Meaning of 16.5 stops of Dynamic Range

[Brian Boyer]

B)It doesn't matter which OLPF/calibration is being used in the tests. Unless people are claiming that there are 3 more stops of DR in the new calibration. Which I would be very surprised to see. And if you read back John Marchant claimed that in his tests with the new OLPF/calibration combo, the overall range was about the same as the old OLPF/calibration used in the tests in question. So there's still a discrepancy between how the Dragon is labeled vs. how most other cameras are labeled. Which should be the point of this thread based on the question that was posed by the OP.

But I understand that people want to wait for the new tests, so I'll just let it go and let everyone use these numbers as they will.

It's not necessarily a case of merely wanting to wait for tests with the new OLPF. It's more a case of that probably being the best course of action.

I disagree with your assertion that it doesn't matter which OLPF/calibration gets used. I don't think I've ever seen a discussion about dynamic range where it didn't inevitably lead to people counting the number of stops above and below middle gray.

Even when talking about the total range, "6 stops over and 7 stops under - so 13 total" (or whatever the numbers are) becomes a part of the conversation.

Since the new OLPF/calibration changed the distribution of those stops and added at least another stop to the top end, that "6 over" measurement is no longer applicable outside of talking about that particular build of the camera, which is not manufactured anymore.

It may be that the total range hasn't changed with the new OLPF/calibration. But, at least waiting for the new data ensures a more accurate endless debate on the subject.

As for RED's claim of 16 stops of DR, John Marchant shot a Xyla 21 chart as well as the Sekonic chart (if I'm not mistaken) that both showed 16 stops and people still choose not to believe it, I'm assuming mostly based on Arri's claims of only 14 stops for Alexa in combination with the CML/UWE results.

Maybe people should be hounding Arri about why they claim so few when they seem to have much more.

My guess is RED counts everything from full blown sensor to the just noticeable difference between signal and noise floor.

I'd still like to see both cameras tested using Arri's contraption for determining dynamic range. They've pretty much dismissed every other method of testing for it on a digital sensor.

 

[Clayton Burkhart]

It's not necessarily a case of merely wanting to wait for tests with the new OLPF. It's more a case of that probably being the best course of action.

I disagree with your assertion that it doesn't matter which OLPF/calibration gets used. I don't think I've ever seen a discussion about dynamic range where it didn't inevitably lead to people counting the number of stops above and below middle gray.

Even when talking about the total range, "6 stops over and 7 stops under - so 13 total" (or whatever the numbers are) becomes a part of the conversation.

Since the new OLPF/calibration changed the distribution of those stops and added at least another stop to the top end, that "6 over" measurement is no longer applicable outside of talking about that particular build of the camera, which is not manufactured anymore.

It may be that the total range hasn't changed with the new OLPF/calibration. But, at least waiting for the new data ensures a more accurate endless debate on the subject.

As for RED's claim of 16 stops of DR, John Marchant shot a Xyla 21 chart as well as the Sekonic chart (if I'm not mistaken) that both showed 16 stops and people still choose not to believe it, I'm assuming mostly based on Arri's claims of only 14 stops for Alexa in combination with the CML/UWE results.

Maybe people should be hounding Arri about why they claim so few when they seem to have much more.

My guess is RED counts everything from full blown sensor to the just noticeable difference between signal and noise floor.

I'd still like to see both cameras tested using Arri's contraption for determining dynamic range. They've pretty much dismissed every other method of testing for it on a digital sensor.

I find this thread hilarious, when I consider that in the beginning, RED was talking about 20 stops on the Dragon sensor.
My goodness, I most certainly can understand some folks dismay here.

Maybe people should be hounding Arri about why they claim so few when they seem to have much more.

I think it's the other way around. RED should begin testing things the way ARRI does because the truth is it has so much less than was initially thought. My guess is that ARRI doesn't gain 4-5 extra stops by using a testing contraption which fogs their output.

I'd still like to see both cameras tested using Arri's contraption for determining dynamic range. They've pretty much dismissed every other method of testing for it on a digital sensor.

And for good reason.

 

[Scott Crawley]

I find this thread hilarious, when I consider that in the beginning, RED was talking about 20 stops on the Dragon sensor.
My goodness, I most certainly can understand some folks dismay here.

The following quote comes from same the thread where this chart first appeared. You will note that he does not say 20 stops.

Conservative... and I think the chart shows 18+ stops. But remember we still have a lot to go in calibration and color science.

Jim

His very next remark was this:

I'm not sure where the 21 stops came from but clearly we see a usable 18+ stops of dynamic range. Maybe more given some tweaking. In any event, it is more than any film stock we have ever seen and certainly more than any digital sensor known to man. This is meaningful.

Jim

 

[Bob Gundu]

True. Never did anyone say 20 stops. Besides, thats not how you read that chart just because you see a 20 up there.

 

[Brian Boyer]

I find this thread hilarious, when I consider that in the beginning, RED was talking about 20 stops on the Dragon sensor.

I think it's the other way around. RED should begin testing things the way ARRI does because the truth is it has so much less than was initially thought. My guess is that ARRI doesn't gain 5.2 extra stops by using a testing contraption which fogs their output.



And for good reason.

From Jim, in that same thread:

It appears that the Dragon will have about 4 stops more DR and less than half the noise of the current Mysterium-X sensor.

Jim

And (referring to the chart)...

I'm not sure where the 21 stops came from but clearly we see a usable 18+ stops of dynamic range. Maybe more given some tweaking. In any event, it is more than any film stock we have ever seen and certainly more than any digital sensor known to man. This is meaningful.

Jim

And, from Graeme:

The chart itself has 21 steps. One stop of dynamic range is measured when you can reproduce the two steps that bracket that stop. The top step of the chart is at clip, so you count the number of steps you can see then subtract 1 (because you need an end step at both ends) to get DR. With the DSC chart, each step is 1 stop apart so the math is easy. Jim's 18+ figure is very reasonable. And tweaking will continue too...

Once the dust settled, it was this:

From Jarred:

Epic Dragon 6K. 6144 x 3160 up to 100fps. 3 additional stops over Epic M-X.

We have kinda given up on stating dynamic range numbers because there are 20 different ways to measure it and they usually all give different results and every method has a somewhat subjective element to it.. so whatever you rated the MX sensor at.. add to that.

From Jim:

Let’s see if we can clear up some Dragon misconceptions.

1. The native Dragon sensor clearly shows 18+ stops shot on our Frankie test bed last December which was posted. Obviously, we were excited out of our skull. Stuffing the sensor’s capability into an electronics package the size of a matchbox always causes loss of performance. That is why we stuck with 3-4 stops of improvement over the EPIC MX. We always measure dynamic range with the engineering definition… as do most others. We said repeatedly 16 stops for EPIC Dragon, which is more than any other digital cinema camera. We stick by this statement… even though we see continuing improvements over time, just as we have always done. For free.

2. If you want 20 stops of dynamic range… shoot EPIC Dragon HDRx 4.

They never claimed 20 stops of native dynamic range.

EDIT: Scott and Bob beat me to it.

 

[Clayton Burkhart]

Sorry guys, the reality is that your lovely Dragon sensor yields around the same, perhaps only a smidgen more DR than the Alexa.
DXO Mark gives it 14.8 and everything that I have seen confirms that.
Not too shabby (DXO considers it the best sensor they have ever tested), but going from 0-20, 18+, 16.5, 16 to 14.8...all these figures bandied around, but generally trending downwards over time..it's funny to watch from the sidelines. (The advantage of renting rather than owning is that one remains uninvested in these kind of tests).
In any case, ain't no way it's 16.5.

By the way, I think that the reason people choose the Alexa had nothing to do with resolution or dynamic range.
It's for the roll-off in the highlights, a curve which has been tweaked to look like film, for the industry standard connectivity, the simplicity of use and the durability of the body.

 

[Gunleik Groven]

Looking good out of the box, highlights and colors were good points for the Alexa.
I agree.

Now the Dragon has those and then some.

 

[Phil Holland]

I'm standing by my measurements in reality with the new tech (calibration/OLPF). 19 visible stop variations via REDlogFilm flat (ain't nobody gonna use it that way!). 16/17 with a "normal" gamma s-curve applied. And I had to measure it the bloody hard way; stop by stop across all ISOs. I'm fine with whatever other folks think, but I've measured the other goodies out there too.

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

Whatever you think the final value is on that DR (and more importantly how you use it) it's real good to acknowledge a few things. We have good color, organic-ish highlight roll-off, great resolution, large format, and industry leading dynamic range right now (though people can keep posting forever about who's right and wrong here).

The only thing that actually happened while testing DR this time around for me is that it's got to be about as controlled as can be in terms of your light source and methods. I found that there's too much cross bleed on some charts to measure accurately because the sensors are so sensitive now. Controlling the source is more ideal and recording and measuring along the way. Using a shadow box massively helped. Clip, in range, plus one is easy enough to find. Shadow density is where things get mysterious because well... it's dark. And mysterious.

Meh. The camera is making images I like. It's easy to get caught up in whatever this is, but in the end all that matters is the image quality and what you can do with it.

 

[Clayton Burkhart]

I'm standing by my measurements in reality with the new tech (calibration/OLPF). 19 visible stop variations via REDlogFilm flat (ain't nobody gonna use it that way!). 16/17 with a "normal" gamma s-curve applied. And I had to measure it the bloody hard way; stop by stop across all ISOs. I'm fine with whatever other folks think, but I've measured the other goodies out there too.

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

Whatever you think the final value is on that DR (and more importantly how you use it) it's real good to acknowledge a few things. We have good color, organic-ish highlight roll-off, great resolution, large format, and industry leading dynamic range right now (though people can keep posting forever about who's right and wrong here).

The only thing that actually happened while testing DR this time around for me is that it's got to be about as controlled as can be in terms of your light source and methods. I found that there's too much cross bleed on some charts to measure accurately because the sensors are so sensitive now. Controlling the source is more ideal and recording and measuring along the way. Using a shadow box massively helped. Clip, in range, plus one is easy enough to find. Shadow density is where things get mysterious because well... it's dark. And mysterious.

Meh. The camera is making images I like. It's easy to get caught up in whatever this is, but in the end all that matters is the image quality and what you can do with it.

I'm actually with you on this Phil, the numbers are fairly meaningless to me.
As I mentioned before though I don't think people rent the Alexa for the numbers either.
Unfortunately, there is a perception in the industry that RED cameras are just too finicky and a bit fragile.
They want something simple that just works out of the box and is tough and I have to say that I have a tendency to agree for some jobs.
My own experience recently with 6 x 8K's in a tight closed lot, was that the Epic we were working on just quit because of the heat.
Took 20 min. for it to come back up again. Thanks god we had a second body.
When the producer/client walks over and tells you that the next time we go Alexa, it's hard to argue.
So once again, it's not just about the numbers.

 

[luigivaltulini]

I'm actually with you on this Phil, the numbers are fairly meaningless to me.
As I mentioned before though I don't think people rent the Alexa for the numbers either.
Unfortunately, there is a perception in the industry that RED cameras are just too finicky and a bit fragile.
They want something simple that just works out of the box and is tough and I have to say that I have a tendency to agree for some jobs.
My own experience recently with 6 x 8K's in a tight closed lot, was that the Epic we were working on just quit because of the heat.
Took 20 min. for it to come back up again. Thanks god we had a second body.
When the producer/client walks over and tells you that the next time we go Alexa, it's hard to argue.
So once again, it's not just about the numbers.

same symptoms ..
Same as when we used the Alexa on a set ... I want it on a gimbal!
oppss .. the producer wanted the Epic ..
And 'it is not a question of numbers but of what you can use at the appropriate time.
For now, Red is the one that offers the most possibilities in comparison to Alexa.

 

[Brian Merlen]

i hear stories like this and think i must have the best luck ever...in like 400 or 500 red shoots i've only had an error message for service once and the camera kept shooting fine after the message anyway...so i just sent it in when i got home... so it always amazes me to hear these stories about problems when i just don't see it happening on the sets I'm on...sheer luck i suppose?

 

[Brian Boyer]

Sorry guys, the reality is that your lovely Dragon sensor yields around the same, perhaps only a smidgen more DR than the Alexa.
DXO Mark gives it 14.8 and everything that I have seen confirms that.
Not too shabby (DXO considers it the best sensor they have ever tested), but going from 0-20, 18+, 16.5, 16 to 14.8...all these figures bandied around, but generally trending downwards over time..it's funny to watch from the sidelines. (The advantage of renting rather than owning is that one remains uninvested in these kind of tests).
In any case, ain't no way it's 16.5.

By the way, I think that the reason people choose the Alexa had nothing to do with resolution or dynamic range.
It's for the roll-off in the highlights, a curve which has been tweaked to look like film, for the industry standard connectivity, the simplicity of use and the durability of the body.

What's funny is how simply correcting a mistake somehow implies being an apologist/fanboy/lovestruck. I wish I had a dime for every time that happened. It just seems like easy money and I could use the cash.

"...and everything I've seen confirms that" sounds a lot like anecdotal evidence to me. At least the folks standing by the 16+ stop range did their own tests and posted the results.

Alas, you're right in stating the numbers alone don't tell you what the image looks like...but they can be useful in determining why the image looks the way it does.

And even Arri disagrees with you on the subject of dynamic range, so much so they published a white paper in which they stated:

"Of course, the objective evaluation of camera systems isn’t just about contrast and sensitivity. Physiologically however, as hopefully conclusively proven in the preceding 4K+ series of articles, dynamic range is the decisive criterion."

Even if some tests suggest Alexa has a little more total DR than Dragon, it shouldn't come as a shock given Alexa has fewer and bigger photosites and its dual readout technique.

However, when taking into account Dragon has considerably more and smaller photosites and has roughly the same practical DR (give or take a little depending on who you ask), and that range hits a sweet spot on the top end that puts both cams in film territory, in addition to the color handling, I'd say that's a win for whoever uses it.

 

[Robert Ruffo New]

What's funny is how simply correcting a mistake somehow implies being an apologist/fanboy/lovestruck. I wish I had a dime for every time that happened. It just seems like easy money and I could use the cash.

"...and everything I've seen confirms that" sounds a lot like anecdotal evidence to me. At least the folks standing by the 16+ stop range did their own tests and posted the results.

Alas, you're right in stating the numbers alone don't tell you what the image looks like...but they can be useful in determining why the image looks the way it does.

And even Arri disagrees with you on the subject of dynamic range, so much so they published a white paper in which they stated:

"Of course, the objective evaluation of camera systems isn’t just about contrast and sensitivity. Physiologically however, as hopefully conclusively proven in the preceding 4K+ series of articles, dynamic range is the decisive criterion."

Even if some tests suggest Alexa has a little more total DR than Dragon, it shouldn't come as a shock given Alexa has fewer and bigger photosites and its dual readout technique.

However, when taking into account Dragon has considerably more and smaller photosites and has roughly the same practical DR (give or take a little depending on who you ask), and that range hits a sweet spot on the top end that puts both cams in film territory, in addition to the color handling, I'd say that's a win for whoever uses it.

This is an urban myth.

Photosite size is not the issue - it doesn't really work that way. Yes, 1000 large photosites will give you less noise than 1000 small ones, but 100 000 small ones might actually do better than 1000 big ones. What matters is total light covering area vs non-photosite (dead) area on the sensor, and sensor size, whether that total light gathering area is made up of many pixels or fewer doesn't necessarily affect the noise floor on way or another.

As to reliability our Epic has had fewer problems than our Alexa and has never failed, whereas the Alexa has shut down due to heat twice. Not sure about Epic not being simple - I find it very easy to use - as easy as any camera (easier because it's RAW so many settings don't really matter).

 

[Tobias Buehler]

visible dynamic range

visible dynamic range

Everybody seems to be about or how many stops the sensor captures. I think it already captures what people wish for. even on mx. imo it's rather a question of software to squeeze that extra bit out of the footage and make it visible in one frame. I haven't found any grading software that can pull back the information in a blown out looking sky or window from a r3d (which is clearly there in redcine when you lower the iso) without having two different iso renders and then lumakey/powerwindow. Why is there nothing working like a "shadow/highlight" filter in redcine and why does the shadow/highlight in Premiere/After effects for example not see the extra info in the files? Okay, it's arguable that there will be horrible looking hdr videos like the ones we all know from stills photography

 

[Tommaso Alvisi]

Tobias,

if you are familiar with PS controls and prefer to manipulate your footage that way...

you can always render 16bit TIFF sequences from RCX PRO and then import these in Photoshop...

Now it supports video too! (you have to be patient tho...he he)

 

[Matt Ryan]

Everybody seems to be about or how many stops the sensor captures. I think it already captures what people wish for. even on mx. imo it's rather a question of software to squeeze that extra bit out of the footage and make it visible in one frame. I haven't found any grading software that can pull back the information in a blown out looking sky or window from a r3d (which is clearly there in redcine when you lower the iso) without having two different iso renders and then lumakey/powerwindow. Why is there nothing working like a "shadow/highlight" filter in redcine and why does the shadow/highlight in Premiere/After effects for example not see the extra info in the files? Okay, it's arguable that there will be horrible looking hdr videos like the ones we all know from stills photography

I agree, we don't have the ability to capture all those stops in a single image at the moment. I mean noise free on the low end as well. It would be nice to have an option to combine multiple ISO's to create a 16 stop image, and I'm not talking about HDRX. In RCX or any program if you lower ISO you can see a couple stops worth of info come back, same for lifting the iso and seeing a stop or so in the low end come back. It would be nice to have all that range from the start. Then let me crush, push, pull, or add a shit ton of contrast to... Just my opinion.

 

[Robert Ruffo New]

I'm actually with you on this Phil, the numbers are fairly meaningless to me.
As I mentioned before though I don't think people rent the Alexa for the numbers either.
Unfortunately, there is a perception in the industry that RED cameras are just too finicky and a bit fragile.
They want something simple that just works out of the box and is tough and I have to say that I have a tendency to agree for some jobs.
My own experience recently with 6 x 8K's in a tight closed lot, was that the Epic we were working on just quit because of the heat.
Took 20 min. for it to come back up again. Thanks god we had a second body.
When the producer/client walks over and tells you that the next time we go Alexa, it's hard to argue.
So once again, it's not just about the numbers.

Again, people have had experiences like with Alexa too, and we have never had a problem with any Red, over 100s of shooting days. We have never slammed our camera against a wall, but I would not do that with an Alexa either. thsi "built solid" stuff is just a lot of macho nonsense. Any digital camera is a delicate instrument, like it or not. It's not a hammer or a power drill, even if it says Arri on the side.

I really don;t see what could be simpler than shooting with a camera where the only settings you need to worry about are shutter, fps and compression, which are right there in the touch LCD, and everything else is just metadata. How technologically challenged are some ops/DPs/ACs!? It's not rocket science to use an Epic menu system, the rest of cinematography is equal in difficulty

 

[David Litchfield]

Sorry guys, the reality is that your lovely Dragon sensor yields around the same, perhaps only a smidgen more DR than the Alexa.
DXO Mark gives it 14.8 and everything that I have seen confirms that.
Not too shabby (DXO considers it the best sensor they have ever tested), but going from 0-20, 18+, 16.5, 16 to 14.8...all these figures bandied around, but generally trending downwards over time..it's funny to watch from the sidelines. (The advantage of renting rather than owning is that one remains uninvested in these kind of tests).
In any case, ain't no way it's 16.5.

By the way, I think that the reason people choose the Alexa had nothing to do with resolution or dynamic range.
It's for the roll-off in the highlights, a curve which has been tweaked to look like film, for the industry standard connectivity, the simplicity of use and the durability of the body.

And I think a lot of people just trust Arri more.

Let’s see if we can clear up some Dragon misconceptions.

1. The native Dragon sensor clearly shows 18+ stops shot on our Frankie test bed last December which was posted. Obviously, we were excited out of our skull. Stuffing the sensor’s capability into an electronics package the size of a matchbox always causes loss of performance. That is why we stuck with 3-4 stops of improvement over the EPIC MX. We always measure dynamic range with the engineering definition… as do most others. We said repeatedly 16 stops for EPIC Dragon, which is more than any other digital cinema camera. We stick by this statement… even though we see continuing improvements over time, just as we have always done. For free.

I've always found this baffling. I've never understood why the size of the box effects how the sensor responds to light. But anyway, if you've got an 18+ stop sensor why put it in a matchbox? Why not put it in a box the size of 2 matchboxes? Then you could have an 18+ stop camera, with decent cooling, full size xlrs, multiple monitor outputs and multiple SDIs. You're only compromising size then.

David

 

[Robert Ruffo New]

Everybody seems to be about or how many stops the sensor captures. I think it already captures what people wish for. even on mx. imo it's rather a question of software to squeeze that extra bit out of the footage and make it visible in one frame. I haven't found any grading software that can pull back the information in a blown out looking sky or window from a r3d (which is clearly there in redcine when you lower the iso) without having two different iso renders and then lumakey/powerwindow. Why is there nothing working like a "shadow/highlight" filter in redcine and why does the shadow/highlight in Premiere/After effects for example not see the extra info in the files? Okay, it's arguable that there will be horrible looking hdr videos like the ones we all know from stills photography

Mystika does this. R3Ds stay in 32 bit float even after debayer.

 

[Nick Morrison]

And I think a lot of people just trust Arri more.

I also find it interesting that a growing number of ICONIC Director's seem to be trusting/embracing RED more. The list is impressive:

1) James Cameron (Avatar)
2) Ridley Scott (Prometheous, Moses)
3) Michael Bay (Pain and Gain, Transformers 4)
4) Steven Soderbergh (All his recent movies)
5) David Fincher (House of Cards, Dragon Tatoo, Gone Girl)
6) Peter Jackson (The Hobbit)

I'm sure I'm forgetting some. By any stretch of the imagination, these are some of the most influential directors in Hollywood. RED must be doing something right to have earned their faith...

I've always found this baffling. I've never understood why the size of the box effects how the sensor responds to light. But anyway, if you've got an 18+ stop sensor why put it in a matchbox? Why not put it in a box the size of 2 matchboxes? Then you could have an 18+ stop camera, with decent cooling, full size xlrs, multiple monitor outputs and multiple SDIs. You're only compromising size then.

David

Hollywood's embrace of the EPIC for 3D cinema should answer your question about size. Clearly, RED saw an advantage and an appeal in a smaller form factor. Being able to place an Epic on a gimbal is another major advantage. A heavier camera could not do that.

RED is clearly committed to a smaller form factor.

That said, the community has been clamoring for a larger body. Whether that happens (or not) has yet to be seen...