Red Dragon: Candle and Match Low Light Tests

[Phil Holland]

There's been a bit of talk and curiosity about Dragon's low light performance and I thought I'd put together a couple of quick tests and share some quick notes.

In the spirit of other tests online that are designed to "break cameras" I shot these with a White Balance 2800K, so this would be considered a stress test. The White Balance will also show one of the ways Dragon separates itself from Mysterium-X. Both tests were shot at 6K HD at 5:1 REDCODE RAW and both exhibit the ISO range of ISO 800, 2000, 3200, and 5000.


Epic Dragon Candle Test


- 6K HD JPG Still
- Vimeo Preview Link
- ProRes4444 1080p Download Link (3.1GB)

*notes: Shot at 35mm T2.1 using straight REDcolor3/REDgamma3 with no additional curve or noise reduction. The light meter is set to 23.976fps w/180 degree shutter at ISO 3200 if you're curious.


Epic Dragon Match Test


- 6K HD JPG Still
- Vimeo Preview Link
- ProRes4444 1080p Download Link (2.4GB)
- 6K HD .R3D Download (1.4GB)

*notes: Shot at 75mm T2 using REDcolor3/REDlogFilm with a fairly standard S-curve to create a pleasing look. At the end there is an additional ISO 5000 sample that I exported with Neat Video to show what that can do to eliminate visible noise at that high ISO.


My thoughts

First up I'm not as handsome as Leo and I didn't use T1.3 Masters, however, this gets the thought across and shows what's going on with Dragon in low light. However, as noted with the Leo/Fincher/Match/Cigar Test ProRes actually falls apart a bit here and the 16bit DPX frames tell the best story, but ProRes4444 is the highest quality I can reasonably get you guys at the moment.

Dragon's specs state "improved low-light capabilities let you shoot at ISO 2000 and still produce a clean image fit for theatres or publication" and that's pretty much what I see here. The theatrical quality range on Dragon to me is ISO 250-2000. If you can handle a bit of texture and you aren't going nuts in post, ISO 3200-5000 is pretty darn usable straight away as well. However, if you are shooting for theatrical purposes and have to stretch that far and want a "clean image" noise reduction is a viable option at those top tier ISOs.

These tests were produced at a White Balance of 2800K, which strays away from the native sensor White Balance of around 5000K. However, it's performing very nicely while pushing the material this far and even further to 1800K, which is not shown here. Side Note. It's good to know the controls in Redcine-X Pro. There is an option in there to "denoise" that will eliminate some color noise when utilized, I mention this as I've met people "who don't know what that does". It's useful for shooting low light and pushing the White Balance around. The older Mysterium-X practices of using a CTB filter still can be employed, but there's far less of a reason to do so with Dragon. This is a good thing.

The key to working with Red cameras (and other digital cinema cameras for that matter) is understanding that there is a noise floor and once you hit it with your exposure rating it's best not to lift anything out of that darkness. The ISO 5000 samples show the extreme of dipping into that dark bucket and bringing out the noise. Dragon does in fact see quite a bit into the darkness and the current calibration shows that. There's nice clear tonal separation at the bottom end and that's going to be fun to explore.

Another note a bit away from low light shooting. I've been stress testing the camera over the holiday and something that is very interesting with Dragon is on well lit and well exposed images that live above the noise floor you have rather insane flexibility in post. Pushing and pulling both luminance and chrominance 4 or 5 stops is totally doable. Especially if you have scenes that benefit from exposing to the right of the histogram. If you have an exposure that lives above the noise floor you can push and pull the ISO a lot. So keep that in mind.

I'd be curious to see if there's potential room for improvement with newer color science optimized for Dragon. I'm already damn impressed by the color straight out of the camera, but since DRAGONcolor and DRAGONgamma were hinted at it's certainly making me wonder where we can go from here.

How does it compare to other cameras on the market? I don't know exactly yet, but based on the ISO 3200 samples from my previous testing and what's been posted elsewhere I can say Dragon looks better than that at the same ISO. Likely when the world starts spinning again after the new year I'll test it out against the usual suspects. It's not a really needed test, but curiosity will likely get the best of me at some point. I know it will bite others who want to see what's going on exactly.

How am I shooting and exposing with Dragon so far? I'm pretty much working within ISO 250-800. 2000 here and there, but I tend to like the balance and distribution of tones around ISO 500-800 the most. Just a trend I noticed myself doing after a few weeks with the camera. One thing I will point out and I've written about it before, is just how nice Dragon handles overexposure and highlight roll-off. It clips, but to me you can feel and see it "burning out" in an organic/analog sort of way. I'm describing this at the moment as a minimal digital footprint. It's such a nice touch and really separates Dragon from not only Mysterium-X, but also other digital cinema cameras.

That's what I got for now.

Here's a rainbow to finish out the year

 

[Bob Gundu]

Hey it's Leonardo DiCaprio! Thanks for the tests! BTW, are these Xenars?

Never mind... I just noticed they are.

 

[Jason Kraynek]

nice colors around the hand and the match (red, purple, pinks). love it

 

[luigivaltulini]

Thanks Phil

 

[Sean Cruser]

glorious stuff. cannot wait for more!

 

[Peter Lyons Collister ASC]

Really appreciate the test Phil. Did you use any de noise in RCX? And what firmware are you using. Back from France on the 11th and maybe we can grab lunch and chat.

Or just meet up at RSH.

Bonne Fete

 

[Phil Holland]

Thanks guys. Hope this one is a useful and informative one. I should also mention, these are best if viewed in a dark room on a calibrated monitor to really see the range.

Really appreciate the test Phil. Did you use any de noise in RCX? And what firmware are you using. Back from France on the 11th and maybe we can grab lunch and chat.

Or just meet up at RSH.

Bonne Fete

That sounds great. I should be done with my "nerd tests" by then, so I'll have a good amount of info. I used the Denoise at 6 in RCX v.22.1.31200 in this case. This cleans up some of the chroma noise only visible at the higher ISOs. It's also available in most of the 3rd party plugin/decoders as well.

 

[Wes Howell]

impressive. Thanks for doing this.

 

[AndreeMarkefors]

Nice, Phil! Thanks!

 

[Dr. Sassi]

Very interesting Phil, thanks for all your efforts and that you share it with us.

My "New Years Resolution" for 2014: 6K ;o) Hehe, I so look forward to it.

 

[David Battistella]

Love the catchlight a single match can produce.

Thanks for the test Phil.

Any plan for an r3d?

Battistella

 

[Phil Holland]

Forgot to add one bit to answer one other question I've seen pop up about the "frequency of Black Shading" with Dragon.

Dragon's calibration seems to be a bit broader than Mysterium-X, likely due to the lower noise floor or some other voodoo.

For instance, typically at 6K HD when I black shade at 1/48th for a 23.976 project I'm able to get 1/24th cleanly as well as up to I think 1/144th or so. Which means I haven't needed to black shade while maintaining a 180 degree shutter at 72fps.

Still doing some testing on that though. Just saying it is "broader". Black Shading is taking about 9 minutes. The temperature is pretty darn regulated, so I imagine you really only need to run a Black Shade in rather extreme temperature changes or massive shutter speed changes. I'd say it's a good idea if your shutter speed escapes 3 stops of base calibration. Maybe a bit more than that. Again, still testing (and this is the boring type of testing).

I shot in some cold weather recently too. Don't know all the details on the in camera sensor heater thing, but camera got up to temp and I was off and shooting nice and quickly. I was doing indoor and outdoor Black Shade tests at the same frame rates and didn't see any difference in noise. So that should answer that question about Black Shade frequency. I'd say if you're dealing with over a 35 degree weather change from last calibration, it's probably a good idea to spend that 9 minutes. I'm not sticking my camera in a freezer to test further though. I think Red does this however.

 

[David Battistella]

Phil,

may I ask?

have you done any testing with the frame averaging/frame summing modes?

i get very clean results on MX, but wondering about dragon.

Thanks,
david

 

[Alain M]

Thank you Phil! once again you providing awesome info to us!!!

 

[Elsie N]

Phil, I would be interested in hearing your preferred settings for 48 fps time base

 

[Brice Ansel]

Thanks Phil for sharing.

Any plan for an r3d?

Battistella

And joining David on the request.
Cheers
B.

 

[Phil Holland]

Any plan for an r3d?

Might upload a 10 second trim later today.

have you done any testing with the frame averaging/frame summing modes?

Frame Processing is currently not enabled on Dragon (it's in the DSMC Ops Guide). Don't know exactly if it's coming soon either. HDRx sounds like it's coming sometime January according to one of Jarred's posts.

Anamorphic has changed since the prototype builds as well. Hoping for 1.22:1 aspect ratio option so that works a bit more "as expected". The prototype build worked that way.

Phil, I would be interested in hearing your preferred settings for 48 fps time base

Only worked on a few HFR things in the last few years. I stick to the 180 degree shutter, so I I'd black shade at 1/96th if at 48fps and 1/120th at 60fps. I will say I've been toying around with a little bit more open shutter technique and theory with HFR, but I'm still doing some research and testing on that subject. Essentially because there are more frames per second the sort of predictive and pleasing effect of the motion blur at slower frame rates is a bit lost. It sounds counter intuitive but in some tests with a 48fps timebase with a 1/72 shutter I saw happy results for a lot of moves.

One thing, and I stress this, when people start shooting more HFR and we start regularly seeing it you got to have your shooting technique down. Every dolly bump, focus hitch, fluid head scoot suddenly becomes glaringly apparent. At least to me. Handheld sometimes looks incredibly bad too. It's a very thin line to make stuff look wonderful in the world of HFR. Seems backwards, but it's a truth for the time being.

 

[Elsie N]

Only worked on a few HFR things in the last few years. I stick to the 180 degree shutter, so I I'd black shade at 1/96th if at 48fps and 1/120th at 60fps. I will say I've been toying around with a little bit more open shutter technique and theory with HFR, but I'm still doing some research and testing on that subject. Essentially because there are more frames per second the sort of predictive and pleasing effect of the motion blur at slower frame rates is a bit lost. It sounds counter intuitive but in some tests with a 48fps timebase with a 1/72 shutter I saw happy results for a lot of moves.

One thing, and I stress this, when people start shooting more HFR and we start regularly seeing it you got to have your shooting technique down. Every dolly bump, focus hitch, fluid head scoot suddenly becomes glaringly apparent. At least to me. Handheld sometimes looks incredibly bad too. It's a very thin line to make stuff look wonderful in the world of HFR. Seems backwards, but it's a truth for the time being.

Thanks Phil... copied and pasted into my notes.

 

[Bobby Farokhzad]

As always, thank you for all your insight Phil. I am trying to imagine the candle light scenes from Barry Lyndon shot on Dragon...... Kubrick would have undoubtedly been impressed.

 

[David Battistella]

Thanks Phil,

You have been very generous here of late. I know how much time this stuff can eat up. I can tell you, being on the other side of the world and reading your posts and seeing how much excellent sharing is going on here lately by many equally generous contributors has made me realize how great reduser is when it is functioning in this open and respectful way.

I'll be in LA in January and I would be happy to meet up or buy you a beer anytime.

Cheers,
battistella