Extreme Purple Fringing on Dragon

[Michael Tiemann]

I'm going to make another call for an R3D file posted to this thread. It's important to see the raw data to be able to give proper advice.
Also, no mention of which OLPF was used.

ONe thing is for sure. When you clip all three channels on Dragon or MX you are going to be looking at these kinds of issues. PLus this was a controlled setting where stuff like that could be avoided. I'm not entirely sure you can blame the camera for this.

David

David, did you see the R3D I posted? I think it illustrates well the sensitivity of the problem with DC2 vs. original DragonColor.

 

[David Battistella]

David, did you see the R3D I posted? I think it illustrates well the sensitivity of the problem with DC2 vs. original DragonColor.

Yes, I did, thank you Michael. I have seen this CA issue myself and it generally happens when the sensor is clipped or on clipped points of light etc. I have some fast but older Nikon Glass that can display these issues on DC2 that is tamed when switching to DC.

It's just my personal belief that when you have these kinds of issues an R3D can be posted to fully understand how the file has been processed.

I have not been following this thread to closely, but when I did figure out that this happened in a "controlled setting" and not say an "as it happens wedding video" then it really becomes a problem of the production more than then camera.

Shots can be cheated to attain the same effect. So this error is more on the production than the camera or color science.

David

 

[Andy White]

Hey Andy, I did something like that. Check it out: http://www.reduser.net/forum/showth...SE-on-Dragon-PHOTOS&highlight=purple+fringing

Hi Kenneth - I'd already seen those (back when originally posted)

Really I'd just like the OP to grab the other camera and lens and do a quick test in the same conditions... as the fringing is way bad on the original 2nd Operator camera.

 

[Andy White]

Yes, I did, thank you Michael. I have seen this CA issue myself and it generally happens when the sensor is clipped or on clipped points of light etc. I have some fast but older Nikon Glass that can display these issues on DC2 that is tamed when switching to DC.

It's just my personal belief that when you have these kinds of issues an R3D can be posted to fully understand how the file has been processed.

I have not been following this thread to closely, but when I did figure out that this happened in a "controlled setting" and not say an "as it happens wedding video" then it really becomes a problem of the production more than then camera.

Shots can be cheated to attain the same effect. So this error is more on the production than the camera or color science.

David

Just an FYI http://www.reduser.net/forum/showth...ng-on-Dragon&p=1645469&viewfull=1#post1645469 - here are details of the setting used on the Dragon.

 

[David Battistella]

Just an FYI http://www.reduser.net/forum/showth...ng-on-Dragon&p=1645469&viewfull=1#post1645469 - here are details of the setting used on the Dragon.

The only other thing I will say is that a common mistake for people to treat the dragon as the same camera as the MX. They are totally different beasts and need to be treated as two different cameras because, while they do share the same body, they do not share the same sensor. When I got my dragon way back I ran it through a whole new set of tests and adjusted how I work accordingly. There was a big big difference between the camera's.

Even though I had the same physical body in my hands it was pretty clear it was not the same old camera. For example, in a situation like the shot above I would not have exposed it that way, knowing what could and could not be recovered.



David

 

[Andy White]

The only other thing I will say is that a common mistake for people to treat the dragon as the same camera as the MX. They are totally different beasts and need to be treated as two different cameras because, while they do share the same body, they do not share the same sensor. When I got my dragon way back I ran it through a whole new set of tests and adjusted how I work accordingly. There was a big big difference between the camera's.

Even though I had the same physical body in my hands it was pretty clear it was not the same old camera. For example, in a situation like the shot above I would not have exposed it that way, knowing what could and could not be recovered.



David

Yes - that's pretty much my thoughts too.

The original exposure/settings of the Dragon have dragged it down a certain path. As you say, MX and Dragon are not the same.

 

[Hugh Scully]

The Dragon sensor has more Total Captured Dynamic Range than Mysterium-X and due to this it can potentially "see more".

In terms of wrangling chromatic aberrations in high contrast shooting situations you ideally would stop the lens down 2-3 stops and you will see a noticeable improvement. If you are shooting at f/1.4 in this particular scene you will 100% have CA if that overexposed/hot window is behind a darker object.

DRAGONcolor will help minimize this, but the overexposure is the general spawning point of this chaos.

I agree. I think it's incumbent on the cinematographer to manage highlights in this situation. Until shooting on a RED it felt like that's all I used to do! ND the windows, stop down to avoid CA, don't blow out the sensor, apply appropriate levels of fill. If you need blooming highlights to achieve your look, try a diffusion filter. Just because the Dragon sensor lets you shoot blown highlights without horrible artifacts doesn't mean you can get away without limits. Cameras ain't eyeballs + brains. Manage your ratios. Use a spot meter or gio scope. Manage the highlights. It's part of your job.

 

[joshua csehak]

Shots can be cheated to attain the same effect. So this error is more on the production than the camera or color science.

I completely disagree. I spent $45k on my Red Epic Dragon. I want to be able to put a Master Prime on and point the thing anywhere I damn well please and be confident I'm not gonna get this kind of weirdness. That's what you expect from a cheapo FS7, not a Red.

FWIW, I've never seen anything this bad, and frankly, I'm surprised the Sigma looked that good -- I tested the 50 Art and found a good amount of CA. I'm inclined to run some quick test of my own right now.

 

[David Battistella]

I completely disagree. I spent $45k on my Red Epic Dragon. I want to be able to put a Master Prime on and point the thing anywhere I damn well please and be confident I'm not gonna get this kind of weirdness. That's what you expect from a cheapo FS7, not a Red.

FWIW, I've never seen anything this bad, and frankly, I'm surprised the Sigma looked that good -- I tested the 50 Art and found a good amount of CA. I'm inclined to run some quick test of my own right now.

Every technology has it's limits. I think that the top DOP's in the world might agree with me. When you have a shot where the limits have been reached, you can do a lot to balance it. More fill, less aperture, on and on. Film might have blown out a different way, but the sensor resolves all the good in a frame and all the bad as well.

A Red Dragon sensor does not give you the luxury of bad decisions but in my experience it has been very forgiving at time with underexposure, with over exposure I have learned to back off, not clip and recover in post.

David

 

[Kenneth C Merrill]

Every technology has it's limits. I think that the top DOP's in the world might agree with me. When you have a shot where the limits have been reached, you can do a lot to balance it. More fill, less aperture, on and on. Film might have blown out a different way, but the sensor resolves all the good in a frame and all the bad as well.

A Red Dragon sensor does not give you the luxury of bad decisions but in my experience it has been very forgiving at time with underexposure, with over exposure I have learned to back off, not clip and recover in post.

David

It just seems to me that when it comes to purple fringing, the Dragon sensor is the most limited sensor I have ever worked with. Arri, Canon, and Sony can usually shoot under the exact same conditions and not exhibit this weird artifact. Yes, technology had its limits, but Dragon is limited in ways that other systems don't seem to be. That's fine, except the price and the advertised image don't reflect this particular limitation.

If this was happening on a Blackmagic camera, I doubt people would be defending it. Again, it seems like a "simple" color science fix. Why aren't we pushing for that? Why is everyone ok with the status quo when it clearly interferes with certain styles of work? (Sometimes you WANT that heavy ratio)

 

[Kenneth C Merrill]

I agree. I think it's incumbent on the cinematographer to manage highlights in this situation. Until shooting on a RED it felt like that's all I used to do! ND the windows, stop down to avoid CA, don't blow out the sensor, apply appropriate levels of fill. If you need blooming highlights to achieve your look, try a diffusion filter. Just because the Dragon sensor lets you shoot blown highlights without horrible artifacts doesn't mean you can get away without limits. Cameras ain't eyeballs + brains. Manage your ratios. Use a spot meter or gio scope. Manage the highlights. It's part of your job.

When end you clip a sensor, it should go to white, not purple. Other cameras go to white...

 

[joshua csehak]

Every technology has it's limits. I think that the top DOP's in the world might agree with me. When you have a shot where the limits have been reached, you can do a lot to balance it. More fill, less aperture, on and on. Film might have blown out a different way, but the sensor resolves all the good in a frame and all the bad as well.

A Red Dragon sensor does not give you the luxury of bad decisions but in my experience it has been very forgiving at time with underexposure, with over exposure I have learned to back off, not clip and recover in post.

David

I totally agree with you there. But it's not about technology here, it's about apparently Red having this issue while other cameras don't. But I'm going to stop opining now, until I can reproduce the issue on my end.

Yeah, it's EXTREMELY forgiving! The amount of latitude it has is one big reason I shoot Red, with its extra expense and weight, over other cameras.

 

[John Marchant]

When end you clip a sensor, it should go to white, not purple. Other cameras go to white...

The purple in the image isn't clipped though. The white highlight adjacent is... The purple is due to the near UV and top of the violet focussing very differently to the rest of the light in the image. The greater the degree of overexposure, the greater the effect. The less apochromatic the design of lens, the greater the effect.

Purple fringes are very often easily mitigated with a haze 2A or other strong UV filter that knocks out the 390-410nm region. Our OLPFs attenuate that zone by 1/2 a stop for this reason, but we don't cut it entirely, as you occasionally want that range of wavelengths.

 

[Phil Holland]

When end you clip a sensor, it should go to white, not purple. Other cameras go to white...

This is not how this works and incorrect based on the OP's original sample images.

I've already posted in this thread, but I'll try to simplify everything.


Lenses.

Lenses are not created equal. However, one surefire way to produce lateral and axial chromatic aberrations is shooting in extreme contrast scenarios wide open. Some lenses exhibit/produce much more of this, some produce minimal levels of CA (especially super modern lenses).

In the case of the OP's example its a lens that is fairly known to exhibit CA wide open.

As you stop a lens down you will notice the chromatic aberrations begin to diminish. Typically you will see a fairly big improvement on most lenses when you stop down 2-3 stops. On a few older lenses you may want to stop down even more.

Here's an example:

RED Cameras

There's a few concepts to keep in mind. All cameras have different levels of Total Captured Dynamic Range. If you have a system that doesn't capture much DR you don't "see" as much when clipping occurs.

Dragon has a lot of Total Captured Dynamic Range. And also we have (at the moment) two versions of DRAGON color science with DRAGONcolor and DRAGONcolor2. Specifically you can "enhance" chromatic aberrations if you have them in your shot when you select DRAGONcolor2. This occurs due to specifically how the blue channel is tuned in DRAGONcolor2. You will see deeper blues through violets and in terms of where in the spectrum purple fringing may occur it will make it pop out much more. Again, that's if and when it's there. DRAGONcolor is a nice solve if you are in a problem situation.

Another note to make regarding "other cameras" is occasionally discussing some of their in camera image processing techniques. For instance some cameras have methods of applying "lens correction". Some even tackle CA in a few ways on the color processing side and occasionally create a different type of chromatic edge artifact which can look rather bad.


Shooting Method

Through traditional shooting methods there are a lot of ways to avoid chromatic aberrations. Most start with sound shooting methods and techniques. Understanding lighting ratios, the lenses your shooting, how/when to utilize lens based techniques, how/when to utilize filtration based techniques, how/when to utilize lighting techniques to tackle things like this are part of the journey as a cinematographer as this knowledge/understanding falls under the blanket of capturing high quality motion pictures.

Exposing methods, stopping the lens down, using specific types of filters (like circular polarizers, low contrast, ND, and UV), and using diffusion on your light sources are all some things to look at when figuring this out.

Also, understanding the camera and what it can capture as well as having a light meter nearby can truly inform you about where trouble areas might occur when you step into an environment. After a while, your eyes and experience become that tool.

Testing will save you time and money when it comes down to it.

One more note as to how top DPs work and more or less eluding to how I acquired some of my own education on these subjects long ago. Many, many DPs shoot/light/expose in fairly low contrast/controlled dynamic range scenes to fight against "weirdness" such as this. The reality is high contrast situations stress all camera systems and is truly the hardest thing for it to do. Years ago, especially in the old days of video people had a limited playground of around 7-9 stops and if you look for well shot stuff on those dinosaur cameras you'll see it's all fairly low contrast and limited dynamic range. With film it's been different as you generally have/had 13-15 stops of freedom, which is a whole bunch more. DPs who shot film had/have lots of techniques whether they did it through rather deliberate or experienced based mindsets. The safe 6 up/6 down - 7 up/7 down method, expose one stop under/push one stop up, and some high key methods can be seen pretty clearly if you look for it. Additionally film clips pretty nicely due to the aesthetics of the medium. Most higher dynamic range digital cinema cameras also clip much more nicely. For instance the difference between say a Dragon sensor and maybe some of the more popular action cameras in terms of how the look of overexposure is captured. Highlight Roll-Off is a touchy subject when it comes to these cameras and linear light, but there is indeed a different on what is retained under common lighting conditions and in the RED world that can also be visually scene by your OLPF selection and what is "held" at the top end. Some cameras are more organic when applying the log or video gamma curve, some less so.


Hope that helps.


Here's something shiny from a fairly old lens, Weapon 8K VV Dragon w/ DRAGONcolor2:

 

[David Battistella]

I hear you on this.

I think what you might be seeing with "other sensors" is circuitry that clips the pixels for you. I am not defending this problem in the color science, I am simply recognizing the factors that contribute to helping it appear and offering solutions when that problem does arise. Much of this might be attributed to how much is being resolved by this sensor and the artifacts are partially due to CA which you might not see resolved on a sensor with less resolution and a higher resolving power.

As the sensors become more "perfect" the glass (or rather the imperfections in the glass) can become more visible and more pronounced.

I love letting things blow out, but there is a limit. I mean in this shot if you change the purple pixels to white ones anyone can tell that it's being hit to hard because it is distorting the shape of the guys face. I think it's the LLO OLPF as well which might tend to artifact a bit more in the highlights than the STH. That is why I asked that question and for the R3D.

I'm not here to defend problems that arise, they arise for me as well and I have seen pretty much all of them and try to offer solutions. Generally, that is when the "i spent so much on this camera and this should not happen" comments start to become the discussion. I try to focus on solutions. It's up to RED to fix these "problems".



David

It just seems to me that when it comes to purple fringing, the Dragon sensor is the most limited sensor I have ever worked with. Arri, Canon, and Sony can usually shoot under the exact same conditions and not exhibit this weird artifact. Yes, technology had its limits, but Dragon is limited in ways that other systems don't seem to be. That's fine, except the price and the advertised image don't reflect this particular limitation.

If this was happening on a Blackmagic camera, I doubt people would be defending it. Again, it seems like a "simple" color science fix. Why aren't we pushing for that? Why is everyone ok with the status quo when it clearly interferes with certain styles of work? (Sometimes you WANT that heavy ratio)

 

[Dustin Wilson]

annnd this is why I'm still using my trusty Epic MX.. Like many have said you shouldn't have to develop work arounds via color science or stopping down or under exposing to avoid problems like this. The OP images are ridiculously bad. Regardless weather your camera or my camera does or doesn't produce issues like this, when purchasing a $30,000 camera you shouldn't have to be "focusing on solutions" for this. It should just work whether you blow your image right out or not (within reason), that's creative freedom. Why would I spend $10k upgrading my sensor to be more worried about my highlights and CA? Isn't that working backwards? I thought Dragon offered the opposite of that? I understand knowing your camera and its limits, but what it seems you guys are saying is, you need to be a lot more careful using this new, better sensor.. isn't that an oxymoron?

 

[Kenneth C Merrill]

Thanks for all of the technical info guys! I already had a rudimentary understanding of what is happening; I was just being a butt head because other cameras I use don't experience it nearly as often or as much At any rate, thank you for taking the time to explain things so thoroughly. I'm definitely learning things!

And thanks for hearing me out David. I understand trying to be helpful and offer solutions to issues, when other people are losing their heads. I guess I just wonder if we might have gotten a color science fix for this if more people had been more demanding/vocal about the issue (like they were with the LLO OLPF, for example). At any rate, all cameras have trade offs, and this is a relatively minor one compared to issues that other cameras have. Would be awesome if a new color science could help us out a bit though.

 

[David Battistella]

annnd this is why I'm still using my trusty Epic MX.. Like many have said you shouldn't have to develop work arounds via color science or stopping down or under exposing to avoid problems like this. The OP images are ridiculously bad. Regardless weather your camera or my camera does or doesn't produce issues like this, when purchasing a $30,000 camera you shouldn't have to be "focusing on solutions" for this. It should just work whether you blow your image right out or not (within reason), that's creative freedom. Why would I spend $10k upgrading my sensor to be more worried about my highlights and CA? Isn't that working backwards? I thought Dragon offered the opposite of that? I understand knowing your camera and its limits, but what it seems you guys are saying is, you need to be a lot more careful using this new, better sensor.. isn't that an oxymoron?

Its not about the sensor and the color science as much as it is about using poor glass at apertures where it performs the worst. You're going to see CA issues, that is not the sensor, it's the glass. It's the sensor out resolving the glass and pronouncing problems with the glass.

To me, Its good to know how and when it happens and also why it happens.

David

 

[Phil Holland]

Why would I spend $10k upgrading my sensor to be more worried about my highlights and CA?

Better color, industry leading IR contamination protection, more dynamic range, larger format sizes, cleaner images, and *gasp* more resolution for those who desire such things.

I'm not even going to mention DSMC2 related things like intermediate codec support, better optical path, lower noise floor, ventilation, auto sensor calibration, better data rates, lighter camera body, and a dozen other things.

MX is great. I have no problem saying that as a former MX shooter, but Dragon in DSMC and DSMC2 is better in nearly every way. This is pretty much reflected by the industry wide acceptance and use of Dragon sensor cameras.

 

[Blair S. Paulsen]

1) Shooting lenses with known CA issues WFO into high contrast scenes with the LLO OLPF will "pop" like a bass note dead on the resonant frequency of a speaker cabinet.

2) It's not a "dodge" to point out that a more accurate imaging system will show more artifacts. When I brought home my first true audiophile level stereo system I was shocked at just how much more you could hear the "noise" created by every bit of dust/damage done to my vinyl record albums. Suffice it to say that my next stop was for gentle yet effective cleaning products ;-)

3) IMO, the idea that the MX is a fundamentally better sensor than Dragon is quite a reach. For starters, based on color grading sessions with footy from each, there are some looks that are easy to achieve with Dragon clips - but require much more manipulation to accomplish with MX material. Sure, there are a few looks that are right in the MX wheelhouse - but for those that are not, it can take some effort to get there. OTOH, I haven't had a session yet with Dragon footy that couldn't quickly be characterized to meet creative expectations. YMMV.

Cheers - #19