First Look: Interchangeable OLPF on Dragon

[Phil Holland]

Rob.
And this is why:

I think in reality this boils down to Rob and myself being two rather different types of people. While I don't know him in real life often his posts are goading and come off as rather insulting while pointing out his opinions. I respect his POV as a business man and Director/DP, but I tend to look at tools from a wildly different perspective. Mainly focusing on what they can do first and also where concerns would pop up. Pretty much how every DP looks at any camera system, and heck why not, film stock. My particular focus is the world of feature films, broadcast, and commercial work as that's where my experience and desires have led me. That also influences my more practical real world tests as well as my more chart/experiment related methods.

My favorable thoughts on Dragon in particular come from the experience I've had with it as well as knowing what I can do with it. The very same opinion that far larger than I Director and DPs also share, mostly because we are "seeing" the same things when it boils down to pure image quality and functionality. Much like literally every camera on the market there's things to look out for. I've pointed them out and have tried to help steer folks in the right direction to get quality results. Other's simply make it sound like it's a metal box that doesn't produce some of the highest quality moving pictures you can at the moment. And in reality Dragon is something rather astounding as a digital cinema camera, especially if you know how to get the most out of it.

Literally every Mysterium-X shooter I know who has seen the color alone "gets it".

Meh. All good. Moving on.

You may be right, but my rep went and talked to someone who would know (he said) and that's what he told me. I even referenced this thread and Jarred's comment. Sigh, I wish there wasn't conflicting info.

From what I can "see" is it looks like some of the IR or AR or a combination of both have been tweaked a bit. Like John said, the improvements seem to be in the optical path and are looking rather improved now. So I'm a happy piglet.



Back to business with a quick daylight skin test, in this case White Balanced to chart I dropped out of frame. Did this with a single sheet of Opal Frost to get a nice transition of highlight through soft shadow, hence the hand in frame and my rather pensive look. Additionally I'm wearing a pure black shirt and with ND it's not turning reddish brown or having a big magenta shift. It's actually doing pretty well. Not really a huge shift on the house wall or window color either.

First no ND:




Next my slightly blue-ish ND 1.8:




That looks pretty damn good to me. Though it reminds me I need to get my hair cut today.

 

[Akin A]

I'm reading a lot about the improved colors/skin in Dragon STH OLPF over MX.
If you get time, could you please post a skin tone comparison of STH vs MX, where it's noticeable? Thanks!

BTW, as a guy new to RED, I like reading responses from both extremes on whether Dragon is living up to the hype. As long as it's not personal and strictly technical, it helps me learn and keeps me from setting myself up for disappointment.

 

[Peter Strietmann]

I think in reality this boils down to Rob and myself being two rather different types of people. While I don't know him in real life often his posts are goading and come off as rather insulting while pointing out his opinions. I respect his POV as a business man and Director/DP, but I tend to look at tools from a wildly different perspective. Mainly focusing on what they can do first and also where concerns would pop up. Pretty much how every DP looks at any camera system, and heck why not, film stock. My particular focus is the world of feature films, broadcast, and commercial work as that's where my experience and desires have led me. That also influences my more practical real world tests as well as my more chart/experiment related methods.

My favorable thoughts on Dragon in particular come from the experience I've had with it as well as knowing what I can do with it. The very same opinion that far larger than I Director and DPs also share, mostly because we are "seeing" the same things when it boils down to pure image quality and functionality. Much like literally every camera on the market there's things to look out for. I've pointed them out and have tried to help steer folks in the right direction to get quality results. Other's simply make it sound like it's a metal box that doesn't produce some of the highest quality moving pictures you can at the moment. And in reality Dragon is something rather astounding as a digital cinema camera, especially if you know how to get the most out of it.

Literally every Mysterium-X shooter I know who has seen the color alone "gets it".

Meh. All good. Moving on.




From what I can "see" is it looks like some of the IR or AR or a combination of both have been tweaked a bit. Like John said, the improvements seem to be in the optical path and are looking rather improved now. So I'm a happy piglet.



Back to business with a quick daylight skin test, in this case White Balanced to chart I dropped out of frame. Did this with a single sheet of Opal Frost to get a nice transition of highlight through soft shadow, hence the hand in frame and my rather pensive look. Additionally I'm wearing a pure black shirt and with ND it's not turning reddish brown or having a big magenta shift. It's actually doing pretty well. Not really a huge shift on the house wall or window color either.

First no ND:




Next my slightly blue-ish ND 1.8:




That looks pretty damn good to me. Though it reminds me I need to get my hair cut today.

Now I'm impressed! Thank you sir.

 

[Phil Holland]

I'm reading a lot about the improved colors/skin in Dragon STH OLPF over MX.
If you get time, could you please post a skin tone comparison of STH vs MX, where it's noticeable? Thanks!

BTW, as a guy new to RED, I like reading responses from both extremes on whether Dragon is living up to the hype. As long as it's not personal and strictly technical, it helps me learn and keeps me from setting myself up for disappointment.

I'll see what I can do. I no longer have an MX body and the last time I shot with one was back in June 2014, prior to that I haven't shot on MX since 2013. After my initial tests with Dragon from way back then, I made the switch and haven't looked back.

One of the big things for me is the need of IRNDs on Mysterium-X and on Dragon you just don't need them. Regular NDs work like a charm.

 

[David Battistella]

Man, I'm not being hostile or anything, not sure why I'm upsetting you. Seems like I am. Just trying to figure out whether there is a physical change in the v1 OLPF vs the lowlight(interchangeable). There is conflicting info.

I think I explained pretty clearly why. Yes I have my battle tested dragon, as I said before. They took v2 out and put v1 in.

Love you.

jake,

didn't mean to come off as hostile. Admittedly, I answered hastily as I was going through an airport terminal.
Its cool. Yup. There's conflicting info and all Gunleik and I suggested was that Jarred's might be the one to go with.

I can appreciate the spot you are in and the info you are seeking.

Based on what I am seeing an reading from Phil if I were in your shoes I might move toward the new OLPF swapable system as it opens many doors and seems to improve the LL overall. I don't know whether that is optically or via FW but Phil seems to be finding it has all the characteristics of the V1 with some optical improvements because the orbs and things are reduced and that may have been more of an optical path issue.

Love back at you.

Ps. I'm cleaning up the posts that lead us here. My bad. Don't want to add to the noise.

Battistella

 

[Antony Newman]

Phil,

I think the skin tone in the second shot looks great.

If the ND and the Opal Frost are not absorbing IR, does that bode well - in your view - that an LUT is all that separates the LLO and STH?

[@RED] I know relatively little about the complexities of what is required to recreate colours of each pix in RedCini X - but I am wondering if there is a 'virtually perfect' engineering sensor that be used to measure the Dragon sensor against - using each of the LLO and STO OLPFs over a range of intensities - and allow this to be imported into Redcini to make processed images look 'perfect'?

AJ

 

[Robert Ruffo New]

Phil, I didn't mean to goad or insult you, or anyone else. I have insulted Elsie a few times, but that's about it.

I am not from the same production culture as you. Here in Quebec DPs look for problems, they don't excited and look for high points in any system (for example, a new light comes out, first thing I do is look for flicker at various frame rates, not try to make cool lighting set-ups). We have smaller budgets here than Hollywood, so I suppose it comes from not being able to afford mistakes or surprises. We are cautious, we temper our enthusiasm.

That's how I was shown to do it by my mentors, that's how everybody does it here I can think of. I suppose that makes me different from other DPs you may know - although I hope what you said about my personality being unlike the DPs you know was not an indirect attempt at discrediting me in my profession, which by the way is director-DP, which I suppose is a different personality type (probably more ego driven and loud mouthed, in my experience. "Only DPs" are much more soft spoken and easy going as a lot than my hyphenated brethren).

In QUebec we also very much respect people who disagree with our point of view, and have the guts to be honest about it. We figure that is the best way to learn. Much more useful than compliments. "You can tell a real friend because he tells you the hard truth" is a deeply held Quebec belief, and my belief.

Anyway, I think it's time to join the many friends who have stopped posting on Red user. The minute you say something that is critical, you get slammed, attacked insulted, and it seems only fanboys are treated with any respect, as they trip over each other to agree with each other and praise Red.

This does not lead to interesting discussion or debate, in my opinion, nor does it impress agnostic users much. It certainly is not an environment that makes me feel comfortable.

 

[Phil Holland]

I think the skin tone in the second shot looks great.

If the ND and the Opal Frost are not absorbing IR, does that bode well - in your view - that an LUT is all that separates the LLO and STH?

Yeah I'm actually pleasantly surprised with what I'm seeing with the LLO OLPF. I think I'm going to run with it for my next shoot at the end of the week.

For an absolute spot on color match between tge two OLPFs a LUT might be the solution for some. Though the color is good enough that you wouldn't really need it. I'm still testing various color temperatures and lighting conditions, but what I've posted here are pretty mich the general broad strokes that are the most informative.

Daylight, daylight diffused, artificial daylight, tungsten, low light, etc...

 

[Phil Holland]

3:05 AM - Saturday..... erm.... Sunday..... erm..... whatever.


Something peaked my curiosity late, late this evening. I was thinking a lot about that Low Light Optimized OLPF shot that I pushed up to ISO 4000 and was trying to quantify the general texture character variation between the LLO OLPF and the Skin Tone - Highlight OLPF across the ISO range.

Jarred mentioned it before being around 1 stops. Previously I measured around 1.3 stops. That's sort of what we are looking at in terms of "matching texture character" here.

This will be one that people will have to test out for themselves, but pretty much how the texture measures up is the LLO OLPF is about 1-ish stops cleaner overall.

So basically what I've done is ran some image processing on some images and also exported out some mirrored tones and watched them in 4K on the big screen.

Here's sort of how it's playing out for me texture-wise.

LLO OLPF @ ISO 800 = STH OLPF @ 400
LLO OLPF @ ISO 1600 = STH OLPF @ 800

You get the idea. I would say this is give or take 1/3 of stop in some circumstances.

Comparing texture and general "energy" on screen is a bit tough, but this is pretty much how I'm experiencing it. It's certainly less visible at the lower ISO range, like 250-500, but after that you can get a clearer A/B comparison when it comes to general character.

Both OLPFs are allowing the sensor to "see" about the same amount of light and the calibration is revealing that the tonal/chroma range is mapped slightly differently between the two. Making the LLO OLPF the cleaner option for those who desire super clean images or the higher ISO performance.

Curious to see other's thoughts when the official Interchangeable OLPFs start flying out soon.

*edit - I'll add since I've been shooting with the STH OLPF I've been shooting mostly in the range of ISO 250-1280. I've found ISO 500 to be my little happy place for a variety of reasons, which makes me wonder if I'll be rolling at ISO 800-1280 more often on the LLO OLPF. Dunno yet, but we'll see. ISO 800-1280 on the LLO OLPF looks time nice in terms of tonality to me, which is what it should look like.

 

[David Battistella]

Phil

I know you have shared a bit but can you share a bit about how you were swapping OLPF's while in the feild and the steps "hardware and software wise" when you do a swap?

Cheers.

I can for certain see the need for both. Looking forward to the stuff other are cooking up right now as well.

Battistella

 

[Phil Holland]

Phil

I know you have shared a bit but can you share a bit about how you were swapping OLPF's while in the feild and the steps "hardware and software wise" when you do a swap?

Yep. Really simple too. This is my "in the field" procedure. I do actually recommend going to a camera cart, just for safeties sake really, but......

Prep stuff you need (flashlight, hand blower, OLPF w/ it's case, Sidewinder) and walk over to the camera and make sure it's level.
Remove the DSMC Lens Mount, I stow the bolts and mount on top of the camera.
Remove the Interchangeable OLPF and stow bolts, I place them on the little container I'm using which has a ridge.
Quickly inspect sensor box and the incoming Interchangeable OLPF and blow away particles as needed.
Install Interchangeable OLPF and DSMC Lens Mount. I take a quick look at the DSMC Lens Mount as well for dust.

In camera go to Menu>Settings>Maintenance>OLPF select the OLPF.

You'd be alarmed at how quick you can get at this.

 

[David Battistella]

Freaky!

cool.

thanks Phil.

battistella

 

[Nick Morrison]

3:05 AM - Saturday..... erm.... Sunday..... erm..... whatever.


Something peaked my curiosity late, late this evening. I was thinking a lot about that Low Light Optimized OLPF shot that I pushed up to ISO 4000 and was trying to quantify the general texture character variation between the LLO OLPF and the Skin Tone - Highlight OLPF across the ISO range.

Jarred mentioned it before being around 1 stops. Previously I measured around 1.3 stops. That's sort of what we are looking at in terms of "matching texture character" here.

This will be one that people will have to test out for themselves, but pretty much how the texture measures up is the LLO OLPF is about 1-ish stops cleaner overall.

So basically what I've done is ran some image processing on some images and also exported out some mirrored tones and watched them in 4K on the big screen.

Here's sort of how it's playing out for me texture-wise.

LLO OLPF @ ISO 800 = STH OLPF @ 400
LLO OLPF @ ISO 1600 = STH OLPF @ 800

You get the idea. I would say this is give or take 1/3 of stop in some circumstances.

Comparing texture and general "energy" on screen is a bit tough, but this is pretty much how I'm experiencing it. It's certainly less visible at the lower ISO range, like 250-500, but after that you can get a clearer A/B comparison when it comes to general character.

Both OLPFs are allowing the sensor to "see" about the same amount of light and the calibration is revealing that the tonal/chroma range is mapped slightly differently between the two. Making the LLO OLPF the cleaner option for those who desire super clean images or the higher ISO performance.

Curious to see other's thoughts when the official Interchangeable OLPFs start flying out soon.

*edit - I'll add since I've been shooting with the STH OLPF I've been shooting mostly in the range of ISO 250-1280. I've found ISO 500 to be my little happy place for a variety of reasons, which makes me wonder if I'll be rolling at ISO 800-1280 more often on the LLO OLPF. Dunno yet, but we'll see. ISO 800-1280 on the LLO OLPF looks time nice in terms of tonality to me, which is what it should look like.

Very informative. Thanks for the research Phil!!

best

 

[Anton Shavlik]

Wouldn't an easy test be to set the aperture so that a source of light is just blown out on one of the stoplights, then switch OLPFs, and repeat. Comparing these two apertures will give you the exact stop loss of one OLPF relative to another. Sitting and comparing grain patterns is important, I understand, since the color calibration is a little different as well, but this might be a more useful test for saying "ISO 800 = ISO 2000".

 

[Phil Holland]

Wouldn't an easy test be to set the aperture so that a source of light is just blown out on one of the stoplights, then switch OLPFs, and repeat. Comparing these two apertures will give you the exact stop loss of one OLPF relative to another. Sitting and comparing grain patterns is important, I understand, since the color calibration is a little different as well, but this might be a more useful test for saying "ISO 800 = ISO 2000".

Yes and that's what I was doing for the table top setup as that light source is clipping. Though it's not exactly how it "totally works". Since the camera was in the same position I can make fairly accurate measurements with the extra footage I shot to match it. However, the true different is being revealed not in how much the camera is seeing (which is approximately not so different) but more in texture/grain/noise performance across the board.

What I'm really seeing is a 1 to 1.3 stop improvement in texture performance in the Low Light Optimized OLPF over the Skin Tone - Highlight OLPF. At it's core people wanted cleaner images at higher ISO ratings and they are going to get that here.

The sensitivity isn't what has changed really, though the Skin Tone - Highlight OLPF reveals a "smidge" more highlight protection which to me is a nicer highlight roll-off, but it's a slight difference.

The real difference here is where the texture is. The LLO OLPF isn't seeing "more" it's just seeing more "cleanly".

 

[Phil Holland]

Just to expand on that post a bit with sample images to show you guys what I'm talking about here are some REDlogFilm ISO 2000 samples that show that both OLPFs are essentially seeing the same data, just the calibration is improved in tonality of the texture by about 1 to 1.3 stops in the Low Light Optimized OLPF:














Things get even more interesting when you incorporate DEB at higher ISO ratings for both OLPFs.






These last two are Skin Tone - Highlight OLPF at ISO 800 and Low Light Optimized OLPF at ISO 1600. Viewing these at 4K in motion helps with the side by side texture comparison and general "energy" and "feel", but this should give you guys the general idea.






I would say in this case ISO 1600/2000 on the LLO OLPF would be a textural equivalent of the STH OLPF at ISO 800. It's pretty much a straight forward 1 to 1.3 stops of textural difference.


One thing I'd like to add about this table top setup is nothing is there by accident. I actually spent a long while setting things up, lighting, taking meter readings of charts and object values, making sure there was tonal variation from clip to crush/noise, etc..... Even the antique roller skate box has the name "Frankie" on it as a nod to the original Red One Prototype "Frankie".

Hence the very "Knick" and Frankenstein's lab setup

 

[Phil Holland]

Oh. And one last post with images as I have some files here I'm inspecting.

Dynamic Range. Dragon has a lot. 16+ stops is a nice figure for reference and generally "is what you want it to be".

Remember when shooting at lower ISO ratings you're tapping into a smaller sliver of DR, though you will have alarmingly clean shadows and a clean image overall. However your highlight detail can get nuked, especially if you have something that rolls to clip in frame. When shooting at higher ISO ratings you will begin to see more texture/noise/grain, but you are also tapping into where a lot of the Dynamic Range available to you is on Dragon.

A very "film-like" image (tonal placement) for me is in the realm of ISO 500-1600 on both of these OLPFs.

Here's 3 samples from the Low Light Optimized OLPF all exposed to 18% Gray. Check out the detail in the light and the image texture/noise/grain:








This isn't new news for many, but I still see a lot of talk about the DR around here and around town.

The 100% real question is what sort of character do you want your image to have? In my personal experience especially now since I've communicated with so many regarding Dragon is that the "film guys" want some texture. The "video guys" and commercial producers don't. Dragon is notable cleaner at an equivalent ISO rating to film ASA comparison across the board with both OLPFs, which is more or less why the noise talk was a bit more "noise" than substance to me. However, in this digital age and more video minded individuals who often have never touched film the desire for clean is very real.

There is a difference between clean and plastic though. Myself and studios are acutely aware of cameras that utilize on board image processing because of that. And in reality any system that does that is eliminating some of the pure resolution that we are all striving so hard to shoot for.

 

[Les Hillis]

Thanks for doing such a thorough examination of the OLPF's Phil, some really clear examples of what's going on.

Am I mistaken though, or are the pictures in post #192 that followed "These last two are Skin Tone - Highlight OLPF at ISO 800 and Low Light Optimized OLPF at ISO 1600." the same 2000ASA ones that were posted just before that?

 

[Phil Holland]

Thanks for doing such a thorough examination of the OLPF's Phil, some really clear examples of what's going on.

Am I mistaken though, or are the pictures in post #192 that followed "These last two are Skin Tone - Highlight OLPF at ISO 800 and Low Light Optimized OLPF at ISO 1600." the same 2000ASA ones that were posted just before that?

Apologies, I was linking to the previous two images. Fixed now.

 

[Björn Benckert]

From my simple, none side by side testings of olpf1 and olpf 2, I came to the following conclusions:

The build,coatings and filtering in the olpf2 provide a more accurate color rendition but it also eats about 1.3 stops of light in comparison with olpf1 (based on how much more noise there is i my testings). The loss of light in the olpf2 RED compensated for within the camera color science simply explained with a "digital gain" this gives the two olpfs about the same exposure result however the noise comes 1.3 stops earler on olpf2 and olpf2 has cleaner colors, So simply it's the "extra filtering" in olpf2 that cuts light and when compensated for it brings up the noise floor and ofcourse also leave more room in the highlights.

Again this is just my assumptions from not so scientific testing on how it all works so please correct me if Im wrong so I dont cause more confusion on this matter.