PDA

View Full Version : Dragon Monochrome dynamic range.



John Marchant
08-12-2014, 12:44 PM
Giving this its own thread.

I shot a Xyla-21 chart with our Dragon Monochrome. The best way to communicate what I see is as a prores file to download. H264 kills the bottom of the image too much.

Here's a 250mb 720p prores, at ISO2000, the camera default.

Download Dragon Monochrome Xyla 720p (http://kippertie.com/red/dragon/mono%20dragon%20xyla%20720p.mov)

Notes:

The xyla chart has a sliding mask to block the stops. I exposed so that the top chip is clipped, and so that the second chip is a full stop below clip.

The best way in my opinion to judge which shadow stop is the last is by watching the mask slide back and forth. The point where you can no longer discern a chip coming and going is the last one. Depending on your monitor calibration you may see more or less of course.

The chart is intentionally slightly soft - it has a textured surface which resembles fixed pattern noise in the chips; shooting soft prevents any confusion.

How many stops do you guys see? ;)

Phil Holland
08-12-2014, 12:59 PM
Downloading now John!


Here's a 250mb 720p prores, at ISO2000, the camera default.

However, it might be a good idea to explain why "Chroma Dragon" is rated at ISO 800 and Monochrome at ISO 2000. I can do it or you can do it :)

Also, just finished the download and watched it. I have only had a brief moment with Monochrome Dragon to date and know a tiny bit about what it can see, however, I would personally like to "see" what ISO 4000 sees as well in terms of tonal range.

Thank you!

John Marchant
08-12-2014, 01:00 PM
I'm eating, you do the 800/2000 thing. I will award marks for clarity and Holland-diagrams.

Martin Stevens
08-12-2014, 01:02 PM
I'm eating, you do the 800/2000 thing. I will award marks for clarity and Holland-diagrams.

Gotta love the love on reduser. Thanks John and Phil.

Phil Holland
08-12-2014, 01:05 PM
I'm eating, you do the 800/2000 thing. I will award marks for clarity and Holland-diagrams.

I'm going to eat too! However, I don't have a Monochrome and can't generate patches. Tummy rumbles.

David Kuhn
08-12-2014, 01:09 PM
As mentioned on the other thread... I saw the chip between 16 and 17 clearly, but couldn't call the 17-18 chip (I thought I saw something but it was just a bit too dark). I tried it on 2 monitors... a 4k Asus and HP Dreamcolor (I also tried it with the dreamcolor brightness maxxed... same outcome).

The Epic Mono was rated at ISO2000... if the Dragon Mono is also rated at ISO2000 what exactly does that mean? My interpretation was that the 'rated' ISO is the ISO at which dynamic range is maximized (but that doesn't seem to be true for my Epic... which seems happier at 800).

John Marchant
08-12-2014, 01:31 PM
A little refresher on Monochrome -

The Monochrome (both Dragon and MX) is a 'naked' version of the sensor. No bayer pattern colour filter array is present - as a result the chip is more sensitive to light. Changing the default ISO makes sense in this context.

The way that the system as a whole responds is also very much helped by the adaptation to monochrome. The grain pattern is more filmic than digital, and of course the issues of colour noise, highlight discolouration etc are all non-existant. Sharpness is excellent, of course - the OLPF needs to be less aggressive and each pixel is truly giving straightforward information. No debayering guesswork (sorry guys, but you know what I mean ;) )

Red have chosen 2000 as the 'default' ISO for the Dragon Mono - I think its a more conservative choice than it was with the MX Mono. Its just a starting point however, as always - it gives a good balance above and below with acceptable noise. I always tended to knock the MX mono back down to 800 and regretted that I couldn't go lower.

On Dragon Mono, anecdotally speaking, I'm happier to shoot at 2000 if it doesn't create issues of its own, but very welcome is the typical Dragon trait of a good image even at 250 ISO.

Shooting Dragon Monochrome is a joy, but does require you to selectively forget and remember certain parts of the Red 'method'. You need to use colour filtration - there's no shifting tones in post as you would when desaturating a colour image. Overexposure is hugely well catered for, and looks great. The rest of the camera is more familiar...

People ask why you'd shoot with it - you can after all desaturate a colour image to good effect. Bear in mind though - if I want to create a dark brooding sky in B&W, I need to hit the image with a heavy red or orange filter. I can do that in front of the lens, or in post with a colour camera by playing with my channels. However when I do that to a bayer equipped colour camera, whether physically or in post, only one in four pixels is doing any work. On monochrome, every pixel counts, whatever you do.

scott devitte
08-12-2014, 01:49 PM
Has anybody with Red Monochrome experience seen the film "Ida"?

It was shot in color in 1.33 on Alexa 4:3 with Arri raw and then to black and white on a Nucoda.

It is a wonderful piece of work.

I am wondering if anybody could compare/contrast it to what the Monochrome epic or dragon produces.

Les Dittert
08-12-2014, 02:56 PM
Dynamic range will be the same at the color version. Same sensor but without filters. Same electronics.

Think of it this way : Does putting a bunch of ND on your camera change it's dynamic range ? Nope. It just shifts the range.

John Marchant
08-12-2014, 03:29 PM
However, it clearly has greater DR. Removing the CFA improves basic sensitivity; ISO 800 on mono has equal illumination to ISO 800 in colour, so we can infer that less gain is required at the sensor A/D stage - hence better s/n ratio for a given ISO.

Technical speculation aside, there's no way I can discern 18 stops on colour Dragon in the exact same scenario - mono is more than a full stop better, maybe two.

Antony Newman
08-12-2014, 06:44 PM
John,

Unless there is optical compression or rarefaction in the light patch - I believe the DR will be identical (as Les said)

The S/N of the monochrome would be superior.

Perhaps we should start specifying the best S/N level at various DR?

Would be a great metric to compare to other camera's - if they didn't have an Non-defeatable NR.

AJ

Edit 1 : If CFA has slightly different absorption levels ... the DR would reduce by in proportion to the max loss/min loss. Maybe RED also chuck some of the super bright where colour becomes non linear, and where the low end luma mush makes R3D encoding inefficient.

Edit 2 : So the sensors have the same DR .. but the Dragon Monochrome is more likely to encode a wider DR.

Marc Wielage
08-12-2014, 09:03 PM
People ask why you'd shoot with it - you can after all desaturate a colour image to good effect. Bear in mind though - if I want to create a dark brooding sky in B&W, I need to hit the image with a heavy red or orange filter. I can do that in front of the lens, or in post with a colour camera by playing with my channels. However when I do that to a bayer equipped colour camera, whether physically or in post, only one in four pixels is doing any work. On monochrome, every pixel counts, whatever you do.
I agree with all your sentiments, John, but note that in many modern color correction systems, you have the ability to use RGB color mixer controls to generate B&W images based on a different initial color balance. This eliminates the need to use any kind of color filter in original photography. No question, as always, anybody trying this needs to shoot tests first to understand the limits of exposure and available grayscale.

Les Dittert
08-12-2014, 09:22 PM
However, it clearly has greater DR. Removing the CFA improves basic sensitivity; ISO 800 on mono has equal illumination to ISO 800 in colour, so we can infer that less gain is required at the sensor A/D stage - hence better s/n ratio for a given ISO.

Technical speculation aside, there's no way I can discern 18 stops on colour Dragon in the exact same scenario - mono is more than a full stop better, maybe two.

Yes John,
As you said, for a given ISO. The loss of sensitivity on the color camera makes it harder to exploit the dynamic range because you have more noise. Pump enough light into the color camera and the DR should be the same, unless the compression is degrading it on the color camera.

John Marchant
08-12-2014, 10:49 PM
I agree with all your sentiments, John, but note that in many modern color correction systems, you have the ability to use RGB color mixer controls to generate B&W images based on a different initial color balance. This eliminates the need to use any kind of color filter in original photography. No question, as always, anybody trying this needs to shoot tests first to understand the limits of exposure and available grayscale.

Absolutely true that shooting colour for post conversion gives flexibility, but the payoff for committing to monochrome at acquisition is substantial. I've shot considerable amounts of landscape b&w material on colour epics (even with custom or deleted OLPF) and it still looks mushy and lacks tonal flexibility in post compared to the Mono camera. You have to be bold to shoot with it, and understand what you're doing, but it gives back in proportion :)

Jarred Land
08-12-2014, 11:37 PM
The Monochrome doesn't have color filters that inherently by nature absorbs light, so the monochrome is much more sensitive than the color version of the Dragon Sensor. The transition and gradations also cannot be matched by a color sensor, so the steps between that enormous latitude on the Monochrome Dragon is much smoother.

Plus.. the Sharpness is significantly greater on the monochrome Dragon as you don't loose resolution through the process. I believe John has shot those tests already.

Karim D. Ghantous
08-13-2014, 03:55 PM
The Leica M Monochrom (18Mpx) was compared to desaturated files from the D800E (36Mpx). The resolving power of both cameras was the same, despite the D800E having 50% more resolution. The Dragon Monochrome, therefore, should have the equivalent resolving power of a 9K sensor with a Bayer filter.

Some photographers don't understand why you would want a monochrome sensor. There are very good technical reasons why, which I do not have to repeat here.

David Kuhn
08-13-2014, 06:04 PM
The Leica M Monochrom (18Mpx) was compared to desaturated files from the D800E (36Mpx). The resolving power of both cameras was the same, despite the D800E having 50% more resolution. The Dragon Monochrome, therefore, should have the equivalent resolving power of a 9K sensor with a Bayer filter.

Some photographers don't understand why you would want a monochrome sensor. There are very good technical reasons why, which I do not have to repeat here.

Yes unless you go 3ccd you give up a lot of resolution. The D800 is a great camera (D810 even more so, I'll probably upgrade next year), but I'd LOVE to own a Leica M Monochrom. The Nikon is what I choose for utility... but the Leica would be for it's single minded excellence.

Gavin Greenwalt
08-16-2014, 09:36 AM
Dynamic range will be the same at the color version. Same sensor but without filters. Same electronics.

Think of it this way : Does putting a bunch of ND on your camera change it's dynamic range ? Nope. It just shifts the range.

If we were talking about ND you would be correct but remember it's an RGB filter + OLPF. As a result the color science is going to by necessity apply separate gains to different pixels. The extreme example would be white balancing a nearly purely blue light. In order to expose the blue properly your red and green channels would be deep into the noise floor. So you would have to overexpose and clip the blue in order to pull the red and green channels up out of the noisy dirt. The end result would be that much of the red/green would be crushed and much of the blue would be clipped and you would have a white balanced image with a very narrow dynamic range.

With monochrome there is no such thing as white balance so you don't have varying dynamic range from pixel to pixel producing a intersecting range of dynamic range. Even at 5600 as evidenced by the "red speckles" Dragon is applying red channel gain to compensate for the OLPF. That means if you want a clean image you have to crush out perfectly usable blue and green channel shadows to match the noise floor of the red.

http://i111.photobucket.com/albums/n134/im_thatoneguy/Ideal_zps12826780.png (http://s111.photobucket.com/user/im_thatoneguy/media/Ideal_zps12826780.png.html)
http://i111.photobucket.com/albums/n134/im_thatoneguy/RealWorld_zps6291a843.png (http://s111.photobucket.com/user/im_thatoneguy/media/RealWorld_zps6291a843.png.html)
http://i111.photobucket.com/albums/n134/im_thatoneguy/RealWorldWhiteBalanced_zpsba5603a3.png (http://s111.photobucket.com/user/im_thatoneguy/media/RealWorldWhiteBalanced_zpsba5603a3.png.html)