AVS Forum banner
1 - 20 of 2069 Posts

· Registered
8,008 Posts
Discussion Starter · #1 · (Edited)
There have been several repetitive questions about how to use a Color Management System (CMS) to achieve accurate color for those who own a display that has one. The purpose of this post is to layout in as non-technical way as possible how to do this. Along the way, I'll explain how to set Color/Tint without using filters and how to set Brightness and Contrast without making subjective judgments against test patterns.

Equipment needed
  • Color Analyzer. This is a USB or serial device that you connect to a laptop computer and point at the display that measures color and light output. The most accurate device under $1000 is the X-Rite i1Pro 2, though it suffers from some practical limitations, primarily limited low-light sensitivity. For most users the X-Rite i1 Display Pro is the best color analyzer available. It costs about 1/4 of the i1Pro 2 offers MUCH better low-light sensitivity and includes a built-in tripod attachment and diffuser that you can easily swing in and out of the light path. For those on a budget, a less expensive option that does not perform as well as either X-Rite unit is the Spyder 5.
  • Calibration Software. You need this to interpret the data that the color analyzer receives. Aside from the very expensive professional tools, there are 3 good choices for amateurs and Prosumers: HCFR, CalMan, and ChromaPure. HCFR is freeware. CalMan is commercial software that costs about $200 and up, depending on selected options. ChromaPure is also commercial software of my own design which also costs $200 and up depending on options.
  • Test Patterns. Finally, you'll need some way to get a test pattern on the screen. The easiest way to do this is with a DVD or Blu-ray. AVS members have created a very nice set of HD patterns for Blu-ray and DVD players. Finally, I have created a calibration disc that provides all the patterns necessary for the steps in this guide that is optimized for ChromaPure. Other AVS users, such as Ted and Masicor have also created great full-featured discs that are available for a small fee. Just search this forum to find more information about these.
  • Light Meter (optional). I have found that a AEMC CA813 illuminance meter is very useful for front projectors and for all measurements that do not require color readings, such as contrast, black level, and gamma. It is an accurate, inexpensive standalone tool that is also very easy to use.
Once you have the items in this list, you are ready to calibrate your display.

First, some basic principles and terminology. Color performance is measured in three ways:
  • White Balance. This is the aspect of color performance that gets the most attention and is arguably the most important. It concerns the display's ability to provide a neutral shade of white. The ability of the display to do this all the way from darkest gray to the brightest white is called grayscale tracking, which is just white balance at multiple levels of intensity. If the display can't do this well, then all of the colors will look very unnatural.
  • Color Gamut. This is the range of colors that the display is capable of rendering. The gamut is popularly represented by a triangular pattern called a CIE chart or chromaticity diagram.

    This chart plots colors as xy coordinates. Color also includes luminance, which does not appear on this chart, so it must be represented separately. The xy coordinates that define the chart are the primary colors (red, green, and blue) and the white point. The secondary colors (cyan, magenta, and yellow) are derived from the primaries and the white point. All of these points have specific definitions for both standard definition, high definition, and Ultra High Definitionm. All commercially available video material is mastered according to these standards. If the display cannot reproduce the gamut accurately, then the image will suffer. If the colors provided by the display are poor, then the only way to fix it is with a Color Management System (CMS). A CMS can make a profound difference to the performance of the display, but most displays don't offer one and of the many of those that do not all work properly.
  • Gamma. This is how we measure the light output of the display throughout the grayscale. It is based on the amount of light the display produces at 100%. This is not a linear relationship (20% video is not 20% as bright as 100%).
It is important to understand that these aspects of color performance--though in some ways interactive--are for the most part independent. For example, you can have good gray scale tracking and an inaccurate color gamut. The bottom line is that each needs to be adjusted separately.

  • xyY - A common method for precisely measuring color performance. x and y are the coordinates that plot out a color on the triangular CIE chromaticity chart shown above. This graphically represents the established definitions of the color spectrum. Y is the luminance or intensity of the color. This is not plotted by the xy coordinates on the CIE chart.
  • Saturation - the colorfulness of the color independent of its luminance. A color's saturation displays on the CIE chart as the distance from the white point. Add saturation to a color and it will appear deeper and richer-red becomes crimson. Undersaturate a color and it will appear less colorful, but at the same level of intensity--red becomes pink.
  • Hue - the primary characteristic of color that allows us to distinguish one color from another. A color's hue is represented on the CIE chart by its angle to the white point. When a color's hue is off, its appearance will seem contaminated by other colors. For example, red that is too yellowish will begin to seem orange. Blue that is too reddish will begin to appear purplish.
  • Luminance - the intensity of color. The luminance of color (or white) can be measured by a simple light meter.
  • Color Decoding - This term refers to a process that is used to lower bandwidth requirements by encoding the native RGB signal into YPbBr (analog) or YCbCr (digital) and then decoding back to RGB prior to display. There are different encoding/decoding standards, so sometimes a poor design may lead to color decoding errors. These errors are primarily seen as primary colors with incorrect luminance and/or secondary colors with incorrect hues. All commercial displays include a Color and Tint control. These are basically color decoding controls, though their effectiveness is extremely limited because Color adjusts chroma (chroma is a combination of saturation and luminance) of ALL of the colors and Tint effects hue of ALL of the secondaries. The problem is that displays with color decoding errors effect the colors differently. For example, you could adjust Color/Tint to get the correct luminance of blue and the correct hue of cyan, but then the luminance of green and red may still be inaccurate. You could adjust the color control to get red right, but then blue and green would be inaccurate. See the problem? A full set of color decoding controls addresses this problem by offering color/tint controls that operate on red/magenta and green/yellow independently. Then you can use the main Color/Tint controls to adjust blue/cyan. With the death of analog television, color decoding has become less of an issue and color decoding controls have become less common, though some digital TVs can exhibit color decoding errors
For all practical purposes the color performance of a display device can be adequately described by the three characteristics defined above--saturation, hue, and luminance. These are abstract concepts and sometimes a picture really does say a thousand words, so here are examples that illustrate the three characteristics of color.




In each of these examples, the green on the right is adjusted by approximately the same amount in the direction of first lower saturation, second yellowish hue, and third lower brightness.

How are these concepts related?
The xy coordinate of a color on the CIE chart establishes its saturation and hue. The Y value establishes its luminance. If a color deviates from the reference point by appearing shifted towards other colors on the chart, then its hue is wrong and needs correcting. If a color is shifted closer to or father from the white point on the chart relative to the reference, then its saturation is wrong and needs correcting. Finally, if the color is too intense or too dim relative to the establish standard (not shown on the chart, but determined mathematically), then its luminance is wrong and needs correcting.

Color Definitions

Of these definitions, only the xy coordinates for the primary colors and white point are absolute. The secondary colors and luminance values are mathematically DERIVED from the primaries and the white point. If your primary colors measure according to these standards, then this list correctly states the proper specifications for luminance and secondary hue/saturation. However, a different set of primaries establishes different luminance and secondary color hue/saturation targets. The math required to figure out these relationships is too complicated to go into here, but good calibration software should take all of this into account.

Although there are no hard and fast rules about this, I would make color adjustments in the following order:

  1. Black and White levels
  2. White Balance
  3. Gamma
  4. Color gamut
When finished, go back and remeasure these parameters, because changes in one parameter may have affected the readings for another.

What's wrong with the ISF description of color?
I am a graduate of the ISF seminar, and I think that the organization has performed a valuable service at educating the public about the importance of accurate video. However, the ISF description of color is not entirely clear.

An often-repeated claim by ISF literature is that the characteristics important to image quality can be ranked in the following way:
  1. Contrast
  2. Color Saturation
  3. Color Accuracy
  4. Resolution
Ranking resolution and and contrast in this way seems about right to me, but what in the world is meant by "Color Saturation" that is different from "Color Accuracy"? As described above, all of the standard gamuts have specifically-defined standards for saturation, hue, and luminance. So saturation is just one aspect of color accuracy. Perhaps, this point refers a to the fact that a lot of people prefer oversaturated primary colors. If so, then this is an endorsement of color inaccuracy! Surely, ISF doesn't mean this. Perhaps it refers to a purely subjective quality of color that has not been quantified by the established standards. However, color is a fairly well-understood phenomenon. There is, so far as I know, no important aspect of color beyond the characteristics of saturation, hue, and luminance that remains unexplained.

Thus, it seems that "Color Saturation" is either a ghost or simply a poorly-expressed reference to something already known. Some of the statements I have heard ISF personnel make suggests the latter. I think that this may just refer to the intensity of color. If so, then the reference is redundant. Intensity or luminance (I use the terms interchangeably) is just one aspect of color accuracy, along with gray scale performance, hue, and saturation.

Thus, the ISF rank of characteristics that are important to image quality really just boils down to:
  1. Contrast
  2. Color accuracy
  3. Resolution
Furthermore, even this revised list is not, I think, quite right. It leaves out two very important aspect of image quality: sharpness and clarity. These concepts are not the same as resolution. Two 1920x1080 displays (equal resolution) can and often do exhibit different degrees of sharpness and clarity. So, what do these concepts mean?

  • Sharpness is the quality of an image that gives it clearly defined boundaries. This should not be confused with the type of artificial sharpening that you often see in poor DVD transfers or with an excessive use of the sharpness control on the display. These only result in ringing and edge enhancement, which makes the image worse instead of better. A good example of sharpness can be had by comparing a good digital display with a good CRT. CRTs can look very nice, but they simply cannot compete with the sharpness of digital displays.
  • Clarity is the quality of an image that appears when the image is free of artifacts. These artifacts come in a wide variety of types and for a wide variety of reasons. They include: poor focus, poor geometry, poor convergence, chromatic aberration, ringing, moire, line twitter, and interference/ghosting/snow (for over-the-air broadcast).
Problems with image clarity are result of the quality of the optics, processing, or mechanical alignment, and involve the display adding something to the image that is not originally there. Problems with sharpness can be influenced by the resolution, optics, and other inherent properties of the display device, and involve the display removing from the image something that was originally there.

The importance of these two factors is perhaps best illustrated by thinking back to the first time you saw a good plasma display fed by a good source. Compared to CRT images, plasmas could produce an image with startling clarity and sharpness that results in an almost scary looking-though-a-window quality that CRTs simply could not match. The important fact to note for this discussion is that these relatively early plasma displays offered lower resolution, much lower contrast, and often worse color accuracy than the CRTs of the day and they still could look better, sometimes much better.

Another important factor to consider, especially when comparing the importance of sharpness/clarity to contrast, is that these qualities are persistent. This means that they are decisive factors in image quality all of the time. The same cannot be said for contrast. Many types of common images--such as brightly illuminated live sports--are relatively unaffected by the contrast of the display. Contrast becomes increasingly important as the image gets darker, and is thus not a persistent characteristic of image quality, though still a very important one. The same can also be said of color accuracy. Color representation is a persistent quality of the image. If faces are red and the trees glow with a neon green, good contrast won't compensate for this.

For this reason, I would rank the elements that contribute to overall image quality in the following way:

  1. Sharpness/clarity
  2. Color accuracy
  3. Contrast
  4. Resolution

Why can't I fix over-saturated colors by simply turning down the main Color control?
This issue comes up often in the context of popular displays that exhibit a strongly over-saturated gamut. The original JVC RS1/2/10/15 front projectors offer perhaps the best example of this.

Lacking a full-featured CMS, one is tempted to try to alleviate the problem of oversaturation by simply turning down the main Color control. Turning it down slightly may help somewhat, but anything more than a very small adjustment is likely to make the color worse rather than better. Why? The reason has to do with the fact that, contrary to popular belief, color controls are not engineered to adjust saturation. They are Chroma gain controls. Turn Color up, you increase the chroma of the signal. Turn the Color down, and you decrease the chroma. Although related, chroma and saturation are not the same.

Perhaps the best way to think of the difference is this: Imagine a red patch of color illuminated under a strong, bright light and then imagine the same patch seen under a dim light. As you change the lighting conditions, the red appears more or less colorful. This is chroma.

Interestingly, the reverse is not true. If you lower the saturation of red, the chroma decreases to approximately the same degree. A less saturated red seems proportionally less colorful, but a less colorful red is not necessarily proportionally less saturated. Consider the two examples below.

Example 1: Chroma change

Example 2: Saturation change

The first example mimics the effect of turning down the main Color control. If you turned the Color control all the way down to zero, the the patch would finally lose all of its colorfulness (and saturation) and retain only some residual brightness, appearing as a shade of gray.

The second example mimics what occurs when we decrease saturation using a CMS. The luminance stays relatively constant, but it loses colorfulness.

This should make clear why turning down the main color control is not a good strategy for addressing over-saturated colors. What this does is similar to what you see in Example 1. It will reduce the saturation of the colors, but it will also significantly reduce their intensity. What we need is what is simulated in Example 2.

However, the main color control CAN be a good tool for addressing color decoding problems. Unfortunately, it works equally for all of the colors, when what is generally needed is color-specific adjustment.

Note: "Chroma" is a term that has somewhat different meanings depending on the context. Those familiar with video engineering will understand chroma to refer to a rather general concept of color. Video signals contain chrominance and luminance. However, in color science "chroma" has a more specific meaning, which is "colorfulness relative to a similarly illuminated white." Color scientists use the term "colorfulness" to refer to what video engineers refer to as chroma.

Luminance vs. Illuminance
You measure the intensity of a display differently depending upon whether you have a direct view/flat panel/rear projection display or a front projector. For direct view/flat panel/rear projection displays, just point the color analyzer at the screen and measure directly. The software will measure either in imperial fL (foot-lamberts) or in metric cd/m2 (candelas per meter squared or nits) units. If you measure in nits, just multiply by 0.2919 to get fL. If you measure in fL, then divide by 0.2919 to get nits. Nits and fL are both units of luminance, which is an emission or reflection of light from a flat, diffuse surface. All color analyzers natively measure luminance.

If you have a front projector, it is a little more complicated. You can still measure luminance by pointing the color analyzer at the screen. However, you can also measure illuminance by pointing the color analyzer directly towards the projector's lamp. Illuminance is a measurement of light that falls on or illuminates surfaces. Thus, while reading light off the screen would be a luminance measurement in nits or fL, measuring light directly from the projector's lamp would be an illuminance reading in Lux. Front projectors are about 1/3 the brightness of a typical SDR flat panel, thus the black level measured off the screen is very, very low. Unless you have an expensive luminance meter, such as the Konica Minolta LS-100 which can accurately measure very low luminance, you will probably get a more accurate reading by taking an illuminance measurement directly from the lamp. The AEMC meter cited at the beginning of this tutorial is a good choice. The i1 Display Pro colorimeter allows you to do this as well. All you have to do is swing the built-in diffuser into the light path. You can adjust the amount of light that reaches the sensor by adjusting the distance of the sensor from the lamp.

Illuminance meters such as the AEMC unit use lux as the unit of measurement. However, you can convert between lux and luminance in the follow way: Just place the meter against the screen facing the projector's lamp and measure a test pattern in Lux. Next, divide Lux by 10.76. Finally, multiply by the real* gain of the screen to get the fL for the projector. To get the lumens of the projector's lamp, just multiply the lux by the screen area in square feet and then divide by 10.76.

* Note: a screen's real gain will often be lower than its advertised gain. Manufacturers routinely inflate a screen's gain rating. Stewart is the only company I know of whose gain ratings are reasonably accurate.

ΔE Color Difference
The purpose of ΔE (dE or Delta-E) is to provide a single number that we can use to grade color accuracy relative to some standard. The smaller the number, the more accurate the color. ΔE can be used for both gray scale and primary/secondary color evaluation.

ΔE is based on one of two color appearance models, Luv or Lab. Both of these models were adopted by CIE in 1976 and they yield slightly different results. Luv numbers scale a little higher and place a greater emphasis on red, while Lab numbers place a greater emphasis on blue. In 1976 when CIE was considering the adoption of a color appearance model that offered a more perceptually uniform standard, CIE had originally wanted to adopt Lab only, but the industries that CIE represents argued against a Lab-only solution. They were concerned by the fact that Lab fails to offer a linear chromaticity diagram, such as the CIE chart shown at the beginning of this post. For this reason, historically many video ΔE values have been expressed in Luv (which does offer a linear chromaticity diagram in u'v' units). However, since 1976 most of the research on the CIE system has relied on Lab only.

In 1994 CIE adopted another even more perceptually uniform standard--based exclusively on Lab--that is referred to as CIE94. It scales much smaller and reduces somewhat the 1976 Lab emphasis on blue. It also offers an easy analysis of the Chroma, Hue, and Lightness components of color error that can be useful. Finally, the CIE94 formula treats lightness (a perceptually uniform version of luminance) very differently than the 1976 color difference equations. Both Lab and Luv 1976 models predict that you can substantially reduce perceived color error caused by over-saturation by simply lowering the lightness of the over-saturated color. According to the CIE94 formula, lowered lightness does NOT mitigate the effect of oversatuturated color. Rather, it just makes the color appear darker. So, which is correct? To my eyes the CIE94 model gets it right, but there are many in video industry that continue to rely on CIELUV.

Since 1994, there has been much additional work, and in 2000 CIE adopted yet another ΔE model, known as CIEDE2000, but it is a VERY complicated formula that has been most widely adopted by the textile industry. Future work points to a new universal standard, the latest version of which is CIECAM02. However, a color difference formula for CIECAM02 has not been officially endorsed by the CIE. So, for now, we are probably best served by the 1994 or 2000 dE models. I use CIE94 for all dE reporting, but to a large extent this is a matter of personal preference.

I provide a spreadsheet at the bottom of this tutorial that allows you to calculate CIELUV, CIELAB, or CIE94 ΔE values using SMPTE-C or Rec. 709 values against your own test data.

So, how do we measure color performance?
For years the most popular method of specifying the color of white has been in terms of color temperature, with 6500K being the specification for neutral white. In recent years, the inadequacy of this approach has become evident. The great weakness of color temperature as a specification of the color of white is that it assesses only the relative strength of red and blue—reddish whites yield a lower color temperature and bluish whites yield a higher color temperature. This ignores green-magenta axis entirely, which means that a very greenish white or magenta white could both be 6500K. Furthermore, color temperature is useless in any case when assessing the accuracy of primary and secondary colors. There is really no reason to continue using color temperature as a metric of color performance of any kind.

ΔE offers a much better approach. SMPTE has established a standard for the color accuracy of Digital Cinema, which is 4 Lab (1976) units or less. (This is approximately equivalent to 1.5 CIE94 units for color.) This seems like a reasonable tolerance. Unfortunately, we are left with the problem discussed above: which ΔE standard are we to use? SMPTE offers no guidance as to why they selected CIELAB.

Consider this oversaturated, but dim, shade of green:

x0.296, y0.678, Y0.535

How far from the Rec. 709 standard does this green deviate? Using ΔE as a guide it is very hard to say. CIELUV reports that this green has a ΔE of 11.4, which, though far from perfect at just under three times the allowed color error, is not horrible. However, CIE94 reports that the same green exhibits a 1994 ΔE of 11.2, which is over seven times the allowed color error. This is a huge error. Two ΔE systems report radically different results for the same color! For me this is an excellent example of why the 1976 ΔE standards are now obsolescent.

Thus, although ΔE remains a very useful tool, it is useful to supplement it with another measurement of color error. % deviation from specification is a good choice. Using this standard, we can add to our ΔE number the following information about the green above is:

Lightness: -10.9%
Saturation: + 17.3%
Hue: +0.1%

Add to this the requirement that no color should exceed +- 2% error in lightness, saturation, or hue. Even this method is not perfect. The human eye is not equally sensitive to lightness, saturation, and hue errors, nor is it equally sensitive to errors in each of the primary and secondary colors. For example, red errors are much more easily noticed than blue errors. ΔE tries to accommodate these factors, but as we have seen the different ΔE formulas yield different results.

Test Patterns
Use window test patterns for CRT, plasma, and OLED. For everything else, you can use either windows or full fields. NOTE: This applies for SDR material. For HDR material you should always use window test patterns.

Setting White Level (Contrast)
The Contrast control determines your display's light output. Set too low you lose image punch and lower contrast ratios. Set this too high and you lose color accuracy and detail in bright scenes and you may suffer from eye strain.

The standard method for setting Contrast requires that you look at a test pattern that has a just-below-white stripe against a white background. You are supposed to set Contrast as high as you can without losing the ability to distinguish the just-below-white stripe from full white.

However, there are a couple of problems with this method.
  • Some displays, especially LCDs, will never suffer from loss of high level detail even with Contrast set to 100%. This method will recommend a setting that is much too high.
  • This method does not take into consideration color performance. Many displays will lose their ability to track a neutral white at high output levels with Contrast set as high as this method recommends.
Thus, I think that a better method for setting Contrast is to just set it at a level consistent with good color performance and reasonable light output for a given display device. What's a reasonable level?
  • CRT tubes: 30-40 fL
  • Plasma: 30-40 fL
  • LCD flat panel: 30-40 fL
  • Digital rear projection: 30-40 fL
  • Digital front projection: 12-16 fL

Setting Black Level (Brightness)
The typical method for setting black level is to use a pluge pattern that displays just above and just below black information against a black background. You set brightness so that the just-above-black is barely visible and the just-below-black is invisible. The closer the just-above and just-below information is to video black, the more precise the adjustment.

Here's another method that works well. Put up a 0% field on`1qqqqqqq the display. Turn brightness up several ticks. Now slowly start turning it down one click at a time. At some point you will find that lowering the brightness control further will not make the screen visibly darker. That is the correct setting. This can also be used to set black level for broadcast sources. If you have a DVR just record a fade-to-black sequence. Then play that back and pause it. Use this as your 0% test pattern.

Setting Gamma
Many displays offer gamma presets. Some displays even offer the ability to adjust luminance at various video levels, which is even better. Gamma simply specifies what level of luminance any color should produce relative to the % stimulus of the signal and the measured luminance at 100% video. For example, if 100% video measures 120 cd/m2, then at 20% video and a 2.22 gamma you should measure 3.37 cd/m2 or 2.81% of 120. Why 2.81% rather than 20%? Remember, I mention early in the article that gamma response is not linear. This is because the human eyes sensitivity to changes in luminance is not linear either. We are much more sensitive to small changes in luminance at low levels of light than we are at very high levels.

Gamma is primarily a choice about shadow detail vs. darkness of blacks. A 2.5 gamma will give a higher contrast ratio and deeper blacks, but it will also reveal less information in dark scenes. I recommend a gamma in the 2.2-2.35 range.

The calibration software should offer the ability to select any gamma point target and more than one method for calculating gamma. In addition to the standard power law method, there is also sRGB and BT.1886.
  • First, ensure that you have already calibrated white level, black level, and the grayscale. These steps alone can often get you a long way towards a good gamma response.
  • Experiment with different picture presets. Use the preset that offers the best gamma response. "Best" is generally defined as falling within a range of 2.2-2.35.
  • Experiment with different gamma presets. Some displays offer a gamma selector that is independent of the picture preset. Just select the one that offers the best response.
  • Best of all, a very few displays will allow you to directly calibrate gamma by changing the output at each level of stimulus. Again, select a value at each level that results in the best gamma response.

Setting Sharpness
This one is simple. Just use the sharpness pattern to look for ringing or faint outlines along the edges of the horizontal and vertical lines in the test pattern. Set the Sharpness control to the highest point you can that minimizes ringing (you may not be able to eliminate it entirely). On some sets, the sharpness should be set to zero. But for many it is somewhat higher.

Setting Color/Tint
The standard method for doing this involves looking at a SMPTE color bar test pattern through a blue filter. This method has 2 drawbacks. First, at best it is an approximation of the correct setting. Second, and more importantly, for some displays it simply does NOT work. On some plasmas in particular I have noticed that this method will recommend a grossly inaccurate setting. Here's a foolproof method for setting Color/Tint that does not use filters.

  1. Point the color analyzer or light meter towards the screen and display a 100% white test pattern.
  2. Measure the Y value (luminance) of white.
  3. Display a 100% Red test pattern, and measure the Y value here as well.
    You will notice that as you move the Color control up and down, the Y value of Red increases and decreases, but white stays the same.
  4. Set the color control at the point where Red measures closest to 21% of the white reading.
  1. If you have not already done so, adjust the white balance and get it as close to neutral (x=0.3127, y=0.329) across the entire range.
  2. Point the color analyzer towards the screen and display a cyan test pattern.
  3. Put the Tint control at its neutral mid setting.
  4. Use the software controls to plot cyan on a CIE chart.
  5. Adjust Tint up or down until the reading places the hue of cyan as close to the target on the CIE chart as possible (it is useful if the software has a continuous reading mode, so you can see changes you make to Tint in real time).
  6. If you had to substantially adjust Tint from the neutral point to get an accurate hue of cyan, then check the other secondaries. You may have to select another setting that gets all 3 secondaries as close to correct hues as possible.

Adjusting white balance

Unlike a good CMS, which is rare, virtually all displays have white balance controls. Sometimes they are in the user menu and sometimes they are buried in a service menu that can only be accessed by a specific key sequence on the remote (this is much less common than it used to be). The goal is to get an xy measurement as close as possible of x0.3127, y0.329. The calibration software will provide these raw numbers and some type of graphical representation of RGB balance relative to the target gamut. Ideally, you would like to see red, green, and blue all balanced equally at 100%. That is the definition of neutral white for the selected gamut.

To calibrate the white balance:
  1. Aim the color analyzer at the display.
  2. Display a 80% video test pattern.
  3. Adjust the RGB Contrast controls until RGB is balanced at 100% or until you read x0.3127, y0.329.
  4. Display a 20% video test pattern and use the RGB Brightness controls to balance RGB at 100% or achieve x0.3127, y0.329.
  5. Repeat the last two steps as many times as necessary until both the 80% video and the 20% video patterns measure neutral gray. This may take several sets of measurements.
  6. Finally, take an entire series of white balance measurements--this is the grayscale--at 5% and then at 10% intervals up to 100%.
Sometimes you may find that even though 80% and 20% are a neutral gray, the mid range 40-60% is not. This means that your display won't track a good grayscale and you have to make some compromises. The general rule of thumb is to focus on getting the mid range to track neutral gray. Then get the low end right. Sacrifice accuracy at the top end if you have to.

Note: There is no industry-wide accepted terminology for white balance controls. You may see RGB Contrast/Brightness, RGB Gain/Bias, RGB Gain/Offset, RGB Drives/Cuts. They all mean the same thing. Contrast, Gains, and Drives are for adjusting the high end of the grayscale. Brightness, Biases, Offsets, and Cuts are for adjusting the low end of the grayscale.

Adjusting Color using a Color Management System (CMS)
  1. Point your color analyzer towards the screen and display a white test pattern, and then take a reading.
  2. Display a red test pattern, and then take another reading.
  3. Repeat the previous step until you have measured all of the primary and secondary colors.
  4. Use the software controls to plot the Lightness, Saturation, and Hue of all of the colors against the selected gamut reference, usually Rec. 709.
  5. Adjust these values for each of the colors, one by one, using the CMS until Lightness, Saturation, and Hue line up as close as possible to the references for your target gamut. It is helpful if the software has a continuous reading mode so your changes can be viewed in real time.
Note 1: You probably won't be able to get all colors lined up perfectly, but get them as close as you can.

Note 2: Some software only plots changes that are visible on the CIE chart. This allows you to get saturation and hue right, but it doesn't tell you how your changes affect the luminance of the colors. Unfortunately, some CMSs automatically change the luminance of a color as you adjust its saturation. This will give you a good looking CIE chart, but you could actually end up with LESS accurate color than when you began.
Note 3: The human eye is not equally sensitive to all colors and all color differences. For example, it is more important to get red and green right than blue. It is also more important to get correct hues than correct saturation.

Note 4: It is probably not a good idea to calibrate your color using 100% saturation and 100% intensity test patterns. These patterns simply do not reflect the vast majority of video content the display will reproduce. Getting adjustments correct for these will often leave a majority of the color gamut substantially inaccurate. It is better to use 75% saturation and 75% intensity test patterns.

That's it. Now you should go back and remeasure the white/black levels, grayscale, gamma, and color gamut. There may have been interaction between the various adjustments and you may have to go through two or three rounds of measurements until all are correct.

What you also need
All displays include adjustments for Color, Tint, Brightness, Contrast, and Sharpness. You can use test patterns from calibration DVDs to correctly adjust these parameters. However, to calibrate a display you must go beyond these basic adjustments and that requires two additional items:
  • Calibration equipment (software and color analyzer), which I discuss at the beginning of this post.
  • A display that has controls that go beyond the basic user adjustments listed above. This is the single biggest factor in getting a good calibration. Without the necessary controls, the best software and color analyzer available will do little good.
In addition to the basic user controls, most displays have white balance controls, though they are sometimes hidden in the service menu and sometimes labeled with obscure nomenclature. Sometimes manufacturers offer even less.

Color decoding and the resulting color gamut errors occur in modern displays and the great majority lack controls to adjust these critical parameters correctly. Thus, before you consider whether you want to get your display calibrated, you should first ensure that it CAN be properly calibrated with the controls it offers.

With this in mind, I thought it would be useful to list the important set of calibration controls that you should look for. At a minimum this includes 2-point white balance for independent control over the bright and dark ends of the grayscale, and at least one of the following:

  • gamma adjustments
  • a full-featured 3-axis CMS for all primary and secondary colors


· Registered
1,324 Posts
Cool, thanks for the step by step. I can't wait to try these method. I have always found it difficult to use the blue filter to adjust the color and tint.

· Registered
8,008 Posts
Discussion Starter · #5 ·

Originally Posted by grinchy /forum/post/0

Can you address the rationale behind using "window" patterns for non direct-view displays?

Some types of displays, CRTs and plasmas come to mind, lose the ability to reproduce higher levels of output across the whole screen as increased voltage is applied. You can get around this problem by using a test pattern in which the gray levels are reproduced in relatively small windows, rather than the full screen.

For other types of displays this is not really a problem. However, since it is absolutely required for some direct views, it was just simpler to describe a single procedure. Also, relatively small gray windows throw less light into the room than full screens of gray, lessening the possibility that room reflections may contaminate the readings.

· Registered
1,324 Posts

Originally Posted by TomHuffman /forum/post/0

Color Definitions (HD)









What about Color Definitions for SD (Rec 601)? I can get xy from HCFR but not Y. Adjusting for SD would have some different percentages wouldn't it? Are the Y's the same for HD and SD?


· Registered
1,324 Posts
I found this:

Rec. 601 (SMPTE C) for 480i, 480p

White (D65) = 0.3127, 0.3290

Red = 0.630, 0.340

Green = 0.310, 0.595

Blue = 0.155, 0.070

Yellow = 0.421, 0.507

Cyan = 0.231, 0.326

Magenta = 0.314, 0.161

but I am still missing Y.

· Registered
1,324 Posts
Thanks, Tom. You rock!! So the percentages would basically be the same for SD and HD, correct?

· Registered
6,804 Posts
Really nice guide Tom, I have one question.

Originally Posted by TomHuffman /forum/post/0

[*]Set the color control at the point where Red measures closest to 21% of the white reading.

Why do you use the saturation control to set Y to spec instead of setting x,y to spec? My display's saturation control changes the spectral distribution of the primary so I use it to set red at the Rec709 x,y point. In other words is the perceived result of deltaY=0 better than deltax=deltay=0?


· Registered
8,008 Posts
Discussion Starter · #11 ·

Originally Posted by zoyd /forum/post/0

Why do you use the saturation control to set Y to spec instead of setting x,y to spec? My display's saturation control changes the spectral distribution of the primary so I use it to set red at the Rec709 x,y point. In other words is the perceived result of deltaY=0 better than deltax=deltay=0?

Zoyd, this is something that may vary among displays. However, in my experience, the main Color control primarily affects color luminance and only secondarily saturation.

· Registered
6,804 Posts

Originally Posted by TomHuffman /forum/post/0

Zoyd, this is something that may vary among displays. However, in my experience, the main Color control primarily affects color luminance and only secondarily saturation.

I'm not sure how one would determine primary vs. secondary effects when using a color control since the two are coupled. Using the i1pro in spectrum view I can see both things happening and on my display, the spectral shape changes in the wings and the intensity changes. Here is a plot using a green pattern;

Note that the % change in the red and blue side peaks are much greater than the change in the green peak.

The before and after green coordinates are:









So there was a 4% change in x, 12.5% change in y, and 16% change in Y. I guess the real question is what is the deltaE with respect to Rec709 in an appropriate perceptual colorspace that includes luminance as well as chroma.

· Registered
11,213 Posts

check out the front projector forum - Hugh2/hughman did that very study of changes to color control impact on the CIELCH measures (same dE as CIELUV) of his JVC - but it is rumoured the firmware version may behave slightly different. I would say it is generally the case that you have to shift color brightness significantly before color saturation moves with video color controls - he posted several sets of numbers on that starting here....


I have observed similar with Sony Ruby/Pearl - the dL hit to move in the primaries with the color control per spec resulted in much worse dE than leaving primaries be and only shorting dL a bit under spec.

Of course Tom's post was how to use a CMS and video decoder that is capable of getting you the perfect result - as his other posts on color demonstrate - sometimes this can be an exercise in futility in picking the worst vs. best compromises.

The only thing I would add is to repeat the measures with 100% patterns - as sometimes displays can have settings impacting bright colors like DLP's BrilliantColor(tm) or any dynamic iris - which maybe can be turned off - but if not - compromise between light/dark colors time!

Also if doing tint with this method of sensoring rather than filtering - I would do it after greyscale not before - because secondaries shift with greyscale - with filters this is irrelevant because you have bluescale not greyscale!

Using red brightness to set color levels will only work if your set has no red push - if it does then the color will probably be too low overall - so again measure blue/green levels to find the compromise in between.

I also use 100% white for greyscale to first establish my clipping limits - I find hot white colortemp blowouts much more annoying than the tweak to contrast they give.

· Registered
6,804 Posts
Great thread, thanks for the link. If I understand your comments regarding dL at different luminance levels it means the perceptual errors are actually a function of where you are in the luminance curve, is that correct? If so then optimizing dE or (dL,dC,dH) will depend on this too.

· Registered
8,008 Posts
Discussion Starter · #15 ·

· Registered
6,804 Posts

Originally Posted by TomHuffman /forum/post/0

As I said, this may vary among displays. In any case, using a ratio of white/red Y is a very easy and accurate way to set Color properly, certainly better than using a filter and a SMPTE color bar, which was my only point.

yes, I agree it's superior to the filter method but as always I'm interested in squeezing dE as much as possible. I repeated your red test from those links and I see similar behavior on my display but not quite as severe a hit in Y to get my red primary aligned to rec709. At color=-2 Yred=0.21*Ywhite, at color=-5 I get alignment in x,y and Yred is 18% low.

As for dE(LCH), here are the results for my display:

color=-2 (Yred=0.21*Ywhite), dE=8

color=-3 dE=4.4

color=-4 dE=5.7

color=-5 (aligned to CIE,Yred=0.17*Ywhite) dE=12.4

This agrees with krasmuzik's comment about shorting the color brightness slightly to get the best perceptual results(assuming you're starting from outside the triangle). I was also surprised to see that dE was such a strong function of L*

· Registered
11,213 Posts

Originally Posted by zoyd /forum/post/0

Great thread, thanks for the link. If I understand your comments regarding dL at different luminance levels it means the perceptual errors are actually a function of where you are in the luminance curve, is that correct? If so then optimizing dE or (dL,dC,dH) will depend on this too.

That depends on what white reference you used - if you used 100% as your white reference then yes for anything less - dE models perception that color error perception is greatly dependent on relative luminance.

You could also use the same white reference as your color bars - make everything 75%. Then it tells you what errors you see in the color bars since you adapt to its white as your reference.

But if you wanted to quickly compare 100% colorbars to 75% colorbars - now 100% white becomes your reference - and the 100% colorbars will appear to have the worst errors.

Literally it is all a matter of perception. When I plot dE for greyscale I could go eitherway. If I look at a grayscale chart I cannot see near black color deviations, however if I look at grayfields sequentially and adapt my eye to near black (making it my "white" reference) - then I can see the errors.

The perceptual model does not consider night blindness - so obviously there is a limit you can see color error.

· Registered
1,324 Posts
I am trying to get to rec601 (just 480p dvd player)

100% stimulus values:

Yred = 4.64 divided by Ygray=23.005 I get 0.201

75% stimulus values:

Yred=2.543 divided by Ygray=9.543 I get 0.266

I adjusted the color down from +9 to -10 and got the following result for 75%

Yred=2.342 divided by Ygray=10.204 is 0.230

It appears I am headed in the right direction. It just seems like a lot of clicks down (for color) to get a slight change in the Y. I guess I should keep reducing the color until I get as close to 0.212, right?

I have also attached the plot of my luminance prior to any adjustment.

Any feedback would be helpful.

I was not sure if I should use the 75% or 100% for color adjustment. I used the 75% based on the guide.

· Registered
8,008 Posts
Discussion Starter · #19 ·

It appears I am headed in the right direction. It just seems like a lot of clicks down (for color) to get a slight change in the Y. I guess I should keep reducing the color until I get as close to 0.212, right?

Yes. Remember, though, you may have a color decoder error, so when Color is set good for Red it may be screwed up for Blue and/or Green.

There is absolutely no advantage to using 100% stimulus and it could potentially cause problems. Use 75%. That's why I included that in the instructions. Also, be sure to use a window, not a full screen. Finally, what type of display do you have?

· Registered
1,324 Posts

Originally Posted by TomHuffman /forum/post/0

Yes. Remember, though, you may have a color decoder error, so when Color is set good for Red it may be screwed up for Blue and/or Green.

There is absolutely no advantage to using 100% stimulus and it could potentially cause problems. Use 75%. That's why I included that in the instructions. Also, be sure to use a window, not a full screen. Finally, what type of display do you have?

My display is a DLP Front Projector (Benq PB6200). Using the Avia color decoder test - it appears that it has a bit of a red push. Will that give me some problems? Maybe I should use blue or green?
1 - 20 of 2069 Posts
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.