13 Stops of DR on an BMPCC? More like 10.

To test the over exposure I exposed a grey card to f22 and set the lens to f22, then proceeded to open up the lens. Hence 0 is f22, +1 is f16, +2 is f11 etc etc.

To test the under exposure I exposed the grey card to f2.8 matched this on the lens and then closed the lens down. Hence 0 here is f2.8, -1 is f4, -2 is f5.6 etc etc.

Even if I had a lens that was capable of f1.0 at its widest aperture and f64 at its smallest, that would only give me 12 stops to test. The lens I had to use had 7 stops from wide open to closed so I split the test into two sections.

I know that my lens has some falloff at the edges when wide open and a few stops down, but I was judging the centre of the image.

I know that my lens has some falloff at the edges when wide open and a few stops down, but I was judging the centre of the image.

Falloff is not an issue, I applied a centered power window.

But still, you have 2 wings of your characteristic curve (Under and Over), but since you don't give a joint point of grey for them, this I'm afraid is exactly as JB said, just a measurement for under and over exposer, not a DR measurement.

You can use either ND filter or break the curve into 3, 4 parts, but the values of the frames should be consequential.

Hope this makes sense.

... but since you don't give a joint point of grey for them, this I'm afraid is exactly as JB said, just a measurement for under and over exposer, not a DR measurement.

Call me an idiot, but this made no sense to me. Can you explain what you mean?

... but since you don't give a joint point of grey for them, this I'm afraid is exactly as JB said, just a measurement for under and over exposer, not a DR measurement.

Call me an idiot, but this made no sense to me. Can you explain what you mean?

Sorry, but I might be as wrong, so let's try to find some common ground:

Imagine you have a perfect lens that may change the aperture by 13-14 stops. You will be able to draw a characteristic curve of the camera's sensor. Sorry if this is totally off-topic here, but have a look at the curve I drew based on my measurements for Canon 50D http://www.magiclantern.fm/forum/index.php?topic=5586.msg95991#msg95991.

I did this in 2 goes: the OVER wing starting from grey point and the UNDER wing starting from the SAME GREY POINT.

If you do the OVER part, you may have the aperture at f/22 and then increase the exposure.
If you do the UNDER part, you may have the lens wide open and then decrease the value.

BUT the reading of the exposure level at the grey point should be the same (in my 50D case the reading was approx. 240).

If you do it this way, you may not even care about IRE -- just do it step by step from the grey point.

HINT: if you do all this stuff at angle 180 degrees and have your lens closed at f/22 as the max closed, start decreasing the shutter angle. Rememeber? You can substitute speed for aperture and vice versa as long as you do it by proper units. 180 -> 90 -> 45 degrees is the same as closing the aperture by 2 stops.

Also, don't look at the f-stops in my graph - they are symbolic of the exposure. My real f-stop was different depending on the lighting. But I was dosing the light STEP BY STEP, judging by the lighting reading within the camera and Resolve.

Please tell me this makes sense ))))

Ok, so my friend and I did a quick test for BMCC DR, and it pretty much reflects what we are told by BMD.
Kudos!

definitely getting less usable DR from the BMPCC than from my NEX 5R (stills). Even then, amazing to have what it has available in video in a small package.

Deals pretty well with high contrast scenarios, here's something I shot last weekend

You all might learn something from this blog post: http://www.ryanewalters.com/Blog/blog.p ... 5844575706

To summary of this person's method yields "Usable DR = 9 5/6" for the BMCC. Which is the same as the Red Epic-MX.

Before you get too excited, the DRAGON measures more than 16 stops (and Alexa more than 13). But regardless of which kit you are rooting for, this test method is very easy to understand and replicate...if you have an accurate light meter.

It also validates the idea of using a 12.5% gray card instead of an 18% gray card.

Deals pretty well with high contrast scenarios, here's something I shot last weekend

I can see you've pulled back the highlights a little on the sun, but no suprise there.
Also, a little off topic, but I'm curious which lens you used, as it's obviously image stabilised and doing quite a good job too...

Cheers. Yeah the sun is where the DR breaks down a bit

6mm 1.8 Kowa machine vision lens, and very gentle warp stabilizer in AE. the lens is so wide that it doesn't need much

Not heard of the 6.5mm... I know the 6mm 1" c-mount kowa lenses cover. Is this a M4/3 version or something?
http://www.rmaelectronics.com/kowa-lm6h ... xel-rated/

sorry - 6mm is the one. it's c-mount, meant for 1" so covers the sensor well. No coatings so flares a lot, that might be a bad thing for some people but I like it

Thanks for the Ryan Walters link to the overexposure tests. I also watched the under exposure tests. Regardless of the measurable hard numbers, just looking at the images confirms for us that the BMCC does perform well and to my eyes seemed better on overexposure than the Epic MX.

Interestingly, it did well on the look of the underexposed image and seemed as pleasant as the Alexa. Granted the hard dynamic range is less, but staying within the usable range provides a great image when underexposed.


Rick Lang
Sent from my iPad using Tapatalk HD

So despite my fairly inaccurate tests, all the other tests of useable DR seem to come out fairly similar...

13 stops when shooting raw at 800asa and exposing to the right.


Sent from my iPhone using Tapatalk

Thanks for the Ryan Walters link to the overexposure tests. I also watched the under exposure tests. Regardless of the measurable hard numbers, just looking at the images confirms for us that the BMCC does perform well and to my eyes seemed better on overexposure than the Epic MX.

Interestingly, it did well on the look of the underexposed image and seemed as pleasant as the Alexa. Granted the hard dynamic range is less, but staying within the usable range provides a great image when underexposed.


Rick Lang
Sent from my iPad using Tapatalk HD

I watched Ryan Walter's test, and the fact that he finds Epic MX to technically have slightly more overhead latitude than the BMCC, but the blowouts on the woman's skin look worse than they do on the BMCC footage, leads me to believe that maybe one of the Epic's color channels is clipping before the others (maybe the red channel?) and at a lower level than the BMCC's equivalent channel.

I think I remember someone on the forums in the past (John Brawley, maybe) mentioning that this sort of thing happens with the BMCC as well, though no one seems to have a definitive numbers on this. Another reason I ETTR with 95% Zebras instead of 100%.

You think as high as 95% is safe? Some people I know recommend shooting with it set at 85%.
But then if I did that chances are i'd probably over expose it a bit anyway as I knew there was more there...

Ignore my question.

You think as high as 95% is safe? Some people I know recommend shooting with it set at 85%.
But then if I did that chances are i'd probably over expose it a bit anyway as I knew there was more there...

Ignore my question.

Matt, I don't have the camera but I'm one of those that feel 85% might be a good idea in some situations to help manage the highlight roll off. If I ever do get a camera in my hands, I'd also test a lot of shooting at 200 and 400 ISO versus the 800. 1600 would only be for true low light situations. Although 13 stops of dynamic range is part of the appeal, you and Ryan Walters may have the right approach: it's the usable dynamic range that matters not the measurable dynamic range. I'm confident there really are close to 13 stops at ISO 800, as measured carefully, but when you trust your eyes to tell you what is useful, the DR is a likely at least a few stops less for all cameras.

Ryan's over and under exposure tests, although probably dated as they don't include the Epic Dragon, are an eye- opener for me. I don't care what the specs say about the Epic MX versus the BMCC, the BMCC performs as well in those tests. Would be great if he updates the videos when he can add the Epic Dragon as that does sound like it has the broadest dynamic range on paper. Very much hope he'll include the BMPC4K too.

Rick Lang
Sent from my iPad using Tapatalk HD

I'd stick with 95/100% on the zebras. There are different types of lights to consider. For specular highlights, where light bounces off like a mirror, that's okay to clip surely, why you need detail in there?

Then with ETTR you can capture the most relevant range and manage noise.

Jules

Ryan's over and under exposure tests, although probably dated as they don't include the Epic Dragon, are an eye- opener for me.

Out of respect for the rules of this forum, I didn't copy Ryan's post over on the RED forum about DRAGON's DR. But using the same measurements that concluded DRAGON has nearly 7 stops more DR than either BMCC or EPIC. FWIW.

According to the specs sheet on the Fairchild Imaging CIS1910F (the sensor in the Pocket or the closest to it) dynamic range is 88dB. That works out to 14.62 stops (88/6.02 - see the Imatest q13 doc for more info on the math).

The BMCC's sensor on the other hand (Fairchild Imaging CIS2521F) has a specification of 86dB or 14.29 stops (or about a third less than the Pocket.)

While these numbers seem high, they seem to be consistent with the low light tests out there, like Tom's. As well, when you use temporal noise reduction (like Neat Video) you can really reach into the noisy bits and extract usable images, which may not be so apparent in unprocessed stills.

According to the specs sheet on the Fairchild Imaging CIS1910F (the sensor in the Pocket or the closest to it) dynamic range is 88dB. That works out to 14.62 stops (88/6.02 - see the Imatest q13 doc for more info on the math).

The BMCC's sensor on the other hand (Fairchild Imaging CIS2521F) has a specification of 86dB or 14.29 stops (or about a third less than the Pocket.)

While these numbers seem high, they seem to be consistent with the low light tests out there, like Tom's. As well, when you use temporal noise reduction (like Neat Video) you can really reach into the noisy bits and extract usable images, which may not be so apparent in unprocessed stills.

Exactly, I have found the usability of those lower levels to go through the roof with even a modest bit of NR. Even if you push to ASA 2000 - with a tiny bit of NR you can get a clean and usable image.

Shot at 800 with no clipping:



Pushed and graded to 2000 with a bit of NR

Exactly, I have found the usability of those lower levels to go through the roof with even a modest bit of NR. Even if you push to ASA 2000 - with a tiny bit of NR you can get a clean and usable image.

Shot at 800 with no clipping:

<>

Pushed and graded to 2000 with a bit of NR
<>

Hi Tom, I have asked in another thread, but perhaps you didnt see it. Are you grading on a calibrated rec709 monitor?

Hi Tom, I have asked in another thread, but perhaps you didnt see it. Are you grading on a calibrated rec709 monitor?

Calibrated, but it is not a proper colour grading monitor.

My Funds are very very low these days.

… My Funds are very very low these days.

No shame in that. Thanks for all the tech info!

OT: Hopefully the winds aren't blowing from the NW to where you are. Be well.
http://www.bbc.co.uk/news/uk-england-cumbria-25975785

-

Anyone know what usable DR for film negative? I thought it was 9 or 10 stops?

Jules

Depends on how you want to define "usable." that's a personal preference-- especially in terms of how much grain you're going to accept and also how you're going to scan the thing in the first place (you'd get a lot more DR off of a 3 pass Arriscan than a Spirit 2K for example).

Also, while the film is the same, you'd get more Dynamic Range off of 65mm than one would off of 16mm since the grain is smaller-- even though it's the exact same emulsion.

Also, while the film is the same, you'd get more Dynamic Range off of 65mm than one would off of 16mm since the grain is smaller-- even though it's the exact same emulsion.

This is another arguement of why the BMPCC has a lower DR than the BMCC. At least in ProRes, as the BMCC does not window it's sensor so I understand. Correct me if I'm wrong...

… My Funds are very very low these days.

No shame in that. Thanks for all the tech info!

OT: Hopefully the winds aren't blowing from the NW to where you are. Be well.
http://www.bbc.co.uk/news/uk-england-cumbria-25975785

-

Cheers Peter, the weather here has been quite windy and rainy for a few weeks, but fortunately I live on top of a hill so I am immune to flooding. But thank you for the concern

According to the specs sheet on the Fairchild Imaging CIS1910F (the sensor in the Pocket or the closest to it) dynamic range is 88dB. That works out to 14.62 stops (88/6.02 - see the Imatest q13 doc for more info on the math).

The BMCC's sensor on the other hand (Fairchild Imaging CIS2521F) has a specification of 86dB or 14.29 stops (or about a third less than the Pocket.)

While these numbers seem high, they seem to be consistent with the low light tests out there, like Tom's. As well, when you use temporal noise reduction (like Neat Video) you can really reach into the noisy bits and extract usable images, which may not be so apparent in unprocessed stills.

Exactly, I have found the usability of those lower levels to go through the roof with even a modest bit of NR. Even if you push to ASA 2000 - with a tiny bit of NR you can get a clean and usable image.

Shot at 800 with no clipping:



Pushed and graded to 2000 with a bit of NR

Spot on. +1

I'm a Neat Video groupie.

How are you applying your noise reduction?

How are you applying your noise reduction?

In resolve, I usually apply it to my last node - using the noise reduction tab. Available in Resolve Full version only - not the free one.

I usually apply more Chroma NR than Luma.

Also, while the film is the same, you'd get more Dynamic Range off of 65mm than one would off of 16mm since the grain is smaller-- even though it's the exact same emulsion.

This is another arguement of why the BMPCC has a lower DR than the BMCC. At least in ProRes, as the BMCC does not window it's sensor so I understand. Correct me if I'm wrong...

If there were any difference at all it would be negligible, maybe 1/4 a stop or so. I haven't noticed any appreciable difference in DR on the pocket or the bmcc. If anything-- and this is non scientific since you can't easily match lenses in real world scenarios-- but when we have used the pocket as a b cam on the two features we shot on the cinema, the pocket seemed to have more usable information on the day. But that's all anecdotal.

Hi Tom, I have asked in another thread, but perhaps you didnt see it. Are you grading on a calibrated rec709 monitor?

Calibrated...

Which device (hardware) did you use to achieve that?

How are you applying your noise reduction?

In resolve, I usually apply it to my last node - using the noise reduction tab. Available in Resolve Full version only - not the free one.

I usually apply more Chroma NR than Luma.

Thanks Tom.

I usually use neat video but i will try Resolve´s.