Dragon/Epic comparison

[Matt Ryan]

Hard to believe this is with Tungsten as well. Not much noise at all...should also mention I added a bit of film grain as well just for fun.

 

[Martin Stevens]

Wow..... Great stuff Gunleik.

 

[Les Dittert]

I think some of you would get a kick out of this 2009 Kodak video promoting motion picture film over 'hd' .
They talk a lot about latitude, and blowing out highlights

http://youtu.be/LJjWZRuPX0U

 

[Alain M]

Impressive DR and Low Noise @ ISO 4000... WOW WOW WOW

Impressive DR and Low Noise @ ISO 4000... WOW WOW WOW

Impressive DR and Low Noise @ ISO 4000... WOW WOW WOW I played with it a lot and I must say REDCINE-X with Dragon is extremely fun...so much latitude and so much everything...Thanks for the stills!!!


GIRL_DRAGON.jpg

 

[Gunleik Groven]

Thanks Phil...

Cool take on the backyard... I should maybe post some of those...

THX

 

[Gunleik Groven]

BTW.

Low exposures obviously works a lot better than on the Epic. Much cleaner and there is actually data in there...

But a low physical exposure (which does not necessarily equal a high ISO exposure) will be a lot more beautiful... for a dark image.

WHich again, only makes sense, and is something to be grateful for

 

[Alain M]

Make a lot of sense...

Thanks Gunleik.

 

[Will Keir]

Does that suggest we gain 2 extra stops using the Dragon sensor?

G, were you able to quantify this?

2 stop?
More?

They look pretty close. Are there any conclustions to be drawn with Epic 4000 ISO being on level footing with Epic MX at 800 ISO?

Processed this at 4000ISO:

This one at 800:

 

[Gunleik Groven]

Actually I think I will stop posting new comparative examples in this thread, and make a new where I post Dragon-only examples, as time allows.

Most points are made.

Epic is lovely
Dragon is more lovely.

I agree with Phil. Comparative stuff like this has some value, and very often no, as people look for the "wrong" stuff in the tests. And the tests (if well done) are basically designed to reveal specific things. Which do not necessarily give a hold on overall performance of either of the cams tested.

Epic is a good "baseline" for many here, though, to show off what "more" Dragon brings to the table, and give an idea of all the new juice the Dragon brings in.

Simply put, I can fully understand why people queue up to buy Carbon Dragons.
And I can likewise that people are happy with their Epic/scarlets.

Both are capable of stunning imagery.

And fortunately for RED and all, Dragon brings a substantial combination of "extras" to the table.


My priority of those are
Fidelity/Colors
Dynamic Range
Resolution
Lowlight performance

But I see how these factors interact and gives it the "it" factor, even though only one of these parameters could be a good reason to upgrade or simply go RED for many.

 

[Gunleik Groven]

Does that suggest we gain 2 extra stops using the Dragon sensor?

G, were you able to quantify this?



They look pretty close. Are there any conclustions to be drawn with Epic 4000 ISO being on level footing with Epic MX at 800 ISO?

I think (and have seen) that that way of thinking leads to many problems.

The difference between the highest and lowest exposure is 2 stops, if memory serves me.

But there is a lot of light there.

These images would develop equally well with a starting-point of 200 and 6000 ISO, given RedLogFilm as the starting gamma.
What I love about them is color- and skin- rendition under tungsten close to clip.
Noise is determined by physical exposure, not ISO rating.
And these images should not have noise of interest under "any" ISO, being exposed high.
ISO is a confusing parameter in that way to many.

You can push the Dragon "further" in both ends of the exposurerange than the Epic.
How far exactly?
You simply have to make your own judgement.

 

[Will Keir]

That's precisely what I aim to do. Because the nature of shots I tend to go for, light is always something I'm glad to have more of. As I continue to gain skill and earn jobs, the packages increase and the need for natural light is less as production values increase.

There is something raw, natural and beautiful about using what you are provided that I have, for many years, used as my stew to acquire my knowledge of the Epic image acquisition system and cinematography. There is something special that happens after three to four hours in the wild, something changes, a window opens, a calmness finds it's work. That is the kernel of what I'm still around doing this, why I come back for more, why I still upgrade, instead of move on to "normal things." There's something out there, something I'm hunting for and, as I did with the R1, I believe Dragon will get me there.

I get it... the Dragon is better. That's why the money as been sitting in the bank for the past four months. I've seen the "better" improvements work from R1 to Epic. The R1 I failed on. The Epic, I failed less but I had more tools and knowledge. When I look back at the shots, I wonder how much was the camera, how much was my growth, how much was just getting the basic tools I needed.

I was hoping for a solid number to rely on. How much more light is there, how many stops can I count on gaining? I am a bit disappointed in a variable answer. I will never have a Epic Dragon and Epic MX at the same time, just as I have never had the R1 and Epic together. I will have to learn from feel and memory. My lenses are different too but I won't give up on the comparisons and looking for the answer to my question. How much light do we gain upgrading?... it's out there.


Shooting a piece in the woods with 45 degree shutter using available light can be dangerous if you swing into a dark patch, noise at the ready to dance into your shot.

You can push the Dragon "further" in both ends of the exposurerange than the Epic.
How far exactly?
You simply have to make your own judgement.

 

[Phil Holland]

Hi Will.

I was hoping for a solid number to rely on. How much more light is there, how many stops can I count on gaining? I am a bit disappointed in a variable answer.

The last time I measured it, it came out about 3.6 stops better than Mysterium-X and there might be more in there. I'll measure again next week. I'm guessing right now things land in the 3.6-4.2 stops of improvement actually. 16+ stops is a nice way to state it though and I get why Red went with that.

 

[Gunleik Groven]

Phil gave you numbers.

I could suggest numbers, problem is all depends on "acceptable noise" which is individual.

To me the gain goes beyond the numbers, because the color part makes a wider part of the image useable.

One thing people tend to forget, is that a total latitude of 16+ stops, will not give you a normally "useable" image at all of those stops, as little as an 8 stop cam, would give you a useable image in all those stops.

It means:
"A much wider window for practical exposure"

I am always trying to be conservative in numbers. Try to be, because I want to not be called out...

But I would say that there is a 1+ (to possibly closer to 2) advantage in the top, actually physically that would depend on what kind of light - source you use, but as I have shown in these shots of my daughter, it is getting academic to discuss Tungsten distortoin.
(hey. Practically I just gave you one more stop on top of those for tungsten use, compared to Epic)

In the bottom. It is "a lot" if you measure signal only, but noise and preferences come into play.

I prefer discussing light - stops over "what ISO", because THAT and not the ISO determins noise/precission...

 

[Will Keir]

Hi Phil.

I read that and pass over it thinking it was not possibly that high! Thanks for pointing it out!

3.6 stops!!! Unbelievable. When is your 2nd test? Is that all possible on the dark end? Or both ends? A dark candle lit scene, or exterior night lit by street lights... I gain 3.6-4.2 stops into the shadows now?

Hi Will.

I'm interested in the same. I'm wondering if this camera makes a set of Super Speeds less relevant. Say I can use T2.8 or Ultra Primes instead. I found that the closes match, to what my eyes see on the Epic is T1.3 in low light. What you see is what you get is something I'm happy to settle for.

I use T1.3 just to control Depth of Field now and use NDs to control my aperture.

Acceptable noise for me is not being able to see any "dancing flies" in my picture. Make black, BLACK. If I film a dark tree, I want to see what I filmed, the tree being dark. Not the tree being dark with noise movement... you get my point. In that range, could you guess how many stops we are gaining?

I prefer discussing light - stops over "what ISO", because THAT and not the ISO determins noise/precission...

 

[Gunleik Groven]

Understood, will. But then there is denoising and other factors comming into play.

But I will upload two sets of file with the difference of 5 stops. That would equal an ISO change from 400 to 12800 iso to compensate for light loss...

THAT is one way I use these controlled stop-by stop exposures. Determining when "unacceptable noise" happens and how lighloss affects saturation and color renderition.

Now, as you will se from the meta, I did not actually change the ISO to get this right, just changed the physical exposure.

I will not publish "stop by stop" the full comparison, as there are many variables which could be "wrong" with my example, and I accept that.

This, however, will give you a general idea on the behaviour.

You will get a few frames of ech exposure to look at the flies...

Give me 15

 

[Gunleik Groven]

Please do some scaling/denoising experiments, too down to a practical deliverable format like 1080, 2k or even 4k before comparing the resulting noise...

 

[Phil Holland]

Hi Phil.

I read that and pass over it thinking it was not possibly that high! Thanks for pointing it out!

3.6 stops!!! Unbelievable. When is your 2nd test? Is that all possible on the dark end? Or both ends? A dark candle lit scene, or exterior night lit by street lights... I gain 3.6-4.2 stops into the shadows now?

I'm interested in the same. I'm wondering if this camera makes a set of Super Speeds less relevant. Say I can use T2.8 or Ultra Primes instead. I found that the closes match, to what my eyes see on the Epic is T1.3 in low light. What you see is what you get is something I'm happy to settle for.

I use T1.3 just to control Depth of Field now and use NDs to control my aperture.

Much like anything involving cameras where that dynamic range improvement will show it's strength is how you choose to expose, which is what Gunleik was alluding to earlier. At the same ISO rating however, I will need to take a precise measurement of Dragon against Mysterium-X via my shadow box and then I'll know exactly how much and where the dynamic range placement via the shipping calibration lands. So stay tuned for that. That will be where the "hard number" comes from. I may wait to do this test when DRAGONgamma arrives, but it will happen either way.

In terms of the validity of Super Speeds due to increased light sensitivity I'm of the school of thought that if you have a more light sensitive camera fast glass become an even more flexible tool. You'll have to make the call on what's more useful for your style of shooting however. Me personally, T2.8 is always a bit slow for primes covering this particular format. I'm more of a T1.3-T2 man for primes in the S35 and Dragon 6K FF format size. If we're talking larger formats that might be a different discussion. Depth of Field for me is a storytelling tool. While out here the rather standard T2.8-4 split is pretty damn common, a set of primes that can't open up past T2.8 only creates a creative wall that I don't want to deal with.

Please do some scaling/denoising experiments, too down to a practical deliverable format like 1080, 2k or even 4k before comparing the resulting noise...

I've done a bit with Neat Video. Very good results if you want to remove the grain at the highest ISOs. I did a couple tests at 6K, 4K, and 2K. I'll do more soon and shoot some low light content to see exactly what can be squeezed out. There are gains because of the 6K NR -> 4K workflow for sure. Particularly for projection Mysterium-X at 4K with Neat Video on ISO 6400 produced good results. You can do a bit more with Dragon

 

[Gunleik Groven]

There are a few things to note.

I had to go down 3 stops from the initial exposure to get an Epic image which was not unpleasantly clipped, the way I see it.

That does not interpret like I think the difference in the top is 3 stops, but just makes my 1+ stops claim "more than likely"

You now have the top exposure. One 3 stops down, one 5 stops down (from initial) and one 8 stops down (from initial)
It should be possible to extract a bit of info of noise and color rendering and general tendencies from them.

If you set the ISO at 400 for the highest exposure and the ISO at 12800 for the lowest exposure, levels should match.

Please look at both stills and moving images when determining "how you feel about" or analyze noise.

Please scale to different output resolutions. 1:1 tells you… well THAT.
But not a lot about practical use.

Have fun and play nice.

http://www.gunleik.com/dragonepic/ISO_EXPOSURE.zip

This is the final comparative images I will release…

Unless I get a very good question that makes me change my mind, like these from Will…


Have fun.

G

 

[Will Keir]

Haven't had a chance to play with de-noising. IMO, if you have to de-noise, you've already done something wrong.

I try to get what I want in camera and only going to post on the shots I have a gameplan for.

Understood, will. But then there is denoising and other factors comming into play.

But I will upload two sets of file with the difference of 5 stops. That would equal an ISO change from 400 to 12800 iso to compensate for light loss...

THAT is one way I use these controlled stop-by stop exposures. Determining when "unacceptable noise" happens and how lighloss affects saturation and color renderition.

Now, as you will se from the meta, I did not actually change the ISO to get this right, just changed the physical exposure.

I will not publish "stop by stop" the full comparison, as there are many variables which could be "wrong" with my example, and I accept that.

This, however, will give you a general idea on the behaviour.

You will get a few frames of ech exposure to look at the flies...

Give me 15

hahaha, glad I could offer something to the conversation, thanks man. Much appreciated. I'll have a look when they load...

Unless I get a very good question that makes me change my mind, like these from Will…


Have fun.

G

You know were on the same page with our love for Fast Glass. Without a doubt, you have more storytelling options. I made some assumptions early on about fast glass and did not realize the relationship more light has to depth of field. Why is all my working happening in this dark places these days...

I'm of the school of thought that if you have a more light sensitive camera fast glass become an even more flexible tool. You'll have to make the call on what's more useful for your style of shooting however. Me personally, T2.8 is always a bit slow for primes covering this particular format. I'm more of a T1.3-T2 man for primes in the S35 and Dragon 6K FF format size. If we're talking larger formats that might be a different discussion. Depth of Field for me is a storytelling tool. While out here the rather standard T2.8-4 split is pretty damn common, a set of primes that can't open up past T2.8 only creates a creative wall that I don't want to deal with.

Looking forward to seeing some of these results with and without the Neat Video.

I've done a bit with Neat Video. Very good results if you want to remove the grain at the highest ISOs. I did a couple tests at 6K, 4K, and 2K. I'll do more soon and shoot some low light content to see exactly what can be squeezed out. There are gains because of the 6K NR -> 4K workflow for sure. Particularly for projection Mysterium-X at 4K with Neat Video on ISO 6400 produced good results. You can do a bit more with Dragon

 

[Filip Orlandic]

It's not all about the noise... Dragon obviously sees far more deeper than Mysterium-X sensor. Also Dragon at REDCODE 6:1 is less compressed than Mysterium-X at REDCODE 4:1 which provides more details at the bottom.