Back To Listings RSS Print

DSC Labs Hawk Chart: The Simplest Color Chart That You Can’t Live Without

Wouldn't it be great if someone designed an easy-to-use color chart that could be quickly and easily used in the field? Well, someone did. And they call it The Hawk.

By Art Adams | June 17, 2011

It wasn't until I worked at the DSC Labs booth at NAB that I discovered The Hawk... and it blew me away. It's a very simple chart, but it offers a colorist (professional or amateur) the most critical information necessary to accurately neutralize your raw, log, or even WYSIWYG images.

At NAB I was approached by a DP who was looking for a chart for her latest episodic series. "I'm shooting on one coast and color grading is happening on another," she said. "Which chart do I need in order to bring a still into Lightroom, color balance it, and then add a look on top of it as a reference for the colorist?"

Initially she'd thought the Chroma Du Monde was the answer, but I've worked on episodic television shows and they move FAST. The Chroma Du Monde is a very powerful chart but it's not ideally suited to an episodic schedule, where time constraints rarely allow more than rolling a few frames on a chart placed in whatever light is available. The Chroma Du Monde should be evenly lit for best results.

The Hawk, however, is designed to provide the most useful grading information in the smallest package possible:



To break it down:

The white patches provide a perfectly neutral white balance reference.

The black patches provide a perfectly neutral black balance reference.

The gray area around the patches provides a perfectly neutral gray (gamma) reference.

And the brown patches provide an accurate reference for... wait for it... flesh tone.


That's really all you need to quickly neutralize a shot in color correction: a white reference, a black reference, a gray reference, and skin tone.

Here's a still that I shot in my living room on my Nikon D7000. I deliberately balanced the camera for 3200K even though the room was lit by ambient window light.



In Final Cut Pro I zoomed in until all I could see was the chart:



Here's what the waveform looked like before tweaking:



You can see that there's a LOT of blue in the image and not a ton of red. We'll need to fix that. After firing up the basic Final Cut Pro color corrector and using the white, black and gray color pickers (in that order) I ended up with a chart that looked like this:



...and a waveform that looked like this:



...and a scene that looked like this:



And that's exactly what the environment looks like to my eye.

There are lots of charts that offer white, black and gray references, but it's the flesh tone patches that really set this chart apart. A look at a vectorscope shows why:



It's easy to balance white, black and gray using only a waveform monitor, but that doesn't tell us much about the accuracy of flesh tone--which is a range of hues that every human being recognizes at a very, very deep and unconscious level. The waveform monitor can tell us the brightness levels detected by each color channel and whether white, black and gray are neutral, but it's impossible to detect how accurately flesh tones are being reproduced.

The YIQ color space is the color space used by NTSC television. "Y" is the luminance channel, while "I" encodes hues on the orange/blue axis and "Q" encodes hues on the green/magenta axis. Quite a long time ago it was discovered that the eye is much more sensitive to changes on the orange/blue (warm/cool) axis than they are to green/magenta, so "I" is given more bandwidth than "Q" is.

The question you're probably asking yourself is: why do I care?

On the vectorscope there exists a line, known as the I+ vector, that is the legacy of the YIQ color space. Here's the trick: every skin tone known to humankind will fall on or near this line. The circled area on the vectorscope above shows you where The Hawk's flesh tone patches fall, and they are well within acceptable skin tone range. If that patch skews any farther counterclockwise from the I+ line then skin will look too yellow; any further clockwise and it will look too red.

Here are a couple of interesting tidbits that might be worth knowing:

Colorists generally balance white first, then black, and then gray if necessary. When I used the white color picker in Final Cut's three way color corrector the results were initially very poor. The white picker brought all the peaks in line with the lowest peak--red--and they all landed around 60 units. Then when I used the gray picker the whites became contaminated.

What I realized is that placing white at 60 units, even just temporarily, puts it too close to the range affected by the gray picker, such that gray balancing has an unwanted affect on the white balance. I discovered that if I used the white picker once and aligned the peaks at 60 units, then increased the gain to boost those peaks higher to between 80 and 90 units, and then used the white picker again, the white peaks aligned high enough on the waveform that they were unaffected when I used the gray picker.

This trick also works in Adobe Lightroom.


Here's what I've learned intuitively about white balancing:

Think of color temperature (measured in Kelvin) as a seesaw overlaid on an RGB parade waveform display. The center post is the green channel. Unbalanced tungsten light sees the seesaw tipping up on red and down on blue; unbalanced daylight shows the opposite. Any time you're manipulating color temperature you're changing the balance of red to blue in the signal. Green is unaffected.

If the camera you're using has a "CC" control (the Arri Alexa is a good example) then you have the ability to raise or lower the green post on which the seesaw sits: CC controls whether the picture skews green or magenta, so raising the post makes the picture green and dropping it makes the picture magenta.

This should sound familiar from the YIQ color space description above. White balance works in the orange/blue realm, while CC works in the green/magenta realm.



A lot of people have asked why DSC Labs charts generally have a glossy finish. I posed this question to David Corley, one of the owners of DSC Labs, and he had this to say:

Given otherwise similar characteristics, ridiculous as it sounds, matte white is more reflective than the glossy. The reason is a little tricky to explain (akin to describing a spiral staircase without using your hands) but let me give it a try:

If you look at a matte chart through a microscope the surface is like a mountain range with millions of tiny peaks and valleys. Consequently light falling on the rough surface bounces from one peak to another and eventually forward towards the camera. Compare this to a glossy white surface where light falling on the surface either goes in and is reflected forwards or, like shining a light on a mirror, some of it bounces off at the reciprocal angle (light falling on it at 45° will be reflected at 45° for a differential of 90°).

Typical bright white matte paper reflects 90% of the light striking it compared to barium sulfate (BaSO4), the reflective reference standard. On the other hand, glossy CamAlign and ChromaDuMonde white reflect 85% towards the camera, 5% being bounced off harmlessly at the reciprocal angle.

So with the matte surface the 5% being bounced around is eventually reflected forward as flare - five parts in 100. Now compare this with matte black where 5% is still bouncing around and coming forward as flare, but the black chip is supposed to reflect only about 1%. This is why we can't achieve a true black with a matte surface.

The question everyone asks is, "How did you come to develop glossy CamAlign and ChromaDuMonde front-lit charts in the first place?"

Indirectly we have to credit Sputnik, President Kennedy and the Space Race. A client in the space industry was using Ambi/Combi rear-lit systems to align cameras, and they were looking for a smaller lighter test pattern to send into space. They loved Ambi/Combi, but asked why we couldn't make a precision high dynamic range front lit chart.

Thirty years ago, a Lambertian Surface was the goal in test charts. Lambert said that light falling on a flat surface should be scattered in such a way that the surface appears to have the same brightness regardless of the angle at which it was viewed.

This was a worthy goal, if (and it's a big "if") it was possible to achieve and maintain its Lambertian properties, which are easily destroyed even by gently wiping with soft cloth.

Let's assume we do produce a Lambertian grayscale, we immediately find that the maximum dynamic range we can make is about 64:1 (six f-stops). So we stick on a piece of black velvet and pick up another f-stop. In the early days of B/W television, a dynamic range of 128:1 would have looked pretty good. Today however, with camera systems boasting a range of 60,000:1 or more it's a different ballgame. Glossy surfaces allow us to create front lit charts with the highest contrast that is printable. Back lit charts allow for a much higher contrast range but require placement in front of an evenly lit, broad spectrum and neutral white light source, which is not always possible.

Another serious drawback of matte charts is that the reflection level tends to change with the angle and type of light source, flood or spot, and sets aren't all lit with two spots at exactly 45°. "Aha!" you say, but how about the reflection from glossy charts? The glossy chart's dynamic range is so much higher than a matte chart that the glossy surface produces more consistent results even with some reflection. If the reflection is huge, then simply drop a black flag below the camera lens and tilt the chart slightly so that the black flag reflects in the surface of the chart.




The back of The Hawk has three additional features: a framing guide, a perfectly neutral "CamWhite" for a neutral white balance, and a "CamWarm" white balance patch that, when used as a white balance reference, warms the image by fooling the camera into adding orange and magenta by removing blue and green.

The Hawk is the perfect chart for productions that move very, very quickly. It has all the information necessary to reproduce an accurate and neutral image, whether this is done in a grading suite, desktop color corrector or photo editing software such as Adobe Lightroom. And it's such a simple tool that it's hard to use improperly. It can be found here and here.

Click here for a slightly longer explanation of the YIQ color space.

Disclosure: I have worked as a paid consultant for DSC Labs.

Art Adams is a DP who likes to use tools that help him work fast. His website is at www.artadamsdp.com.
Editor's Choice
PVC Exclusive
From our Sponsors

Share This

Back To Listings RSS Print

Get articles like this in your inbox: Sign Up

Comments

Chris Lawes: | June, 18, 2011

Would this neutralize any differences in color between different non-matched lenses? Not talking about resolution/bokeh/etc. just color.

Would my shots match between olympus zooms and zeiss primes if I color matched them all with the same chart first?

Please forgive the stupid question.

Charles Angus: | June, 18, 2011

Short answer is yes.

Art Adams: | June, 18, 2011

You should be able to match two lenses with this chart as long as you’re only looking at overall color shifts.

The one thing you can’t do with this chart is match cameras that may have different matrix settings. You can’t see how the matrix affects color by looking at white, gray or black. You’d need the full Chroma Du Monde in a controlled environment for that.

But for neutralizing overall color casts and dialing in flesh tone between multiple lenses or multiple cameras with the exact same matrix settings, this chart should be perfect and easy to use.

Chris Lawes: | June, 18, 2011

So using multiple AF100’s (same scene files) with say, Olympus zooms mixed with Zeiss zf.2 primes would work for color-matched shots if matched later in post with a chart like this, correct?

Makes buying color-matched lens sets less important I suppose.

Thanks for the valuable information, hope it applies to somebody who reads PVC besides me smile

Chris Lawes: | June, 18, 2011

Another question I’ve wondered about. In the post you say after using the eyedroppers in FCP the image looked “..exactly what the environment looks like to my eye.”

However my understanding has been that using a greycard in post neutralizes the effect of the lighting and creates perfect whites (not warm white or cool whites.) So if you have warm lighting (sunset) balancing off a grey card in post will neutralize it and remove the color warmth.

Am I correct in my thinking? Is there a way to color correct a sunset-lit shot to a greycard or some kind of color chart so that it will look just as it did to the eye in the field?

Art Adams: | June, 18, 2011

Hmmmm…. probably not. cheese The background colors in the still above are very warm, so although the gray card looks pretty neutral there’s nothing else in the shot that’s a neutral gray or white.

Any grading you do to a gray card shot at sunset will remove the sunset effect, because the role of a reference like this is to allow you to neutralize overall color shifts. This is good for environments where there are overall color shifts such as fluorescent-lit offices (green) or open shade (blue), etc. It’s also important for color correcting with a consistent look, as this card allows you to “zero out” the look before applying a grade. If you apply a grade to non-neutralized footage you’ll get a lot of different looks where you only want one, so The Hawk allows you to find a baseline from which to work.

As for sunset… the problem is that none of the light in a beautiful sunset is a neutral color, so you don’t buy anything by shooting a reference chart. It can give you highlight and shadow color references so you can give other footage a sunset feel, but it won’t make the sunset match what you see by eye. You’re probably best off picking a preset white balance, like 5600K, and sticking with it as the sun goes down. If you can a color accurate monitor you can skew the color temp as the sun goes down and enhance the image by making it warmer or cooler. (Sometimes adding blue to a sunset can make the most amazing purples.)

In the case of the episodic DP that I mentioned in the article, she wanted to be able to pull a frame out of shot ungraded footage, neutralize it, and then apply a look as a reference for her colorist. The colorist can then neutralize the footage in the same way and then build her look from a neutral palette.

I used the eyedroppers as an example, although I’m not really promoting their use as a serious color correction tool. For quickly neutralizing a still in Lightroom it works very well, but for tweaking HD footage you’ll most likely want to use more sophisticated tools. The eyedroppers get along quite well with this chart, though, so you can do a rough grade very quickly for a rough cut and then go back and finesse it later.

Stephen Taylor: | June, 19, 2011

Hey Art,
Firstly thank you for continuously sharing your wealth of knowledge with everyone.

Hopefully this isn’t a stupid question but it’s bothered me for a while.

So i’m used to seeing nice primary reds and blues etc. in a set of camera generated colour bars that align perfectly with the respective dots on a vectorcope.
So how come the colours on say the Chrome Du Monde chart are what appear to be muted pastels but they still hit the marks?

Thanks Stephen Taylor

Art Adams: | June, 19, 2011

Hi Stephen-

My understanding is this:

Color bars are an electronically generated test pattern, and as a result they are almost impossible to reproduce in reality. It’s impossible to print colors that are that saturated, at least on a front-lit chart. (This may not be true for back lit color charts, but I haven’t played with those yet.)

The goal with the Chroma Du Monde is to use the waveform monitor to set levels and then verify that their dots on the vectorscope fall on the right vectors. (A vector is a line, in this case drawn between the middle of the scope and each color’s box.) If the dots fall on their vectors then they are at least accurate in hue, if not in saturation—and saturation can usually be fixed later.

I’m doing this by memory so I may not have the numbers exactly right, but the theory is correct: the reason the colors are muted on the Chroma Du Monde is that they are not pure colors, but mixed colors. For example, the “red” patch is 80 units of red (as seen on a parade waveform) with 40 units of blue and 40 units of green. The same breakdowns hold true of the “blue” and “green” patches. In this way the signal never drops to black in a color channel when a color is not present; if it did then you’d have no way to know whether it was present in the shadows as it would be crushed.

On modern cameras these signals don’t always appear at 80 units and 40 units due to the various gamma curves in play, but if the camera is set to a true (crippled) Rec 709 gamma curve this is where the colors should fall. Otherwise they should appear a bit flatter as the Rec 709 standard is only designed to hold six stops or so and modern cameras can usually see ten stops or more, so you have to cram them into the six stop space using some creative math.

Also, while the Rec 709 standard exists, not all the manufacturers hit it exactly or use the same formulation of dyes on the sensor to get there. As a result some colors don’t hit their boxes at all. That’s typically not too important as the distance from the center of the vectorscope indicates saturation, and the rotation describes accuracy. Accuracy tends to be more important than proper saturation levels as you can almost always tweak saturation, and Chroma Du Monde charts are designed to be used with a vectorscope magnification of 2x anyway as the colors will never be saturated enough to hit the boxes at 1x. And even at 2x some of the colors won’t hit their boxes: green is a notoriously hard color for cameras to hit and is usually the one color that isn’t saturated enough to hit its box.

There are also colors that don’t fit into the video color space. You can get a rich bright blue in video that you can’t get in film; you can have bright blue or saturated blue on film but not both. Similarly, you can get oranges on video that film can’t see.

I hope that helps…

Adam Wilt: | June, 19, 2011

The roughly half-saturated colors on the chart are much easier to render using reflected-light pigments; fully-saturated colors are very difficult to create in print with good accuracy. The ChromaDuMonde colors are set up to hit the target boxes (assuming Rec.709 colorimetry) on the 75% saturation vectorscope if you boost the vectorscope gain to 1.875. This corresponds to boosting gain on the ‘scope until the colorburst (assuming, of course, that your signal has burst!) extends out just below the outer limit of the 10% target box for green.

If you don’t tweak the ‘scope, the ‘scope pattern for the CDM colors will be about half as big as the corresponding pattern for electronically-generated 75% colorbars.

Adam Wilt: | June, 19, 2011

...And I find it amusing that we’re talking about color patches that don’t exist on The Hawk!  [grin]

Art Adams: | June, 19, 2011

DSC rounded 1.875x to 2x years ago for some reason I don’t remember.

Adam Wilt: | June, 19, 2011

2x is a much easier computation, and required only a minor tweak to the colors… My documentation is from my CDM which is, I fear, a wee bit out of date (blush).

Stephen Taylor: | June, 20, 2011

Thanks for the explanation guys. much appreciated.

IEBA: | June, 21, 2011

Great concept.
I did something very similar to this matching a mid-level point-n-shoot 720p30 video to a higher end 3-chip HD camcorder by using a black/white/grey pop-out reflector. It didn’t get me _all_ the way there, a flesh-tone bar across the middle would have helped greatly, but it did get the cameras to look so much closer together in 1 minute than I thought was possible at all.

I find blurring the image a bit before color picking makes a great difference because he color picker is _very_ specific in what it selects.

I wrote this up for EventDV but the article doesn’t appear to be online.

IEBA: | June, 21, 2011

Oh, here’s the article. It’s embedded in one of those new “digital Magazine” thingies.

http://www.nxtbook.com/nxtbooks/infotoday/eventdvlive_2010autumn/#/23/OnePage

Please login or register to comment