-

Jump to content



Photo
- - - - -

Must Read re Red Push and Grayscale (cut & drive) Enthusiastic Amateur Calibrators


This topic has been archived. This means that you cannot reply to this topic.
36 replies to this topic

#1 of 37 OFFLINE   Guy Kuo

Guy Kuo

    Supporting Actor

  • 581 posts
  • Join Date: Mar 06 1999

Posted March 18 2002 - 07:05 PM

Begin Rant Mode


I keep seeing this time and time again when people learn about red push problems, cut and drive controls in the service menu, put the two together in a flash of brilliance and try to fix red push by cutting down the amount of red using the red drive controls. Even if you are not currently thinking about doing something in the service menu, tuck the following tidbit away for later.

==================================================
Grayscale and Color Decoding are Not the Same Thing

Do not confuse grayscale (aka white balance or color temperature) with color decoding problems like red push. You must not change controls for one to fix the other. I frequently read of someone trying to correct red push (excess reddish color in red color objects) by displaying color bars and turning down red drive to fix the red. If you are doing anything like that or are contemplating such a manuever, don't. You've fallen into the newcomer trap of confusing the amount of red in the grayscale with the amoutn of red in red colored objects. You've gotten grayscale and color decoding confused with each other.


The video image is basically a grayscale one upon which colorization is added. The color of the underlying grayscale is a specific ratio of red, green, and blue. That ratio ideally creates a color of gray known as D65. If you alter the cuts and drives of a display, you are altering the color of the underlying grayscale image. That grayscale must be kept completely neutral or else any tinting you create would affect all images. If you drop the amount of red you decrease the amount of red globally. Even things which are supposed to be neutral gray would lose the amount of red in them.

Chroma or color information is decoded and used to colorize the underlying grayscale image. The user controls saturation and hue adjust how color decoding is done. If you visualize it in terms of a color wheel. Saturation changes the intensity of colorization. Hue basically alters the angle of the color wheel used to assign colors. Color decoding which causes red push simply means that colors near red are intentionally overemphasized. Note that altering chroma controls does not change the underlying grayscale's color. It changes the colorization which is added.

So you see you have two very different things which shouldn't be mixed up.

Grayscale
Color Decoding

The color of gray is examined using things which have no colorization like gray windows, ramps, and field patterns. The idea is to set the grayscale to a neutral D65 using the cuts and drives. Cuts affect mostly the dark end. Drives mostly the bright end. One uses a colorless test pattern because you don't want anything to change the on screen color except the color of gray. BTW, it is quite difficult to set grayscale accurately without some sort of instrumentation.

Color decoding is how the colorization information gets interpreted and then added atop the neutral grayscale image. The user controls can be used with color bar patterns to make adjustments. Color bars are useful because they contain fixed amounts of colorization which are equal for red, green, and blue (for each portion of the pattern that contains that color). Since the gray portion of the pattern has zero colorization but has a brightness of gray exactly composed of the amount of red, green, and blue which should be in the colored portions of the pattern. One can use the gray as an unchanging point of comparison as color decoding is adjusted. In some displays there are color decoding axis and gain controls which can be used to make a color decoder behave in a more NTSC standard fashion and eliminate red push. Not all sets have such controls.

If the difference between grayscale and color decoding is still unclear, try this analogy.....

You are in charge of art displays across the country and want to ensure that the art looks the same in every one of your galleries. To make that happen you choose a standard color of lamp (D65) for all your galleries.

Think of the color of gray (the grayscale) as the light of the lamp which is illuminating a painting. If the artist chose paint with too much red for faces (decoder has red push), a newbie art corrector could try "fixing" the excess red by reducing the red the lamp light. That would fix the excess red problem in faces but you no longer neutrally illuminate the painting. Everything would be badly tinted because you have changed the underlying white balance. That is what adjusting the cuts and gains does.

The real solution is to convince the artist to always use paints that actually look the right color under the neutral standard lamp (D65). That's basically what you do by adjusting the saturation, tint, and color decoder axes & gains correctly. You have selected paints which are the same color used by artists everywhere when they paint under a D65 colored lamp. This fixes the problem reds without ruining everything else.

A few artists still insist on using the wrong color of paint. In such cases you force the audience to stay in a bright red room to make them less sensitive to seeing red when they actually see the incorrtly colored painting (turn down the saturation). This hides the flaw, but doesn't really correct it.

I hope this makes it a little more clear why cuts and drives are not used to correct red push.


End Rant Mode
Guy Kuo
Director - Imaging Science Foundation Research Lab

#2 of 37 OFFLINE   Pep Guidote

Pep Guidote

    Agent

  • 27 posts
  • Join Date: Sep 08 2001

Posted March 18 2002 - 10:32 PM

Hello Guy,

I've seen threads where people use some kind of attenuator on the Red lead of their component DVD inputs. Does this strategy have any more merit than turning down the red drive controls?

My thinking is that this is almost the same thing, you'd be attenuating all the red uniformly therefore encountering the same tinting problems.

-Pep

#3 of 37 OFFLINE   Guy Kuo

Guy Kuo

    Supporting Actor

  • 581 posts
  • Join Date: Mar 06 1999

Posted March 19 2002 - 12:27 AM

The colorization data for the image is carried on the Pr and Pb component lines while the luma (grayscale picture) data is on the Y line. That means the red attenuator on the Pr signal is a reasonable fix for those displays which have red push. It alters the magnitude of the red difference signal and allows reduction of the red chroma problem without altering underlying grayscale. Unfortunately, it can only correct that particular component input. You still have to contend with red push on the set's other inputs and tuner.
Guy Kuo
Director - Imaging Science Foundation Research Lab

#4 of 37 OFFLINE   SteveK

SteveK

    Supporting Actor

  • 521 posts
  • Join Date: Jan 10 2000

Posted March 19 2002 - 09:14 AM

Guy- I understand your very helpful and informative post. I may have fallen guilty to the same manipulation in an attempt to reduce the red on my new Sony Wega, in that I've changed the AXNT from its default value of 1 to zero. I also fiddled with RDRV and RCUT but set them back to their defaults after not really noticing any difference. If I read your post correctly, I should set the AXNT back to its default value of 1 and find another way to limit the red that seems overstated on my Wega.

Do I interpret your post correctly?

Thanks.
Steve K.

#5 of 37 OFFLINE   Guy Kuo

Guy Kuo

    Supporting Actor

  • 581 posts
  • Join Date: Mar 06 1999

Posted March 19 2002 - 09:32 AM

Uh. Maybe. That is right if by the "red" you mean a general redness in the grayscale and not red push.

If you meant redness in fleshtones, then I think you have the controls backwards. Fix red push with AXNT which I believe adjusts the color decoder.
Guy Kuo
Director - Imaging Science Foundation Research Lab

#6 of 37 OFFLINE   Tony_Pat

Tony_Pat

    Agent

  • 44 posts
  • Join Date: Dec 28 2001

Posted March 19 2002 - 12:18 PM

Well, I used AXNT to fix the red push problem on my Wega (as you noticed in a different thread), but then here's the question: are there controls on higher end TVs that will allow a person to adjust color decoder accuracy? Is it even possible to resolve color decoder problems (without the sorts of tools used by ISF techs)?

For example, on the FV27 series of Wegas there is a service menu called VP, where the DRV and CUT controls are located. But there's also a PALETTE menu and a CLR-TMP menu. PALETTE would imply chroma type adjustments, and CLR-TMP would imply gray scale adjustments, but then why are the DRV and CUT controls not on the CLR-TMP menu? Posted Image

#7 of 37 OFFLINE   Guy Kuo

Guy Kuo

    Supporting Actor

  • 581 posts
  • Join Date: Mar 06 1999

Posted March 19 2002 - 12:48 PM

Color decoder adjustment options are pretty non-standard. We aren't really supposed to know they exist nor that there is non-standard decoding going on. One would have to know specifically for each model whether there are controls and how they are named and accessed. For the non-professional, some help might perhaps be found on another HT discussion site (the one with a spotted dog and broken down into about six pages of forums). Here is where an ISF professional's experience can help. A calibrator also has access to the ISF's body of knowledge about specific machines - something you can't find every day. Then again, not all ISF calibrators will deal with color decoding issues.

There is one more ignored portion of the color correctness story. The primary colors of the display also need to be standard. Otherwise it is still impossible to get the color right even with perfect color decoding and grayscale at D65. Orangish reds are all so common. It's been a while since I've played with it but a color correction filter over the red lens of an RPTV was useful in getting red closer to red rather than orange. I couldn't find the right intensity of filter gel so it tooks some time punching holes in the filter to get the right amount of effect. Would I do it for someone else's machine. Probably not, but for my own it was worth the effort to get real reds. The overall light output dropped after redoing white balance so light cannon fans might not appreciate the change.
Guy Kuo
Director - Imaging Science Foundation Research Lab

#8 of 37 OFFLINE   Tony_Pat

Tony_Pat

    Agent

  • 44 posts
  • Join Date: Dec 28 2001

Posted March 19 2002 - 01:39 PM

Well, I appreciate all the effort you put into explaining this stuff to us, Guy. One of my (serious) hobbies is digital imaging. I've spent countless hours calibrating my scanner, computer display, and printer, and know how frustrating it can be. And this with software and hardware tools (Q60 targets, spectrophotometer)!

So, if I stick with the instructions on Avia, is it better to adjust Tint using the blue bars if I turn off the red and green guns when I do it? Then, when I use the Color control on the red bars, should I turn off the blue and green guns?

#9 of 37 OFFLINE   Tony_Pat

Tony_Pat

    Agent

  • 44 posts
  • Join Date: Dec 28 2001

Posted March 19 2002 - 01:45 PM

Quote:
For the non-professional, some help might perhaps be found on another HT discussion site (the one with do and broken down into about six pages of forums).


BTW, is it rude of me to ask here on HTF which site you are referring to?

#10 of 37 OFFLINE   Guy Kuo

Guy Kuo

    Supporting Actor

  • 581 posts
  • Join Date: Mar 06 1999

Posted March 19 2002 - 01:48 PM

Yes, turning off the other two colors while reading color bars is more accurate than viwing through filters. The various versions of color bars in AVIA have the same color content in the patches and bars. They are simply rearranged to make it easier to compare the primary being examined against the 75% gray.

The included filters aren't 100% effective in eliminating the other two colors. Actually, even good quality glass color separation filters aren't perfect. We don't typically tell people to turn off the other two colors because most users simply don't have that ability, but you do so go ahead and do it that way.

BTW, dont' turn off all three guns - you'll have fun trying to turn them back on blind.
Guy Kuo
Director - Imaging Science Foundation Research Lab

#11 of 37 OFFLINE   Guy Kuo

Guy Kuo

    Supporting Actor

  • 581 posts
  • Join Date: Mar 06 1999

Posted March 19 2002 - 01:54 PM

The usual preference is not to lead people easily off this site. There are times when it is smart to do so because another site offers info unavaialable here. Sending people back to this forum is also something one does with indirect links for the same reason. Nobody wants to lose traffic. Unfortunately nobody has everything so some cross pollination is tolerated.

Sorry, I missed typing out the full "clue" to the site. I've since corrected my post.
Guy Kuo
Director - Imaging Science Foundation Research Lab

#12 of 37 OFFLINE   Tony_Pat

Tony_Pat

    Agent

  • 44 posts
  • Join Date: Dec 28 2001

Posted March 19 2002 - 02:05 PM

Ah, yes, well I'm a member of that site as well.

#13 of 37 OFFLINE   SteveK

SteveK

    Supporting Actor

  • 521 posts
  • Join Date: Jan 10 2000

Posted March 20 2002 - 01:24 AM

Guy- By redness, I am referring specifically to skin tones; or at least that is where it is most noticeable. It seems to have subsided somewhat lately, but previously skin tones were EXTREMELY red. For example, when I watched "Tombstone", almost everybody looked like they were badly sunburned! Hair color also seemed off, as nearly everybody had red or at least reddish-brown hair. The effect was most pronounced with the component input/DVD player, less noticeable on other inputs. After switching the AXNT from 1 to 0, skintones seem perfect when using the S-video input (DishNetwork) and quite good using the component input (DVD). Red is still more noticeable on dark scenes, however.

I'm thinking about recalibrating using the red filter with your Avia software to see if that will help. I currently have it calibrated using the default blue filter. I may also decide to have Greg Loewe calibrate my set in a couple months when he does his Nashville area tour. For now, AXNT set to 0 instead of 1 seems to be a significant improvement. Or maybe my set is just now getting sufficiently broken in. Either way, it's better than it was but not yet perfect.

Steve K.

#14 of 37 OFFLINE   RyanDinan

RyanDinan

    Second Unit

  • 251 posts
  • Join Date: Oct 25 2000

Posted March 20 2002 - 06:55 AM

Quote:
There is one more ignored portion of the color correctness story. The primary colors of the display also need to be standard. Otherwise it is still impossible to get the color right even with perfect color decoding and grayscale at D65. Orangish reds are all so common. It's been a while since I've played with it but a color correction filter over the red lens of an RPTV was useful in getting red closer to red rather than orange. I couldn't find the right intensity of filter gel so it tooks some time punching holes in the filter to get the right amount of effect. Would I do it for someone else's machine. Probably not, but for my own it was worth the effort to get real reds. The overall light output dropped after redoing white balance so light cannon fans might not appreciate the change.


Hi guy,

The color correction filter idea really interests me. I've noticed that Toshiba advertises red/green lens filters on their sets, which I assume do exactly what you're talking about.
What is the "standard" color of red/green filters we should be looking for? I imagine that not every "red" filter out there would be benificial. Do you want to get to (as close a possible) the SMPTE-C CIE xy coordinate of red? Maybe a photo store would carry such filters/gels? Also, why not filter blue as well? Is it just the fact that red and green are more visible, thus, need the correction?
Also, after such a filter is applied, shouldn't the grayscale be re-calibrated?

Thank you very much!

-Ryan Dinan

#15 of 37 OFFLINE   ThomasL

ThomasL

    Supporting Actor

  • 972 posts
  • Join Date: Mar 13 2001

Posted March 20 2002 - 07:23 AM

Steve, you can use Avia to figure out exactly which "red problem" is bothering you. After calibrating user color and tint settings using the blue filter, check, the color decoder pattern. Take a look at red and see how much red push you've got. If its above 10 percent, then this could be the problem. Besides setting that particular service parameter from 1 to 0, you could also just trying ticking the color setting down a bit. I don't know what the scale is on your tv so this may be one tick or 4 or 5. See what it does on the color decoder pattern.

In addition, you can go into the Grayscale pattern section and take a look at the vertical ramp grayscale pattern. Also, look at the individual window patterns for 10 IRE to 100 IRE. See if you can see any type of purplish tint to any of the vertical ramps and/or windows. If the grayscale is noticeably red tinted you'll see it with your eyes. I've also read that the older a set gets, the more its grayscale tends to drift toward the red. I've found this to be true. I took a look at a 11 year old 20 inch Panasonic set and it suffered from both red push in the color decoder and red tint in the grayscale. It was impossible for me to watch anything on it. If you see a noticeable red tint in the grayscale, then the correct way to adjust things is using a color analyzer and its associated software. But one can also use one's eyes to make the red a little less annoying. One simply has to be familiar with the grayscale parameters you can adjust for both the low and high end of the grayscale (they interact and are not discrete). I've found for my eyes, the low end always affects things more than the high end. I think perhaps our eyes notice differences in the lower end more so than the high end - i.e. it's harder to tell differences in shades of white. Next you need to know how the basic colors interfact on a palette. If you have too much red, what amount of green and/or blue do you need to introduce to neutralize that.

good luck,


--tom

#16 of 37 OFFLINE   Tony_Pat

Tony_Pat

    Agent

  • 44 posts
  • Join Date: Dec 28 2001

Posted March 20 2002 - 11:42 AM

Quote:
I'm thinking about recalibrating using the red filter with your Avia software to see if that will help. I currently have it calibrated using the default blue filter


Steve, please elaborate. Are you saying that - having already done the blue bars calibration with Tint - you will now do the red bars calibration with Color? That's what I did, and it seemed to work out okay. I'm just wondering if you had some alternative technique.

#17 of 37 OFFLINE   SteveK

SteveK

    Supporting Actor

  • 521 posts
  • Join Date: Jan 10 2000

Posted March 21 2002 - 12:24 AM

Tony- I did my current calibration using the blue filter. As you know, Avia also has the color bars test pattern for use with the green filter and for use with the red filter. What I'm thinking about doing is using the red filter and the appropriate test bars to set the color and tint level, rather than the blue filter with its color bars. My only concern is that since red appears to be oversaturated while other colors are correct, if I set the color and tint levels using the red filter that the reds will be correct but other colors will be understated. Theoretically, it should probably make little or no difference which filter/test pattern is used, but it could if one of the colors is inaccurate.

So I guess my best bet would be to go ahead and recalibrate it using the red filter and see what happens. If I don't like it (if colors other than red are too muted), I can always do the calibration again with the blue filter.

Actually, either I'm getting more accustomed to the redness or it is less pronounced than it was before, as skin tones now seem fairly accurate. Somebody commented in an earlier thread that too many people are accustomed to whites that are too blue, and perhaps that is the case for me. I've only had my Wega for about 6 weeks, so perhaps I'm adjusting to what may be a more accurate color presentation.

Steve K.

#18 of 37 OFFLINE   Tony_Pat

Tony_Pat

    Agent

  • 44 posts
  • Join Date: Dec 28 2001

Posted March 21 2002 - 12:01 PM

Interesting thought, Steve. I wonder what Guy would have to say about this approach. I remember reading somewhere why blue bars is the preferred method for setting tint, but I can't remember exactly why right now.

#19 of 37 OFFLINE   Guy Kuo

Guy Kuo

    Supporting Actor

  • 581 posts
  • Join Date: Mar 06 1999

Posted March 21 2002 - 12:08 PM

If you adjust using the red bars and red filter you'll get hue and saturation just right for red axis decoding.

If you adjust using green bars and green filter you get it just right for green axis decoding.

Basically, if the color decoder is NTSC accurate getting it right for one would get it dead on for the other two. By doing it with red on a set with red push, you will end up dropping overall saturation lower than if you do in blue. Somewhere between the two will be a "best" compromise. You can do it this way. I wanted a more systematic approach to be provided on AVIA so the Color Decoder test pattern was born. That lets you find the compromise saturation setting in a more quantifiable method than just going for somewhere between where saturation look perfect in red only vs blue only.
Guy Kuo
Director - Imaging Science Foundation Research Lab

#20 of 37 OFFLINE   Tony_Pat

Tony_Pat

    Agent

  • 44 posts
  • Join Date: Dec 28 2001

Posted March 22 2002 - 09:16 AM

But Guy, I thought there was some reason that blue was selected as the "best" choice for Tint calibration.





Forum Nav Content I Follow