Modular Dragon OLPF sneak peak

[Gavin Greenwalt]

It's funny - all the BMCC owners (myself included) are begging for an OLPF (Mosaic is slowly working towards that). Though I guess a 6K sensor would alias quite a bit less than BMD's 2.5K one...

There's also a question of coverage. If you have tiny pixels spaced out you'll get worse aliasing than a hypothetical sensor without any gap between the sensor sites (or microlenses).

 

[Ryan Hamblin]

can I order a no LP mx filter? How much would it be? does it still filter some of the IR? I would love to grab one if possible.

 

[Jarred Land]

Play nice kids..

 

[Elsie N]

sorry.

 

[jacob.schwarz]

I would love a filter that acted like the Schneider Hollywood Blackmagic 1/8. that would be choice.

People might have already asked but i'm assuming no, but can you stack filters? or is that going to cause back focus issues?

 

[Jarred Land]

Honestly while I was annoyed by the red dot noise in the footage, I don't mind having some noise, as long as it's not colored like red in dragon or blue in mx, so I can't wait for D.E.B to come out to test it out.

But the idea of swapping olpf's is exciting, specially if I'm filming a dark neo-noir scene.

I do wish we would get a lowlight version of the new olpf (same great color, IR protection, no flares, but moving the stop from the top to the bottom) it would be great! But realistically I don't think it's worth RED's resources and time, which I understand.


I do want to say thanks Jarred and others at RED, listening to our opinions and helping us means a lot!

Its not that simple.. the new OLPF isnt just an ND that shifts the bottom to the top. So you can't get the best parts of every filter in one filter with no penalty.

That said.. we already plan on making one that lands somewhere in the middle. And I really do like the Low-Con option that people are discussing, but that is going to take some time.

The options really start to open up as we open this option up to both individuals like John Marchant that have their own unique take as well as to the big filter companies to make modules as they may.

 

[Hugues Dagallier]

That's more or less the concept. It's all about how it's "mapped" and where things are placed. Each ISO rating has implications towards how much texture your final image has.

I would say in terms of tonal response and what the sensor has to offer you would be targeting ISO 800-1280 for the most "film-like" image.

The highlight roll off effect seems to always be there to me. It just doesn't clip hard like Mysterium-X, and that's a visible difference even in extreme scenarios.

The difference in textures is really how much texture you actually want. Texture is there across the ISO range, 250-2000. As you rate your midtones for a higher ISO value more general "activity" will be seen in the texture and you will see that across the image.
.

Thank you Phil.

Now let's imagine I am shooting at t1.3 and I want to say at this aperture and I set the dragon at 1000 iso. Let's say I am over exposing thé scene by two stops with this setting. Let's say I do no want to touch the lighting set ups (Natural or not).

What would be the major difference in terms of rolloff and tonal response between shooting at 250 ISO or keep my 1000 ISO and put an nd filter on it ? In my opinion, staying w/ this iso would provide a more film-like image.

 

[Gunleik Groven]

Jarred.

This sounds like the answer to a lot of issues.

Most of the "issues" are pretty constructed, though...

But if people can have a "lowlight" Dragon, and a "lit/cinematography" Dragon in the same unit, someone should start looking at the images the camera creates...

 

[David Battistella]

I would love a filter that acted like the Schneider Hollywood Blackmagic 1/8. that would be choice.

People might have already asked but i'm assuming no, but can you stack filters? or is that going to cause back focus issues?

I like that filter too. Maybe not in every situation, but then again if I put it in a mattebox, it's still probably simpler than having it in the OLPF. Plus, this level of filtration does not account for different focal lengths where you want to change filter densities to match shots based on the focal length of the lens or the subject/actor.

If you love that filter at that strength, it is not equal at all focal lengths, so a mattebox and different densities is still important to consider. Plus sometimes I want a locon, and sometimes a digicon is better and lately it's been a locon, black promist combo on one film.

I don't think there will ever be a "one ring to rule them all" OLPF, unless PJ order one,of course,

battistella

 

[Phil Holland]

Thank you Phil.

Now let's imagine I am shooting at t1.3 and I want to say at this aperture and I set the dragon at 1000 iso. Let's say I am over exposing thé scene by two stops with this setting. Let's say I do no want to touch the lighting set ups (Natural or not).

What would be the major difference in terms of rolloff and tonal response between shooting at 250 ISO or keep my 1000 ISO and put an nd filter on it ? In my opinion, staying w/ this iso would provide a more film-like image.

The tones are "mapped" towards whatever gamma (like REDgamma4) or manual curve you apply (like using REDlogFilm + custom cuve). Meaning if you expose to meter or chart at ISO 250-2000 individually you will see "the same image", however, with increasingly more texture towards the higher ISOs values as that's essentially bringing the midtones closer and closer to the noise floor.

It's tricky to say exactly what happens for every given situation as you might have a light that doesn't fully clip and maintains detail for instance even at T1.3. If you have like a crazy daylight back lit window it will clip (like we're talking 9 or 10 stops over exposed), but it's not as instant or harsh compared to Mysterium-X. Even more interesting is Dragon's published specs of ISO 250-2000. That's the given usable range, but I did start noticing some good things when shooting at various ISOs.


For instance:

This washer in frame here is reflecting the sunlight during sunset. I shot this at ISO 800 and graded my in Redcine-X Pro. Nothing was clipping. I brought the image down to ISO 50 to see how much Dragon was retaining and there's just detail in there. So basically what I've saying is if working within the recommended range the highlight is "rolled off" because Dragon sees and retains a lot of highlight information. This is with the newest OLPF, I'm going to start calling that V2 OLPF I guess. The V1 OLPF is about 1 to 1.3 stops more sensitive in the shadow information, but you lose those stops on the highlight retention. Interestingly the roll-off is similar-ish though. It's just where the dynamic range begins and ends really.


In terms of adding ND variables get a little bit more interesting as the strength and formula of that ND come into play. Generally speaking you will see any ND effect the contrast of the image. Higher densities often "feel" a lot like a low con filter even. Utilizing a low contrast filter can "squeeze in" a bit more information in contrast situations, which is why I think this might be a fun option for part of the future Modular OLPFs.

Back to your original question about film in relationship to Dragon and why I recommend ISO 800-1280. Film is 13 or 15 stops of captured dynamic range. 13 typically. However if you shoot, process, and scan flawlessly you're looking at about 15 stops. Dragon at ISO 800-1280 out of the published spec of 16+ stops of dynamic range gives you approximately equal stops above and below middle gray. This to me, combined with the texture, gives off the most film-like image. This is one reason why many films are actually shot at ISO 800.

If you are looking at pure "DR" in relationship to film somewhere between ISO 500-1280 overall would be about right. That's getting you approximately a similar image at film's measured 13-15 stops. Dragon at any given time is capturing all of the available dynamic range. It's basically how you choose to use that DR that defines the texture, look, and feel.

 

[John Marchant]

For those of us thinking entirely on the creative end of OLPF usage - although we all have favourite diffusion / low-con options, unfortunately its not as simple as transferring that material to the OLPF position. There's learning to be done in scaling diffusion to match the position in the optical path - and manufacturing will be another story again. Exciting stuff though - a well thought out and tested low-con giving organic highlight bloom is going to rock. Especially when you can whip it back out for the next shoot, or even the next setup.

 

[Phil Holland]

For those of us thinking entirely on the creative end of OLPF usage - although we all have favourite diffusion / low-con options, unfortunately its not as simple as transferring that material to the OLPF position. There's learning to be done in scaling diffusion to match the position in the optical path - and manufacturing will be another story again. Exciting stuff though - a well thought out and tested low-con giving organic highlight bloom is going to rock. Especially when you can whip it back out for the next shoot, or even the next setup.

I might try to bake a filter. Made a few notes over the last couple years regarding certain things. I remember during my "net tests" I ended up suspending metal flecks in liquid and put that in front and behind a lens. Utter madness and the glow was holy or unholy depending on how you look at it. This should be fun stuff.

 

[Nick Morrison]

That said.. we already plan on making one that lands somewhere in the middle. And I really do like the Low-Con option that people are discussing, but that is going to take some time.

Exciting to know a "mid" OLPF is in the works. Low-Con too. Thanks for always staying ahead of the curve. Impressive.

 

[Cid J Salcido Uyarra]

Does this apply to Dragon only or does this also include Epic/Scarlet MX

Say if we wanted Infra Red or other specialty filters?

 

[D Fuller]

Its not that simple.. the new OLPF isnt just an ND that shifts the bottom to the top. So you can't get the best parts of every filter in one filter with no penalty.

That said.. we already plan on making one that lands somewhere in the middle. And I really do like the Low-Con option that people are discussing, but that is going to take some time.

The options really start to open up as we open this option up to both individuals like John Marchant that have their own unique take as well as to the big filter companies to make modules as they may.

One of my favorite softening tricks in 35mm was always a bit of black nylon stocking taped to the back of the lens. (The ones that worked best we're always the most expensive French genuine silk variety I could find. Usually Dior, IIRC) it worked like black ProMist, only more... Natural, or subtle or something. is that kind of filter possible in this OLPF form?

 

[Hugues Dagallier]

The tones are "mapped" towards whatever gamma (like REDgamma4) or manual curve you apply (like using REDlogFilm + custom cuve). Meaning if you expose to meter or chart at ISO 250-2000 individually you will see "the same image", however, with increasingly more texture towards the higher ISOs values as that's essentially bringing the midtones closer and closer to the noise floor.

It's tricky to say exactly what happens for every given situation as you might have a light that doesn't fully clip and maintains detail for instance even at T1.3. If you have like a crazy daylight back lit window it will clip (like we're talking 9 or 10 stops over exposed), but it's not as instant or harsh compared to Mysterium-X. Even more interesting is Dragon's published specs of ISO 250-2000. That's the given usable range, but I did start noticing some good things when shooting at various ISOs.


For instance:

This washer in frame here is reflecting the sunlight during sunset. I shot this at ISO 800 and graded my in Redcine-X Pro. Nothing was clipping. I brought the image down to ISO 50 to see how much Dragon was retaining and there's just detail in there. So basically what I've saying is if working within the recommended range the highlight is "rolled off" because Dragon sees and retains a lot of highlight information. This is with the newest OLPF, I'm going to start calling that V2 OLPF I guess. The V1 OLPF is about 1 to 1.3 stops more sensitive in the shadow information, but you lose those stops on the highlight retention. Interestingly the roll-off is similar-ish though. It's just where the dynamic range begins and ends really.


In terms of adding ND variables get a little bit more interesting as the strength and formula of that ND come into play. Generally speaking you will see any ND effect the contrast of the image. Higher densities often "feel" a lot like a low con filter even. Utilizing a low contrast filter can "squeeze in" a bit more information in contrast situations, which is why I think this might be a fun option for part of the future Modular OLPFs.

Back to your original question about film in relationship to Dragon and why I recommend ISO 800-1280. Film is 13 or 15 stops of captured dynamic range. 13 typically. However if you shoot, process, and scan flawlessly you're looking at about 15 stops. Dragon at ISO 800-1280 out of the published spec of 16+ stops of dynamic range gives you approximately equal stops above and below middle gray. This to me, combined with the texture, gives off the most film-like image. This is one reason why many films are actually shot at ISO 800.

If you are looking at pure "DR" in relationship to film somewhere between ISO 500-1280 overall would be about right. That's getting you approximately a similar image at film's measured 13-15 stops. Dragon at any given time is capturing all of the available dynamic range. It's basically how you choose to use that DR that defines the texture, look, and feel.

Thank you Phil for your detailled answer.

 

[paul schefz]

sorry to keep bringing this up...but tobias straka posted another dragon raw file in this thread
The ones who have loved - Short shot on DRAGON
perfect example of what i use my camera for, so obviously i downloaded.....
at iso 400 exposed to protect that hotspot on the shoulder (i might have opened up just a tad, but histogram still looks ok)....there is a ton of noise all over...especially in the skin...i am not even talking about the mush of hair or the "light trap" area of where the hair meets the skin on her cheek/neck....i just don't have that kind of noise on my epic....and definitely not at 400....this is not red specs....this is color noise....
it does not seem to be user error to me? i agree that this is probably not an issue for 4k delivery or 2k.....but i need stills out of this camera and they have to be better then epic...this file is not printable larger then maybe 5x7....after that the noise would start to show....and please no mention of NR...my pic just does not have this kind of noise....the light trap areas go bad at about 1000 iso with my epic....
i am not trying to stir something here....i just would like to know if this is considered normal (since it goes pretty much hand in hand with my other dragon findings of raw files downloaded here)? will there be a olpf to take care of this?

 

[Blair S. Paulsen]

Bayer CFA on CMOS tech is both brilliant and problematic. AFAIK it was developed exclusively for still capture which requires a lot less optical AA (anti aliasing) than motion.

The RedOne and other early examples of Bayer mask motion capture devices tended toward using strong OLPF filtration to control potential artifacting. As you might imagine, putting a strong OLPF filter in front of a native resolution (1080 for 1080) sensor typically led to rather soft images and color registration issues. The Madman, and some other very bright folks decided the way around that problem was to super sample (capture more pixels than the destination format). Sounds good on paper, but how in the heck can you record all those pixels without being tethered to a big RAID? Enter RedCode compression. This, IMO, is the fundamental strength of the Red Digital Cinema program.

Refining the characteristics of the OLPF to hold more resolution without aliasing is a key part of getting the best performance from Bayer CFA on CMOS. When one OLPF design has to work for every possible shooting challenge you're kind of forced to be pretty conservative. The "Swappable Era" should offer a tremendous playground for creating OLPF/debayering algorithm combinations that perform very well on some material without great concern for how bad it might look on other material.

In the short term I expect the current shipping OLPF (Dragon V2) to be the "standard" set up. I do expect that an OLPF designed to eat as little light as possible (Dragon V1 the sequel) will be the second option if they have time to bake one up. I love the idea that the designers could get much more aggressive in creating an OLPF that eats very little light and would be allowed to have some moderate moire and aliasing - after all, if the results are too funky you can just swap in the stronger one.

Cheers - #19

 

[Wayne Morellini]

Wayne, like you foreshadowed, I don't think the small size of the Dragon camera design could accommodate the switchable or rotatable design for OPLFs. Cameras that provide something like that for filters seem to be much bigger. Of course camera size is relative to your needs.

For my use, I'll take the smaller Dragon size over the options available. A recent job with an Alexa M on a Movi M10 and backpack recorder almost required an exoskeleton to walk. The higher end Sony's with switchable NDs are much bigger than the Dragon as well.

lael

I don't know Lael, I think an add on (if the Epic had been designed to accommodate it, could be added at minimum size, even in case. We are talking about two+ relatively small slots. Also why I mentioned a slider, for size. But I would have to research these systems to know anything for sure. Hmm, just thought, thinking of flip down 4 filters around a cavity but requires a sizable length of gap, or slide in that can put 2-9 filters on the same plane, but the case starts to get bigger when not in 2 dimensions.

It's funny - all the BMCC owners (myself included) are begging for an OLPF (Mosaic is slowly working towards that). Though I guess a 6K sensor would alias quite a bit less than BMD's 2.5K one...

There's also a question of coverage. If you have tiny pixels spaced out you'll get worse aliasing than a hypothetical sensor without any gap between the sensor sites (or microlenses).

Ooh, don't say that Gavin, remember my discussion with Graham.

Would I get banned if I offered a group buy for an OLPF replacement kit for the MX ?
Some of you know that I have been shooting for over a year now with my Epic that has IR cut but no low pass filter.
I out resolve Dragon that way. Has aliased so rarely in the real world that it is not a problem.
Just pondering !

What was that. Whose that phantom ex member to be?

I think you are alright Les, but probably a good idea to do it through Red.

The reason you probably don't see much aliasing is the over resolution recording mentioned by Kevin, as the gaps often fall in pixels, but do you get detail flash with movement? I'd prefer 100% FF BSI myself.

You're an engineer and would know more about the cameras physical layout than me. Do you think my swappable filter idea is doable Les? I find it fascinating, to get that little bit extra quality by exchanging OLPF's. I had an ultimate variable ND filter design years ago. I could see how to stack many filters in at once and do 20 stop+ attitude.

 

[Christopher Probst]

I may just have to have a separate little pelican case fitted with all of these different OLPF's at my disposal. Truly awesome to custome tailor your OLPF to the given scene's requirements. I'm in.