points gives you 5% increments.

Even more helpful, if this is your real interest, is the set of Near Black patterns on the AVSHD disk. Again, in the Measures->Parameters dialog, you can increase the number of near black levels from 4 to 10. Then if you select the Near black scale in the View, the Go button starts the measures, and you get values from IRE0 through IRE10. The neat thing about the Near Black Scale chart in HCFR is that by right clicking on the mouse you can enable logarithmic mode, which gives you the gamma at each of these points. When you measure this, you may find it is already pulling out nicely at the bottom. I measure gamma on my set starting at 2.0 at 1% and pulling up from there.

You can try first taking the Near Black measures and seeing what is really happening at the low end. Certainly you want to take the measures if you change the gamma control to -2 to try to tweak the overall gamma higher. One of the few uncertainties about your calibration is the apparent increase in gamma from IRE20 down to IRE10. Either the 20 point grayscale measure or the near black measures will help you decide whether you have more work to do.

great info.

do you think my EyeOne Display LT is sensitive and accurate enough to give useful measures that i can trust in that low range?

great info.

do you think my EyeOne Display LT is sensitive and accurate enough to give useful measures that i can trust in that low range?

It may depend somewhat on how low you set the Backlight. If you set Backlight to zero, it may be difficult; I have not tried calibrating mine lower than 3. And you want to be careful of ambient light perturbing the readings. When I am doing this, I do enable the "Average many reads on dark measurements" box in the Measures->Sensor->Configure dialog to give me a little more confidence, but it does make it time consuming.

Looking at the several threads on display probes, one of the areas where the EyeOneDisplay is perhaps more reliable than the EyeOne Pro is luminance at the low end. I am not so much concerned with color shift at the very low end, as my eyes cannot detect that anyway. Trust? I still use my eyes to see that the measures match my perceptions, that I do see the steps on the screen from IRE0 through IRE10, and afterwards that I do see enough shadow detail in dark scenes.

So it's just a suggestion to try. What I like about it, is that it can tell me the calibration is good enough and I should stop and enjoy. Without the measures, I would worry that I was not seeing enough detail, or that I was seeing too much detail and it was too flat.

if i want to tweak my new calibration to come out of black slightly more quickly and show more detail, i don't really want to touch the gamma control, right? because if i lower it to -2, that will bring up the average gamma, which will yield even less shadow detail. but if i raise it to zero, it will lower the average gamma well below 2.2, which will make everything washed out, and true black will only look dark gray. (couldn't i then just lower the brightness to fix that?)

Dave, increasing the gamma control will boost luminance across the entire between 0% white (black) and 100% white (lowering gamma). So, while doing that and ticking up brightness may make the very low end better, your gamma above 30% will most likely suffer.

so, you're saying i should somehow try and tweak only the low end of the gamma curve to try and lower the gamma at 10%IRE, while leaving the rest of the curve where it is. and i can do that by increasing my offsets in unison (which should add some shadow detail to dark scenes) and simultaneously lowering the brightness (which will keep true black looking black and not dark gray)?

As Bill said, yes, that is the idea. You would increase your offsets in unison and then re-check both Brightness and Contrast. Lower Contrast a bit (probably only one notch) until you match your previous Y reading at 100% and then tick down brightness if you can easily see the 1% black bar. It may be that you don't need to change Brightness at all. But before trying that type of tweak, I'd simply follow Bill's advice and check the Near Black measures. I think you had said Brightness was set so that the 2% black bar was just visible - I'm not sure which calibration patterns disc that you're using but the solution to getting a bit more shadow detail may be to tick brightness up one notch - enough to allow you to just barely see the 1% black bar. GetGray has 1% and 2% black bars on it's pattern while I believe AVSHD even has a pattern that shows 17 (black is 16, 1% is 18, 2% is 20).

hope this helps,


--tom

will try that, thanks.

yesterday i set my PS3 to output in RGB range and set the samsung hdmi black level to low, and my calibration readings were completely different. 100IRE Y was about 10 ftL higher than before. tried a calibration from scratch and it came out poorly.

my plan is to enter my old picture settings (which had awesome black detail), then measure grayscale and colors. i'll use the 100IRE Y reading and the gamma curve from that reading as my goal. from those settings, i'm going to try to fix the luminance and xy of the colors but without messing up that overall 100IRE Y, contrast, brightness, and gamma. hopefully that can be done.

you guys have been really helpful!

will try that, thanks.

yesterday i set my PS3 to output in RGB range and set the samsung hdmi black level to low, and my calibration readings were completely different. 100IRE Y was about 10 ftL higher than before. tried a calibration from scratch and it came out poorly.

my plan is to enter my old picture settings (which had awesome black detail), then measure grayscale and colors. i'll use the 100IRE Y reading and the gamma curve from that reading as my goal. from those settings, i'm going to try to fix the luminance and xy of the colors but without messing up that overall 100IRE Y, contrast, brightness, and gamma. hopefully that can be done.

you guys have been really helpful!

If you saw a higher Y at IRE100, it sounds like you left the PS3 in RGB Extended Range (0..255) while the Samsung at HDMI Black Level Low was expecting studio (16..235) levels.

I think you had said Brightness was set so that the 2% black bar was just visible - I'm not sure which calibration patterns disc that you're using but the solution to getting a bit more shadow detail may be to tick brightness up one notch

Dave, I just realized, assuming I am remembering correctly, that your color balance showed a trailing off of green at the low end. Putting aside the issue of the meter's reliability with dimmer measures, I wonder if not re-working your grayscale so you increase green at the low end which at the same time will help to boost the luminance at the low end (since green has the most luminance of the 3 colors in the mix), is not a possible solution to your shadow detail/low end gamma issue.

hope this helps,


--tom

If you saw a higher Y at IRE100, it sounds like you left the PS3 in RGB Extended Range (0..255) while the Samsung at HDMI Black Level Low was expecting studio (16..235) levels.

the ps3 is definitely set to 16-235 range now. (if it weren't, then the hdmi black level control on the tv would be grayed out.)

thing is, i simply can't remember how i had the ps3's range and the samsung's black level set the first time i calibrated. but i ended up with great blacks (they only look great though if the hdmi black level is set to low). so, although i don't know how i got to those settings, the fact is i did. they should be a great starting point and target. from there, tweaking the primary and secondary xy and even minor gain and offset adjustements (a tick here, a tick there) should not spoil my great gamma. but i do need to fix the luminances of the individual colors, as that is very important for overall color balance. i worry that doing that will affect the 100IRE and 10IRE Y readings and require me to start changing contrast and brightness, all of which will spoil my gamma. we'll see.

hope you don't mind me thinking out loud (figuratively) here. kind of learning as i go. very interesting stuff.

the ps3 is definitely set to 16-235 range now. (if it weren't, then the hdmi black level control on the tv would be grayed out.)

...

Wouldn't it be nice. I don't think that's how my Samsung works. The set does not know what range the source is using, all it knows is that it is encoded in RGB. That's the reason for the user option at the set. So the box is grayed out if YCbCr, and not grayed out if RGB. (I have seen in another thread a plausible exception, that perhaps the option is also not available on HDMI2/PC even in RGB, that the set assumes 0-255 in PC mode. I have not tested this hypothesis myself.)

BTW, I should have written "Full" instead of "Extended" for the PS3 0-255 range. Not having either a PS3 or an Xbox360, I am always confusing the terminology.

i think i am making progress.

using the calculator i think improved my colors a lot (skin tones more realistic). and i think i've hung on to most of my black detail (though i had to reduce the contrast a bit so that i could keep the gray scale under control).

all of you suggestions have helped about how to take readings, etc.

newest calibration file is attached.

bill, chris, or thomas, if any of you wants to take a look and give me your feedback, that'd be fantastic.

using the calculator i think improved my colors a lot (skin tones more realistic). and i think i've hung on to most of my black detail (though i had to reduce the contrast a bit so that i could keep the gray scale under control).

Dave, it looks pretty good.

As far as color measures go, everything looks very spot on on the CIE chart - the only two things I'd do is add a bit more blue to magenta and add a bit more green to yellow. But I guess this depends on what your target is. Plugging in your measured primaries and measured white point from your color measures data, I see that you're pretty much spot on with both magenta and yellow. I initially assumed D65 as the white point. As for color luminance values, if you're targeting your custom gamut based on your measured primaries and measured white point at 100% then they all look good and in fact, are all close to the hd709 standard as well so I don't think you'll see a difference tweaking them to match hd709 as opposed to your custom gamut.

On the grayscale/gamma front, the one thing I still see is a dropping off of green below 25%. I think it you tick up the green offset a bit, you should be able to improve the color balance at the low end. On the RGB Graphs, if you right click and choose Scale/RGB, you'll see 80-120% - this will give you a more detailed view. You may need to tick down green in the high end a bit to compensate. Think of it as a see-saw since the low end controls invariably have an affect on the high end some (as does the high end controls on the low end). This will add more luminance to the low end so make sure to re-check brightness/gamma after doing this. While it may seem tedious, I find it good practice to redo all my grayscale/color measures after making any change. As for gamma, it looks fairly flat now with just a few dips/bumps. If you find the picture looking "flat", you could now try ticking up the gamma control which should raise the gamma up about .1 (if your Samsung behaves as mine does) and put you at about 2.25.

hope this helps,


--tom

Dave, it looks pretty good.

As far as color measures go, everything looks very spot on on the CIE chart - the only two things I'd do is add a bit more blue to magenta and add a bit more green to yellow. But I guess this depends on what your target is. Plugging in your measured primaries and measured white point from your color measures data, I see that you're pretty much spot on with both magenta and yellow. I initially assumed D65 as the white point. As for color luminance values, if you're targeting your custom gamut based on your measured primaries and measured white point at 100% then they all look good and in fact, are all close to the hd709 standard as well so I don't think you'll see a difference tweaking them to match hd709 as opposed to your custom gamut.

On the grayscale/gamma front, the one thing I still see is a dropping off of green below 25%. I think it you tick up the green offset a bit, you should be able to improve the color balance at the low end. On the RGB Graphs, if you right click and choose Scale/RGB, you'll see 80-120% - this will give you a more detailed view. You may need to tick down green in the high end a bit to compensate. Think of it as a see-saw since the low end controls invariably have an affect on the high end some (as does the high end controls on the low end). This will add more luminance to the low end so make sure to re-check brightness/gamma after doing this. While it may seem tedious, I find it good practice to redo all my grayscale/color measures after making any change. As for gamma, it looks fairly flat now with just a few dips/bumps. If you find the picture looking "flat", you could now try ticking up the gamma control which should raise the gamma up about .1 (if your Samsung behaves as mine does) and put you at about 2.25.

hope this helps,


--tom

very helpful as usual. yes i am actually finding the picture a bit "flat". it doesn't have quite the shadow detail i had before. i know black is not crushing, because i can see the 1% window. but there are areas of just inky black on the screen, and if i pause it and tick up the brightness, i can see detail. (but that makes things a little gray instead of black, and probably throws off my whole calibration.) by "ticking up the gamma control" you mean lowering the gamma setting, right (from say, -1 to -2)?

I thought I understood gamma correction, but could you tell me in a nutshell how doing that (ticking up the gamma control) differs from upping the contrast, brightness, gains (in unison), offsets (in unison), and/or backlight? That's probably a really stupid question, but hey...

lack, and probably throws off my whole calibration.) by "ticking up the gamma control" you mean lowering the gamma setting, right (from say, -1 to -2)?

Dave,

Yes, sorry I meant tick down the gamma control. But if you feel that the low end is too "milky black" without detail then ticking down gamma will probably make what you perceive to be worse since what it will do is make things somewhat darker in relation to 100% white but it'll preserve your black level.

Contrast, brightness, gamma, gains, offsets controls are all doing things that affect the luminance response of your display (although that may not be their main function such as the gains/offsets). Contrast mainly affects the high end but it can have some affect on the low end all depending on how it is implemented - e.g. it may affect the entire range equally which would have little/no effect on your measured gamma. Brightness mainly affects the low end but once again, what it actually does will depend on the specific implementation within the display. A gamma control is typically one that is specifically designed to affect the luminance response between 0 and 100 by applying some calculated offset. On the Samsung, moving the control lower (< 0) will make everything between 0 and 100 appear darker, raising it will make everything appear brighter between 0 and 100. Gains and Offsets are specifically designed to affect the grayscale color balance (and not the luminance) but as you add/subtract color (with green being the one with the most luminance), you also affect the luminance. How these controls, when moved in unison, affect the luminance will be specific to the display once again. They may behave exactly like alternate versions of contrast/brightness or they may behave somewhat differently. I've found the best way to see how they are behaving is to just do a few experiments by measuring perhaps a 11 point grayscale run - make a baseline run and then for the next run, tick brightness down/up 5 ticks and re-measure. See what it does to your luminance response. How much brighter/dimmer did it make 10, 20, 50%? Put the Brightness back to it's baseline setting and repeat the experiment for Contrast, etc etc. Assuming the controls are linear (i.e. each tick has about the same affect) then this should give you a good idea of what the controls are doing to luminance.

Backlight is the one control that is different - it is affecting overall light output. It is the light bulb that supplies the light to your LCD panel. Think of it as a x-way light bulb - as you move the backlight control, you can adjust the amount of light emitted by it. The backlight will affect your luminance across the entire range hopefully equally and will have miminal affect on your color balance (but it can as the actual color of the light emitted from the backlight changes). The general recommendation for backlight that I give is to set brightness/contrast accordingly (i.e. using the steps many have outlined in other posts) and then use the backlight to find a decent overall light level for your viewing conditions with the thought in mind that the lower you can get it, the darker black will be.

hope this helps,


--tom

very helpful as usual. yes i am actually finding the picture a bit "flat". it doesn't have quite the shadow detail i had before. i know black is not crushing, because i can see the 1% window. but there are areas of just inky black on the screen, and if i pause it and tick up the brightness, i can see detail. (but that makes things a little gray instead of black, and probably throws off my whole calibration.) by "ticking up the gamma control" you mean lowering the gamma setting, right (from say, -1 to -2)?

I thought I understood gamma correction, but could you tell me in a nutshell how doing that (ticking up the gamma control) differs from upping the contrast, brightness, gains (in unison), offsets (in unison), and/or backlight? That's probably a really stupid question, but hey...

"Flat" is frequently a term associated with having a gamma level that is too low. You've done an admirable job of making the gamma curve flat, As Tom suggested, I would probably tick the gamma control down a notch to try to raise it. If you look at the luminance graph, the effect of a higher gamma is to make the luminance graph more of a curve, a lower gamma makes the luminance curve more of a straight line. If the gamma is higher and the luminance is more of a curve, you will perceive more depth of field, more difference between light and dark.

I have seen where Tom Huffman recommends a gamma between 2.2 and 2.35. I recently came across an extended discussion by Poynton (http://www.poynton.com/notes/PU-PR-IS/Poynton-PU-PR-IS.book.pdf) recommending that Rec709 should incorporate a display gamma of 2.4. There is a complete thread discussing this subject, sometimes heatedly, started by Chris Wiggle, titled 2.5 is display gamma, NOT 2.2 (http://www.avsforum.com/avs-vb/showthread.php?t=1008297). As you will see if you look through Poynton's writings about gamma, the right choice depends on the environment, but 2.2 to 2.35 seems to work well for a home environment that is not completely dark. I have tried a higher range, but it did not work for me as we usually have some ambient light on in the room when we watch movies.

By the way, I'm glad your skin tones are now working out for you. You did spark an interesting discussion by getting Greg involved in the thread about his Accupel tool. It is fascinating that he does not agree with how we use his program. It has given me some food for thought. Thanks.

This is all interesting stuff. Although sometimes tiring. (I can only devote so much time and thought to this. But there is a desire right when you get a new tv, like i have, to work hard to optimize it off the bat so then you just enjoying it going forward.)

You all have been really really helpful to me. Thanks a lot for your time and insight.

By the way, I'm glad your skin tones are now working out for you. You did spark an interesting discussion by getting Greg involved in the thread about his Accupel tool. It is fascinating that he does not agree with how we use his program. It has given me some food for thought. Thanks.

I'm more comfortable thinking out loud in this thread than the other one. (Kind of intimidating over there. I have a lot of experience with digital audio, not much with video.)

But what do you think of this reasoning... Assume a particular skin tone falls somewhere within Red, Yellow, and White (not a crazy assumption). If you look at the CIE chart though and try to point out a skin tone, it's much closer to White than to Red and Yellow (because skin is rather pale). So, White is an important component in skin tones.

And RGBYCM all affect white. They can either be calibrated to the computed orthogonal values (straight lines through white), or they can be calibrated as close as possible to the Rec709 values.

I think there is something about calibrating them to the computed orthogonal values that makes White more accurate. Maybe Magenta itself is not as close to Rec709 as possible, so someone's magenta umbrella or something may look a little off (if you could even tell). But, this will leave White looking truly white. Which means the important White component of skin tones will be accurate.

Maybe it's as simple as... by aiming your secondaries for Rec709, you move your white point? So, by doing that, very saturated colors will be more accurate, but skin tones (which are very unsaturated) will suffer.

I'm not explaining this well.

... They can either be calibrated to the computed orthogonal values (straight lines through white), or they can be calibrated as close as possible to the Rec709 values.

I think there is something about calibrating them to the computed orthogonal values that makes White more accurate. Maybe Magenta itself is not as close to Rec709 as possible, so someone's magenta umbrella or something may look a little off (if you could even tell). But, this will leave White looking truly white. Which means the important White component of skin tones will be accurate.

...

IMO: Yes.

btw: Explanation is as good as i have seen. :)

Interesting, Dave. I posted a comment in the other thread, showing the results of my Excel experiment of the day. Somewhat along the lines of your suggestion, my experiment seemed to show that calibrating using Y values and secondaries for the new gamut may give a better compliance to Rec709 colors at the intermediate points, even if it gave poorer measures at the 100% saturation points. After working through the numbers, I am reluctant push any argument by analogy. I do think when I did my calibrations that following the advice to make the secondaries be the complements of the primaries gave me better skin tone. As I suggested in the other thread, this may be a better calibration, but a complete accident with no theoretical justification, that just happens to give us better color around skin colors.

I, too, like you, have thought about the effect near white. Because we have a white balance control to set grayscale, it is hard to demonstrate that our calibrating the primary Y values and the secondary points differently has much effect here. Intuitively, I think there may still be an effect. To the extent that calibrating all the colors using the actual primaries as defining the new gamut makes the red-green-blue add nicely to white, one might expect that there would be less adjustment needed to pull white to the right D65 point. And so, perhaps the grayscale controls end up with a more even performance from IRE100 down to IRE30, and a better mixing of colors. But that is still an intuitive hypothesis, I have little to back it up, and it is hard to support given that I start my calibration using Warm1 and go to a lot of work to pull the white point back to D65.

The next time I review the calibration on my set, I may try adjusting the color Y values, etc., to the Rec709 coefficients to see if my grayscale errors come out any worse. If so, that might explain why I thought I saw better results calibrating to the actual primary gamut.

Bill, saw what you posted over in the other thread. Crimony. :) Could you possibly explain it in 2-3 sentences? Or better yet, just give instructions on how to calibrate using your new sheet.

An update... I've told you that I have one set of settings that give me just phenomenal black levels, and I just want to tweak that to fix the color (green shadows in flesh tones is the biggest problem).

So, I started with those settings, took a 100IRE Y reading, and typed that into the accupel calculator along with the primary xy values that i know are the closest my tv can get to Rec709. I figured if I then just calibrate to the calculator's output, I'd be golden. (i.e., wouldn't really change my luminance, just adjust colors and maybe improve the white component).

Well, it did improve the color somewhat, and mostly preserved my black levels. But neither is perfect. So now I have (i) my original calibration with perfect black levels but bad skin tones, (ii) a calibration I did from scratch aiming for the accupel calculator values which resulted (maybe a coincidence) in perfect skin tones but bad black levels, and (iii) this newest calibration with pretty good black levels and alright skin tones.

I need to find a way to combine the black levels from (i) with the skin tones from (ii).

Bill, saw what you posted over in the other thread. Crimony. :) Could you possibly explain it in 2-3 sentences? Or better yet, just give instructions on how to calibrate using your new sheet.

An update... I've told you that I have one set of settings that give me just phenomenal black levels, and I just want to tweak that to fix the color (green shadows in flesh tones is the biggest problem).

So, I started with those settings, took a 100IRE Y reading, and typed that into the accupel calculator along with the primary xy values that i know are the closest my tv can get to Rec709. I figured if I then just calibrate to the calculator's output, I'd be golden. (i.e., wouldn't really change my luminance, just adjust colors and maybe improve the white component).

Well, it did improve the color somewhat, and mostly preserved my black levels. But neither is perfect. So now I have (i) my original calibration with perfect black levels but bad skin tones, (ii) a calibration I did from scratch aiming for the accupel calculator values which resulted (maybe a coincidence) in perfect skin tones but bad black levels, and (iii) this newest calibration with pretty good black levels and alright skin tones.

I need to find a way to combine the black levels from (i) with the skin tones from (ii).

Did you download the spreadsheet and look at the notes at the front? I tried to leave hints in case someone wanted to try to use it. The idea from Greg was that neither the Rec709 Y coefficients nor the coefficients from calculating a new gamut were right. Rather, one needs to optimize each Y at the best point to minimize the deltaE measures, and trust that minimum error gives the best calibration. So the spreadsheet was designed to try this idea out, to use Excel's Data Solver add-in to figure out the Y values that give the minimum dE measures. Then, rather than the trial-and-error, one can target these calculated Y values as the ones to use with our best achievable x-y coordinates. That's the simple idea, and the rest is just detail.

Obviously, the best position in the x-y domain is still somewhat trial-and-error, to see what the Samsung will do, and what yields the best result under the delta E measure. In my case, I am still leaning toward CIE94, as suggested by the Calman person in the other thread. I did note that Greg is pretty set on CIELUV76, the more traditional measure.

See if the comments in the spreadsheet are enough for you to see how it works and let me know.

Cheers,
Bill

P.S. I have now tried using the same spreadsheet during a calibration, beyond just the experiment to verify which strategies work better. As a calibration aid, it leaves a lot to be desired, as some key information is not close to each other, and a lot of internal details of the calculations could be hidden for normal use. Not yet "user friendly." If you try to use it for its new purpose as it stands, you are really only interested in the stuff in the first columns, related to Strategy 1, and not the other hypothetical alternatives. I still need to watch some films with the new calibration to see whether the improvement by numbers yielded an improvement to the eye.

Obviously, the best position in the x-y domain is still somewhat trial-and-error, to see what the Samsung will do, and what yields the best result under the delta E measure. In my case, I am still leaning toward CIE94, as suggested by the Calman person in the other thread. I did note that Greg is pretty set on CIELUV76, the more traditional measure.

Yes, dE1994 definitely tells you different luminance targets. dELUV basically says you can raise/lower luminance of a color to compensate for it's over/under saturation in order to minimize error. dE1994 really says that you mainly want to eliminate hue error in order to try and minimize error. I think from a color science perspective this has been debated a lot and there are folks in both "camps." I can't say which is better - other than to say that you can't compare the numbers since the scale is very different as far as what is acceptable error.

cheers,


--tom

Yes, dE1994 definitely tells you different luminance targets. dELUV basically says you can raise/lower luminance of a color to compensate for it's over/under saturation in order to minimize error. dE1994 really says that you mainly want to eliminate hue error in order to try and minimize error. I think from a color science perspective this has been debated a lot and there are folks in both "camps." I can't say which is better - other than to say that you can't compare the numbers since the scale is very different as far as what is acceptable error.

cheers,


--tom

I was following your exchange with Greg. You had me worried for a while that you were using a spreadsheet in which I had made a mistake. So I did the same as you, plugged the numbers into Huffman's dE-All spreadsheet and saw the same thing. I think I remember seeing in the discussion of the various measures that red was one of the areas where they tried to tweak things in CIE2000. Just a hypothesis, then, that perhaps CIELUV76 leans one way around red, CIE94 leans too far the other way, and in 2000 they tried to find the happy median.

It is fascinating, though, that the two metrics recommend going in different directions with Y to reduce the perceived error. I was seeing the same effect in my experiment as you were seeing. Like you, I am guessing this is a side-effect of the different weighting in CIE94, and that Lightness impacts the other variables differently in Luv and Lab. It does make sense, once you have seen it, if the differences were large both might say one of pair of colors is much brighter than the other when compared to a third, but at some point you would get you would get close enough with two distinct measures that one would say you go one way and the other says the reverse. If two measures gave the same answer everywhere, they would be the same measure.

I like your simple explanation that Luv says one can compensate for saturation error with brightness.

So I did the same as you, plugged the numbers into Huffman's dE-All spreadsheet and saw the same thing.
In one of Tom and my heated discussions, he calculated a dE error that did not match any of my models. He might have made a mistake, or the error might even be on my end, but someone should verify the math given the disparity in the results reported in Greg's Calculator thread.

Bill

In one of Tom and my heated discussions, he calculated a dE error that did not match any of my models. He might have made a mistake, or the error might even be on my end, but someone should verify the math given the disparity in the results reported in Greg's Calculator thread.

Bill

I found a different small error in my spreadsheet, where my results differed from Tom's and other spreadsheets, which did not affect the experiment I was doing. So I think it is just a difference between the two standards, and perhaps, as Greg says he doesn't use CIE94, he isn't used to a different set of reference limits.

Tom Huffman was very insistent about determining the right error limits for each standard. It doesn't much matter whether you look at a CIELUV error of 7.2 and recognize that it is significantly higher than a max acceptable of 5, or look at a CIE94 error of 2.1 and recognize it as higher than a max acceptable of 1.5. No one would consciously compare error measures under one spec to error values under another, but that is what is implicitly done if one thinks of an error of 2 as small. From what I've read so far, the only thing in common across them is that an error less than 1 is not perceptible.

By the way, Bill, it occurred to me that the misconception I had is partly encouraged by the breakdown of deltaE 94 numbers into %L, %C, and %H components. HCFR does the same by giving delta xy and delta luma errors. For we novices, it is easy to think that Y is correct and needs no change when the numbers show no or little error in lightness, without realising that by deliberately introducing error we might reduce the measured error even more elsewhere. I'm sure an experienced calibrator would not fall into this trap. For a person with a math background, this might be obvious if one could see the partial derivatives of delta E relative to x, y, and Y. But that would not be easy to explain to the less mathematically inclined. I suppose one might illustrate it graphically without using the intimidating words "partial derivative", if one could solve the problem of scaling to make the numbers comparable. Maybe one could use the approach that economists do when calculating elasticity with respect to multiple variables.

Oops. That sounds like another "exercise left for the student".

The idea from Greg was that neither the Rec709 Y coefficients nor the coefficients from calculating a new gamut were right. Rather, one needs to optimize each Y at the best point to minimize the deltaE measures, and trust that minimum error gives the best calibration. So the spreadsheet was designed to try this idea out, to use Excel's Data Solver add-in to figure out the Y values that give the minimum dE measures. Then, rather than the trial-and-error, one can target these calculated Y values as the ones to use with our best achievable x-y coordinates. That's the simple idea, and the rest is just detail.

So, I just input my primary xy values (as close as I can get them to Rec709) and my white xyY, and it'll give me the optimum Y values for RGBYCM and optimum xy for YCM that will minimize dE values from Rec709? And you think that'll give us the "best" (i.e., most natural or whatever) color.


Obviously, the best position in the x-y domain is still somewhat trial-and-error, to see what the Samsung will do, and what yields the best result under the delta E measure.

I wonder if, since Green xy cannot hit Rec709, there are better Red and Blue xy positions for us than the Rec709 positions.

Let us know how it looks to you when you watch some content.



I have another dumb question... there is no way to calibrate the actual White xy point, right? But by calibrating the other colors, that should get White as close as possible to Rec709?


Also, did you actually go through the exercise of using the calculator to ensure proper color decoding? What did your Color control end up at? (Mine is at 45 from previously just trying to use it to Red Y to equal .21 of White Y. But I wonder if that's throwing everything off and it should be left at 50.) Have you changed your Tint from 50/50?

So, I just input my primary xy values (as close as I can get them to Rec709) and my white xyY, and it'll give me the optimum Y values for RGBYCM and optimum xy for YCM that will minimize dE values from Rec709? And you think that'll give us the "best" (i.e., most natural or whatever) color.

Not quite but close. So the revelation from the discussion with Greg is that we should ignore completely the intuitive idea of trying to get the lines straight from the secondaries to the primaries in the x-y (or u-v) domain. So you have to optimize in the x-y plane as best you can, to see what the Samsung lets you do. If you use CIE94 as an error measure, or follow its tendancies, this becomes easier as it is most important is to reduce hue error, to line up the actual points between the target Rec709 points and white, where this does not cost too much in the other dimensions. For example, one can easily add red to get green in line with the target, without losing too much saturation. Adding blue to red to reduce hue error gives up too much saturation error, that's my guess anyway. For the secondaries, you just try to get them as close as possible to the Rec709 targets, with an emphasis to getting them in line between white and the target.

When you have the x-y points as close as possible, Excel can calculate the optimum Y values for all six points to minimize the calculated error. This reduces the number of times one has to measure and recalculate the numbers, and avoids the trial and error process of finding the optimum Y value with Greg's calculator. But, with the measurement variation in the Y of 100% white, measurement variation in the xy of all the colors, and the fact that on the Samsung changing the Y values moves the xy coordinates, there is still an iterative process of trying to calibrate to the calculated optimum targets, take all the color measures again, see what the new calculated targets are. When I tried this last night, it was still four or five iterations of dropping the controls down, tweaking them up, nope they're up too high, drop a little again, until I stopped and decided I could not do any better.


I wonder if, since Green xy cannot hit Rec709, there are better Red and Blue xy positions for us than the Rec709 positions.

Fascinating question, but no. Now that we have given up the falacious idea of lining up the secondaries with the primaries, with the 6-control CMS there is no reason we cannot calibrate cyan outside the line from green to blue, for example, if we can get there. Greg's clear advice is that you try to get each of the six points as close as possible, measured by the chosen delta E measure, to the Rec709 target. My experiment seemed to confirm that doing this resulted in lower or equal delta E values throughout the CIE triangle, not just at the six extreme points.

The tricky part of getting close, though, is this point that matching the Rec709 Y value when the x-y are different is not the closest possible point. Rather, because Y interacts with the other components in the error calculation, not just Lightness, some other Y value may give the best fit. Thus, the opportunity for trial-and-error or using the Excel Data Solver to calculate the optimum point.


Let us know how it looks to you when you watch some content.

Will do. I tried last night, but I was too tired after calibrating to stay awake through even one blu-ray. As far as I got, skin tones still looked good.


I have another dumb question... there is no way to calibrate the actual White xy point, right? But by calibrating the other colors, that should get White as close as possible to Rec709?

Questions only seem dumb after you know the answer. In my hypothetical 3 point CMS with no grayscale control, your intuition is right that the red-green-blue colors would determine where white is. But because we have white balance/grayscale controls, the set of gain and offset controls determine where white lies in the x-y plane at high and low ends of the IRE scale.

As no one who has instruments is trying to calibrate to a different white point than D65, in my spreadsheets I have assumed the x-y white coordinates when performing the other calculations, except for calculating the delta E of the measured white. I assume that one will calibrate the grayscale separately, and will end up with a set of measured points clustered around D65 white. I suppose one could consider taking an average of the 11 or 21 gray points, or all the ones above some lower limit, e.g., IRE20, but why bother.

If one was trying to deliberately target a different white point, that would impact the calculations. You can see this in HCFR, for example, in the RGB grayscale graph if you change the reference white point. The red/green/blue lines keep the same shape, but where they are on the scale changes a lot.


Also, did you actually go through the exercise of using the calculator to ensure proper color decoding? What did your Color control end up at? (Mine is at 45 from previously just trying to use it to Red Y to equal .21 of White Y. But I wonder if that's throwing everything off and it should be left at 50.) Have you changed your Tint from 50/50?

Given we have a CMS, the Color control is just the rough control to set the general range Y values before calibrating the colors individually. It doesn't really matter whether we use it to set red at .21 of white, green at .72 of white, or use the Rec709 coefficient of one of the other colors, it just gets us in the range where we can do fine tuning. When I decided to start my calibration from Warm1 instead of Warm2, I set the Color control based on green instead of red, because I knew I would have too much blue, not enough red, and I wanted a starting point that would give me adequate range for both.

I actually did try to do some interesting things with my Color and Tint, only to deal with a special situation. For my upscaling SD DVD player, I calibrate the set to the Rec601 colors. But that is also my multiregion player, where I would also like to use the PAL color gamut, without re-cabling the player to a different HDMI port. So I switched to Auto color space and adjusted the Color and Tint controls to minimize the errors relative to the PAL primaries and secondaries. Afterwards, in the custom color space, I can use the CMS to pull them back accurately to the 601 color points.

For the blu-ray player, I just left Tint at 50, and set the Color level to get the Y values in about the right range. In principle, the Color control would change whenever one adjusted the Contrast control, to keep the levels at about the same ratio. But there is no need since it provides just the rough level and the custom color controls provide all the fine tuning.

By the way, in the next few days I will try to strip down my experiment spreadsheet into something that is clearer and more user friendly, that just compares the actual measures to what the results should be with the optimized Y numbers.

FYI, this is the calibration that resulted in such perfect skin tones. (Not perfect black levels though.)

FYI, this is the calibration that resulted in such perfect skin tones. (Not perfect black levels though.)

Tom pretty much focused on the same areas that I saw when I first looked at this chart. The only thing he did not mention, but we may have talked about it before, is the blue is a little high on the low end of the grayscale. You have that characteristic downward slope in the blue level. I don't know if you can make that a little flatter by dropping the blue in the offsets, and if need be raising the blue gain slightly.

I say characteristic downward slope as I have spent hours trying to reduce errors in my grayscale measures, and I ended up calibrating starting from Warm1 instead of Warm2 just because that seemed to reduce the slope in the blue component of the grayscale above IRE30. But it has been a battle.

I do like how close your red point measures to the Rec709 red. Mine does not do that, even at 100% stimulus; I get close but not that close. I don't know if the difference represents manufacturing variation in the set or in the EyeOne probe.

Tom pretty much focused on the same areas that I saw when I first looked at this chart. The only thing he did not mention, but we may have talked about it before, is the blue is a little high on the low end of the grayscale. You have that characteristic downward slope in the blue level. I don't know if you can make that a little flatter by dropping the blue in the offsets, and if need be raising the blue gain slightly.

I say characteristic downward slope as I have spent hours trying to reduce errors in my grayscale measures, and I ended up calibrating starting from Warm1 instead of Warm2 just because that seemed to reduce the slope in the blue component of the grayscale above IRE30. But it has been a battle.

I do like how close your red point measures to the Rec709 red. Mine does not do that, even at 100% stimulus; I get close but not that close. I don't know if the difference represents manufacturing variation in the set or in the EyeOne probe.

I actually don't want to touch the color on this calibration. I'm thinking of taking these primary and secondary xy points and Y ratios and just making them my target, then trying to start from my other calibration that has great black levels and aim for these targets.

Like you say though, the real issue is probably white balance. But if I start from my other calibration that has great black levels and aim for these RGB (white balance) levels, that will just throw off those great black levels.

Dave, if you would like to play with it, I now have a more user-friendly spreadsheet to allow one use Excel Solver to optimize the Y values for colors. From my experimental model, I removed the comparisons against other strategies, and it represents just two situations, the status quo as measured, and what set of Y values will minimize the combined delta E of the primary and secondary colors. I did leave in the overall comparison of these two alternatives over all the saturation points, so one can get an idea of the overall effect, but this is lower down in the layout. All the critical information is now near the top, on the left, so one can see the actual and recommended information easily.

I did include my idea to show the sensitivity of each color's delta E with respect to Y, saturation and hue changes. These do illustrate why an optimizer is needed to find a solution. For the secondary colors, it is assumed that one can make Y be the desired value to minimize the error. But among the primary colors, if they are all a little bit off Rec709 in the x-y domain, adjusting Y to minimize the error on one does increase the error on the others. This follows from white luminance being the sum of the other three.

There is an option for which error measure to use. As Tom highlighted in a separate discussion, CIELUV76 seems to give very different advice for how to use brightness to adjust for saturation errors. In playing with the alternatives, it recommended significantly higher Y values for red than did CIE94. I have mostly implemented CIE2000, to see what impact that choice would have. Using CIE2000 gives optimized values close to CIE94, but not quite the same -- which is not surprising as it is based on CIE94 with tweaks in certain parts of the gamut.

I tried to implement something for your idea of evaluating using the actual measured white, instead of assuming the white point is D65. It is very tricky to try to keep straight when one uses the actual white and when one uses the D65 white to compare to Rec709. At least on my calibration, this did not have a significant effect on the recommended Y values of the 100% saturated colors, although it did change the errors on the colors of less than 100% saturation. It is interesting that it calculated mappings of some of the 0% saturated colors at points other than white, something that matches what I have sometimes measured on my set.

If you try it, please let me know if you find the instructions at the top of the spreadsheet sufficient.

Enjoy,
Bill

EDIT: The spreadsheet has been moved to a better discussion thread. It can be found here (http://www.avsforum.com/avs-vb/showthread.php?p=16381922#post16381922).

Yeah i'll check it out. But basically you input your readings, run the solver, then try to calibrate to the recommended targets, right?

I'm wondering (I hate to always throw problems into the mix)... even though this will preserve your White Y value, won't you then have to re-do your grayscale, then re-do this, going back and forth a few times? (And these affect optimal Contrast:Brightness setting.)

Yeah i'll check it out. But basically you input your readings, run the solver, then try to calibrate to the recommended targets, right?

I'm wondering (I hate to always throw problems into the mix)... even though this will preserve your White Y value, won't you then have to re-do your grayscale, then re-do this, going back and forth a few times? (And these affect optimal Contrast:Brightness setting.)

Yes, absolutely. Changing the color Y values can affect grayscale, and vice versa. If you are already close, I expect that Contrast and Brightness can stay the same.

With the normal measurement variation of the EyeOne, that is one of the things that frustrated me last night, going back and revising the grayscale then checking the colors again. Also, misunderstanding something Greg had written, I had temporarily lifted the restriction that RGB add up to white, which left me trying to fine tune one color to its target Y and not realizing that I was also de-tuning another color. So there are still iterations, you just don't have to iterate to find the perfect Y target, a difficult task to do by hand when one has to optimize all Ys together as a set.

Another use for the spreadsheet, of course, would be to calculate the effect of the ideal Y values, decide that those did not make enough difference to be worth the trouble, and just leave the set as it is. That cuts out a lot of iterations.

FYI, this is the calibration that resulted in such perfect skin tones. (Not perfect black levels though.)

Dave, in re-looking at this calibration, I'd say that it's very good measurement wise. The only thing I see which I think I mentioned before is the trailing off of green in the low of the RGB grayscale color balance. When you say "not perfect black levels", what do you exactly mean? Do things appear washed out in the very low end or crushed into black? If you add a bit of green to the low end color balance, this will increase luminance in the very low end in addition to adding some green thus will make things a bit brighter if the problem is crushed/too dark low end.

hope this helps,


--tom

Dave, in re-looking at this calibration, I'd say that it's very good measurement wise. The only thing I see which I think I mentioned before is the trailing off of green in the low of the RGB grayscale color balance. When you say "not perfect black levels", what do you exactly mean? Do things appear washed out in the very low end or crushed into black? If you add a bit of green to the low end color balance, this will increase luminance in the very low end in addition to adding some green thus will make things a bit brighter if the problem is crushed/too dark low end.

hope this helps,


--tom

In my other calibration (the one I feel has perfect black levels), black is truly jet black, but I also have amazing detail in all shadowy spots (like there are a million different shades of gray). Just seems to be the perfect balance - nothing sacrificed (black is black, not dark grey, but I still see all detail). I posted those settings in the calibration thread for my set. Others have tried them and commented that they cannot believe how good black looks - "like a Kuro". Maybe I just stumbled on magic settings, but I need to fix the skin tones, it's killing me.

In this calibration (the one with perfect skin tones), black is getting crushed - there are scenes where too much of the screen (like someone's whole black suit, or every part of the tires in a car commercial) is pure black, with no detail. i guess that is black crush. if i pause it and tick up the brightness, then i do see that detail, but of course then blacks become dark gray and it seems like the whole dynamic range is narrowed (no longer a million different shades of gray).

I'm hesitant to up the green offsets because my biggest gripe about skin tones is having a green tint on dark areas of faces (either shadows or facial hair). and my set seems very prone to do that (others have reported the same thing).