Dragon vs Alexa color accuracy?

[Ron Michaels]

Hey guys/gals,
Between the dragon camera's and their corresponding color spaces and the Alexa line, which of the 2 have better/more accurate color reproduction under any given light source? From my understanding, color has never been RED's strong suit, especially under tungsten light. I have also read that the blue channel can clip with vibrant blues. Now, I know all digital camera's favor daylight, however I was wondering what everyone's professional opinion was regarding which of the two is most versatile under varying lighting conditions.
Thanks,
Ron

 

[Blair S. Paulsen]

The Alexa has solid color science and credibly mimics the spectral response of Kodak 5219. RED Dragon with RG4/DC2 on STH OLPF can produce quite accurate colors, perhaps even slightly more technically accurate than Alexa in terms of scene referred values. That said, which is more important - prettier or better matched to the real world?

FWIW, in the early years, colorists used to find RED footy frustrating to grade, but with the latest color science and using log encoded DPX or OpenEXR files in the leading grading tools - both Alexa Log-C and RED Dragon RLF (or the latest WideGamut iteration) are very gradable. It's important to remember that R3Ds are Bayer data sets - not color images - those require construction via algorithm. How they look as RGB pictures is highly dependent on the choices made in the RAW development process.

Jarred recently posted that a new iteration of Graeme's color science is coming in October - for both Helium and Dragon. Additionally, if I understand it correctly, Panavision/LightIRON is promising a new color science for the Millennium XL - presumably tied to newly developed OLPFs - that is claimed to better isolate skin tones from IR pollution and reproduce brighter colors more like film does by remapping saturation to better mimic a subtractive vs additive color mode.

Cheers - #19

 

[Marc Wielage]

Between the dragon camera's and their corresponding color spaces and the Alexa line, which of the 2 have better/more accurate color reproduction under any given light source? From my understanding, color has never been RED's strong suit, especially under tungsten light. I have also read that the blue channel can clip with vibrant blues. Now, I know all digital camera's favor daylight, however I was wondering what everyone's professional opinion was regarding which of the two is most versatile under varying lighting conditions.

I think there's a lot of "it depends" in the answers.

I honestly think both cameras can provide usable pictures under optimum conditions. Where I've seen Red pictures go bad is when shooting under tungsten light (nominally 3200°). I've also seen problems with mixed lighting conditions, particularly certain combinations of florescents and tungsten.

The best answer for you might be to see if you can borrow or rent both cameras over a weekend and put them through a torture test. Shoot at varying exposures, different kinds of lighting, look at what happens to the picture 2 stops over or 2 stops under. I'd also look at what happens to color charts (like the DSC charts) under varying light conditions and under optimum light conditions.

Geoff Boyle on the Cinematography Mailing List (CML) has done a bunch of tests of different cameras and drawn different conclusions:

http://cinematography.net/

It's fair to say there are subjective areas that boil down to camera handling, available accessoies, cost, rental availability, picture quality, ergonomics, dynamic range, color accuracy, file size, and other issues. You can make a good argument that there is no one camera that hits 100% in every category.

I have to say, when I'm handed an Alexa project to handle, I seem to run into fewer problems; my take is that the camera works well even under problematic conditions. But I have also seen Red projects that came out beautifully, because the DP really took the time to learn where the limits were and exposed right in the sweet spot. I've honestly had more problems with low-end Canon projects (like the C100) than I have Red or Alexa. No problems with Sony F55. In truth, even 35mm negative can be very, very challenging under some conditions.

 

[C. Burkhart]

. Where I've seen Red pictures go bad is when shooting under tungsten light (nominally 3200°). I've also seen problems with mixed lighting conditions, particularly certain combinations of florescents and tungsten.

This. ^
Alexa has more tolerance in those situations.

 

[Daniel Stilling]

The accurate color issues are something from the MX sensor days. With the Dragon those issues have gone away, but people keep perpetuating this stigma that no longer holds truth.
Dragon renders both daylight and tungsten beautifully. Mixed color temperatures is not an issue at all as I mix the color temperature in my lights constantly for some color contrast.
Both Alexa and RED are today the top in digital imagery creation. They are different from each other and I see the Dragon sensor has an edge over the Alev in image quality.

 

[Hrvoje Simic]

I have also read that the blue channel can clip with vibrant blues.

It can.

Doesn't have to, though.
Things turn out depending on the process.

 

[Bob Gundu]

The vibrant blue issue isn't the sensor but the colour science. If blues go strange at times, switch to DragonColor1 instead of 2. The beauty of shooting raw, all this can be changed after the fact.

 

[Brett Harrison]

Here's the most neutral comparison you can get under tungsten light (x2 redheads), using the same delivery colour space from respective manufacturers' ACES debayer option. I completed this test. The purpose was to take the RAW data and put it through the same display prep, to remove any 'look' imparted by the manufacturer's LUTs or 'beauty transforms' as they've been called, such as RG4 or -2Rec709 Please see the below chart for the workflow:

For individual whole frames please see this reduser post: http://www.reduser.net/forum/showth...eking-advice&p=1640211&viewfull=1#post1640211

 

[Graeme Nattress]

I think you can have accuracy and beauty. It's also important to read this thread: http://www.reduser.net/forum/showthread.php?148272-RedWideGamutRGB-discuss!&p=1654778#post1654778 about REDWideGamutRGB and see where things are heading.

Graeme

 

[Ron Michaels]

Based on Brett's test, the Arri all the way. The RED seemed to be having trouble resolving all the colors under tungsten. Whereas the Arri looked to over saturate a bit.

 

[Hrvoje Simic]

Based on Brett's test, the Arri all the way. The RED seemed to be having trouble resolving all the colors under tungsten. Whereas the Arri looked to over saturate a bit.

You are judging camera performance based on results which are a product of camera performance and transformations of their material.

 

[David Mullen ASC]

Color "accuracy" isn't the end-all and be-all -- when I used to shoot film, sometimes I preferred Fuji or Agfa, even though Kodak was always the most "accurate". What you don't want, however, are surprises, something turning out oddly that wasn't expected. And you want consistency, you don't the colors to behave differently in different scenarios and settings within reason.

 

[Corey Robson]

The vibrant blue issue isn't the sensor but the colour science. If blues go strange at times, switch to DragonColor1 instead of 2. The beauty of shooting raw, all this can be changed after the fact.

Agreed. For night / tungsten scenes I use Dragoncolor1 / LLO OLPF as a starting point for matching our Alexas. It doesn't oversaturate highlights / colored LEDs as much as Dragoncolor2, skintones still look great. No idea if it's technically possible but I hope Alexa and Weapon can eventually achieve more natural looking highlight saturation and rolloff characteristics similar to 5219. Also Weapon shadow areas are a bit undersaturated for my taste with both DC1 and DC2.

Can't wait to try RedWideGamutRGB in camera when it's ready. More excited about that than Helium!

 

[Peter Lyons Collister ASC]

I have to agree with David on this. Conor accuracy is not the end all. I used to love Agfa and some Fuji reversal films. They were really hard to shoot and had to be tested ... What is most important is skin tone reproduction. Alexa was always the leader and now I think Red has come through and levelled the playing field. We, as DPs, are judged many times by how the leading lady looks.
Graeme has listened and I think Red has reached parity.

 

[Ron Michaels]

Would have been interesting to see Brett's scopes....

 

[Marc Wielage]

Agreed. For night / tungsten scenes I use Dragoncolor1 / LLO OLPF as a starting point for matching our Alexas. It doesn't oversaturate highlights / colored LEDs as much as Dragoncolor2, skintones still look great. No idea if it's technically possible but I hope Alexa and Weapon can eventually achieve more natural looking highlight saturation and rolloff characteristics similar to 5219.

That's basically what I do in correction. After using Dragoncolor1, I have some curves and chroma/hue settings applied to RedLogFilm that generally wind up as a great starting point. I came at this using the DSC Chroma DuMonde chart a couple of years ago, and it helped me realize where Red color science was going and how I needed to make my adjustments to get it in the ballpark of normal. It worked fine.

I do the exact same thing with Alexa, and those adjustments have stayed pretty accurate most of the time.

 

[Ron Michaels]

The RED definitely shows far more noise in the shadows under tungsten. Don't think the noise issue makes much difference with dragoncolor 1 vs 2. As far as clipping goes, dragon color 1 seems to be better in this regard but de-saturates the whole image. Whereas dragoncolor 2 brings EVERYTHING up; blues, reds, greens and everything in between. I know color accuracy isn't the most important aspect of the image, however I am going to be working on a very colorful project with plenty of tungsten.

 

[Hrvoje Simic]

In the context of creative choice color accuracy is more relative factor with more room for offsets.

In the context of camera profiling and defining and shaping its base response, there is less room for offsets.


In a way, film "absorbs" the effects of light, whereas digital "adds" them (loosely put), by its linear principle and other principles of digital signal processing, so tends to enhance some things which film can reduce and hide. New digital cameras are much more sensitive to light and compared to digital cameras of the past era the image is sampled in much greater precision. Film is an analog medium with random chemically joined and responsive elements, sensor pixels are separated and uniform light gathering elements practically independent and individually quantized. Also, with digital, as the signal values increase the offsets tend to enlarge because it becomes more difficult to accurately quantize the signal as it reaches the limitations of electronics.

Those factors combined when meeting light variations lead to more prominent and less natural-looking tonal offsets. This is why there has to be a more accurate basis from the start of the image chain, because those offsets are not of the same type as with film and they can complicate further post work.

Alexa "limits" particular signal properties which reduces some of the offsets. This helps with certain image aspects but that has a price of limitations in others. Alexa is also not immune to problems with effects of certain light scenarios, which can be seen more rarely because it has less online exposure of dirty laundry.

With lighting choices, light quality and post only one improper part is sufficient to mess up the image properties while all three done properly are necessary to make it shine. Red has a wider market coverage and Arri route more often defines level of all of those three elements, together affecting the final outcome.

There is a tendency to focus on the image capture device properties when seeing the final results while disregarding those three key factors. One example is a great underestimation of the effect of poor light quality or light pollution, especially when bathing a human skin, captured by a highly sensitive digital camera. Brain "irons" white balance and other offsets perceptually adapting to the scene but cameras are less forgiving.

 

[Brett Harrison]

You are judging camera performance based on results which are a product of camera performance and transformations of their material.

I think I get what you mean Hrvoje, but the truth is also, that without transforming the image to a display-ready format (REC709 monitors don't care if it's linear to light, or tone-mapped for our sake) we can't make a subjective evaluation, which is what we're doing when looking at a side-by-side comparison. Therefore the most neutral subjective comparison is to transform from ACES using the most modern display transform maths available from SMPTE for REC709 monitors. This way we will see the color difference for a given scene between the cameras, given that ACES is a calibrated colourspace for both cameras.

The scopes were all very small and no-where near clipping. I'm planning to repeat the test with the DSCLabs chart I got at NAB this year to go further and do this test in more depth with some varied scenes.

 

[Ron Michaels]

It's about as agnostic as you can get between the two.....