Meaning of 16.5 stops of Dynamic Range

[John Marchant]

Hey Mark. Its probably too long a post for now, as I should be wrapping for the day... essentially though, all cameras will have a noise floor, and all humans will have a tolerance for that noise in the image. The picture is everything above the noise floor, so it makes simple logical sense to test camera systems for comparative dynamic range by benchmarking a visually acceptable noise floor and judging what lies above it on that basis.

Every manufacturer's implementation of ISO varies, and testing at identical practical settings of lens / software setup may hamper or advantage any given system.

My suggestion is that to compare two cameras for dynamic range, the most illustrative starting point is the one where the shadow noise is most visually comparable in motion. If one camera is cleaner than another, it implies that the chosen ISO, mid gray, gamma curve, or whatever other developmental implementation is used is comparatively biased towards shadow. If the two cameras were otherwise identical, the one with cleaner shadow would consequently have to show less highlight range.

When you then conceptually write off the shadow difference, you draw a false conclusion about the highlights.

 

[Paris Remillard]

I don't mean to sound argumentative, but none of these explanations match the tests that have been shown so far. Geoff's tests show that the Alexa has more DR. The Dxo score says 14.8 stops for Dragon, not sure what their methods for testing are. And the Sekonic profile that John posted says 16.3 stops DR with 13.6 clipping point/usable for Dragon. Where Luigi's Sekonic profile of the Alexa gives it 17.6 stops DR with 14.1 clipping point/useable.

So it appears that we either need to say that the Alexa has 17.6 stops of DR or we need to say that the Dragon has 13.6, based on which value we decide to go by. Either way, all of the tests posted so far show that Dragon is still behind Alexa (and likely f65 since it seemed to match or better alexa in previous tests) in DR. And the general consensus is that the Alexa is a 14+ stop camera. So 16.5+ stops for Dragon just doesn't make sense, in my opinion.

 

[Christian Jadot]

I do a lot of work inside to outside military aircraft. HDRx has been a godsend in this regard to increase the dynamic range. Several times I have "broke" the MX sensor and have retreated to HDRx for the extremes. I am gushing over the 16.5+ stops. Here is a still I took with a Nikon D4 trying to find a good compromise. It is nearly impossible to get without HDRx, but my hopes are the Dragon will change that.

 

[Gunleik Groven]

Now seriously, What I don't undersand is why looking at the test done here:

http://www.cinematography.net/UWE/index.html

in the interiors scene I can see from the ungraded files very easily Alexa getting better dynamic range than the Dragon

on the exterior scene they look very close to me probably because there is not anything to make reference to (like the grayscale bar on the studio)

http://blog.gunleik.com/2014/03/19/...d-scaling-dragon-alexa-f55-c300-pt-1-preface/

1st.
I subscribe fully to Phils and Johns statements, and also to what we did in Bristol.

2nd
By now the new OLPF and color-science is here + ADD...

Each of these three factors change the landscape, united they change a lot.

As to the exteriors, they were shot with the same exposure (T-stop/shutter), but they were not exposed to break either camera, thus they do not.

As to practical DR, it seems to me that... (hahaha)

Simply put:
Shoot a scene (not single frames, even though that is what we all publish) with Dragon and Alexa side-by side and display it. Even at 1080 you should be able to note some significant differences.

What these tests are not too good at showing is color, saturation of channels and color precision.
Also they are not very good at showing off "gentleness" and "robustness". They are analytical.
Good or bad so. They still are. They do not communicate too well "how it feels".

 

[luigivaltulini]

I don't mean to sound argumentative, but none of these explanations match the tests that have been shown so far. Geoff's tests show that the Alexa has more DR. The Dxo score says 14.8 stops for Dragon, not sure what their methods for testing are. And the Sekonic profile that John posted says 16.3 stops DR with 13.6 clipping point/usable for Dragon. Where Luigi's Sekonic profile of the Alexa gives it 17.6 stops DR with 14.1 clipping point/useable.

So it appears that we either need to say that the Alexa has 17.6 stops of DR or we need to say that the Dragon has 13.6, based on which value we decide to go by. Either way, all of the tests posted so far show that Dragon is still behind Alexa (and likely f65 since it seemed to match or better alexa in previous tests) in DR. And the general consensus is that the Alexa is a 14+ stop camera. So 16.5+ stops for Dragon just doesn't make sense, in my opinion.

Paris , was with OLPF old ..
Now we have added 1.5 stops above and removed about 0.5 stop under + - the new OLPF .
For what I can see it.
If we want we can also do a test with Add plugged in and give a plus even in noise DRagon .
But surely Phil and John are much more technical and subtle in their representations .
So now as I'm concerned we are more than a little to Alexa, the DR (new olpf).
I remind you that Alexa has half the resolution , we can talk when she arrives at 6k .
But maybe I'm boring and I repeat, when looking at the monitors side by syde Alexa and Dragon , Dragon is much more soft, clean, you have to see it to believe it , comparisons with the still " test" does not make it reality.
So to yield the Dragon is now made ​​even better in the colors of Alexa and much more manageable in post, you can push more .
Even here you have the same feeling, Dragon goes further.
Having said that Alexa is a wonderfull camera but now DRagon is better.
Expected to see something to NAB by Arri but for now nothing .

only my opinion.

 

[Paris Remillard]

Thanks for the comments and for the testing that you've done so far Luigi. And to everyone else who has taken their time to do so. It's very much appreciated and I hope not to seem ungrateful by any of my comments or observations.

It would be great to see an revised Sekonic profile with the new calibration. Everyone uses that as a disclaimer when looking at these tests. But I've yet to see any visual proof of the numbers stated.

And people are mentioning noise floor and noise tolerance, but based on the raw files from Geoff's tests, Adrian Jebef posted these frames halfway down the page on this thread:

http://www.reduser.net/forum/showth...Vs-Alexa-Sony-threads-are-belong-to-us/page31

and he states on the next page "Initially, I'm seeing the ALEXA able to pull down highlights and dig deeper into the shadows than DRAGON V1." And Gunleik states, "As of "now", I see more DR in the Alexa.
Currently, from what I see, the alexa records both higher and lower luma values. No "but" to that."

Now, I of course didn't perform the tests. I'm only going by what I'm seeing.

And with the new calibration, it appears that there is indeed more overexposure latitude in Dragon based on Luigi's and Phil's tests. But it comes with the trade off of more noise at the bottom. So it seems that we wouldn't really be able to stretch the useable DR by starting lower in the noise floor.

I eagerly await more tests and more information. And I hope I'm wrong. But speaking of DR alone, not taking into account all of the other great features of the Dragon sensor, the Alexa is still on top in this one respect.

The point of all of this, other than gaining knowledge of this sensor and how to use it to it's fullest potential, is to say, the 16.5+ stop claim either needs to be revised, or other camera manufacturers numbers need to be revised to the same scale to show that the Dragon does not have 2.5+ stops of DR over Alexa or f65. In my opinion.

 

[Brian Merlen]

the reality is all the tests need to be redone cuz too much has changed now... and once that is complete we can have a better understanding via the various measuring systems. red releasing the new color science is a big deal. now we can truly compare the two cameras to our hearts contents! see where the cards fall, but for the price one is paying per stop of dynamic range via red, even if it were measured less then alexa, which i won't believe until all the tests are redone by multiple supposedly unbiased sources clearly red is still the less expensive option for owner ops. I have to admit i am happy looking at the way arri and sony have been operating lately, firmware updates, module upgrades, red has warn off on the competitors and customers are getting more value, thats a win win for everyone guys. forget the partisan approach to camera choice, we have a real competition going on now on who can deliver the most quality, efficient workflows, and have had real success in retailers not cannabilizing last years models via rerelease tweaked models. things are better then ever for us, the customers, and i think a major part of that is due to reds success. and on that heartfelt note if the 65mm rumor is true i am robbing a bank and buying an arri still lol

PS what about HDrX on dragon? does that count too? cuz if we add that to the situation i am pretty sure it can beat alexa...

 

[luigivaltulini]

Thanks for the comments and for the testing that you've done so far Luigi. And to everyone else who has taken their time to do so. It's very much appreciated and I hope not to seem ungrateful by any of my comments or observations.

It would be great to see an revised Sekonic profile with the new calibration. Everyone uses that as a disclaimer when looking at these tests. But I've yet to see any visual proof of the numbers stated.

And people are mentioning noise floor and noise tolerance, but based on the raw files from Geoff's tests, Adrian Jebef posted these frames halfway down the page on this thread:

http://www.reduser.net/forum/showth...Vs-Alexa-Sony-threads-are-belong-to-us/page31

and he states on the next page "Initially, I'm seeing the ALEXA able to pull down highlights and dig deeper into the shadows than DRAGON V1." And Gunleik states, "As of "now", I see more DR in the Alexa.
Currently, from what I see, the alexa records both higher and lower luma values. No "but" to that."

Now, I of course didn't perform the tests. I'm only going by what I'm seeing.

And with the new calibration, it appears that there is indeed more overexposure latitude in Dragon based on Luigi's and Phil's tests. But it comes with the trade off of more noise at the bottom. So it seems that we wouldn't really be able to stretch the useable DR by starting lower in the noise floor.

I eagerly await more tests and more information. And I hope I'm wrong. But speaking of DR alone, not taking into account all of the other great features of the Dragon sensor, the Alexa is still on top in this one respect.

The point of all of this, other than gaining knowledge of this sensor and how to use it to it's fullest potential, is to say, the 16.5+ stop claim either needs to be revised, or other camera manufacturers numbers need to be revised to the same scale to show that the Dragon does not have 2.5+ stops of DR over Alexa or f65. In my opinion.

good comments Paris .
You could redo the test SEKONIC a day.
The way I see it, if the ends for you ' are the important thing then I can say that Alexa has something more in high light, low light and also something more , dig more .
Perhaps Gunleik can confirm , my English is bad.
But you have to keep in mind that the dynamic range is also in the rest of the image, the sweetness of the nuances of Dragon images are better and less noisy for what I keep repeating that you have to see the same images on the monitor next to wall.
I have many DP who regularly use Alexa and now they see Dragon fell in love ..
So I think that Dragon has the most dynamic Alexa in the complex but not in the end extreme high and extreme low, if that 's what you want to hear .
I do not believe that you or others are going to dig up 5/7 stop of light in order to extract something, it would almost always noise .. it would probably be most wanted by the DP a dark area ..

sorry for my bad english .. maybe I mean bad.

 

[Paris Remillard]

I'll just reiterate for clarity. I'm not trying to say who wins or loses or which camera is better or worse overall. I think they're all great at this point. If a DP can't make great images with any of the available digital cinema cameras, it's operator error. It has nothing to do with the camera.

I just think that the numbers used should be revised. Because as it stands, saying 16.5+ stops for Dragon is misleading when cameras that claim 14 stops clearly out perform it on both ends of the scale in this one measure of DR.

 

[Robert Ruffo New]

Andrey what you are seeing in those tests is hard to read and interpret, but Dragon indeed exceeds Alexa in dynamic range and that isn't really being argued.

You would almost never want all 16 stops within the same region of an image because that region would likely look flat and washed out.

What you likely will do is choose a range of 11 stops or so within that for your final grade, and pro graders do not often use Redgamma 1 2 or 3, they use logfilm. (So why do these tests seem to always use Redgamma 3?…)

Wonderful thing about RAW is that you can use power windows and have one side of the image (or say a bright window outside) choose a different range than another side (say the dark parts of the image).

It should also be noted that good noise reduction like Neat Video - which those tests never try - works better on Dragon than Alexa images because Dragon's are higher resolution, so you get an extra half to full stop in the darks, sometimes even more, depending on the material, than without noise reduction.

I would also note that 16 stops is just past the dynamic range of 99.9% of real world scenes, lit or not, whereas 13.5 stops (MX) is just under once in a while. This is significant. It means you have headroom no matter where you are or what you are shooting, and can use that headroom for sexy falloff curves at both ends without hitting white clip or the noise floor.

Cams that have 12 stops are very often clipping somewhere, because they are under most often seen real-world scenarios.

 

[luigivaltulini]

I'll just reiterate for clarity. I'm not trying to say who wins or loses or which camera is better or worse overall. I think they're all great at this point. If a DP can't make great images with any of the available digital cinema cameras, it's operator error. It has nothing to do with the camera.

I just think that the numbers used should be revised. Because as it stands, saying 16.5+ stops for Dragon is misleading when cameras that claim 14 stops clearly out perform it on both ends of the scale in this one measure of DR.

this is a good discussion Paris,
I actually would remove all the numbers
We take the same pictures and go to color them we can judge what we have.

 

[Paris Remillard]

Andrey what you are seeing in those tests is hard to read and interpret, but Dragon indeed exceeds Alexa in dynamic range and that isn't really being argued.

All of the tests I've seen so far show that the Alexa still exceeds Dragon in DR. And it's been confirmed by some of the testers, who are trusted Redusers, as I posted above. I'd love to see any tests that show otherwise. And I say that not as a challenge, but out of a genuine desire for information and understanding.

 

[Jon MIchael Puntervold]

What I find kinda interesting, is that most of the new cams released, still just (and that is by announced specs) tangent the 2007/2008 RED ONE M in specs.
Somehow it is understandable why RED and ARRI are the two main "cinema" cameras.

But with the Dragon, there is a real choice, even if easy access to colour and fidelity is what you look for.
I konda am not too fond of the "resolution" argument in itself. But when everything else gets there, it is of real importance.

As it is for stills
And...

Geoff and DxO found about the same numbers for "useable" DR on the Original Dragon.
And that number is already not rivaled by many cameras.

I have said this all the time since I got to play with the prototype in august:

the numbers ARE impressive, but they are really not what made me change my mind from my plan to stick with Epic.
The FEEL and colours of the damned thing, changed my mind.

(Gunleik, unfortunately logged in with Jon Michaels computer I discovered.
We are not one, though... Hahaha)

 

[Robert Ruffo New]

Paris this not really true. What they have shown is that a single Red Gamma curve output at a single ISO has less dynamic range - they never even try anything else, or moving ISO up and down to see what is really in the image. If you have seen tests done in some better manner, please point me to them.

If you raised or lowered the ISO of any of the images he shows, you would get detail back in most of them from darks and from highlights. - But what he show is ISO set to 800 for all of them, and RedGamma 3 applied, whereas the Alexa images are in LOG mode. That doesn't tell you much and isn't technically correct nor fair.

 

[Paris Remillard]

The raw frames are posted with Geoff's tests. And as I linked before a few posts ago Adrian Jebef did take the raw data and reprocess it at different ISOs. You can do so yourself if you'd like.

http://www.reduser.net/forum/showth...Vs-Alexa-Sony-threads-are-belong-to-us/page31

And both he and Gunleik, who performed the tests and has been testing Dragon against Alexa for months, have said that there is more detail in the highlights and in the shadows on Alexa.

In addition to that, anyone who has posted Sekonic DR profiles of Alexa and Dragon, they have shown Alexa to have more DR than Dragon.

But as I've been saying, and I'm assuming what was being asked by the OP is, why is the Dragon rated at 16.5+ stops of DR when it appears to be less than other cameras that are rated at 14 and when other measurements also show it to be less than that?

This is just about that number. I'm not debating the other merits of the camera system. It looks to be quite good on many other levels, and in DR, even, if taken on its own. But it's simply not 2.5+ stops improved over Alexa. So that 16.5+ stop claim seems to be misleading. Or else, Alexa needs to be rated at 17.5+. That's the only point I'm trying to make

 

[Björn Benckert]

The raw frames are posted with Geoff's tests. And as I linked before a few posts ago Adrian Jebef did take the raw data and reprocess it at different ISOs. You can do so yourself if you'd like.

http://www.reduser.net/forum/showth...Vs-Alexa-Sony-threads-are-belong-to-us/page31

And both he and Gunleik, who performed the tests and has been testing Dragon against Alexa for months, have said that there is more detail in the highlights and in the shadows on Alexa.

In addition to that, anyone who has posted Sekonic DR profiles of Alexa and Dragon, they have shown Alexa to have more DR than Dragon.

But as I've been saying, and I'm assuming what was being asked by the OP is, why is the Dragon rated at 16.5+ stops of DR when it appears to be less than other cameras that are rated at 14 and when other measurements also show it to be less than that?

This is just about that number. I'm not debating the other merits of the camera system. It looks to be quite good on many other levels, and in DR, even, if taken on its own. But it's simply not 2.5+ stops improved over Alexa. So that 16.5+ stop claim seems to be misleading. Or else, Alexa needs to be rated at 17.5+. That's the only point I'm trying to make

Are you talking new OLPF or old? I have not seen any test with the new OLPF and the new color science, have you? As all tests with old olpf is beta tests and the new color science came out yesterday, I doubt anyone ha been testing the dragon, in it's final version, at least not for months.

 

[Paris Remillard]

It was the old calibration/olpf. But the tests that Luigi posted of the new calibration/olpf do indeed show more range in the highlights but much more noise in the shadows. So it appeared to be a wash in terms of increased overall DR. More of a shift from one end to the other. I'd be happy to be wrong about that.

And I fully understand that this has been a work in progress and that things change. I'm a Red One early adopter so I understand the drill : )

But do we really think that there's going to be 3+ more stops of DR in the new calibration/olpf/color science? That's what it would take to be 2.5+ stops greater than Alexa, as the 16.5+ stops claim suggests. I'd be thrilled if that were the case. But I'll remain cautiously optimistic.

 

[luigivaltulini]

It was the old calibration/olpf. But the tests that Luigi posted of the new calibration/olpf do indeed show more range in the highlights but much more noise in the shadows. So it appeared to be a wash in terms of increased overall DR. More of a shift from one end to the other. I'd be happy to be wrong about that.

And I fully understand that this has been a work in progress and that things change. I'm a Red One early adopter so I understand the drill : )

But do we really think that there's going to be 3+ more stops of DR in the new calibration/olpf/color science? That's what it would take to be 2.5+ stops greater than Alexa, as the 16.5+ stops claim suggests. I'd be thrilled if that were the case. But I'll remain cautiously optimistic.

the speech should be, Arri is more conservative in his numbers.
Let's look at the reality, Sony F55 claim 14 stops, is equal to Alexa, NO.
Blackmagic 13.5 or stop, I do not think there is only 0.5 difference, so it is NO.
Take a C300, 12 stop, only 2 stops difference from Alexa hehe NO much more ..
So as you can see is Arri most conservative states or only the usable range!
Going back to your speech, as you see in the post John, http://www.reduser.net/forum/showthread.php?111895-Dragon-dynamic-range-profiling-for-Sekonic, we are around 13.6 stops usable .. . adding something new calibration could see the 14/14.5 ... you have to try to take the test again.
So no one is wrong, it depends on how the tests are interpreted, and how the results are interpreted by various companies.
Definitely will come the test of comparison from someone soon, and there will still be those who will discuss this better and better each other.




is just my thought.

 

[John Marchant]

To be fair to the results of mine Luigi links to there - 13.6 stops was me consciously throwing away a full stop of highlight range to guarantee good rolloff, and consciously throwing away more than that in the shadows to avoid talk of noise. The DR figure was 16.3, 13.6 was how I was calibrating my meter for Dragon, allowing headroom for myself.

 

[luigivaltulini]

To be fair to the results of mine Luigi links to there - 13.6 stops was me consciously throwing away a full stop of highlight range to guarantee good rolloff, and consciously throwing away more than that in the shadows to avoid talk of noise. The DR figure was 16.3, 13.6 was how I was calibrating my meter for Dragon, allowing headroom for myself.

john've tried to take the test again with the new OLPF?