Dragon/Epic comparison

[Detlev Eller]

I was paying with some the R3D and I think I'm in love with this camera :drool5:

LOL ... wait until you hold it tight, you´ll feel the heat .... ;-) ...

 

[D Fuller]

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

I can understand that perspective, but I think it's a bit foolish to use a camera that shoots raw and not use the tools that are available to you. You do realize that other cameras have noise reduction built in? The power of raw is that you have this great, honkin' Power Mac or Windows machine that can do stuff like noise reduction when it's needed. That's kinda the point of the whole r3d thing.

 

[Mark L. Pederson]

You do realize that other cameras have noise reduction built in? The power of raw is that you have this great, honkin' Power Mac or Windows machine that can do stuff like noise reduction when it's needed. That's kinda the point of the whole r3d thing.

Amen. MOST cameras do in-camera sharpening and noise reduction. There's nothing wrong with using either. It's a subjective, creative choice - just like choosing a filter. Better to have the choice IMO.

 

[Alain M]

It is amazing that at ISO 12800 can look so clean based on proper well exposure in the physical footage. I never imagine that would be the case since I always related ISO with Noise. Lots of learning here.

Thanks Gun and Phil

 

[Adrian Jebef]

When I went to film school(s) (I was kinda thrown out of two of them) ...

There's gotta be a good story or two in there, Mark. C'mon, spill the beans.

 

[Mark L. Pederson]

There's gotta be a good story or two in there, Mark. C'mon, spill the beans.

Let's just say that it's a good thing there was no Facebook, Twitter or Instagram back then, because ... I deny everything

 

[Mark Wilkinson]

Hahahaha

Know what, I agree. Actually you reverse-state my point by including "metering" in the argument!

Take the filmstock equivalent...
I guess you would hardly rate a 200 ASA stock at 12800
At least you would know that IF you chose to do that, you would create a very noisy image...
To achieve a somewhat useable image you would swap to a more lightsensitive stock. Right?

Similarily I know that I do not swap the sensor or add any analogue gain to it to increase sensitivity by sliding the ISO slider.
So I would do what most everyone who have showed off "nice" 12800 images here have done.
Overexpose like hell, and "recover highlights" in post.

Now to me that is not "magic" but just a very impractical way of aquiring images.

That said.
I am capable of doing it...

A lightlevel of X will allways create the same Signal-/noiselevel on the same sensor calibrated the same way.

The difference ISO does, is where in the visual specter (where in the curve) that signal/noise is visible.

People playing tricks with high ISO "overexposure" and redlogfilm "recovery".

Seriously.
If it gets them what they want, I think the approach is good.

Still I find it to be a slightly complicated way to work.

Ok, please post an image. I don't think it will work. 12800 too many stops past what the sensor can do.

And if you are "overexposing the image" you are just rating it slower.

But still, I don't think you can do it. Please prove me wrong (without post noise reduction, I might add).

 

[Robert Ruffo New]

Well ... I hear this all the time but I think it's a sad statement that just points to how lazy and resistant to learning some people are today.

When I went to film school(s) (I was kinda thrown out of two of them) we had to learn and understand how to expose motion picture film. We learned how to use a light meter. We tested various film stocks, under exposed, over exposed - we experimented and learned. When I first learned how to use RAW it just felt like digital film stock to me.

RAW isn't exactly the most complicated thing in the world. The guys in stills managed to learn it - leverage it - and use it without "misunderstandings, misconceptions, pseudo know-how and badly exposed camera files" every day all over the world.

Sometimes I feel like there is this "point and shoot" generation that has never even touched a light meter or learned the basic, fundamentals of cinematography / photography.

Making the tools "dumber" might get you a bigger market share - but ... we'd all have better images if more people paused to learn the fundamentals.

Again, ask any still photographer who makes a living shooting images if he or she finds RAW difficult.

I've always liked you - now I like you even more.

 

[Gunleik Groven]

Ok, please post an image. I don't think it will work. 12800 too many stops past what the sensor can do.

And if you are "overexposing the image" you are just rating it slower.

But still, I don't think you can do it. Please prove me wrong (without post noise reduction, I might add).

Of course I overexpose it... But that is the point, because you do not "really" as the sensor is not and will not ever be 12800 based.

I think this is available as my post 3 in this thread, no?

Try it yourself....

12800 with RedLogFilm and a "slight curve adjustment"

effect would be the same "when shot".



"in camera preview at 12800":

This image should not be confused with "shooting at 12800 is smart and works out of the box", though...
On the one hand, of course it does. On the other, I am shooting high over "rating" but I can see that I do not technically clip, as RG3 (and all the other gammas now, it seems) are based on RedLogFilm. If the highlights are not actually blown, they will be available.

Graeme is a very smart man. In a good way.


Add RLF:

The cool thing about this example is that I could say (and document my statement to be true) that

"xxxxx ISO equals YYYY ISO"

Because, exposure-wise, that is actually true and verifyable.

ISO is easily a misleading concept as a source for info on how the REDs exposure works.
(unless of course, you understand how the ISO on the REDs work...)

Not that I think using ISO changes to make life easier is bad.
It is not.
ISO is just not where you should measure stuff to get "real" exposure and SNR info on the REDs...

 

[Gunleik Groven]

should BTW just mention that my way of underexposing (through raising the shutter) could increase noise.
I am aware of it and it is a trade-off compared to using NDs or changing the optical path in other ways which may affect color and contrast.

Yup.
The Aperture changes change contrast, too, but at least none of these introduce extra flares or tint...

There is no test without possible fault-sources.
Thus one should do many tests where one test for different aspects of the image and particular scenarios relevant to the production at hand.

But still as a general first, I think this might not be totally at fault.
I will accept it if you think otherwise!

 

[Phil Holland]

should BTW just mention that my way of underexposing (through raising the shutter) could increase noise.

Did you notice that calibration remains within spec with a broader range of shutter speeds yet?

That was a happy surprise. Can't recall right now, but it handled a 180 degree shutter I think at 72 fps within the same calibration at 24fps at 180 degrees as well. That might have improved even.

 

[Gunleik Groven]

Did you notice that calibration remains within spec with a broader range of shutter speeds yet?

That was a happy surprise. Can't recall right now, but it handled a 180 degree shutter I think at 72 fps within the same calibration at 24fps at 180 degrees as well. That might have improved even.

There you go.

Have not tested for that.

That sounds good and likely!

"Getting the picture" of the Dragon is cool.

THX!

 

[AndreasOberg]

The closeup indicates to me that Dragon has roughly one stop less noise, possibly a bit bigger than one stop.

Look at:
- the noise spots are smaller
- the color variations are smaller, basically its not as red or green
- more detail is showing through, look at the hair.

However what I would do myself to really test is to
- open both of the images in a program like Neat Video
- do noise reduction and sharpening to the best level,
do a color grading pass
- then scale both to 4k resolution.
Then you compare both of the results.

I never, ever use the raw image from the camera, but always push them in color grading so its often a bit meaningless to compare the "raw" image. I have found that often 2 cameras can create a raw image that looks nearly identical, but one can be pushed much more. Its a bit like comparing color grading jpegs and raw still images. They may look similar but grade very different.

Looking at the images what will really make a difference for me is:
- the increase dynamic range. I rarely have a problem with noise in my images, however dynamic range is always a challenge with every camera that I have owned. Dragon seems to be the first sensor that has very little problem with this. Incredible.

I just looked at the histograms. Wow, so much more information in the top areas of the histogram. The Epic just cuts off while the Dragon smoothly becomes overexposed. Indicating that its not only more sensitive but more "clever" in the brighter areas. Very interesting.

/Andreas

 

[Gunleik Groven]

agree on all points.

THAT is why it is nice to have R3Ds to run through your individual workflows and compare the resulting image...

You guys all have different workflows...
Now you can test them on different scenarios.

I have another batch of R3Ds with a similar methodology but under different lighting conditions, if you want...

 

[AndreasOberg]

The more the merrier Gunleik.
For me I think the most interesting images to see is where one can study:
- how much more range the image has. So bright situations, often outside. I'm just sitting and grading HDRx images because it was impossible to capture higher ranges in Costa Rica sunshine. Love HDRx, but would be even better if we didn't have to use it quite as much.
- Areas were its just too dark but you had to shoot anyway. For example you had to "shoot in ISO 2000" (I never like that terminology since we are never really shooting in any specific ISO, better would maybe be "using ISO 2000 to see anything in viewfinder" because otherwise it would just be underexposed.
- Would be interesting to see how much more detail one can pull out in dark areas. I instantly thought of a forest or a city where you have dark areas. I really like selectively grading shadow areas to bring them up a bit but they can get a bit "smudgy" if its too dark on Epic X.

Well anything is interesting of course. Myself its the dynamic range that is the killer feature, and oh yea, higher resolution/sharpness and better skintones too, hehe. Great job Gunleik!

On a separate topic. Doesn't it seem like there is almost the most confusing regarding "image quality" when it comes to cameras? The Red forum has a much higher understanding of image quality than elsewhere on the net. But never have I seen so many confusing articles and comments when it comes to image quality on the net. I shiver when I think of the Scarlet/5D comparisons we have seen and our investigations of the Dragon images remind me of these discussions. There seems to be a great need of a very thorough article/tutorial that explains the whole concept.
/Andreas

 

[Rudi Herbert]

To me, a lot of the blame falls on production companies, period. Not post houses. In this day and age, I strongly believe you need to own your image. To me that means always delivering GRADED shoot footage to clients, accompanied by the the raw elements so they can re-grade if they need to. We agree on a look on set, and plan to deliver that way. Otherwise the workflow in post can get messy. Regardless of what camera you shoot on.

That, is one of the most insightful things I have ever read on this forum, and it is true. I see it on my chosen medium, underwater, all the time, where people go through huge budgets to travel o far off destinations, with raw cameras, housings, dive gear, boats, support infrastructure, etc, etc, etc, only to deliver ungraded footage in the hopes that somebody down the post production path will know what to do with it. They usually don't. In our case, most people, including top grade colorists, don't know what good underwater looks like, so they tend to leave it as is or do very little to improve it. But it isn't much better in many cases topside either. You're right, you own your footage, it behooves you to make sure it is appreciated in all of its glory. That's why on most cases, I insist on delivering ProRes or Cineform that has already been graded from my raw files, and I fight tooth and nail not to deliver raw, unless it is to people whom I know will know what to do with it.

 

[AndreasOberg]

That's why on most cases, I insist on delivering ProRes or Cineform that has already been graded from my raw files, and I fight tooth and nail not to deliver raw, unless it is to people whom I know will know what to do with it.

Good lessons, I will take them to heart.

/Andreas

 

[Mark Wilkinson]

Ok, I tested it and stand corrected. Comes out pretty crunchy though… But the grain is roughly the same. I suppose all this is proving is that the data path is robust.

 

[Blair S. Paulsen]

DIT's value add

DIT's value add

Five years ago I put a Scratch seat in my Ninja van (calibrated monitor, controlled viewing environment) to facilitate creating an "authored" baked in look. To me, its a virtual darkroom where one "develops" the RAW into the intended RGB image. Obviously a well coordinated team can send the R3Ds directly to post where a dailies colorist can dial in the look the DP/Director/Producer is going for, but as mentioned by others - that is not always viable.

Whatever pipeline one chooses, IMO, working with RAW files is by design a two stage process - 1) capture (exposure) and 2) development ("one light" grading). What I refer to as "one light" grading (essentially a dailies grade) is different from final color grading where power windows, shot to shot matching, beauty passes, etc take place. In a perfect world that final color pass would be done from the R3Ds with the "one light" version as a guide to the DP's intent. That said, if the baked in RGB files are fairly close to the desired look, not too compressed and have decent bit depth; finishing from them should produce very good results, especially for TV.

Cheers - #19

 

[Gunleik Groven]

i have worked a bit in tv-drama...

at first we went to log dpx for final grade.
then "raw" log + linear exr for vfx

no way using the dailies for finnish, would have gained the productions at all.

dp has allways been part of grading process, btw