VistaVision and Monstro and cinematography?

[Nick Morrison]

Dan & Patrick, 100% agree this GIF from Wikipedia incredibly helpful:

As is the accompanying text from wikipedia's entry about perspective distortion

In photography and cinematography, perspective distortion is a warping or transformation of an object and its surrounding area that differs significantly from what the object would look like with a normal focal length, due to the relative scale of nearby and distant features. Perspective distortion is determined by the relative distances at which the image is captured and viewed, and is due to the angle of view of the image (as captured) being either wider or narrower than the angle of view at which the image is viewed, hence the apparent relative distances differing from what is expected. Related to this concept is axial magnification -- the perceived depth of objects at a given magnification.

What I love is the distinction btw ANGLE OF VIEW and FIELD OF VIEW.

In the example of the GIF, the Field of View remains the same - of the house. However, as the ANGLE OF VIEW changes (by going from wider to tighter lenses, and by starting close snd then moving back), so does the perspective distortion.

More specifically, the blurb under the GIF goes into greater detail:

Simulation showing how adjusting the angle of view of a camera, while varying the camera distance and keeping the object in frame, results in vastly differing images. At narrow angles and large distances, light rays are nearly parallel, resulting in a "flattened" image. At wide angles and short distances, the object appears foreshortened or distorted.

This to me is key, and why Monstro is "special" and felt "muscular" when we tested it. By always using longer focal lengths (ie tighter angle of view) for our standard FOV's, we suddenly had more parallel lines (what Dan Sasaki would say are more ortogonal lines), a more flattened image, and of course...more fall off.

Perhaps what we didn't realize is we were seeing a more "natural" image for the first time - again, to invoke Dan Sasaki.

 

[Karim D. Ghantous]

This to me is key, and why Monstro is "special" and felt "muscular" when we tested it. By always using longer focal lengths (ie tighter angle of view) for our standard FOV's, we suddenly had more parallel lines (what Dan Sasaki would say are more ortogonal lines), a more flattened image, and of course...more fall off.

Unless I have misinterpreted what you are saying, it seems that you have misunderstood what is going on. I'm not having a go at your or anyone - I'm just saying that you seem to have misunderstood a certain principle (which will not make you a worse DP, by any means). You quoted this: "At narrow angles and large distances, light rays are nearly parallel". No, the light rays are not more parallel. That's not a scientific statement. I really should get off my lazy butt and draw a diagram.

 

[Nick Morrison]

Unless I have misinterpreted what you are saying, it seems that you have misunderstood what is going on. I'm not having a go at your or anyone - I'm just saying that you seem to have misunderstood a certain principle (which will not make you a worse DP, by any means). You quoted this: "At narrow angles and large distances, light rays are nearly parallel". No, the light rays are not more parallel. That's not a scientific statement. I really should get off my lazy butt and draw a diagram.

Karim, I'm not a DP, so it's OK.

Second, that quote is from Wikipedia, not me.

And thirdly what it's referering to is something simple, that I'm sure you can agree on, no?

If you take a CU portrait of a model with a 16mm prime, and match that FOV exactly with a 300mm prime, the two shots will have the same "frame" but look completely different - she'll be distorted in the 16, and flattened in the 300.

Why? Because with the 16mm prime, you'll be super close to her, putting her in the foreground, with everything else thrown way behind her in background. Being this close to a wide angle lens will also subject her to the 16's higher level of distortion, and it's extreme angle of view.

Whereas with the 300mm you'd need to be another 10-15 feet further back, in effect compressing that 10-15 feet of foreground, and effectively placing her in the mid-ground. The background, proportionately, would feel much closer to her than in the 16mm, by virtue of the model and background together being so far away from the camera.

I think most people can agree to the above phenomena. Wide lenses distort. Long lenses don't. Using a sensor that encourages longer lenses, means you'll generally have footage that feels...less distorted.

I don't get what the issue here is.

I think what may be going is, as Dan pointed out, is an issue of nomenclature. I think people are using "Persepctive" and "FOV" as the same thing, not taking into account that while FOV is constant (say a CU on a model), perspective can change by virtue of how close or far the camera is, and what lens is being used.

 

[Aaron Lochert]

From that very same Wikipedia entry:

Note that linear perspective changes are caused by distance, not by the lens per se – two shots of the same scene from the same distance will exhibit identical perspective geometry, regardless of lens used.

 

[Nick Morrison]

From that very same Wikipedia entry:

Right, and this makes total sense. On Monstro, you're either using longer lenses from the same distance as you were on s35, but taking advantage of their greater magnification and shallower depth of field.

Or you're adjusting your position and getting new perspectives out of old lenses. You'll need to step in a foot or two on your 85mm prime to get the same tight FOV as you did before on s35 - and this will have a noticeable effect on your image as your subject comes more into the foreground.

Filming on Monstro also has affects on wider FOV. In our studio, intead of using an 11mm on s35 to get the full view of the space, the Monstro could accomplish the same on an 18mm prime, which made a noticeable improvement to the experience as the 18mm has so much less distortion in it's optical design.

 

[Timo Alatalkkari]

More mess to the discusion, LOL... A good article about the benefits of a big sensor: https://www.redsharknews.com/production/item/5079-why-large-format-14k-is-the-perfect-future

 

[Aaron Lochert]

which made a noticeable improvement to the experience as the 18mm has so much less distortion in it's optical design.

So then we're not talking about perspective distortion at all. We're talking about barrel/pincushion distortion? I think that's why everyone is in a fury here: the word "distortion" has several uses. So we're saying the magic of Monstro is that it will yield less of the top image and more of the bottom example? We'll achieve more rectilinear images with Monstro?

If that's what we're discussing then I'm extra confused because I think the barrel distortion in the first Panavision anamorphic example above is what makes that image pleasing to me; I like how the edges of the poster aren't straight. The spherical has strong, straight angles that feel clinical. Not to mention volume deformation which makes his face fatter since it is near the edge.

I'm also a bit confused by the term orthogonal. Because if any of you have ever worked in 3D modeling, you'd know that orthographic views lack any leading lines at all, everything is parallel:

And the only way to achieve that is move your camera away as far as possible, in fact, to infinity. I would hope an orthographic view isn't what we're aiming for, anyway. Some of my favorite shots from the Revenant are because the trees lean into the frame from the left and right sides, this happens a lot in the movie:

I also wanted to take the time and say thanks to everyone's input thus far. I know it may seem like I've got a bone to pick, but this is an intriguing discussion and I'm learning quite a bit.

 

[Nick Morrison]

All good Aaron, I'm digging this convo too.

I think maybe what's confusing is we're talking about 4 different formats at once:

1) Spherical s35
2) Anamorphic
3) Spherical VV
4) Spherical 65mm

The optical realities of each format also determine their look, not merely their large FOV.

For example, it's my impression that high end s35 primes are highly corrected for distortion, etc - even when compared to many full frame photogaphy lenses. So this may explain why some spherical wides still look "good" when compared to say anamorphic.

Anamorphic may have a wider FOV, but is well known for it's optical abberations (which is why David Fincher for example HATES it).

Vista Vision/Monstro has similar FOV to anamorphic, however presumably much of the spherical glass that covers should have much less "funk" than anamorphic. This format should indeed feel more rectilinear, especially as new lens designs get more and more corrected.

Finally 65mm is a whole other beast. To cover such a wide image circle lenses have their own challenges - many are slow (f4, etc) and not particularly sharp (they don't have to be). That example you posted is confusing - it was shot I believe on a 30 or 35mm rectilinear lens, which is close to a 10mm in s35 - the FOV is insanely wide, and when paired with that massive sensor it does wild things as the background is STILL COMPRESSED even though the FOV is mega wide.

So each format seems to open up new creative possibilities from the glass you can now use - and I think thats what makes all of this so exciting.

However what binds anamorphic and all the lage formats together is the reality they generally encourage you to use longer focal lengths than in s35.

Lets go back to your Revenant example there. That super wide angle lens thats pushing that sensor to the limit is still in the 30mm range. Think about that for a sec. That's a normal focal length in s35, the most natural, and here its bonkers. And in s16 this focal length would be a tight!

So I think the FOV of each format, paired with the type of glass each will encourage you to use, is what brings out the magic of each platform.

 

[Michael Tiemann]

Dan Kanes gets a gold star!

[...]
Now - the key (to me) here is that the NODAL point of a lens can be DRASTICALLY different.

THAT is what explains one of the major differences both in the Panavision anamorphic videos (Magnification and Perspective as well as Bokeh). Namely, the 35mm Primo Anamorphic is 29.2cm long, whereas the spherical 17.5mm Prime is 10.5cm. That 19.3mm difference is likely a substantial fraction of lens-to-subject distance. For the metrically challenged, that's almost 8". Shifting the lens 8" (not the camera, mind you, which can remain locked down, but the front element of the lens) makes a HUGE difference in cinematography's "game of inches". Other factors that come into play:

• Model leaning toward/away from center (notice overlap of shoulder and light pattern)
• Angle of light pattern (due to new location of front element with respect to scene perspective)
• Barrel distortion vs Rectilinear distortion (round faces magically stay round in corners instead of warping toward the corner)
• Depth of field (already mentioned)

Anamorphic does have some cool features--bokeh for sure, as well as much shallower depth of field for a given field of view and aperture. But two dimensions of magnification leading to perspective changes--I think nodal distance is the key. Software focus rolloff--that's just a function of the de-squeeze process working in concert with our eye's tendency to overly focus on vertical lines. Flares? That's a matter of personal preference/taste. And of course the fact that the anamorphic aspect ratio is closer to a square than a 2:1 rectangle means that the de-squeeze process better preserves the integrity/consistency of barrel distortion--should that be a feature rather than a bug. (When shooting 2:1 or 2.4:1 spherically, the extremes of barrel distortion are cropped off top and bottom. When shooting square or 6:5 the circle remains unbroken.) I am watching with interest the progress of Atlas lenses.

I have a Fujinon Premier Pro 24-180 zoom lens. It's length is 40.5 cm. When shooting a chart from 20' away, it's just like using my TS-E 24L. When shooting scenes with distinct fore-, mid-, and backgrounds, it's perspective is RADICALLY different than my TS-E 24L, despite whatever "perspective corrections" I can make with the Canon lens.

When subjects are far enough away that the nodal point differences are immaterial, then a FOV/Crop sensor tool (such as Phil Holland's) work perfectly. If the lens is literally going to intrude on the scene, then everything's case-by-case, and a good DOP is going to adjust all the variables she can to get the best shot she can. And if she switches lenses, she's going to recompose as necessary, moving the camera, the lens, and possibly the subject--to suit.

 

[Nick Morrison]

Dan Kanes gets a gold star!



THAT is what explains one of the major differences both in the Panavision anamorphic videos (Magnification and Perspective as well as Bokeh). Namely, the 35mm Primo Anamorphic is 29.2cm long, whereas the spherical 17.5mm Prime is 10.5cm. That 19.3mm difference is likely a substantial fraction of lens-to-subject distance. For the metrically challenged, that's almost 8". Shifting the lens 8" (not the camera, mind you, which can remain locked down, but the front element of the lens) makes a HUGE difference in cinematography's "game of inches". Other factors that come into play:

• Model leaning toward/away from center (notice overlap of shoulder and light pattern)
• Angle of light pattern (due to new location of front element with respect to scene perspective)
• Barrel distortion vs Rectilinear distortion (round faces magically stay round in corners instead of warping toward the corner)
• Depth of field (already mentioned)

Anamorphic does have some cool features--bokeh for sure, as well as much shallower depth of field for a given field of view and aperture. But two dimensions of magnification leading to perspective changes--I think nodal distance is the key. Software focus rolloff--that's just a function of the de-squeeze process working in concert with our eye's tendency to overly focus on vertical lines. Flares? That's a matter of personal preference/taste. And of course the fact that the anamorphic aspect ratio is closer to a square than a 2:1 rectangle means that the de-squeeze process better preserves the integrity/consistency of barrel distortion--should that be a feature rather than a bug. (When shooting 2:1 or 2.4:1 spherically, the extremes of barrel distortion are cropped off top and bottom. When shooting square or 6:5 the circle remains unbroken.) I am watching with interest the progress of Atlas lenses.

I have a Fujinon Premier Pro 24-180 zoom lens. It's length is 40.5 cm. When shooting a chart from 20' away, it's just like using my TS-E 24L. When shooting scenes with distinct fore-, mid-, and backgrounds, it's perspective is RADICALLY different than my TS-E 24L, despite whatever "perspective corrections" I can make with the Canon lens.

When subjects are far enough away that the nodal point differences are immaterial, then a FOV/Crop sensor tool (such as Phil Holland's) work perfectly. If the lens is literally going to intrude on the scene, then everything's case-by-case, and a good DOP is going to adjust all the variables she can to get the best shot she can. And if she switches lenses, she's going to recompose as necessary, moving the camera, the lens, and possibly the subject--to suit.

Michael this is fascinating and awesome. I've been learning A LOT of new things the past several days, which I much appreciate.

 

[Aaron Lochert]

That example you posted is confusing - it was shot I believe on a 30 or 35mm rectilinear lens, which is close to a 10mm in s35 - the FOV is insanely wide, and when paired with that massive sensor it does wild things as the background is STILL COMPRESSED even though the FOV is mega wide.

See this is where I have problems, because the notion still persists that the focal length is doing anything to how perspective is represented.

I just shot this. Zeiss Contax 50mm f/1.4 @1.4 and ISO 250 at 4K resolution on Epic-W. The camera is 40 inches from the salt shaker and the pepper shaker in the background is 18" from it. Then without moving anything, I switched to 8K resolution, 100mm and f/2.8:

When lined up (one of the lenses is not exactly the focal length it reports), you'll see that the same perspective, the same DoF, and the same noise characteristics are all possible to recreate across both formats. So where is the 3D voodoo of the larger 8K format over the smaller 4K format? And keep in mind, this is with a 2X crop. Monstro is only 1.37 bigger than Helium, so even if there were any differences going on, they'll be even less apparent than what I just showed here.

What larger formats gets us is allow lens designers to give us wider fields of view at shallower depths of field than previously possible. For example, the 18mm Tokina at T1.5 on Monstro is going to have equivalent DoF and FoV to using a 13.5mm T1.1 on Dragon and a 11mm T0.91 on S35 film, that's nuts.

The funny thing, while we're on the subject, is that Alexa 65 is to Monstro as Monstro is to 6K Dragon, about 1.3 times larger width. So if Monstro is more magical than 6K Dragon, then Alexa 65 is just as magical when compared to Monstro. However; we really aren't going to see the benefits of shallow DoF or FoV increases unless we use a single lens, the 14mm Panavision Primo 70 at T3.1, which is equivalent to 10.6mm t2.3 on Monstro (the Laowa 12mm not far off, but still narrower and slower). The next widest, fastest lens for Alexa 65 is the Primo 70 24mm T2, which is the same as shooting Monstro 18mm at T1.5 - conveniently the same as what Tokina makes. So Alexa 65 is only magical when shooting on the 14mm. If shooting any of the other focal lengths, you might as well just shoot Monstro at T1.5 on the Tokinas.

And we can go even further with that logic. Because Master Primes exist at T1.3 as wide as 12mm, then Monstro isn't any more magical than S35 until you go wider than 20mm or open up faster than T2.1.

Of course, this is not to discount inherent lens qualities, like as can be seen in my example, chromatic aberrations, barrel distortion which bubble the middle of the frame out, giving the illusion that the background is closer, etc. etc. And also we're ignoring better sensor design in this process as well (Monstro is really impressive in this regard). But as far as 3D representation of the scene and subject-to-background size, the focal length and sensor size do nothing. This is literally the reason why we can define what a "normal" focal length is for a given format. If the space was somehow more compressed on a 60mm on VV than a 35mm on S35, then we couldn't define "normal."

So what matters is lens design, which is what has always mattered. And larger formats can be a bridge to that goal - that's the "magic."

The nodal point discussion is an interesting one. If any of us have tried to do a pan with no parallax or stitch a panorama together, you would have noticed that you need to center your pan axis around the entrance pupil.

So this gets me thinking, since we know perspective changes with subject-to-camera distance, is it safe to say that the true definition here is nodal point-to-subject distance? Can we negate these effects by re-positioning the camera forward or backward to compensate for different nodal points?

 

[Karim D. Ghantous]

the FOV is insanely wide, and when paired with that massive sensor it does wild things as the background is STILL COMPRESSED even though the FOV is mega wide.

That's not correct. See Aaron's examples above.

8K/VV does have advantages as far as image quality goes. No question there. 8K captures finer distinctions in colour and tonality. Surface area usually does matter. Look at how awful a photo from a new iPhone looks compared to an older Micro 4/3 sensor. Big difference.

 

[Nick Morrison]

That's not correct. See Aaron's examples above.

8K/VV does have advantages as far as image quality goes. No question there. 8K captures finer distinctions in colour and tonality. Surface area usually does matter. Look at how awful a photo from a new iPhone looks compared to an older Micro 4/3 sensor. Big difference.

I think you're confusing advances in technology with differences btw negative/sensor sizes.

S16/4K Helium smokes an old 5D MKI, but that doesn't change the fact that the helium @ 4K will be using lenses with a much, much smaller image circle - and this also has an impact on the look of the image.

What we're talking about here in this thread is the consequences of Monstro requiring larger image circle lenses - and what that does to its photography.

 

[Jacek Zakowicz]

I think you're confusing advances in technology with differences btw negative/sensor sizes.

S16/4K Helium smokes an old 5D MKI, but that doesn't change the fact that the helium @ 4K will be using lenses with a much, much smaller image circle - and this also has an impact on the look of the image.

What we're talking about here in this thread is the consequences of Monstro requiring larger image circle lenses - and what that does to its photography.

The only look of the image that's different is the depth of field. That's only because there are very fast FF lenses made for film still photography. David explained that and Aaron demonstrated it. That's it, the only difference.
And the puck stops there because there are no fast medium format lenses so FF offers the shallowest depth of field , practically speaking. Not to mention the practical implications of trying to keep subject in focus with razor thin depth of field available.:coolgleamA:
The only other advantage may be less noise due to bigger pixels (all things being equal) but, with more advanced tech like 10nm processors that advantage is offset by the tech that provides better s/n ratio, so no advantage there practically speaking...
What else is there?

 

[PatrickFaith]

More mess to the discusion, LOL... A good article about the benefits of a big sensor: https://www.redsharknews.com/production/item/5079-why-large-format-14k-is-the-perfect-future

This article is very similiar to what I'm finding with my testing:

"Vertovec went on to explain how high-resolution acquisition provides more opportunities in post production. He described an 8/7/6 workflow: 8K capture with a 7K extraction area within that, downscaled to 6K for mastering. Footage from David Fincher’s Mindhunter series was shown, revealing the extra image outside the 7K extraction. This was used for the camera stabilisation which Fincher wanted to be applied to almost every shot and digital bowing to simulate the look of anamorphic glass."

So as I'm shooting with the monstro but with arri master prime anamorphic, I'm starting at 10k in compositing ( 4320 height with a width in compositing of 10512= 4320*6/5*2), with a 9k to 6k extraction depending on framing needs, which then goes to a 4k DI. Having this virtual 50mm backing width is also pretty sweet surprise (I noticed with some of the panavision DXL shots, they must be doing something very similiar but probably not extracting so heavily).

Dan & Patrick, 100% agree this GIF from Wikipedia incredibly helpful:

...
This to me is key, and why Monstro is "special" and felt "muscular" when we tested it. By always using longer focal lengths (ie tighter angle of view) for our standard FOV's, we suddenly had more parallel lines (what Dan Sasaki would say are more ortogonal lines), a more flattened image, and of course...more fall off.

Perhaps what we didn't realize is we were seeing a more "natural" image for the first time - again, to invoke Dan Sasaki.

I'm seeing exactly this with my setup too. Another thing like I commented before, is this is looking especially nice on medium shots that need the context of a wide shot(Bjorn nailed those examples with his westerns shots at the start of this thread). Also exactly like Dan Sasaki commented, this allows me to move the distance between the camera and the face so I can get the facial angles I'm looking for to tell the story. Using this squishy square as is shown in the gif, I can either make the person better looking or worse looking by just changing the distance between the camera and the focus/face, then crop to what ever I need - be it a close/medium/medium-wide shot. Again I do this purely by moving the camera toward or away from the face, and what the large backing is giving me is enough pixels to frame whatever I need to tell the story(again also nuthing to do with the T-stop). The arri master prime anamorphic is near perfect lense, so this "flexing" the face pretty much exclusevly based on the distance and angle from the camera to the face that I want. I btw just don't see how anyone can debate this, but a lot of people are thinking to hard about this with starting with the crop &t-stop (which have nuthing to do with this effect), so they are not flexing the face by adjusting camera distance to the subject to get optimal facial ratios.

 

[Nick Morrison]

The only look of the image that's different is the depth of field.

Perhaps in a lab, that's correct, but not in the real world where you have to use lenses - and lenses themselves have ADDITIONAL properties besides depth of field, namely distortion.

On s16mm, a normal lens is a 16mm.
On s35, its a 35mm prime
On VV, it's a 50mm prime
On Imax, it's an 80mm

All of these formats, on a normal lens, at the same distance, will effectively give you a similar FOV on a CU of a model - however as the lenses get longer, the distortion gets smaller and smaller, till by the time you get to the 50mm and 80mm there's almost no distortion at all, leaving our model's face looking the flatest.

This...is one of the advantages of large format photography, the more natural, undistorted imagery given to it by longer lenses.

 

[Jacek Zakowicz]

Well I've projected thousands of lenses and , believe me , distortion on 16mm format lenses is no different than on 35mm format lenses. Definitely it is not distortion that makes them different...

 

[Nick Morrison]

Well I've projected thousands of lenses and , believe me , distortion on 16mm format lenses is no different than on 35mm format lenses. Definitely it is not distortion that makes them different...

Yes, s16 and s35 primes sets have similar amounts of distortion because both feature high numbers of highly corrected wide angle designs.

However, once we get into large format photogtraphy (the focus of this thread), things start to change.

Lets take the "normal" lens for all four primary formats (s16mm, s35mm, Vista Vision, and 65mm) and look at their levels if distortion shall we?

To glean distortion numbers, we'll pull from the Zeiss download center, a treasure trove of techical info, and use the best of Zeiss's Milvus line as our baseline:

https://www.zeiss.com/camera-lenses/us/service/download-center/datasheets/milvus-lenses.html

Super 16
Normal Lens is a 16mm prime
Distortion of 15mm 2.8 Milvus (considered best in class): -2%

Super 35
Normal lens is a 35mm prime
Distotion of Batis 35mm 1.4:
-1.8%

Vista Vision
Normal lens is a 50mm
Distortion of Milvus 50mm Macro (highly regarded too):0.5%

65mm
Normal lens is an 80mm
Distortion of Milvus 85mm 1.4: +0.3%

So there you have it, using normal lense here are our distortion levels by format, using Milvus glass as baseline:

s16=-2.0%
s35=-1.8%
VV=-0.5%
65=+0.3%

I rest my case.

 

[Jacek Zakowicz]

Haha that's a nice spin. You are showing distortion numbers for a 35mm format lens and apply them to 16mm frame? These numbers do not apply nfortunately. The distortion changes as you change the radius . Not to mention the fact that whether is .5% or 2% distortion it has zero effect on face features. I suspect you will not see that difference at all.
BTW I heard CNN was hiring haha...

 

[Dan Kanes]

Show us the salt shaker with a non-orthographic shot.