OLPF "stack" standards

[Blair S. Paulsen]

I consider the optical path to be from first filter (or lens front element) to the sensor surface. Stray light entering the optical path via internal reflections can be a significant issue in some rigs but for the sake of relative simplicity my post will ignore it.

Let's call anything in front of the lens (including coatings on the front element) like Polas, NDs, diffusion, etc stage 1.

The lens itself is stage 2, with resolving power and telecentricity the key characteristics in terms of OLPF performance.

The space between the exit pupil and the sensor surface where the OLPF "sandwich"/filter stack lives is stage 3.

One could argue that IF external filters and/or the resolution limits of the lens itself reduce spatial resolution enough an OLPF right in front of the sensor is just eating light for no reason. This theory could be applied to IR or UV rejection as well, though there more variables than makes sense to dive into here.

In any case, if the lens plus any front filtration can out resolve the "pitch" of the sensor then the aliasing issue needs to be addressed. Bayer CFA on CMOS cameras designed for motion typically use an optical low pass filter (OLPF) to combat aliasing. I refer to it as a "sandwich" or filter "stack" as it usually includes rejection of wavelengths outside the visual spectrum like IR or UV via coatings, etc. To reduce the chance of creating a new place for internal reflections, these various layers/coatings are typically laminated together into a single "stack".

Optical theory can get deep very quickly, but the short story on the space between the exit pupil of the lens and the surface of the sensor is that everything - even air - can impact the airy disk (which I think of as the maximum potential resolution point). In the days of film, every lens designer could just assume the properties of light passing through air were a constant (relatively speaking). Now that digital motion cameras are 99% of what's being used we have introduced a significant variable - the OLPF "stack". As Jarred rightly pointed out, each OLPF stack is different based on the sensor it is matched to and the choices made in its design.

What has been bandied about lately on CML-Glass and some other places is the concept of standardizing the physical thickness of the "stack", while still allowing each manufacturer to tune their OLPFs to best suit their sensors and other properties of their systems. One potentially valuable upside of standardization would be that properly collimated lenses (proper for digital rather than film) could be mounted on cameras with in spec FFD (flange focal distance) with a high degree of reliability.

My friend Mako pointed me to an article that looked at MTF curves (ostensibly for the MetaBones units) and discovered some pretty significant variations based on the thickness of the filter stacks in various cameras.

http://www.lensrentals.com/blog/2014/06/the-glass-in-the-path-sensor-stacks-and-adapted-lenses

I am starting this thread on RedUser in part to encourage RED to be a part of an industry wide effort to create a viable standard. A spec that gives filter stack engineers enough physical thickness to get their part of the optical path optimized and lens designers (or techs doing adjustments) a known target. This seems particularly important to consider in light of RED's modular swappable OLPF stack initiative. Whatever the final physical parameters/attachment mechanisms of the modular OLPFs turn out to be, I'd rather not have to worry about resetting the back focus if I'm WFO on fast, sharp glass.

I hereby respectfully request that RED make an effort to lock down the "standard" thickness of the filter stack both industry wide and in the modular system. To allow for future advances in OLPF tech (cultured nano crystals perhaps…) I advocate a thickness at least 2mm greater than anything in use today as long as it does not create other issues. BTW, I do know what a prep bay is for and a full day of prep should be SOP - that said, more and more UPMs are truncating or eliminating prep days for budgetary reasons and at the end of the day if the footage you're getting paid to make is soft focus you are the one who will be blamed, fair or not.

Cheers - #19

 

[Patrick Tresch]

Thanks Blair for starting this OLPF Stack thread.

I was wondering how critical is the position of the OLPF to the Sensor. As the new OLPF module is User swappalbe some slight distance variation (<1/10th mm) could have some inconsistant impact on the optical path. Any input?

Pat

 

[Blair S. Paulsen]

There are a number of variables, some interdependent, that need to be accounted for. Air space between the back of the OLPF filter stack and the front surface of the sensor is certainly something I would investigate during the design phase. The whole subject of how the materials in the filter stack act as wave guides is another potential variable. Perhaps the spec we need is based on light transmission formulae rather than physical thickness. I don't claim to know every aspect of this issue, just that having a known spec that lens designers/lens techs can use as a reference should be implemented. Nailing focus on large high resolution sensors is tough enough without having to fuss with the back focus more than absolutely necessary.

The article I linked to in my OP leads me to believe that ignoring the impact of the filter stack could really bite us, especially if you typically shoot on rented (hired) lenses that may have been set for a different camera. A full prep day at a proper rental house with proper collimation gear and an experienced lens tech is certainly the ideal way to avoid such issues. That said, is there anybody who doesn't want to reduce the number and severity of potential issues?

Cheers - #19

 

[ericyoung]

Interesting. I can see how some sort of standardisation might help future lens designs. Having said that, trying to get all the camera manufacturers to come together and agree standards like this would likely take years, not least because no one knows how digital camera capture will develop and no manufacturer wants to be restricted more than necessary.

 

[Phil Holland]

I hereby respectfully request that RED make an effort to lock down the "standard" thickness of the filter stack both industry wide and in the modular system.

Different sensor designs may benefit from different thicknesses and designs of the OLPF. Not exactly sure what standardizing a thickness is going to do other than put people in a box that might not be necessary.

I do think that we might see a different "revolution" of OLPF and sensor design in the near future however. It's already happening elsewhere. It's a matter of time before it ends up in digital cinema cameras.

 

[Blair S. Paulsen]

I guess the tradition of cinema lenses being precisely set to perform optimally at a specific FFD (like 52mm for PL) is old school.

Guess I'll just have to set my lenses for my camera. When I rent glass I'll have to insist on a full prep day, and if they hire somebody who is willing to go forth without a prep day to save money I'll be rooting for them to screw up?

Ugh - #19

 

[ericyoung]

I don't see why FFD of 52mm for PL should change. Every camera manufacturer will take their own OLPF into account to set their sensor position so that lenses can stay setup for FFD of 52mm, and all the high end cameras will have user/tech adjustable sensor position adjustment as well.

Having said that I've found that every rental house has slight tolerance variations where their cameras and lenses all work fine with each other, but cross hiring there can be some mismatch.

 

[Phil Holland]

Well you guys know me by now, but what's been the most truthful thing in regards to optics on digital cinema cameras is quality glass that favors a telecentric (or telecentric-type) design tends to lead to better resolving power and image quality. Though there's a few other ones out there, spherical and aspherical, that produce impressive images just by having a more "ideal" angle of incidence that favors certain optical designs.

This is where measuring line pairs per millimeter via projector works well to inform you about the lens' potential, but not always what is true when used with a particular camera system.

Slight variables like this mean a lot of things towards your final image in reality.

 

[Lee Saxon]

Different sensor designs may benefit from different thicknesses and designs of the OLPF. Not exactly sure what standardizing a thickness is going to do other than put people in a box that might not be necessary.

I think Roger's article he linked (and maybe better this more recent one http://www.lensrentals.com/blog/2014/07/sensor-stack-thickness-part-iii-the-summary) answers this question. It's going to improve lens performance. Exceeding the thickness a lens was designed for (0mm, in the case of the designed-for-film lenses which are still pervasive especially in the motion world) by much has a profound effect on that lens' performance. The idea would be to set a standard (or maximum) thickness below this threshold.

I use manual focus film-era lenses exclusively, so I've been chatting with MaxMax.com about reducing the stack thickness of my Sony A7. They can reduce it enormously if you're willing to give up the AA filter. I didn't ask about also removing the CFA (going Monochrome) would do, but based on some of Roger's comments I think the CFA might be permanently cemented on this particular sensor.

 

[Blair S. Paulsen]

I'm with you on the importance of telecentricity Phil, I've been singing that song on RU for over 3 years.

My agenda is simply to advocate for standardization of some elements/parameters of the optical path to facilitate predictable outcomes when mounting lenses on digital cinema cameras.

I appreciate Eric's note about the camera manufacturers building in an offset that compensates for the impact of the filter stack. I'd be curious what a proper optical engineer would think of that approach. I'm not trying to make this hard, quite the opposite in fact, just looking for confirmation that such a strategy would not introduce other issues.

As more and more gigs start having mandated UHD/4K deliverables in the contract the issue of soft focus is going to a new level of importance for camera crews. The more variables we can lock down, the better chance we have of nailing focus. If there's a slight issue with back focus, then even if you hit the optimum point on the focus ring, it may not be dead tight. Even worse, it means if you are off even a hair combined with the slight softness created by the back focus error it won't be pretty.

If we really want to deliver genuine 4K resolution it will require tack sharp focusing. Period.

Cheers - #19

 

[Les Dittert]

My OLPF stack has 4 glass to air interfaces and I was concerned that it would have some reflection problems. No such problems materialized from bright bulb testing that I did.
I enjoy out resolving Dragon kits even with a cheap Canon zoom lens. If anyone wants to do some quick tests to see, bring your Dragon with your choice of glass this weekend, in Los Angeles, it would be fun !
Yes, I am brewing up a kit for MX users, before anyone asks.

 

[Patrick Tresch]

I don't get why people don't get your point Bair...

Hello Les,

Will you sell your stack in a future "Epic modular" compatible package? I m interested in your design.

Pat

 

[Les Dittert]

Yes, Modular system, swapable with no screws to deal with.

I don't get why people don't get your point Bair...

Hello Les,

Will you sell your stack in a future "Epic modular" compatible package? I m interested in your design.

Pat

 

[Blair S. Paulsen]

Thanks Patrick. I was expecting to get more support from the RU community to push forward on what I see as a critical issue.

I'm really not looking to put OLPF designers "in a box" as Phil suggests, really. I simply believe that I should be able to mount a properly collimated lens on a dozen different cameras without having to tweak back focus (or shim the mount) to get the full resolving power of the lens. I also think giving lens makers a known target will lead to greater precision/resolving power. Yes, character is more important from a storytelling perspective, but buzzed focus takes rarely make the edit. Just sayin'

Cheers - #19

 

[Bill Anderson]

Seems like an excellent idea Blair, good timing too considering how early in the game it is for cine lenses designed for sensors: that is a vast future and it would be nice to head into it with such a standard, should it be feasible.

 

[Blair S. Paulsen]

The optical path as a whole

The optical path as a whole

What really woke me up was the MTF charts in the Lens Rentals article that Roger Cicala wrote. Many of us have particular goals when picking which glass to use for a specific show, scene, look, atmosphere, mood, etc… If the filter stack is altering the MTF curve, then one has to wonder if the rendering of a particular lens will meet expectations. One could argue that selecting a particular sensor/OLPF combination is the rough equivalent of choosing a film stock so its part of the deal no matter what. From that perspective the "Swappable Era" will be even more valuable and I am excited about the possibilities.

If you think 4 screws is a PITA, how committed would you have to be to an OLPF change to also take on trimming the back focus each time as well? High resolution motion imaging capture systems continue to improve their precision. As they do, little technical issues like how the filter stack impacts lens performance and critical focus need solid, consistent solutions. BTW, lens telecentricity and how that interacts with the physical properties of a particular OLPF matter as well. If you wanted to shoot on a particular lens that had poor telecentricity would you choose a filter stack that was more forgiving?

Cheers - #19

 

[Les Dittert]

You have actually seen loss of sharpness due to slightly off backfocus ?
Is this only when attempting to focus using the scale etched on the lens?
I've always been able to get razor sharp focus with visually finding best focus, even if back focus is off.

 

[Mark L. Pederson]

I simply believe that I should be able to mount a properly collimated lens on a dozen different cameras without having to tweak back focus (or shim the mount) to get the full resolving power of the lens. I also think giving lens makers a known target will lead to greater precision/resolving power. Yes, character is more important from a storytelling perspective, but buzzed focus takes rarely make the edit. Just sayin'

Cheers - #19

Lens makers already have a known target.

If a lens is "properly collimated" it's witness marks will come up correct and there's no "buzzed focus". if the back focus of the camera is not set to the known target - you discover this in the camera checkout and adjust back-focus. It's done every day, all over the world - on all kinds of camera systems.

Do you think anyone publishing a fllter stack spec is going to change how we do a camera checkouts and we confirm lens collimation & backfocus??

 

[Blair S. Paulsen]

Not fuddy

Not fuddy

I hear you Mark, a proper prep/checkout can get a particular lens/camera combo to line up spot on. That said, what percentage of the people shooting high resolution wide aperture footage today did a rigorous checkout in a decent prep bay? How many of them will blame the camera if there are focus problems, even if their own lack of preparation is the real culprit?

Until I started reading up I didn't consider the effect of the filter stack significant enough to worry about. Like Les I would just check sharps by eye in the 1:1 zoom magnification and rock on. If Cicala and the crew at Lens Rentals hadn't published their findings I would never have guessed how much effect it could have.

I think you know me well enough by now to know that spreading fear is not my schtick, if it seemed like I was, I apologize to all. Just trying to keep our powder dry.

Cheers - #19

 

[Rolf Kestermann]

I think the days of the OLPF are numbered anyway. BMC 4k is as sharp as Epic 5K, Nikon is getting rid of theirs, any piece of glass behind a wide angle in the days of film was a no no, (nylons were OK), like any glass
in front of a tele past 100mm(unless designed for it). Also, not sure how a 'telecentric" lens design would help, It produces an orthographic picture which is useless for motion picture.