Back To Listings RSS Print

Color Matching a Canon C300 to an Arri Alexa

They said it couldn't be done. Maybe, but I got close. How close? You judge...

By Art Adams | January 26, 2013

I'm shooting more projects with the Canon C300 but none of the built-in looks greatly appeal to me. While trying to build my own custom C300 look at a local rental house I spied an Arri Alexa in the next stall over... and asked, "Can I put that next to this C300 for a while?" The results are really interesting...

IMPORTANT UPDATE: I've been receiving reports on Twitter that my daylight matrix may be turning reds a bit blueish in some cameras. I've replaced the daylight matrix settings with a new set that should help with the blueish-reds issue. Version 2 will appear sometime early next week.

My buddies at Chater Camera let me come in and experiment with their toys whenever they aren't busy. I've been trying to build a new look for the Canon C300 for a while and I'd developed one that was fairly color accurate, but when I saw a a prep tech checking in an Arri Alexa in the next stall I asked if I could borrow it for a bit.

Adam Wilt (background) and I try to match a Canon C300's color to an Arri Alexa. I'm staring at a Leader 5800 waveform/vectorscope.

I use a DSC Labs Chroma Du Monde 28 chart for my color tweaking. I know that their charts are the most accurate available, and so do camera manufacturers: they generally use this chart, or one of its variants, in the creation of their color science.

The Canon C300 offers a variety of matrices, each of which bend red, green and blue in ways that Canon deems pleasing. The Norm1, Norm2, Norm3 and Norm4 matrices are basically the same, as best I can tell. Cinema1 is meant to create a filmic look for viewing on Rec 709 displays. Cinema2 creates a filmic look when shooting for film output. Canon Log captures as much information as possible for post color grading and looks desaturated and flat during shooting as log is a storage format only and isn't meant to be viewed directly. EOS Standard matches the look of Canon's EOS still camera line.

Color science can be really complicated or really not. I've heard that one manufacturer's color science requires 10,000 lines of software code while another requires only 100 lines. It all depends on the taste of those creating a camera's look, and each manufacturer has their own idea of what constitutes pleasing color. On the one hand this is good: a camera's colorimetry is as distinctive as a film stock, and it's possible to select a camera for its look the same way we used to choose film stocks for their looks. On the other hand, if the client dictates which camera you're going to use on a project it can be awkward if you don't think the camera's overall look is appropriate for the project... or if you just really, really like the look of a camera your client can't afford.

I love the Arri Alexa look, even when shooting in Rec 709 WYSIWYG mode. I live and work in a secondary market--the San Francisco Bay Area--where clients can't always justify the expense of an Arri Alexa. The next best thing is to try to replicate as many elements as possible of the Alexa's look in another camera. This is impossible to do perfectly, but if I can achieve some percentage of this look then I'm happier on those occasions when I can't get an Alexa. Fortunately most cameras give us the tools to tweak their looks somewhat so that we can bend their base look into something we find personally pleasing.

Canon does some very interesting things with their color science. I've noticed that most of the matrices (Norm1-4, Cinema1, Cinema2 and Canon Log) push red toward yellow so that red becomes an orange "fire engine" red, and blue is pulled toward green such that bright greens, like grass, become cooler in hue. One interesting side effect is that bright green grass becomes a little less distracting, so if you're shooting people running through a grassy field the cool-hued grass doesn't draw attention from the action. On the other hand, if you want to capture bright green grass you're going to have a hard time doing so using several of Canon's built-in matrices.

On the plus side, the one color Canon nails in every single matrix is flesh tone. That's important as flesh tone's hue is burnt into our brains and we know instinctively when it's not right. If you're going to get one color right, it should be flesh tone. Everything else can be nudged one way or the other to create a unique look as no one will ever notice that greens are a little cool or reds are a little orange unless they do a side-by-side comparison. Meanwhile there's a subconscious part of our brains that either likes this look or doesn't. Clearly a lot of people like this look or it wouldn't exist.

The one matrix that doesn't show a lot of extra manipulation seems to be the EOS Standard matrix. When I look at this matrix on a Chroma Du Monde chart the colors are very close to accurate, and there doesn't appear to be any secondary color correction going on behind-the-scenes. (Secondary color correction, in this sense, refers to manipulating a selected range of hues separately from the overall look. It's a correction that affects only a small color range that sits on top of the overall color science.) For example, all the other matrices show a strange inward curve in the secondary hues between blue and cyan, and no matter how I try to tweak those matrices I can't get rid of that curve. It's a fundamental part of the underlying color science and it can't be defeated. EOS Standard is very straightforward by comparison.

Above: The Norm1-4 matrices and Canon Log matrices have an interesting "notch" where the intermediate hue between green and cyan is noticeably desaturated and distorted. There's nothing I can do with the tools provided to remove that notch, and it impacts blue and green in interesting ways when I try to manipulate those colors in the user matrix.

Above: In the EOS Standard matrix the "notch" between green and cyan is gone: the hue between green and cyan falls on a line directly between those colors. All the primary and secondary colors fall on or close to their vectors, but red is farther from the center of the vectorscope than the other colors which means it's really saturated by comparison. I find this very distracting in an image.

The matrix controls I work with are part of the "Matrix" submenu found in the camera's picture profiles, accessed by pushing the CUSTOM PICTURE button on the side of the camera. Here you'll find several useful controls such as preset matrix selection, gain (saturation), phase (global hue adjustment), and a bunch of numbers labeled R-B, R-G, B-R, B-G, G-B, G-R. These abbreciations stand for "red minus blue," "red minus green," etc. These controls add and subtract color signals from each other in order to create a look. Subtracting one color from another makes it more pure, or saturated, while adding one color to another makes it less saturated. As pure colors are generally more desirable than muddy colors, and subtracting one color from another results in greater color purity, positive matrix values result in color signal subtraction and negative matrix values result in color channel addition. (It makes sense if you think like an engineer.)

There are a number of elements that influence a camera's look: the dyes in the color filter array that pass light to photosites on the sensor play a very large part in the look of the camera as they provide the basic building blocks from which a look is created, but the final look comes from the math of adding/subtracting those color signals to/from each other using matrix math.

The R-G, G-R, etc. settings are generally referred to as the "user matrix" as this is how Sony labeled these settings in their F900 camera and the name stuck. They sit on top of all the other color science that drives the camera. No matter what you do with those controls you're always adding a look on top of another underlying look. That's why I chose the EOS Standard matrix as my base look: it appears to be the simplest matrix the camera offers, therefore it is the easiest one to push around predictably.

Canon's description of how their user matrix works, from the Canon C300 manual.

One important thing to understand is that the user matrix has no effect on neutral tones: white, grays and black. The color matrix is completely separate from white balance. You can zoom into a gray scale chart, white balance, throw crazy random numbers into the user matrix and nothing will change... until you zoom out and include color in the frame, at which point you'll see that every color is completely wrong.

For a rough description of my matrix creation process, and to download my settings, turn the page...

Page 1 of 3 pages 1 2 3 Next »

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

Jeff Regan: | January, 26, 2013

Thank you Art, for the good information.  I did a shoot a few days ago with our C300 PL and a C100 where the scene from the two cameras is supposed to be black and white, except for a red rose.  I used EOS Standard matrix with Norm 3 gamma to try to get close to a REC709 look.  The red rose was over saturated, compared to the other color points, as you pointed out regarding EOS Standard and the overall saturation was pretty high with matrix gain at zero, but other color points were pretty accurate. 

I did find that the C100 matched pretty closely, just had to adjust some blue gain and bring detail up a couple of clicks, most likely due to lensing differences.

I will load your matrix settings into the C300 and see how it looks!

Art Adams: | January, 27, 2013

Hi Jeff- I’d love to hear what you think about them. I’m open to tweaking based on what people like yourself discover. I’m hoping to come up with similar settings for other cameras as well.

Peong1936: | January, 28, 2013

Fantastic input here. I have a C300 and was looking for ways to make the results different from what I’ve produced so far, until I read this article. Thanks for showing us to go way beyond what’s given and experiment the C300 with the Arri Alexa.

Rob von Münchhofen: | January, 29, 2013

Hi Art
Thanks for that - I always enjoy reading your tests and articles.
I’m preparing for a one day commercial shoot with the c300 right now and I do have a question regarding this article and another one, you wrote nearly one year ago, in which you describe how to get rid of the green tint in the WB presets by using the WB Gain under the C300s custom menu. The article is accessible here: http://provideocoalition.com/aadams/story/canon_c300_trimming_white_balance_plus_a_look_at_daylight_vs._tungsten_colo/
Now as you clearly point out above, this here is the attempt to match the C300 as closely as possible to the Alexa. Both Cameras have got some sort of green shift in their preset whites.
So am i right in assuming that in the case of matching both cameras by using the color matrix, you did NOT include any kind of anti green shifting?
Is it possible to use your matrix settings (as above) and also on top of that make adjustments in the WB gain of the camera to get a clearer white under tungsten and daylight respectively, without weighting any colors?

I am really looking forward to your answer and testing your matrix settings.
Thank you

Brian C. Weed: | February, 01, 2013

Thanks for your incredible work on this, Art. One question - would it help to use the C300’s Color Corrector CP setting to tame the reds back a bit instead of decreasing overall saturation? Could that also be used to boost the cyans a bit, since there are A & B options?

Art Adams: | February, 03, 2013

Hi Robert- I sent you a note privately about this as your comment hadn’t shown up yet here when I was notified of it, but within a month or two of my reporting that green white balance issue I played with a C300 at NAB and it was gone. They’d apparently fixed it with a software update. That’s one of the fastest responses I’ve ever seen from a camera company. It white balances perfectly now.

Hi Brian- Interesting idea, but I get a little nervous using secondary color correction for something this basic. Sometimes unexpected things happen because you’re bending one part of the color response in a way that’s inconsistent with the rest. It works great for product shots but I worry about doing it for everything.

This “bending” is what happens internally already if you look at the range between green and cyan on any of the other color matrices and it looks a bit strange. I’m going to try to tweak some more on these matrices later this week and I’ll give this a try when I do.

Hi Peong- No problem, it’s what I do for fun. I’m such a geek. smile It’s fun being my own lab.

If anyone has any comments about these matrices let me know, especially if you get any unexpected results. So far I’ve been fairly happy with them but there’ll be more tweaking and a version 2.0 sometime in the near future.

Please login or register to comment