July 30, 2014

Dueling with DRAGON: RED's 6K Sensor Takes on Older MX in This EPIC Camera Shootout

With DRAGON now available for purchase or upgrade in RED's EPIC and SCARLET cameras, it's about time to see how well the new sensor outperforms its predecessor, the Mysterium-X. Cinematographer Ryan Walters has recently done just that at Indie Cinema Academy. His battery of tests compare everything from overall latitude, highlight handling, recoverable underexposure, low light performance, IR pollution, and more between two EPICs -- one old and one new. Needless to say, if you're thinking of upgrading or simply wondering how all the DRAGON hype stacks up to reality, you definitely want to check out his results.

Headroom & Footroom: Exposure Latitude/Dynamic Range

Ryan's first series of comparisons evaluate what may be the most important -- and one of the most hyped -- of DRAGON's improved capabilities: exposure latitude. He also dips the DRAGON's toes into the icy ponds of underexposure, where things get more subjective than in the case of clipping highlights.

While DRAGON's greater highlight shoulder is undeniable, in this case the MX seems to have a leg up on the newcomer in recoverable shadow details. Look no further than in the underexposure examples Ryan provides in his post, wherein it appears the MX preserves darker areas significantly better than DRAGON does. Even in the relatively tiny crops below, the difference is immediately apparent with a glance side by side (courtesy Ryan Walters & Indie Cinema Academy):

So why is DRAGON going so dark so quickly? Possibly as a by-product of the new OLPF for DRAGON, which does improve several things over the old OLPF. RED's Jarred Land had this to say over at REDUser (my emphasis):

There still people that prefer the old OLPF because even though skintones and highlight protection isn't as good as the new OLPF, they want a noise free image at higher ISOs that the old OLPF gives you by letting in more light. We of course will put whichever OLPF you want in your camera now -- but the new OLPF module will make things a lot easier to change.

RED is currently in private REDCINE-X testing phases for a feature called DRAGON Enhanced Blacks (or DEB), which DP Phil Holland says, "is tackling the 'red specks' that effect midtones, darks, and shadows." DEB doesn't directly address new-OLPF DRAGON's dynamic range -- or doesn't seem to, from the sound of things -- but at least some remedy is on the horizon for noise in darker areas of the image. It's also possible that improvements in image processing for new-OLPF DRAGON material could further aid shadow detail sometime in the future.

Filtration: Diffusion, ND, & IR

Next up, Ryan test drives the DRAGON's ability to cope with lots of ND filtration, which would normally start to introduce infrared pollution depending on the camera. While the ND does exactly that in the MX, this is an arena in which DRAGON soars high above:

DRAGON's robust insulation from IR means you have to worry a bit less about choosing the right neutral density to knock down exposure, allowing greater flexibility than the more vulnerable MX.

Quality & Accuracy: Fill, Compression, & Color

In his final video, Ryan tests remaining factors that influence the final image's beauty: skintone and color reproduction, shifts due to white balance, effective fill ratios, and compression artifacts.

With DRAGON, it seems less tweaking is necessary for natural skin and color balance out of the gate than with MX. Plus, with more spatial resolution than any prior RED sensor, EPIC DRAGON can oversample for its targeted destination even more so than the MX. This should make for less perceptible compression artifacts when downscaling to a given delivery format. Once again, however, DRAGON's shadow-area noise hurts its performance -- in this case at fill ratios which the MX handles just fine.

Incubating Some DRAGON Eggs

The issue of how DRAGON is 'better' than MX is a complicated one -- different configurations can affect the final image, and preferences will vary by shooter. Some may be disappointed with how quickly DRAGON drops into darkness, especially considering some of RED's optimistic earlier claims about its sensitivity. For those users, RED's modular OLPF program may come in handy. After all, Jarred Land has also said that "you can choose whichever [OLPF] you want or all of them and change at will."

Others may count new-OLPF DRAGON's greater headroom as a fair trade, taking shallower shadows as a sacrifice for stronger highlight handling. As Ryan indicates above, if you're mainly looking for a camera that sees in the dark, you may be looking in the wrong place. Whatever the case may be for you, hopefully now things are a bit more clear. There can certainly be some ambiguity separating a camera/sensor's specs from its actual imagery -- Ryan's tests help cut a greater contrast out of that gray area in between.

Be sure to check out all three of Ryan's original posts at Indie Cinema Academy in the links below. There you'll find pixel-peepable downloads, further discussion in the comments, and a place to thank Ryan for all his hard work!

Your Comment

31 Comments

This is a great series, glad someone did this, there were a lot of threads on Red's own message board that just got shut down because people would bring this up. Have had a Dragon for months now and it's been a battle working with it compared to the Epic. The 2000 ISO Dragon equal to 800 ISO MX is a pretty big bait and switch on their part, you really have to shoot it lower ISO than MX. As well as the measured 13.5 stops of DR versus the 16.5 that they claim, such a let down. The past couple years have been rough on Red's reputation after making a camera like the Epic which was so far ahead of it's time.

July 30, 2014 at 2:52PM, Edited September 4, 11:56AM

0
Reply
Phil McTaggins

the problems with the dragon remind me of the issues when the Epic first came out and the Red One first came out - this time though they are dealing with it without Jim Jinnard. This is again further reason why no one should ever buy a camera especially an expensive one until they have been out for a while and many tests have been done. You are going to get a lot of surprises.

July 30, 2014 at 3:49PM, Edited September 4, 11:56AM

0
Reply

For sure. My only beef is that now this is their "3rd" camera, you would think that with the first two they would have learned...

This is still some great tech, but there is definitely a price to pay if you want to be on the bleeding edge and have the latest piece of kit...

July 30, 2014 at 4:14PM, Edited September 4, 11:56AM

0
Reply

RED has consistently set deadlines for themselves and rarely, if ever, hit the mark. You'd think with all the delays they'd at least have actually fixed some of these issues beforehand.

I think it's about time the company stops doing the minimum viable product (MVP) testing on their consumers, and releases things that are actually usable.

....Want a REDmote? It doesn't work most of the time, but now we have a REDlink!
....Want to travel light and compact? Got it! You just need 8+ REDvolts, which last 25 mins each, and 3 of these chargers to keep you going for a day.

Let's not even bring up the Meizler module, which never saw the light of day and is now being broken down into separate components so RED can increase their revenue...

July 30, 2014 at 4:27PM, Edited September 4, 11:56AM

4
Reply
Marco

no doubt- and that is an expensive price to pay... ugh...

July 30, 2014 at 6:02PM, Edited September 4, 11:56AM

0
Reply

The Scarlet deadline was like 2 years in the making!

July 31, 2014 at 5:40PM, Edited September 4, 11:56AM

0
Reply
Kurt

Thanks- I'm glad you liked it. We had fun testing it for real, and putting the results out there- instead of hiding them. :)

It is unfortunate that Red hasn't been able to deliver on the hype that they gave this new sensor. It is GREAT, and I like it A LOT more then the MX. But it isn't living up to what they said it would do.

July 30, 2014 at 4:12PM, Edited September 4, 11:56AM

0
Reply

Great job Ryan! As a recipient of a recent Dragon upgrade, the whole 2000iso is the new 800iso is the most disappointing part of this ride. I will admit that the highlight roll-out and color science seem drastically improved but the supposed improvements in the low end and latitude definitely come across being"over promised, under delivered". Still love the camera and their customer service is awesome, but yeah, too much rah rah without the accompanying substance. Feels like this new replaceable OLPF thing is to make up for the fact that they realized they can't squeeze what they had originally intended into one package.

July 30, 2014 at 5:10PM, Edited September 4, 11:56AM

0
Reply
Ryan Sauve

Not to mention what they did at NAB this year- a bunch of women in bathing suits to distract from the fact they had nothing real to show...

Yeah, the "new" replaceable OLPF seems to be more of s fix then a real solution.

Red's customer service IS FANTASTIC. I have always had a great experience with that. I had my camera go down one day before a shoot, and they overnighted me a replacement while mine went back to them for service, and it didn't cost me anything extra. Who does that? I would never get that from Sony or Canon.

Unfortunately their customer service doesn't line up with the rest of the company- if it did, they would really be on top. IMO anyway...

July 30, 2014 at 5:30PM, Edited September 4, 11:56AM

12
Reply

I agree Ryan. Great customer service but seems like RED jumped the gun with the dragon. The new Standard OLPF seems to be a pretty great fix though.

February 5, 2016 at 3:05AM

0
Reply

Ryan, Thanks for doing this. I haven't gotten my hands on the Dragon til yesterday actually, a friend of mine just got his hands on a Dragon Carbon. I got to play for about 5 minutes in RedCine X and was noticing some surprising results.

One thing I noticed is that their was far less noise then the MX on Mid-Tones and Hi-lights from the Dragon, I did however notice overall there was more lost in the darks, and more noise in the darks, and didn't seem to easy to bring up. Maybe i'm off and i'm not sure how the camera had been blackshaded before the shots I looked at. But it was a bit underwhelming at 800 ISO to be completely honest.

I've got a place in line for the Dragon, and have contemplated and come close to pulling the trigger a few times, but i'm extremely happy with My Epic, and think I will stay the course awhile longer.

Thanks for the time you put into doing this!

July 30, 2014 at 6:17PM, Edited September 4, 11:56AM

0
Reply
Jeffrey Loewe

Welcome. :)

Yeah, as others have noted here i the comments- "ISO 2000 is the new 800" is definitely not true with this camera. Even at ISO 800 there are noise issues that people don't like. (I find 800 on the dragon to be fine.)

The Epic MX is a solid camera that can produce great imagery when there is talent behind the lens. :) So I'm glad that you're getting that imagery. No sense in laying out that extra cash if it doesn't get you what you want in the end.

July 30, 2014 at 8:54PM, Edited September 4, 11:56AM

2
Reply

Epic videos (pun intended). Well done that man.

July 30, 2014 at 6:44PM, Edited September 4, 11:56AM

6
Reply
ADC

Thanks. :)

July 30, 2014 at 8:51PM, Edited September 4, 11:56AM

0
Reply

My thing with the 2000ISO as 800ISO, was that they basically said they accomplished it with smaller photosites, which is an engineering impossibility. I should have know something was up when they were claiming that. Anyway, great video series, lots of hard work went into these. Hopefully this keeps everyone educated and camera manufacturers honest when it comes to delivery.

July 30, 2014 at 9:08PM, Edited September 4, 11:56AM

2
Reply
Phil McTaggins

I saw several comments from people knowledgeable about CMOS processes disputing the claims made by Red years ago. Red never had the tech they claimed to have. Or, to be more specific, they did not have sensor manufacturer with that tech.

Sony had that level tech (= see F65, F55), but obviously would not sell it to Red.

July 31, 2014 at 8:14AM, Edited September 4, 11:56AM

0
Reply
Dusty

And this is the reason why the Alexa and the C300 have been so successful. They work, out of the box and you don't need to do a not of work in post to get accurate color.

I generally don't get into these discussions about which camera is the awesomest. But I had high hopes that Dragon would be a compelling alternative with the early buzz, but from this is just seems clear that it doesn't deliver anywhere near promised. That's not to say that it's not a very capable camera in the right hands. But it seems counter intuitive to pick an instrument that requires this much post work.

That's just me.

July 30, 2014 at 9:49PM, Edited September 4, 11:56AM

0
Reply

And Ryan, didn't you post about how you sold your RED for a C100? How is that working out? Have you moved on or still rocking it. I love my c100. it does what I need and makes it pretty simple to work with.

July 30, 2014 at 9:50PM, Edited September 4, 11:56AM

0
Reply

Looks like a lot of upset customers who spent $50K + to get a RED Dragon up and running for the advancements it would have over the MX sensor. 6k isn't needed for one...as 4k isn't even really needed. An Alexa proved that time and time again and Arri spent there time manufacturing the science of film; the colors, the range and reliability. I'm a fan of the Epic and the range is already 13 stops give or take and if you know how to light, that's all you will need. If you a Blackmagic 1/8 filter in, which will help get rid of the sharpness of digital as well as give an illusion of smooth highlights due to blooming, you're golden. 5k, 96 FPS, compressed RAW, all wonderful, and more then enough for the time being. I had similar issues with an MX though, especially with underexposure being very grainy and ugly. Hopefully they figure out the issues with the dragon because I really had high hopes and was rooting for RED to overthrow Arri as we all are fans of an underdog. German engineering is just superior in many ways though. One good thing though, the prices of MX Epics have fallen and rental rates have gone down as well.

July 30, 2014 at 11:51PM, Edited September 4, 11:56AM

0
Reply
Brad Watts

great points brad. Thanks for mentioning the black magic 1/8 filter. I am going to try that with my scarlet.

February 5, 2016 at 3:18AM

0
Reply

The latitude test is weird. The latitude graph says that Dragon has a stop and a half more in the highlights and is equal in the shadows. Maybe it has more noise there, but it should show the same shadow detail as the MX.

The pictures, though, show like two stops less of shadow detail in the Dragon.

July 31, 2014 at 5:25AM, Edited September 4, 11:56AM

5
Reply

It looks like the test stops too early in the highlights with the Dragon.

It should have gone on at least until +7, since the Dragon is like a stop less sensitive than MX and has a stop and a half more headroom in the highlights.

July 31, 2014 at 6:01AM, Edited September 4, 11:56AM

3
Reply

I'm glad someone else saw the distinction between actual shadow detail and if your still getting a "signal." Shadow detail was a strength of the MX, contrary to what Brad said, underexposure was superb if you know what your doing. Coming from MX, and working with Dragon, underexposure now is a weakness. Without overexposure or rating for lower ISO, the shadows do not carry fine detail.

The bigger budget guys don't care, overexposing a scene two stops while maintaining aperture is easy. For Scarlet Dragon guys like me where lighting budgets are tight. Its probably going to be more common to be shooting at 2.0 instead of 4.0. My AC is not happy. I don't really care about the 2000/800 promises. The only thing that really erks me with the Dragon is that I thought I was going to have more options of where I expose middle grey both over and under at 800 iso. I actually feel more constrained on Dragon. Rating for 800 iso in some instances yields an unacceptable image. We will see what DEB can do.

July 31, 2014 at 12:34PM, Edited September 4, 11:56AM

0
Reply
ryan

But if there's signal, there should be detail. I mean, the camera can 'see' a stop less.

Epic MX was an underposure beast, that's true. I didn't understand why many movies systematically let the fluoerescent fixtures burn out, for example. With minimal lighting and underexposing you could get them in range.

Anyway, I've seen samples where MX and Dragon are much more closer in shadow detail, like this one:

https://www.flickr.com/photos/123436930@N07/14457270568/in/photostream/

I don't really understand it. Maybe there are different sensor batches?

July 31, 2014 at 1:09PM, Edited September 4, 11:56AM

5
Reply

I tested differences myself now - definitely Dragon crushes shadows - MX was great, is great, and an Epic MX is still a great bargain and really no cameras are simply better than it, only different from it.

Dragon I would not shoot with at this time.

August 6, 2014 at 7:21PM, Edited September 4, 11:56AM

0
Reply
Robert Ruffo

Man, I'm happy with my Scarlet and am in line for the Dragon upgrade, but now I think maybe I would prefer to just upgrade to an Epic MX.

July 31, 2014 at 12:49PM, Edited September 4, 11:56AM

7
Reply

Epic MX is still a helluva camera. Dragon is great but this just shows you how "ahead of its time" the Epic MX was!

July 31, 2014 at 1:44PM, Edited September 4, 11:56AM

0
Reply

Like I said, Epic MX is still the better camera in my opinion, overall. It was amazingly ahead of its time - it would be if it were released today. There's really nothing I can't do with out little wonder. Dragon is extremely disappointing.

August 6, 2014 at 7:24PM, Edited September 4, 11:56AM

0
Reply
Robert Ruffo

For whatever reason (it was surmised it may have been the daylight balanced lights used), there's something wonky going on in the shadows in this test that is atypical of what I've seen from every other new OLPF Dragon .R3Ds I've downloaded, including those with daylight sources. I encourage everyone considering buying or upgrading to download as many .R3Ds they can get their hands on to get a better understanding of what Dragon is doing in the shadows. Scientifically speaking, you should never draw conclusions from a single test or a handful of data points.

No, the new OLPF Dragon is not quite as clean as MX in the shadows and the red specks are a problem but this test is by far the worst I've seen it. There's something just not right (typical) about the way it's behaving.

Ryan, how do you think the results from your test compare to what you've seen in other new OLPF Dragon .R3D files? I'm not saying the test was flawed. I am saying that this looks to be a worst case scenario. While informative in itself, it's not necessarily representative of what people would normally experience, based on the dozens of new OLPF Dragon .R3Ds I've looked at.

July 31, 2014 at 1:35PM, Edited September 4, 11:56AM

3
Reply
Brian

In the end, it is all about a silicon surface absorbing light. You can tweak this or that parameter, but it all comes at a price. True progress in this field is rather rare; I think that the last one was the invention of back-lid sensors.

The next real advance may come from introducing a completely different light absorbing surface, like quantum grains.

August 1, 2014 at 1:14AM, Edited September 4, 11:56AM

1
Reply
Thyl Engelhardt

The disappointing thing is that even with 6K res and all that they have done , its still off the Alexa . The resolution definitely helps in making up for that but i doubt RED will be able to get the same detail at 2K which the Alexa manages to extract from that sensor . Secondly the ISO performance was not that impressive , the noise on the C100 at 20,000 seemed much much better then the Dragon . 6K n all that just seem like a distraction , a combination of a slightly larger sensor combined with more resolution to give you an illusion of better DR .

August 1, 2014 at 3:36PM, Edited September 4, 11:56AM

0
Reply
Karan