All the Epic/Dragon Vs. Alexa/Sony threads are belong to us...

[Marc Wielage]

I just have to say that I fundamentally agree with Gunleik and Phil, we are looking at the same things.

Peace at last! Peace at last! Great godamighty, peace at last!

It's a good day when Geoff, Gunleik, and Phil can all agree.

 

[Gunleik Groven]

It is not all surprising, though.

We basically look at the same things with the same cameras available.

And I really do not envy Geoffs work on looking through and preparing all this.
From the few "normalizing" examples I have made here, I can see that that will get a bit noisy... (pun intended)

It is still very informative in a good way, if you understand what you see.
Amongst other things, this test contains a lot of the answers to the eternal "what ISO would you rate the Dragon at" question, which is confusing as Mark T says he prefer to shoot it at 250, Mark P.. suggests 640, and I am kinda more vague, except saying I guess I will more often shoot it at 250 than 2000...
RED suggest 800 (if memory serves me) and I generally says that THAT is a good place to start...


[I]The answer to all those questions, are at least partly in the test published, if you know how to use the raw files relased.

Hint:
the "1 over" shot, when normalized equals "correct exposure measured at" @ ISO 400, the "2 over" equals ISO 200, the "1 under" equals ISO 1600 and the "2 under" equals ISO 3200.

If you download the files and start off at those 5 ISO settings when you grade, you can see a bit of how the camera performs at the different relevant ISOs.
Same can be done with the Alexa files.


The real bomb would be if you find that the image from the Dragon works (for different uses) at all those 5 exposures, that you end up with a "full" image, from all those base exposures and ISO settings, that tell a little of the story of how far the Dragon has come compared to anything I know.

Now... Do the same with the other cameras. [/I]

My hunch is that you'll get a good hunch as to what Phil, Mark, Mark, Christopher and maybe even Geoff sees, through the little test I suggest here with the published material....

Look forward to seeing Geoffs results.
And material with the new OLPF/ColorScience...

If you are curious about the current state of things, take the "base" exposure (there a are lots of colors and levels in that exposure), export TIFFs and scale the cameras you are interested in into UHD, and look at the areas of interest.
Also take that same "base" exposure and grade it in whatever way you prefer, before/after scaling.

My guess is that you will find a lot of relevant info as to performance by doing that. (neglecting operator fuck-ups, of course).

Anyhow. I'll log off for a while now.
have some other stuff to get done...



Cheers

G

 

[David Litchfield]

The Old OLPF and Calibration do measure up to a practical 16.3 stops of usable dynamic range. The key to Dragon's Dynamic Range and Color is how you choose to unlock it.

I was under the impression that the Dragon sensor gave us all the DR regardless of ISO? Can you explain how I 'unlock' the dragons DR in redcine-x so that I can see these 16 stops because they are currently AWOL in the R3ds.

First Red post the 21 step chart and everyone gets excited that we are going to get 18 stops. Then it's an official 16.5. Now it looks like it does the marketing claim of the MX and is still less than the Alexa.

David

 

[David Battistella]

I have a tiny suggestion around all this DR talk.

Can we, just from now on, just refer to whatever the images are like NOW, like, now that the NEW OLFP is shipping with new Dragons and maybe just only refer to the current state of things?

I think all of this "OLD OLPF, NEW OLPF", "when the NEW color science comes, when the NEW calibration comes". I think that this kind of talk tends to cloud the issues and sound like excuses.

I get that there needs to be a way to highlight, "this is how it was then (two months ago) and this is how it is now (today), and I know that RED is always a moving target, but it think it might be helpful to just start only talking about how it is TODAY and going forward, otherwise it might just contribute to the overall FUD arguments that become quite pervasive outside of Reduser.


Just the fact that this major test was conducted on the OLD OLPF, OLD calibration clouds things, because most people are not following this as fanatically as we few redusers and fortunately or unfortunately, this is going to be the side by side people look at. I wish this could have been done on the new OLPF, ext because now the whole thing is clouded by..."no wait, there is actually something new and better".

I know that, "such is red" but it would be nice to assume every Dragon has the NEW OLPF (or will soon) and that is what we could be talking about. Because, for me, not upgraded yet, (still waiting for that e-mail ) and probably the majority of people what ever the OLD color science OLPF combo was, does not really matter.

Just a thought,


Battistella

 

[Gunleik Groven]

Quickly logging back on...

I agree, to some extent.

Geoff has said he will add F65 and "NEW" OLPF to the data when available, and I trust he will!

This test is good for what is currently available, and there is frankly no need to excuse the current Dragon.

When "New" is available, "new" will get up as part of the test.

As I said, this is a very repeatable test, so I see no real issues in adding data with more cameras down the line.

It is (DR-wise) based on the greylevel chips on the chart, the top 2 and the bottom 2, and was measured to the levels we agreed to.

I see no problems in repeating that and adding to the test as time goes by.

But your main point (or mine) is:

No point in making the Dragon "sound bad" because there is something better down the line.
It is already excellent.

To that, I agree 110%

NOW I am off...

 

[D Anderson]

No point in making the Dragon "sound bad" because there is something better down the line.
It is already excellent.

I think Battistellas point isn't Dragon sounds bad - but the CONSTANT drumming on about the new OLPF is what is getting in the way of the test results. The bulk of conversation regarding this test has been about how it doesn't feature OLPF but discussion on DR comes a distant second.

 

[Gunleik Groven]

I think Battistellas point isn't Dragon sounds bad - but the CONSTANT drumming on about the new OLPF is what is getting in the way of the test results. The bulk of conversation regarding this test has been about how it doesn't feature OLPF but discussion on DR comes a distant second.

LOL.
Agree!

It is what it is, and it is pretty good...

 

[Matt Ryan]

I think Battistellas point isn't Dragon sounds bad - but the CONSTANT drumming on about the new OLPF is what is getting in the way of the test results. The bulk of conversation regarding this test has been about how it doesn't feature OLPF but discussion on DR comes a distant second.

I agree, and in this specific test, Dragon has one stop less DR than Alexa. Usable may be equal based on noise of the Alexa low end. I'd like to see with the "new olpf" that we get to that 16.3 stops (usable) in a side by side like this with Alexa. At that point the DR advantage with Dragon should be obvious when compared directly to Alexa...

 

[David Battistella]

I agree, and in this specific test, Dragon has one stop less DR than Alexa. Usable may be equal based on noise of the Alexa low end. I'd like to see with the "new olpf" that we get to that 16.3 stops (usable) in a side by side like this with Alexa. At that point the DR advantage with Dragon should be obvious when compared directly to Alexa...

In terms of DR, and what is usable, etc. I think the real test is how a DP lights based on the knowledge of the camera. In this way I think you would find all things pretty much equal (no matter what OLPF) as any DP would not light and treat any of these camera's equally, just as they would not treat every film stock the same. Every sensor has it's "tendencies" and knowing those are the job of the DP. the "problems" are not "problems" they are an understanding of limitations.

On any given day each camera could "outperform" the other in certain situations.

What I am seeing in Dragon footage is low noise/grain and very clean blacks. That means pushing the lowlight is going to create creative opportunities, especially when it comes to intimate performances and certain kinds of films. Overall, it could mean faster setups and more pages on a shooting day, because the base lighting can come down by a stop or so. That can have huge implications compared to whether you want to blow out a window.

The thing that drove film companies into producing higher speed stocks was partially to speed up the filming process (getting rid of the bulk) and take cameras out of studio environments. Location shooting (then) saved money on sets (because you had ready made sets you did not have to build)

So having a compact camera that shoots a very noise free image in the lower stops can have an amazing affect on filmmaking. I'm not talking about "no craft" on talking about establishing a lighting style that does not compromise quality while allowing new kinds of methods to get shots lit and get through pages.



Battistella

 

[James_Mills]

IN terms of DR, and what is usable, etc. I think the real test is how a DP lights based on the knowledge of the camera.

Of course you should light to maintain the range of the camera's latitude. Excuse my terseness, but that's entirely tantamount to saying "You shouldn't overexpose or underexpose your camera".

This is the kind of cop-out that we fanboys get made fun of for; Red claims 16.5 stops of DR, a test is conducted with results that questions their claim, and then their loyal users proceed to retort that "It doesn't matter, you should use more lighting techniques to suit your camera." It makes us all look bad, RED included.

 

[Matt Ryan]

IN terms of DR, and what is usable, etc. I think the real test is how a DP lights based on the knowledge of the camera. In this way I think you would find all things pretty much equal (no matter what OLPF) as any DP would not light and treat any of these camera's equally, just as they would not treat every film stick the same. Every sensor has it's "tendencies" and knowing those are the job of the DP. the "problems" are not "problems" they are an understanding of limitations.

On any given day each camera could "outperform" the other.

What I am seeing in Dragon footage is very clean blacks. That means pushing the lowlight is going to create creative opportunities, especially when it comes to intimate performances and certain kinds of films. Overall, it could mean faster setups and more pages on a shooting day, because the base lighting can come down by a stop or so. That can have huge implications over whether you want to blow out a window.

The thing that drove film companies into producing higher speed stocks was partially to speed up the process and take it out of studio environments. Location shooting (then) saved money on sets (because you had ready made sets you did not have to build)

So having a compact camera that shoots a very noise free image in the lower stops can have an amazing affect on filmmaking. I'm not talking about "no craft" on talking about establishing a lighting style that does not compromise quality while allowing new kinds of methods to get shots lit and get through pages.



Battistella

I agree with you David, but I'm talking purely in a side by side setup. If Dragon has 2 stops more DR than Alexa, then it should show an obvious 2 stop advantage when shot side by side.

I've never blamed the tools for my mistakes and a camera only gives you what you give it. But my quarrel is in everyone saying it has 16.3 stops when clearly in this side by side it does not have 16.3 usable. Theoretical, maybe, but I'd like to have all 16 stops available in every shot I'm giving it that amount of exposure latitude.

 

[Björn Benckert]

I agree. From these tests if Alexa has 14 stops then Dragon has slightly less. You guys can throw around technical numbers all day but when you compare them side by side, Dragon has slightly less DR. If all the new OLPF did was move the calibration up slightly to give 1.3+ stops more in the highlights, but in return lose a bit on the low end, then were still less (practical) DR than Alexa. While all of Phil's and others theoretical DR numbers are nice, it's not practical at all. If Dragon is indeed a practical 16.3 stops, then it should have an obvious advantage over Alexa regardless of the test. Unfortunately, that does not seem to be the case yet. I do agree by far that Dragon is cleaner than the Alexa noise wise, but the DR isn't there yet from these tests...

Long story short is if Dragon has 2+ more usable stops over Alexa, then it would be more obvious in shootouts...

I'm confused...
I have not shot the two side by side. But we used the Alexas quite a bit lately and to me it's quite obvious that the dragon we just got delivered (I think it has the new OLPF) got far more to give then what I seen in any Alexa material. Again I have not done any scientific tests. But my eye tells me Dragon downscaled to 3k looks shining in comparison.

 

[D Anderson]

This is the kind of cop-out that we fanboys get made fun of for; ....................It makes us all look bad, RED included.

Exactly. It is hard not to get excited by the upcoming OLPF but letting that discussion overshadow this test in the way it has does nothing for how Red users (and REDUSER) are perceived.

 

[Logan Smith]

But my quarrel is in everyone saying it has 16.3 stops when clearly in this side by side it does not have 16.3 usable. Theoretical, maybe, but I'd like to have all 16 stops available in every shot I'm giving it that amount of exposure latitude.

Yeah I think the 16+ stops of DR claim is not accurate, for whatever reason. They should have said "16+ stops of theoretical dynamic range" or something. It looks close to the Alexa but it's not an outright Alexa killer that you would expect when hearing claims of 16+ stops of DR. Seems similar to when car manufacturers claim that their new model has 600 Horsepower but their interpretation of what constitutes a "horsepower" is different from the rest of the industry's interpretation who would've rated the engine at 500 HP. I imagine the new OLPF will give it a slight edge over the Alexa but it's still not the undeniable superiority we were promised on paper. At least this is how it appears to me right now.

 

[David Battistella]

James,

Your point is well taken. I should have been a bit more clear. I was not being dismissive of this test which clearly shows 1 stop more in the highlights on Alexa. And does that make it better in the highlights for this test, YES?

We all want a no compromise camera that you never have to light around, but it is a reality of what we do. I could have also said, "you need more light in the shadows to get rid of the unusable DR range in the noisy shadows of the Alexa". We tend to only look at the top of the scale. This is a fair test and it is what it is. What I take away is that I have to modify my lighting according to how the highlights respond if I use this camera. If I use another camera I might have to employ a different strategy. I'll obviously do my own tests with my own camera.

My overall point was about noise/grain in the shadows (which also counts as usable DR) and what that might mean for filmmaking (seeing better into the blacks) or using the camera in lowlight situations.

Your point is taken I understand your point of view. The point of exposure tests is to know how far you can go and how you want to operate within that range. It's not fanboyism. I think I would be being a fanboy, if I personally tested 16.5 stops of DR then tired to defend that against someone else's test. The Alexa (in this test) won in the highlights. Fair and square, side by side, in this test which was conducted by professionals in a controlled environment, i don't think I disputed that, I just was saying that if this is what the camera does, this is what it does.

Have to distinguish the marketing form the images. Ultimately the images speak.

Battistella

Of course you should light to maintain the range of the camera's latitude. Excuse my terseness, but that's entirely tantamount to saying "You shouldn't overexpose or underexpose your camera".

This is the kind of cop-out that we fanboys get made fun of for; Red claims 16.5 stops of DR, a test is conducted with results that questions their claim, and then their loyal users proceed to retort that "It doesn't matter, you should use more lighting techniques to suit your camera." It makes us all look bad, RED included.

 

[Phil Holland]

I was under the impression that the Dragon sensor gave us all the DR regardless of ISO? Can you explain how I 'unlock' the dragons DR in redcine-x so that I can see these 16 stops because they are currently AWOL in the R3ds.

First Red post the 21 step chart and everyone gets excited that we are going to get 18 stops. Then it's an official 16.5. Now it looks like it does the marketing claim of the MX and is still less than the Alexa.

Read this: http://www.reduser.net/forum/showthread.php?111400-Dragon-Flare&p=1322384&viewfull=1#post1322384

And also read what I'm writing next.

Can we, just from now on, just refer to whatever the images are like NOW, like, now that the NEW OLFP is shipping with new Dragons and maybe just only refer to the current state of things

I get that for simplicity's sake. However, I just tested the new tech and it does hold some rather extreme relevance for all Dragon cameras I believe shipping at the end of this month.

Of course you should light to maintain the range of the camera's latitude. Excuse my terseness, but that's entirely tantamount to saying "You shouldn't overexpose or underexpose your camera".

This is the kind of cop-out that we fanboys get made fun of for; Red claims 16.5 stops of DR, a test is conducted with results that questions their claim, and then their loyal users proceed to retort that "It doesn't matter, you should use more lighting techniques to suit your camera." It makes us all look bad, RED included.

This test isn't questioning how many total stops the system captures. Geoff agrees with that and that's not the point of the test.

This test shows where the stops essentially clip and crush in terms of highlights and shadows. It's also very useful to show how much you can "work" each system.

I agree with you David, but I'm talking purely in a side by side setup. If Dragon has 2 stops more DR than Alexa, then it should show an obvious 2 stop advantage when shot side by side.

I've never blamed the tools for my mistakes and a camera only gives you what you give it. But my quarrel is in everyone saying it has 16.3 stops when clearly in this side by side it does not have 16.3 usable. Theoretical, maybe, but I'd like to have all 16 stops available in every shot I'm giving it that amount of exposure latitude.

My best advice. Take the baseline exposure from all of the cameras tested. Measure patch value, I'll post an example in below, and take readings from a low ISO (like 250) through the something around ISO 6400 for instance.

To be clear, I'm not saying Dragon isn't clipping when it clips here. I am saying there's more Dynamic Range to be measured across this frame. You have to read the density/values of every exposure. Also measure each of these ISOs to see where things are "placed" and stressed. The reason I mention the new calibration specifically in regards to this test is how this baseline exposure was achieved and the fact I've measured how much more highlight retention is kept.

You have to see where each system places it all.

For instance, here's an example from the base exposure:

It's about what each system "does" with their total captured dynamic range everywhere.

Which of those images shows more dynamic range, workable tonal latitude, and color latitude? How many stops would you say this is? Both of these images aren't clipping, this isn't maximum highlight or shadow dynamic range, where does this increase in Dynamic Range and Latitude performance land?

Interesting stuff if you look closely.

*Edit - I did white balance that above example, as there's a part of the usable color range and purity to be found there. Just yet another thing to really look at.

 

[James_Mills]

To be clear, I'm not saying Dragon isn't clipping when it clips here. I am saying there's more Dynamic Range to be measured across this frame.

But are you also saying that the same cannot be attributed to the Alexa's images here as well?

Your point is taken I understand your point of view. The point of exposure tests is to know how far you can go and how you want to operate within that range. It's not fanboyism. I think I would be being a fanboy, if I personally tested 16.5 stops of DR then tired to defend that against someone else's test. The Alexa (in this test) won in the highlights. Fair and square, side by side, in this test which was conducted by professionals in a controlled environment, i don't think I disputed that, I just was saying that if this is what the camera does, this is what it does.

Have to distinguish the marketing form the images. Ultimately the images speak.

Battistella

I understand what you're saying better now, thanks for clarifying.

 

[Phil Holland]

But are you also saying that the same cannot be attributed to the Alexa's images here as well?

See the above image and you make the call.

 

[David Battistella]

Read this: http://www.reduser.net/forum/showthread.php?111400-Dragon-Flare&p=1322384&viewfull=1#post1322384

And also read what I'm writing next.

I get that for simplicity's sake. However, I just tested the new tech and it does hold some rather extreme relevance for all Dragon cameras I believe shipping at the end of this month.

This test isn't questioning how many total stops the system captures. Geoff agrees with that and that's not the point of the test.

This test shows where the stops essentially clip and crush in terms of highlights and shadows. It's also very useful to show how much you can "work" each system.

My best advice. Take the baseline exposure from all of the cameras tested. Measure patch value, I'll post an example in below, and take readings from a low ISO (like 250) through the something around ISO 6400 for instance.

To be clear, I'm not saying Dragon isn't clipping when it clips here. I am saying there's more Dynamic Range to be measured across this frame. You have to read the density/values of every exposure. Also measure each of these ISOs to see where things are "placed" and stressed. The reason I mention the new calibration specifically in regards to this test is how this baseline exposure was achieved and the fact I've measured how much more highlight retention is kept.

You have to see where each system places it all.

For instance, here's an example from the base exposure:

phil,

Just want to make sure I am understanding this clearly
i take this this to mean that if one:

exposes each camera equallly (subjects them to exactly the same lighting)
WB and post process dragon at a different ISO thehighlight DR is recoverable.
the WB step is important because correct bw will have an effect on highlight DR.

So the top image in you spilt screen is dragon adjusted in post to match the Alexa highlights?

or

when exposed correctly the DR is equal but when over exposed (like in this test stressing the sensor and hitting it with more and more light) Alexa holds highlight detail by an extra stop , but only when overexposing?

battistella

 

[Phil Holland]

Just want to make sure I am understanding this clearly

Short and sweet. All that above image is is taking both files, in this case both "base exposures at middle gray", and manipulating them equally. In this case ISO adjustments.

Long and the short of it. Look at these files closer. Push them around hard until they all fall apart. You can see a lot more in there across the whole dynamic range and how it's used. This mainly is showcasing the increased workable latitude and also where the DR is placed on this calibration. Don't just use ISO, adjust the gain and various other settings in your programs of choice. Keep it 1:1 in your adjustments and you can see a lot.

The honest most telling thing for me from this comparison is perhaps why the newer calibration was tuned a bit differently. I see where Red put the weighting before and now after. It makes more sense now.

And yes I apologize about going on and on about the new calibration, but really soon that will be all that is relevant. I've already measured where the gains are and that will be important to those who are merely seeing that Dragon is clipping before Alexa and think that there's less total DR because of that, which is not the case but this test is also not an apt way of testing precise DR in reality. Things have now changed though on that front.

That doesn't diminish the results from this camera comparison though. All of this is gravy.