Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Dragon/Epic comparison

That could be done two ways.

Either exposing for the highlights (which is a plan I generally advice against)
Or exposing for the girls face (which is my general advice in situations like these)

I think I know what the outcome will be in both of those scenarios, though...
 
Gunleik Groven said:
That could be done two ways.

Either exposing for the highlights (which is a plan I generally advice against)
Or exposing for the girls face (which is my general advice in situations like these)

I think I know what the outcome will be in both of those scenarios, though...
Click to expand...

That sounds good. Instead of breaking the image, maybe just pointing out where you what the best/optimal exposure in the frame and then letting everything fall off where it wants too. This points it all toward creative choices, like sometimes you want things to blowout, but you just want to know kinda how it's going to blowout.

The other test (in my post above) helps indicate what you would have to do to match because the X won't just go where the dragon does.

It's all good because it points us toward creative flexibility and creative choices. Like shots you want all the DR and shots you don't and how you want to turn it on and off.

David
 
True

Dragon is not like Epic.
It should also not be exposed like Epic.

If you expose for Dragon and not for Epic, everything good I have seen in my previous test and in peters images all of a sudden makes sense.

;)
 
Gunleik Groven said:
True

Dragon is not like Epic.
It should also not be exposed like Epic.

If you expose for Dragon and not for Epic, everything good I have seen in my previous test and in peters images all of a sudden makes sense.

;)
Click to expand...

Gunleik,

Setting ISO to 800 was one way to sort of "protect highlights" with the Epic. I used this concept as a general guide when shooting outside / harsh light.
This is not to say I exposed for highlights but I would definitely be watching them and trying to leave a little wiggle room.

With that said, how would you expose the Dragon differently? I'm guessing you're less concerned with blown out highlights on the Dragon in general - but I know that is probably over-simplifying it. Can you elaborate a bit more?
 
I think there are a lot of different views in general on MX (and I am not going to speculate on Dragon but this is related in a round about way). Is MX natively 320 and one should expose for "raw" at 320? Is it better to go for red color gamma 3 and shoot a wide negative at 800 iso and a 180 shutter? Different people have embraced differing view points on the MX in general when it comes to best practices and whether your exposing for your Red gamma and color presets, LOG, or "raw' (and there are also obviously other guides, false colors, and what not you can expose for as well)

Red color and gamma is essentially rec 709 LUTs and don't show full dynamic range to begin with is my understanding. So upon chatting with many colorists, DITs, DPs, ACs and such about Red I have encountered many different thought patterns on why things should always be approached from such and such a way. Depending what is exposed for in acquisition has a domino effect, and I hope people collaborate well, because if the DP exposes for one technique but the colorist is going off a different Red theory, things sometimes can run awry I imagine...

Personally I think the fat negative coupled with the power windows of post is pretty decent path. But going for specific in camera looks is a perfectly viable choice as well, assuming your monitoring is dead on accurate of course since your using it to base your look and ignoring other potential paths..

Different ways of going for something doesn't mean one thought pattern is any better then the other. Hell you could shoot wide and bake in dailies that are "the look" too. Theres a million workflows and many ways of approaching image creation. Some want as much in camera as possible, some want a fat negative and play with tracking in post.

So going back to Dragon my understanding is whether I am at 200 or 2000 ISO I still will have the entire dynamic range available. That sounds really awesome. So I don't believe anyone has said "this ISO" gets the most range in all contexts yet. Or "this ISO is Dragons native ISO and should always be adhered to along with what the raw image is seeing, not your Rec709 representation"

Such view points don't seem to yet be established with Dragon. Supposedly the range will be there with any thought pattern, in either case I look forward to doing some tests once it arrives. Wish I had cash for Carbon Fiber! Thank you early purchases for sharing what you can with us and dealing with all constant bewilderment!
 
I agree with Gunleik, expose each camera it's own optimum way.
That is, don't let it clip the highlights that you care about. Then, for looking at tonal 'purity' and noise we can look at different ranges in the images to see how we like it. Digging into the darks and seeing detail there will show the dynamic range.
That's how I see dynamic range : Set exposure on both cameras to expose highlights almost clipping detail out. Then look at the darkest things in each picture. If a camera has more DR it will show the details in the darkest darks with less noise. Noise is what kills the DR. If the darkest objects are lost in noise they are gone. This is why I nit pick about noise.
 
I have learned so much from this thread. Thank you Gunleik for this test and describing why it is important. I can tell one is flatter and see where over exposure has occurred, but that is about all. (Well the magenta highlights, but I knew that.) But I have learned a lot after reading this.
 
Good thread

Good thread

Like Gunleik, when I first got a good look at some Dragon material there was something beyond DR or resolution that made the images sing. What is it? I have a theory.

Think about all the image information in a log image. When you grade that log file into an image with solid blacks and highlight separation its easy to lose some of the nuance in the critical middle stops that are the core of the image. The Dragon files are so elastic with so much image data that they seem able to retain more subtlety in the middle stops even after parking the blacks and shaping the rolloff in the highlights. Perhaps most critically, this quality seems to offer more control over skin tone transitions into the highlights - something I have had to fight with sometimes in high key images on MX.

My point is, that beyond just determining ideal nominal ISOs for each sensor to perform its best, there is also more real image information in the stops between the extremes. Part of that, as Gunleik noted, is more genuine chroma in the highlights and shadows on the Dragon. From an artistic perspective I love the ability to plant skin tones above or below the "sweet spot" to suit mood/context without them going a bit gray.

Cheers - #19
 
There are very many things at play here.
And I am not into giving specific advices, except when working with specific persons or scenarios, and it is possible to follow an idea of an image and test that what I claim is understood and hopefully verifyable.

Just to use myself and David (as we are kind of vocal) we have a bit of different general exposure approach, but our general idea is pretty similar:
Both have tended to stay at one base ISO and learn the camera from there. How it responds in different situations.
ISO is metadata.
The real exposure of the sensor is determined by how much light hits it, not what ISO we shoot at.
Still, ISO and the visual response we see on the monitor affects how we expose. I have yet to see anyone not being affected by the image they see.

So here is the first good news on Dragon.
At "whatever setting" the preview image "looks a lot better" than on Epic.
Even when the resulting files may look pretty similar to the similarly exposed Epic in post.

NOW THIS IS GOOD.
This creates confidence in the tool.
And a confident shooter often makes better decissions.
I have seen this again and again.

I love the Dragon already here...
But this is really just the icing on the cake...

Now. Back to David and mine approaches.
I rarely deviate from 800/5600 when shooting the Epic, but I see that the 800 over 320 (in MY opinion) has two real advantages.
1. Protecting highlights
2. Leaving headroom for automatic whitebalancing (where the red channel tends to be pretty clamped when at a high exposure)

This has a downside, which David thinks is more important than the upside:
You force the exposure 1,25 stops down.
One stop dpwn means half the precission lost (theoretically)
So he wants to live more on the edge than I do.

Nothing wrong with that, as long as you controll it.
I do think as a general advice to a person not familiar with the concepts of digital image capture, it can lead to some bad and unresolveable situarions.

So what did I just say?
Avctually, we agree, we just approach it and value things differently.

Then again there are others playing with ISO and WB all the tiome and learns to "see" the camera and judge the image that way.

To me it makes no sense to take multiple exposures at different ISO/WB, because I have all that info in a single exposure.
For others it makes sense and gives them the data they need.

And we may all end up with the "right exposures for the situation".
Funnily enough, I actually guess that the actual result for a given exposurechallenge with a given wanted result will not differ too much between these approches.
If you know your tool you do, and there are no ultimate rights or wrongs, except for those that give bad results.
My approach to learning about a new cam, goes in the direction of looking on how it breaks... :)

I KNOW it is a capable tool that can make beautiful images, given that it gets a beautiful exposure.
But within what limits is it beautiful in what ways?

SOme look at that as I "tear down" the camera. That is not the intent and not the result I want. :)
Thus I am happy to post Peters lovely frames, and hopefully I will be able to contribute a bit in the "good" image department, too. I know many others will.

So I will not say "Dragon should be shot at ISO X"
Or "2000 is the new 800"
Because that depends 100% on what lights you let the sensor eat.
And from what I see, that is a very individual approach to that.
May make sense to think that way, for others it is just confusing. (for me, as one example)

I will approach this a bit more generally.

I am a bit doubtfull of which R3Ds to release, NOT because I try to "protect", but because I want to avoid confusion. Not because people do not see what they see, but because they are looking at a frame and look at the noise, when THAT is not what I am testing for in that particular scenario.
More will come, but I see that they need to be presented to make sense.

Now.
I have been doubtfull about that -16 stop exposure.

But it will come now as a jpg... :)

First thing to check is "range". Right?

What is the sensors capacity.

I cannot say anything about that, but that I did not in my first test manage to "blow" the Dragon on either side of the range.

Now. This is how I performed it.

I shot at T1.3 to t5.6 on a zess superspeed 65 with a shutter at 1/25
then I shot at 5.6 with a shutter at 1/50 - 1/6400
And then I stopped down to T 16

That is a total of 16 stops tested.

Now. It was night outside with drawn curtains.
Lights were a daylight LED, a 300W Arri fresnell pointed directly into the camera and Espens right chin, and finally small halogen tablelamp.
Mixed lights to see how that affected the colors.

Now. The real telling here is the first and last exposure., so I will repeat the first and add the last and discuss them in the next post.

I would appreachiate no comments to them until I am done.

Here is the first: (repeat)
1476212_10152064395735126_341797619_n.jpg


And here is the last:

1469861_10152071185715126_752137062_n.jpg
 
Now... this blew me frankly away.

But let me start with hosing down some fires.

1. Point here was NOT to check if I can make a nice image at 12800 ISO.
I know I can.
just add some contrast and develop as RedLogFilm.

I could even develop this image at 2000 ISO and get a similar result. Using grading and curves.

Point is;
The last image was kinda not supposed to be there...

and here is why...

But first, let me discuss possible faults in this test. They are mainly 2
Wrong indications of T-stops on the lens.
Wrong calculation of shutter-angle in the cameras.

Both are possible, but for the sake of the argument, let us assume that they are not way off.

Now... My intent was to fully overexpose both cameras, and this is what they looked like (more or less) on my LCDs:

1475805_10152064395445126_948966310_n.jpg


Allready there I noticed that I probably did not manage to break the highlights on the Dragon, so I expected the test to give me a lot of info, but not "real" range info, as I had not managed to break the Dragon.

OTOH, I did not have any more lights available. LOL

Anyhow.
Given that this is supposed to be a 16+ stop sensor and I took away 15 stops of light as good and controllable I could, not by dimming the light and affecting the reflections, but through choking the optical path to the sensor for light, I fully expected to see just black in the last frame where there was image data still available in the first.

Now. I clearly see both colors and contrast in that last frame.
That shows me that the 16+ claim is more than likely.

And that, gentlemen is extremely good news for all of us and something totally unexpected for me, really and should lead to champaign popping in every home that has ordered a Dragon.

This is a biggie.

Even before Dragon Gamma.

just to be clear. The test here was NOT :
"Can I get a useable image at 12800 ISO", as I already know I can. On both cameras. Given that I wanted to use that approach.
The test was:
What happens if I take away 15 stops of light from a high exposure. Will there still be image data in there.
And it was.

Now I have had people giving me numbers on exactly how much more data there is up to clip in the 1st image and down to black from Espens face in the last.
I will not quote those, but they are strong indications that the 16+ statement is NOT a radical marketing hype, but rather with some safe-zones.
I like that approach a lot!

(NOW I am done with discussing these two exposures. So comments are welcome, but I am still typing on other elements of the comparison)
 
So the Dragon has "more" range in both ends of the spectrum.

First I wanna say one thing...

Does this suddenly make the Epic bad?
Not at all. WITHIN its range it is an excellent camera and I plan to use it a lot in both the near and distant future, but the Dragon def has more range, and it would have been kinda scandalous if it did not.

And if the Epic had been capable of following the Dragon in the extremes of this test, something serious would obviously have been wrong, either with the test or with the Dragon.
THAT is all good and as expected.

All happy camping.

Now:Noise
What RED has claimed is that there is less noise and better fidelity at any given ISO than on the Epic.
I will come back to the fidelity part, as it is harder to discuss...

In one of the examples given in this thread, it has been pointed out that the noise at 1:1 pixel level at a given ISO and exposure is very similar on the two, and I have said that I agree, but that that does not falsify REDS claim.
NOW, I can see how that comes out as a paradox, so I will discuss and exemplify a bit.

First I will start off with the scenario where the noise will look similar, and that would be if you did a 1:1 unscaled print.
As I am not aware of many 5k or 6k screens around, that is the only scenario where oversampling would not be a factor in the resulting noise from any of these cameras.

For most of us, that is not the most relevant scenario.
We mostly deliver 1080, 2k and some 4k to normal screenings. Of course there are exceptions, but they are relatively few. Scaling should normally be taken into account.
And given either you record 6k or 5,5k (which AFAIK has the same physical size as the Epics "full" sensor) the Dragon will have the advantage of scaling down the pixles "more" than the Epic to any given deliveryformat, and thus give a cleaner image, given that everything else is similar.

AND as all who have commented on the image and spoken with their wallets after testing it has said, and now finally is there for everyone to play with through Peters lovely images, there is a very special "stretchability" and "flexibility" to a Dragon exposure. If you have it, it is hard to break. Also compared to the Epic. (or any other cam I have had the pleasure to play with).
Actually THAT more than DR or any "number" was what had me come over. It is just so damned fun. But fun does not take away visible and undeniable noise, right? ;)

Exposing differently, does, though...

The images I post in this thread so far are all very brutal.
They are not meant to be "nice" but are designed to dig into "what happens if" on both cameras. In an at least semi-controlled manner.

And they tell me that I would not do what I did here in a real-life scenario. I would NOT shoot the Dragon like I would the Epic, given that I had half a chance to shoot for the Dragon.
And actually, that is what you intuitively do when you do not have to compare, like I do in this thread...

So... For the Dragon image, it is not really too relevant what it looks like when shot and lit for exactly as you would with the Epic.
We asked for "more" and we got "more", but if we do not use it, we will not see as much gain as we hoped for.
Right?

What I have shown in the above examples is that the Dragon has "significantly" more range in both sides of the exposure.
Now, what happens if we start to use that extra meat to get a better image.

This tells me that in a challenging exposure, I would be able to expose an important subject (like a face) higher on the Dragon then the Epic without losing critical highlights (like snow) and thus get better skintones in that face.

Now. Just for fun, let us say that the difference is one stop (I do think it is more...). One stop more light in that face, just by taking away an ND .3 which I would have put in the Epic, may not sound that significant, but according to theory, it would actually double the precision in the exposure of that face. Just by having to worry one nd 0.3 about the highlights and not accounting for bit-depth, compression and AD (which are factors I can have speculative opinions about, but cannot really document), I have simply made that face "twice as lively" as it would, AND I would have lovered the noise like going from 1600 to 800 ISO in the full frame AND I would be able to look a stop further in the dark.

All this would make the image look less noisy, have better blacks, and have a lot of more colorfidelity. Just by removing an ND .3.
THAT is not insignificant.
If it is 2 stops (I will come with a little test here further down...)
That would be 4 times the theoretical precission and fidelity and all the above.
Without even touching upon all the other underlying tech which differ in an epic and a Dragon. Just through pure physics.
I would say that was a rather nice gain... And that + higher oversampling, would definitely account for the lower noise and better precission all who have used it comment on.

All in all
These two factors account for two good reasons we are all seeing "a lot less noise than we are used to" from comparable scenes.
Then add the added lowlight sensitivity, and I think it starts to make sense why we all keep on smiling.

(this part is done. More to come)
 
Great example Gunleik! Glad you decided to share. For anyone worried about noise (me not one of those), I did this to Gunleik's heavily compressed JPEG, so you'd have better results with the R3D frame than this!

Do not look at this test as an example of noise. The idea that there is so much more information there than MX is what shows off the increase in dynamic range. That is all you should take away from this sample... This is pushing both camera's to the extreme, something that should never happen on a real set. To me this is a great eye opening dynamic range test. So much more there in the Dragon image...Amazing...

NRphoto.jpg
 
I'll chime in for a bit. Essentially the fatter workable latitude comes from Dragon's increased dynamic range. 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.

I'm sure we'll see a bazillion Dragon VS camera brand X, Y, and Z in the coming months as well. Often those tests freak me out a little. While rating the camera at the same ISO reveals the manufacturer's accuracy and essentially what the camera sees at that given exposure, each individual camera may benefit from a different exposure method to maximize the potential and aesthetic desires for a given scene. This indeed is sort of the goblin in the gears in these tests. Which is fine honestly, but sometimes when thinking about doing an apples to apples comparison people somehow forget that there are indeed different types of apples out there that all taste rather differently.

Gunleik's underexposed and lift test reveals two things. One. How much more light sensitive Dragon is compared to Mysterium-X. Those are details that are literally gone at an equivalent exposure in comparison (and this is a rather extreme scenario of course). Two. Where the dynamic range gains have been placed and focused on. This new found sensitivity should provide a denser feel to the image as it captures more above the ceiling and below the floor boards of Mysterium-X. Essentially more room to play with. This is all good news. It's better all the way around.

On the top end of the exposure, it's glaringly clear that there have been massive improvements with Dragon. Highlight Roll-Off is exactly that. It's got a nice organic feel. To me this eludes to one of the larger differences in the general character and image quality when compared to Mysterium-X. Mysterium-X on Epic, Scarlet, and Red One make and will continue to make lovely images for some time to come. That's something all need to be aware of. As Gunleik famously said, Dragon is more and that's exactly what the difference is. The "more" concept comes into play at nearly every level of image quality.

Dynamic Range and 6K are the two big selling points of Dragon, but Color to me is a huge one as well. Dragon operates entirely differently compared to Mysterium-X. You can see it across all the channels and tonal ranges. It's real nice actually. Pushing the REDCODE around reveals a lot of potential. And it's going to get even better too.
 
Thanks Phil.

This is a very good sum-up and a great cue to the third thing I want to discuss...
Which is colors and stretchability and fidelity.

I will start off by giving you 3 R3Ds
These are also not beautiful images, but no bad comments on my daughter for contributing to this test.

They are three "high" exposures lit with tungsten and at three different stops, to look at one of the things that has been a constant nightmare for me with all digital cams:
How they handle whiteballancing at non-native colortemperature when approaching clip.

Download and enjoy, while I write.

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

feel free to post grades. while I write and prepare examples of why I think this is important.
This is NOT an AB. This is a test designed to look at some stuff that puzzled me in a good way when doing "Sweet", and probably the most important.
Color-handling and white-balancing.

(I will continue to write in the next post...)
 
Color/Fidelity/White Balance

Now to me, this is THE BIGGIE with Dragon.

Even without all the other goodies described, this alone would have had me over. Even with less rez than the Epic and the same DR.
I would have been drooling...

I spent a lot of the time when I had Proto 77 for "sweet" throwing "bad light" at the sensor to see how it reacted.
And a lot of amazement in post seeing how the Dragon sorted it out.

We have seen it in Phils "paint" film, and in Peters lovely piece, and I would say to some extent in Sweet (I really humbly think the examples I give on WB/ISO handling in that discussion is quite valid, so I will just quote myself in all simplicity:
http://www.reduser.net/forum/showthread.php?105573-Sweet&p=1259888&viewfull=1#post1259888)

But to make any sense of this, I will still have to use a few words. Sorry.

WHiteballance in a digital camera is pretty simply done (in theory) and you can watch it happen by sliding the color-temp slider in f. ex. redcine X.

If you go from say daylight to tungsten, the green chanel rameins untouched, the blue channel is amplified and the red channel gets a clamping curve on the top.
The harder the deviation is, the more brutal effects you get from the alterations in the red and the blue channel.
In addition tint is adjusted by changing the saturation values in the different channels.
All to give you a lovely balanced image.

And that is all sweet... If it worked well.

Which it up to Dragon basically not has done according to my taste...

I will explain...

This actually works kinda ok, unless you are in the top of the exposure range or the adjustment is "too big".
but what happens in those two scenarios, is that you get color artifacts.

Most common is flat/yellow, metallic/patchy skin when approaching clip.
And frankly, I thought that was close to impossible to solve in a digital came, unless you underexpose skin a bit under tungsten.
Well, that kinda solves it, but on the expense of tonality...

On the Epic, I think of the 800 ISO advice as a way to give headroom for white ballancing, so you do not have to change ISO according to colortemperature, which would be rather confusing for many. (and is a weird way to operate a camera...)

Simple explanatioon is;
Now it works...

hahaha

have not figured out how Graeme has done it, and simply does not give a shit if there is elaborate float maths or 20.000 gremlins in the camera, but now colors just works.
But I do think that the added fidelity is due to a number of things.

more precise colorsampling
More overhead/Less noise for color operations.
More saturated colors, so more of the operations are done through desat instead of gaining saturation.

And definitely an oscar-worthy achievement for Graeme and whoever gazillions who have tried to break the puzzle of digital white balancing of a daylight responding sensor.
I just did not think it was possible!

And now it "just works"

I have never had any real issues with RED colors. Or skin.
But handling it has always been a matter of knowing signalpath.

And from the workflow descriptions I have read from f. ex. lightiron, I have seen that I am not all alone and mad. Take it "flat and full" into grade, and work the magic on a developed full RGB signal.

With the developments of Dragon, that has all of a sudden become a lot easier and the results are astonishing.

And by that. I will go to sleep.

I will post some R3ds and stills over the next few points, but all in all...

As Phil said I said...

Dragon is just "more".

G
 
I know some people here struggle with Gunleik's sophisticated references to technical terms, so I will sum up an important advantage:

1 - Dragon does not clip as fast, so you can expose an image with higher light levels (i.e., more open f-stop, less ND, or simply more light hitting the scene). On MX, if you throw too much light at an image, it will clip and/or show color infidelity faster, so you have to let less light hit the sensor to avoid this.

The benefit of exposing an image with higher light levels overall, is that the darker parts will also get more light, not just the brighter parts, therefore the darker parts will be less noisy, even assuming (likely incorrectly as black shading and other firmware aspects will improve quickly in coming weeks/months) that the native noise floor of Dragon and MX are the same in a 1:1 crop).

Yes you can control contrast ratios with lighting - and really you should - but you don't have to be as fussy with that on Dragon, and sometimes resources are not available to do that, and sometimes it's just impossible, even if you wanted to take the time to do so.

2 - Comparing noise at a 1:1 crop is not fair, because for the 5K MX field of view, you are using 5.5K of resolution and throwing that many pixels at the same final deliverable result. A fair comparison would show:

A) A 1:1 crop of MX vs a crop for Dragon that is 5.5 / 5 so contains 10% more pixels, then scaled down to be exactly the same size on screen as the Mx 1:1 crop - this would show exact noise differences for the same field of view/lens coverage.
B) a Full 6K image, scaled down to match the on-screen size of the Epic 5K image. Remember that the wider field of view of full 6K will make some lenses look better in myriad ways- so this comparison will reveal that effect too. Matching the same framing might be tough, but not impossible. You could do it with a good zoom that covers full 6K. Then you could downsample (not crop) both images to 4K, 2K and 1080p.
C) 6K 1:1 vs MX 5K 1:1 would need to have the 6K 1:1 taken from a wider swath of image (20% more pixels in either direction) then scaled down by a factor of 5/6 = 0.833 in both directions. This would be a relevant "Max possible pixels" test, useful for those using lenses, such as Canon Ls, that cover the full 6K sensor.

Make sense?

Sensor calibration should be handled with great care in both cases to be fair to both cameras.
 
You must log in or register to reply here.
Back
Top