The Panavision Millennium DXL: Light Iron and RED create a winning new look for large format cinematography

Please welcome the world’s newest digital film stock. You’re going to like it.

The author trying to look cool at the Panavision/Light Iron press conference. Photo by Adam Wilt.

My invitation to a mysterious press event was nearly one of the biggest surprise announcements of the year. That morning Panavision showed their new camera, the Millennium DXL, to the ASC, and by early afternoon pictures of the camera flooded Twitter and Instagram. Panavision officially announced the camera by early afternoon as the cat was not only out of the bag but loudly playing bagpipes in the street outside the CineGear trade show, where Adam Wilt and I were setting up a booth for DSC Labs. The “surprise” press event that we’d signed up for the week before was no longer a surprise.

When I heard that this new 8K camera was a co-venture between RED and Panavision, my first thought was: “Of course, this makes perfect sense.” RED has an 8K sensor ready and waiting, they’ve given outside vendors unprecedented access to the internal workings of their cameras, and Panavision needed a response to the Arri Alexa 65. (It’s interesting to see RED pivot from being the “bad boy” of the industry to the most cooperative and open camera company on the planet. I like it.)

I’ll touch briefly on camera design, and skip the specs as those are easy to find. Then I’m going to focus on what I’m most interested in as a cinematographer: color, dynamic range and resolution. Along the way I’m going to teach you a bit about how cameras render color, and show why that’s important to you.


DXL_Rt_Profile_Panavised with Lens

The DXL (“Digital eXtra Light”) camera is exceptionally well designed. The camera head is lower than the rest of the body, which keeps the center of mass low on the operator’s shoulder and allows them to concentrate on framing rather than preventing the camera from falling over. The dual displays, one on each side of the camera, are a brilliant touch, as is the fact that adjusting one automatically updates the other. The RED Xenomorph camera is vaguely similar in shape but seems stripped down by comparison: pictures show few buttons, which implies that the camera is controlled primarily via touchscreen. The DXL is a little boxier and less streamlined but has many well-placed buttons for camera control, which I much prefer to touch screen operation as they offer positive feedback and tend to be easier to find and use in dim environments or on moving cameras.
I’d rather work with buttons than a touch screen any day. Buttons are easy to find and press, especially on a moving camera.
The camera design is completely modular, with the body consisting of five basic modules that can be swapped out based on need and future updates. For example, there’s an audio module with XLR inputs that can be installed to record sound internally, but it can also be removed to save weight.

Any port in a storm...

 The camera’s modular design is a natural fit for Panavision as their rental-only system allows them to update cameras as desired. It’s unknown how many of these cameras will populate Panavision’s fleet but they’ll always be up-to-date with the latest technology. Modernizing a camera should be as simple as replacing a module when it lands back in the shop.
Panavision showcased new lenses with internal focus motors, which are brilliant as camera assistants can spend less time adjusting motors and calibrating distances and more time on getting focus marks and rehearsing. Focus in 8K is not a simple thing: the increased size of the sensor results in a 1.8x multiplier to S35 focal lengths. A 90mm lens in 8K captures the same angle of view as a 50mm lens in S35.
On a 50mm lens, at T2, I could wrack between the cheese trays in the foreground and the popcorn in the background and see a noticeable focus shift.
On a 50mm lens, at T2, I could wrack between the cheese trays in the foreground and the popcorn in the background and see a significant focus shift.


Camera specs are freely available and easily quantifiable. I’m more interested in intangibles, such as color. Over the years I’ve asked numerous camera companies for a brief summary of their color philosophy. The look of a camera is a highly subjective thing, and someone makes those aesthetic decisions. I’ve always wondered what they’re aiming for when they make those decisions.

This is the very first time that I’ve received a coherent response. According to Ian Vertovec, the supervising colorist at Light Iron (a Panavision company) who lead development of the DXL’s color science, he had three things in mind during this process:
(1) Years of sitting in grading sessions taught him what DPs did and didn’t like about the look of digital cameras;
(2) There are common digital photography color challenges that most DPs choose to correct in post;
(3) Rather than create a “look,” he focused on building a digital negative that avoided the most commonly disliked digital camera color issues and made for faster and easier grading.

His key concern was maintaining consistent flesh tone. It’s not unusual for digital skin tones to shift hue toward green or magenta in extreme highlights or shadows. Light Iron’s color science works aggressively to eliminate these shifts.

All stills are courtesy Light Iron and Michael Cioni, and were pulled as TIFFs from their 4K presentation. I created the JPEGs myself, using as little compression as possible, as TIFFs are not a supported web image format. Due to PVC’s 640px width restriction, and the inability to publish TIFFs as embedded images on the web, you are not seeing these images in their true glory.

Mixing daylight and tungsten light can tough on most cameras. Less sophisticated cameras render 5500K light as electric blue or reddish blue, and skew tungsten light toward green. In this case, flesh tone looks great (albeit a little cool) and the background lamp is warm but not greenish (which would make it appear yellow).
This is a pixel-for-pixel enlargement. Flesh tone is very neutral here: an 8-bit color picker app shows that the green and blue channels match, while red is about eight points higher. Its hue is consistent regardless of tonality.

Another example is front-lit trees in daylight: green leaves often look oversaturated and “electronic” under these conditions, so Vertovec sought to roll off the saturation in a pleasing way that preserved the color of the leaves without making them appear video-like.

The full frame, reduced from 4K.
The full frame, reduced from 4K.
A pixel-for-pixel enlargement. The cool skylight reflected in the leaves is very pretty. Notice how the brighter areas are simply that—brighter—but not more saturated than the rich green stem. Increasing brightness without increasing saturation is a “filmic” response, whereas the “video” response is to make brighter hues more saturated.

As we spoke at the press conference it became clear that Vertovec was pursuing the holy grail of digital color: emulating the look of a subtractive color process (film) in an additive color medium (digital projection and TV displays). The most saturated hues that a TV can produce are also the brightest because TVs use additive color, where three strongly saturated light sources mix to create all possible hues. The most saturated hue comes about by turning one of those colored lights full on and the others full off: the most saturated red, for example, is also the brightest.

In short, TV highlights are where you can find the most highly saturated color. Most traditional video companies play to that “strength” because it seems like a natural fit: this is how TVs work, so why not take advantage?

In the additive color system, the brightest hues are also the most saturated. Creating a strong red hue, for example, simply means turning the red light up full.

The film look, however, is the opposite. The projector emits white light, and dyes in the film emulsion carve bits of it away to create the desired image on the screen. I like to compare this process to watercolor painting, as the two are very similar: the brightest portion of a watercolor painting is the white paper on which it resides, and the application of paint blocks portions of that paper’s reflected light spectrum and allows only light of the selected wavelengths to reach your eye. The application of paint creates a filter on top of the reflective white paper. Print film does the same thing: it filters light to create color images.

The darkest hues are the most saturated because watercolor paints, and film dyes, must be quite heavy to block enough of the spectrum to create a saturated hue, while the brightest hues are the least saturated because brightness requires passing more white light. This results in a natural roll of color saturation as brightness increases.

The subtractive color process starts with white light and then filters out unwanted hues until the proper hues remain. The most saturated colors are always the darkest.


Once I learned that the sensor came from RED, my next question became: does this camera look like a RED? During the intro, Light Iron CEO Michael Cioni spoke about how aggressively Light Iron sought to expand the sensor’s color palette so it was better able to discriminate between subtle hues. That’s what I love about the reigning hue discrimination champion, the Panasonic Varicam 35. While that camera has a broadcast camera “richness” to it, it also discriminates wonderfully between very fine hues.

This is my favorite Varicam 35 demo reel shot:
Screen Shot 2016-06-06 at 4.42.07 PM
Screen Shot 2016-06-06 at 4.41.33 PM

Notice the subtle differences between jackets of similar hues. The man’s sweater is a very complex and sophisticated green that I’m not used to seeing in footage from any other camera. I see several similar but different hues of purple and blue, and several fabrics whose hues are close to neutral gray but have very slight casts to them… there’s a mix of subtle and strong hues here that I find dazzling. (Michael Cioni consulted on Varicam’s color as well.)

Mr. Cioni emphasized that this was one of Light Iron’s goals in devising the DXL’s color science. He spoke of starting work on the camera’s color as close to the sensor as possible and then manipulating it in a very large color space to push the various hues as far apart as possible.

Every camera has a native color gamut that is defined by the most saturated red, green and blue hues that it can detect. This color gamut doesn’t match any of the standard color spaces (Rec 709, Rec 2020, DCI-P3) because the dyes on the sensor aren’t that precise: none are completely pure, and their spectral overlap affects the size and shape of the color gamut. At some point in the process this color gamut is mapped into a standard color space for viewing, but that’s often the last step and is best handled by a LUT.

A variety of color gamuts, courtesy of Sony.
A variety of color gamuts, courtesy of Sony. In Sony’s case, the camera’s native color gamut is mapped into a huge color space and graded down to DCI-P3 or Rec 709 from there. (A “color space” is a mathematical model through which colors are manipulated. A “color gamut” is the area within that color space that a camera can capture.)

Integral to nearly every camera design is something called a color correction matrix. A color matrix is a brute force method of converting image color data from one color space to another, and the effect is linear: the data is basically remapped, but without any control over how saturated a hue is at a certain brightness level. A hue goes in and another hue comes out.

LUTs, on the other hand, can isolate specific hue ranges and brightness levels and manipulate them in very subtle ways. Rolling off color saturation in a highlight using a matrix is hard. Doing so with a LUT is easy.

Light Iron states that their look is based on color science that affects image data as close to the sensor as possible. This is not unusual as most cameras do this: there’s an order of operations when processing image data, and my guess is that most of their work is done in the RGB matrix that maps white-balanced and deBayered pixels into a working color space, where other operations are then performed. The trick, though, is that Panavision does not say that they are creating a look in camera: rather, they are ironing out the bugs that prevent them from having the maximum amount of grading flexibility so they can achieve the goals that DPs commonly pursue in a grade. They’ve not set out to create a look, but rather an easily-graded “digital negative.”


Every sensor has underlying color characteristics based on the physical nature of how it sees light. The DXL is no different. I’ve certainly had my complaints about the look of RED cameras in the past, but the Dragon sensor, in conjunction with the skin tone filter, seems to be a mature combination. (It’s hard to argue otherwise when the majority of great looking Netflix and Amazon shows are shot on RED Dragon cameras.)

The Millennium DXL looks great. I can still see some of the underlying RED look, particularly in sequences with muted browns and greens that seem to mush together a bit, but the overall feel was very filmic. Flesh tones are very neutral: not green or magenta, or excessively warm, but solid and realistic. One shot, of Keanu Reeves sitting on a motorcycle overlooking a cloud-filled valley backlit by the setting sun, featured beautifully-rendered cyan skylight that nicely offset the warmth of the sunset. I didn’t see a wide range of blue hues, but they were very pleasing, and the dynamic range of the scene was impressive.

This poorly-shot iPhone photo of a slide from Michael Cioni’s presentation shows the dynamic range of the DXL to be about 14 stops, although I’m slightly concerned that the white chip at the far left is labeled “over” and not “clipped.” It’s meant to be just barely clipped. (Disclosure: I consult for DSC Labs, who makes this chart, and I’m responsible for its wedge shape, which is meant to reduce flare from the brightest chips so blacks aren’t unduly lifted.)

Let’s look at some images pulled directly from Light Iron’s 4K grade:

Reduced from original 4K TIFF.
Pixel-for-pixel from 4K.
Pixel-for-pixel from 4K. This is extracted from the far left edge of the frame. Impressive resolution, no?
That brownish tree in the center of the frame has a little bit of the “RED look” to it.

Part of what I’ve come to know as “the RED look” has to do with muted greens and browns blending together. I never understood why until I saw this article whose author did some backwards engineering to plot the primary colors that define the RED camera’s native color space. His results show that the green primary appears shifted a bit toward red, implying a lot of red/green overlap which then pushes some greens and reds toward yellow. This makes sense to me as it reflects what I’ve seen in RED footage in the past, where some greens and browns shift toward yellow or golden brown and merge together. Light Iron seems to be squeezing a lot more separation out of the sensor in this regard. Still, the brown tree feels a tiny bit mushy but the greens that surround it are quite vibrant.

I don’t see the same hue discrimination that I saw on the Varicam 35, but the hues I see are very nice indeed. And, where the Varicam has a bit of a video broadcast look to it, the DXL feels more filmic, like a low contrast film stock such as Kodak 5277 or 5287.


The following scene, placed in a 1920s speakeasy, mixes warm tungsten light with cool daylight. Flesh tones are unaffected by the mix and don’t shift toward magenta as they sometimes do under such circumstances. Practicals add some yellow and pale orange hues to the mix, and the daylight in a background window is a pleasant cyan hue. This is the only shot where I noticed camera noise during the screening, and at ISO 1250 it was very mild: there was some mild speckling on the overhead ceiling lamps and in the shadows but it felt very much like film grain.
Full frame reduced from 4K. It’s hard to see the level of detail in this image, as the small size against PVC’s white background removes most of its wonderful subtlety.
Pixel-for-pixel from 4K. There’s a little noise here but it’s not objectionable. It feels a bit like film grain. Reported ISO is 1250.
That yellow light is actually some sort of yellow-filtered light, but this is often how tungsten light fares in a mixed daylight/tungsten situation. Warm colors can easily skew toward green, and in this case that results in yellow. Here the only yellow light in the shot is actually meant to be yellow.
This is the way a tungsten light SHOULD look in a mixed-light situation. Warm, but with enough red to prevent the fixture from turning green. Note the texture in the glass!
I love that the “daylight” is clearly cool but not electric blue. The Sony F55 renders daylight similarly in mixed light circumstances, and I’ve heard that some DPs who use that camera regularly don’t bother gelling windows with CTO gel when shooting indoors with tungsten light.
This section of the frame shows several strong hues that don’t always work together. Tungsten and daylight both look right, with neither overpowering the other or shifting hue (tungsten often shifts toward green, and daylight skews magenta). That dark green is luscious.
Flesh tone is, once again, very neutral and pleasant. As the camera develops I’d like to see more flesh tone samples: while neutral skin looks good, it’s also nice to see a bit more nuance (pink highlights, blue veins, etc.).
Flesh tone retains its neutrality throughout the entire tonal range. Lighter hues contain a tiny bit more red, but they’re also facing to the side and the light from that direction might be warmer. I can see some of the lens’s character here. (I don’t remember whether this was shot in spherical or anamorphic.)
Full frame image downscaled from 4K.
I like this frame as it shows more variety to skin tone. An 8-bit color picker shows that the reddish areas on the side of her nose and top of her cheek contain about 30 points more red than green or blue, while green and blue are balanced. The blueish skylight reflection surrounding her eyebrow and under her eye shows red dropping to 10 points above green, and blue coming up four or five points.
Flesh tone reads as very neutral to cool, but it’s fairly consistent throughout the tonal scale. The only variations are likely due to the mixture of the actual light sources.
The red hues in this image are wonderfully subtle and sophisticated.
Red is a tough color for a lot of cameras. This red is particularly nice. It is ever so slightly orange, although it tips over into a light blueish hue in the daylight-balanced specular highlights. Many cameras (I’m looking at you, Canon) would emphasize the green in this image to create an unmistakable orangeish red. (I see this a lot in car colors, where reds are often either slightly orange—to appeal to men—or slightly blue—to appeal to women. See The Wagner Color Response Report.)
The highlight in this image is a brighter version of the cyan at the top left of frame. It’s interesting that the cyan hue carries through even as it approaches white clip.
I don’t see as many transitional hues as I might on a Varicam, but the hues I do see are pleasant. The backlit leaves alternate between green, red and golden brown, and the shadows are a reddish blue, likely due to the extra warmth from the setting sun. There’s a little cyan fringing that looks to be some lens “character.” I like it.
Full frame image downscaled from 4K.
I think I see a little more noise and/or compression artifacts in this shot. That’s not surprising, though: we’re only seeing about 1/8 of the image width, tops. This pixel-for-pixel sample, from the image above, encompasses the top of the mountain at the far left of the full frame and extends down to the point where its shadow starts to disappear. Kinda puts things into perspective, doesn’t it?
I wish we saw more footage shot in open sun, as that would better show off the dynamic range of the camera and add some snap to some of these shots. Still, this is a rich vista with lots of depth, thanks to the low hanging clouds and the way the sensor’s 8K resolution gradually reveals detail through the mist.
Another pixel-for-pixel extract. I limited JPEG compression to 97%, to keep the file size down while retaining as much quality as possible, but this still resulted in noticeable loss of detail compared to the original TIFF image. There’s probably twice as much detail in the rocks in the original 4K TIFF. (The gentleman looks a little wider than he should, probably due to his placement at the edge of a wide angle anamorphic lens’s frame. That’s okay, we pay extra for that look!)
I love the sophisticated greens and browns in the tunnel walls. I’m assuming that the car headlights are a little extra warm and red because the grade favored the overheads by removing their green cast. They don’t look bad, though: removing green usually causes a magenta shift, but this is more of a reddish shift. The headlights remain warm instead of shifting toward reddish blue. (The motorcycle headlight appears a neutral color, likely because it is made of cool LEDs or a xenon bulb.)
I love the deep blueish-red accent in the faceplate of the helmet, and the way the edges around the headlight highlights clip smoothly, with no sharp edges. Notice that the orange logo on the motorcycle is fairly sharp but the helmet isn’t! (This sequence was shot with the new anamorphic T-series lenses.)

This sensor does have a look, and it carries through th grade. In spite of this, it’s clear that its footage can be graded in a myriad of ways, with neutral flesh tones and beautifully rolled-off highlights. It appears to have a lot of dynamic range, and it’s very quiet at ISOs 800 and 1250. If it were a film negative, it would be a great all-around low-contrast medium saturation stock that offers a lot of flexibility in the DI suite.

Light Iron has done an impressive job fixing the image issues that most DPs complain about. The difference in subtle hues is not as dazzling as I expected based on the build up to the demo reel, but it’s by far the most filmic look I’ve seen out of a RED sensor to date. I don’t see the DXL taking the world by storm simply because of its look alone, but it’s likely going to be popular due to ease of grading, crazy high resolution, and the fact that it records 8K to a tiny RED SSD drive.

And now, a note on resolution. Mr. Cioni’s presentation contained a very interesting section where he stipulated that resolution is not the same as sharpness. He pointed to black and white prints by Ansel Adams, that were photographed on huge negatives, and showed us that they weren’t sharp, but rather smooth. The reason, he said, is that once a sensor achieves 30 megapixels the pixel density is high enough that pixelation effectively goes away. Here’s a bad photo of the presentation slide that illustrates this concept:


In HD, and even 4K, curves and diagonals are stair-stepped, and the additional edges result in an artificial sharpening of the image that’s not a good approximation of how our eyes see the world. Beyond 30 megapixels, resolution becomes so fine that, on a big screen, these extra edges disappear, and the image becomes smooth without losing detail. The contrast created by hard edges fades away, leaving softer edges around finer textures. We still see everything in the image, but the sense that it has been artificially sharpened—even at a low level—vanishes.

I think this is true. It makes sense both logically and intuitively. And… I’ve seen the 8K images, on a big screen and in 4K, and they looked super smooth.

Once again, JPEG compression is preventing you from seeing all the detail I can see in the TIFF original of this image. Still, there’s quite a lot of texture here, and none of it seems forced. It just IS. It’s smooth and rich with detail, all at the same time.

In addition to the elimination of stair-stepping, I think 8K looks smooth because it captures a much more subtle range of tones than do lesser formats. Lower resolution images should contain higher local contrast because transitions between light and dark become more abrupt. The additional resolution afforded by an 8K sensor captures more shades of gray across smaller areas, which reduces contrast by adding tonal complexity.

Was I blown away by the look of the camera? From a resolution perspective, yes. From a color perspective, I was impressed but not stunned. That’s fine, though: there are a lot of great looking cameras on the market and they all fill their particular niches. Panavision now owns the highest resolution large format motion picture camera system on the market. It makes beautiful images and the design is very functional and smart. While it uses a RED sensor and electronics, they’ve pulled new characteristics out of a known sensor design. They’re also working on a new OLPF/IR cut filter package which will likely further change the look.

The footage I saw was shot with prototype cameras, and I’ll be very curious to see what kind of images these cameras produce when they are released late this year or early next.

Postscript: I asked Michael Cioni to look over this article for inaccuracies. In his emailed response he included the following information:

1) We are still in the very early stages of this system and developments are moving rapidly, which is to say the pictures you’ve seen thus far are the worst they’re going to ever be

2) We want to be transparent with the cinematography community and want feedback from users so we can improve the system based on user feedback

Panavision has enacted a “rapid response team” and it’s specifically designed to make changes fast so that we don’t have to unnecessarily extend our development timeline.  What I’m hoping is that with each passing season, people will be impressed as to how much faster the DXL roadmap is as compared to other top flight cinema systems.  Announcements will be made as upgrades and updates are ready.

And this, ladies and gentlemen, is how you run a camera company.

DXL_Rt_Side_Panavised with Lens

Was This Post Helpful:

0 votes, 0 avg. rating

Support ProVideo Coalition
Shop with Filmtools Logo

Share Our Article

Art Adams is a cinema lens specialist at ARRI, Inc. Before that, he was a freelance cinematographer for 26 years. You can see his work at http://www.artadamsdp.com. Art has been published in HD Video Pro,…