Resolution vs Sharpness: A Fresh Perspective

[Hrvoje Simic]

I don't know about focal reducers and other unusual elements introduced to the optical path...

This has nothing to do with focal reducers, imagine S35 and FF cameras.

But it IS actually possible to match geometry for different focal lengths for corresponding formats but it's not easy and you have to be super precise.

No it isn't.

These spatial geometry principles have nothing to do with framing. They have everything to do with format size.
Framing just affects how well you can notice them. And distance.

So as mentioned before (the Steve Yedlin article http://www.yedlin.net/170504.html)

Steve Yedlin doesn't address this at all in that article. He talks about perspective in general.
Zero focus on specific geometry in the shot and proportions of humans and space, so the shots are not designed to even look for it.

The one in post #97 is.


You can notice human proportion differences in this example, though.
If the shot was designed to focus on proportions the actor would stay put and you'd see the nose enlarging and more discrepancy between the size of the head and shoulders and background elements.
And if the hand was reaching towards the camera it would enlarge more with a smaller format.


Also, when some people talk about "perspective" and other about this factor in this context, which is related to perspective, limited terminology leads to cognitive dissonance and running in circles. So forget about "perspective" in this context, it will only confuse you, and think about proportions.

In addition, even when you cannot consciously notice something and rationalize what is happening, subconscious registers the subliminal very well and intuitively resonates more with what is more natural. And proportions are more natural with a larger format.

 

[Eric Lange]

This has nothing to do with focal reducers, imagine S35 and FF cameras.






No it isn't.

These spatial geometry principles have nothing to do with framing. They have everything to do with format size.
Framing just affects how well you can notice them.





Steve Yedlin doesn't address this at all in that article. He talks about perspective in general.
Zero focus on specific geometry in the shot and proportions of humans and space, so the shots are not designed to even look for it.

The one upstairs is.

You can see facial proportion differences in this example, though.
If the shot was designed to focus on proportions the actor would stay put and you'd see the nose enlarging.

Also, even when you cannot consciously observe something and rationalize what is happening your subconscious registers the subliminal and intuitively resonates more with what is more natural. And proportions are more natural with a larger format.

No it isn't.

lol...


So I can take a massive 100mm Biogon lens (symmetrical + low distortion) for a super large format film surveying camera like ZEISS UMK 1318 which has glass plate 'Literally" 130 mm x 180 mm and match the central portion of the image with a Wild P-32 (swiss made) extremely accurate also Biogon (Symetrical layout + low distortion) camera / lens-cone (with a 60 mm calibrated lens) on a format that is 90mm x 65 mm (co-incidentally) half the format size (also film camera).

Lets say we wanted to play "Franken camera" and butcher these lens cones and put the 100 mm biogon on the smaller format and the smaller Wild p-32 lens onto the lens-cone/camera back of the 130mm x 180 mm Zeiss UMK format the format sizes do not have the capability to affect how each lens handles light.

There is no magic voodoo where format size completely alters the ray path of the lens?

[ I admit there are other effects with larger formats in terms of internal reflection and how light bounces around in that chamber and off the rear element of the lens and sensor/OLPF etc. for digital systems. ]



I'm happy to re-visit this in the future with measured examples as we have to calibrate some more off the shelf lenses and we can map that all out maybe 8 months from now with measurements on optical benches and inspection slabs of the order of microns.

_____________________________________________________________________________________________________________________________________________________________________________________

I'm a huge fan of large format but some of the visual examples have more random and less controlled set ups and the desired effect sought from each random outcome is then selectively cherry picked and presented as fact (in some cases) to help flog or sell cameras... I think Phil Holland's comparison shots are pretty dead on/ really good. Post # 62.

_____________________________________________________________________________________________________________________________________________________________________________________

In the above flickering example @Hrvoje (that's nicely put together BTW) the smaller format shows with the smaller focal length in that instance seems to typify what is usually extolled as being the benefit for larger format... I.e. more 3d look and enhanced bokeh. [in that particular case he (Mr Yedlin's head) is not positioned in exactly the same place as the shots were taken at different times].

_____________________________________________________________________________________________________________________________________________________________________________________

 

[Hrvoje Simic]

There is no magic voodoo where format size completely alters the ray path of the lens?

No magic voodoo.

Basic trigonometry when the task is setup properly.

 

[Jacek Zakowicz]

Actually- since you tried to trick the viewer with zoomed in and reframed image , the nose on the smaller format is actually smaller proportionally to the size of the face by about 5%. So the smaller format (extraction) with the shallower DOF actually looks "flatter" than the large format extraction.
I think that Steve did this on purpose - perfect irony in this thread- good one Steve!!! rrr:rrr:rrr:

 

[Hrvoje Simic]

Actually- since you tried to trick the viewer with zoomed in and reframed image ,

That's not "actually", that is your wishful thinking.

 

[Eric Lange]

No magic voodoo.

Basic trigonometry when the task is setup properly.

If you / one takes the shots such that the rear nodal points of each comparative lens system intersect in three dimensional space at the same point in 3d space (and you have done your math right for selecting equivalent focal length for different format size) the shots WILL match The rear nodal point is the effective perspective center of the ray geometry for the scene and the sensor. Normally it is customary to use the projective transformation equations to compute that, as those also can take into account rotation around the nodal point of the optical system in three dimensional space. Small differences in position and orientation (tilt and swing/pan) will make stuff in the scene appear to 'Move about". Helps if you are using well calibrated high quality / well engineered primes also; zoom lenses are very very difficult to calibrate well and introduce a host of other spatial errors too.


Where people are fuffing up (slightly) here is they are not determining where the rear nodal point of the lens system really lies. If you measure from a long lens from the front vertex and then measure from the front vertex of a short lens then you are going to introduce spatial errors as the rear nodal points of each different optical system are not co-incident nor are they necessarily at the same position in 3d space relative to the scene. This is where the physical length of the different lenses can make a real difference especially for small scenes of close range shots/work.

Another source of error is if you match the position of the film plane on the camera... That's also a minor flump - up... As that does not necessarily correctly align the perspective centers (rear nodal points) for each different comparison shot either.

A lot of the manufacturer's data on where these nodal points basically reside are usually quoted for infinity focus so your principal distance (distance between focused film plane and rear nodal point of lens) will be different depending on where you decide to focus to. ... That decision of where to focus too may also be making some of the differences here too (in these more general set ups) other than minor alterations in optical layout, for example if one was comparing a Distagon layout with another Distagon layout of different focal lengths matched to different format size.

Take a close look at Phil's shots on post number #62 there are subtle differences but for a quick test those are pretty damn well matched without having to go to a three orders of magnitude higher precision and accuracy for a particular set up.

 

[Hrvoje Simic]

Take a close look at Phil's shots there are subtle differences but for a quick test those are pretty damn well matched without having to go to a three orders of magnitude higher precision and accuracy for a particular set up.

Page 6, post 55.

 

[Eric Lange]

How to sell larger formats ?

How to sell larger formats ?

I feel we have gone round the mulberry bush 50 times on certain themes.

Michael Cioni made a really interesting point that rarely gets mentioned. [I definitely have a vested interest in a dynamic larger format, but my own application needs are not exactly main stream right now, but more good/well resolved pixels per shot is a definite need and want.].

So Michael mentioned the concept of increased precision (spatial precision). In other words for having more well resolved line pairs / better MTFs that is possible to capture finer spatial features in a single shot. So therefore this brings about greater precision and potentially greater spatial accuracy in two dimensions as well as for three dimensional work. So I thought that was a VERY good point to make.

But that's where the notion of various compression schemes comes in and the possibility of certain (possibly) invisible compression artifacts to subtly manipulate the spatial locations of certain pixels versus corresponding spatial positions in "Object space" or the scene being imaged. I am VERY impressed with some of the resolution target tests that have been carried out for VV.

Back in the early RED ONE days I think there were a few moans or queries about difficult chroma key subjects and compression patterns near stark borders, has a lot of that been resolved or cleaned up over the years..? And I am assuming VV is much better for that? Than lets say sensors with Lower K from different manufacturers.

___________________________________________________________________________________________________________________________________________________________________________________________

If anybody develops a dynamic system with half decent frame rate or less at 50MP with a large sensor I'm totally jumping on that train. So fingers crossed maybe that Panavision does indeed pursue higher bandwidth/higher photo site/ larger sensor based systems. ... But looking at the other thread about the Alexa LF 4k full frame ... It seems kinda depressing lol.

 

[Eric Lange]

However it hardly seems like a significant difference, it’s not like one screams “large format” at you compared to the other. I can’t imagine someone saying that they want their movie to look like one of those but not the other.

+1 / exactly...

It just seems odd to me that a lot of time the "wordy words" somehow don't match the extremely obvious pictures.

The one thing that really really impresses me about the film industry and in particular folks that work in that industry is that they are (on the whole IMO/IME) extremely practical and very pragmatic. A lot of the time there is not a LOT of time just to be "Thinking about" stuff rather than just "DOING" Production and shooting schedules are absolutely relentless and it's a miracle stuff gets done / executed and released in the time frames they do. Not everybody can take that or do that on a consistent basis without loosing their sh*t On rare occasions there are some really interesting development/pre-production work where a certain look might get tested, but on the whole some of these decisions are made very quickly.

_______________________________

The thing that was interesting was Michael Cioni in his presentation at approx. 55 minutes in said he used the same lens...on the smaller format as he did on the larger format. But it was pretty clear he pulled in/dollied in a bit as Mr Tiemann also pointed out (looking at edge of lap top in frame)... Hard to say if there is certain amount of artistic "License" to visually communicate an idea. But I'm still scratching my head as to how best one would "sell" people on Larger format in general and personally I think it has more to do with deeper digital capability than certain "Optical" arguments (Outside of the possibility better MTFs and possibility of better engineering).

 

[Aaron Lochert]

An experiment for you all...

Open each of these in a new tab so you can switch back and forth and have them in the same place.

Which of these two is the larger format?

 

[Eric Lange]

An experiment for you all...

Open each of these in a new tab so you can switch back and forth and have them in the same place.

Which of these two is the larger format?

What are we choosing between ... ? Just for a clue... Also helps to gauge where T/f stops fall.

 

[Misha Engel]

That is impossible to say.

 

[Michael Lindsay]

Unless you are tricking us?.... the 1st one...

 

[Eric Lange]

Unless you are tricking us?.... the 1st one...

Yeah I was wondering about whether this was "Trick question" too...

But looking at these images I'm thinking much more about the lenses than the format. So to my eye where the bokeh kinds breaks up/ fringes and fragments in the background seems to be a marginally lower quality lens in the top view, and yet in the top frame the foreground with the King/chess piece looks smoother / more pleasing than the foreground in the bottom image...

Whereas the Bokeh in the bottom image looks a little smoother possibly nicer lens.

Not much between them to write home about YKWIM?


Also wondering how it was graded as there is apparent almost film grain added/speckle/possible sharpening in the bottom image ? And that may be more to do with the sensor or processing/grading.

So I would say maybeeeeee the Top image has more photosites / inch of final output as the image is smoother, and the bottom image seems more "Choppy" in terms of texture and micro contrast (for want of a better word).

________________________________________________________________________________________________________________________________________________________________________________________

It's not a massive stretch of the imagination to say those images could be created by the same sensor but at different levels of compression or different compression and coding schemes YKWIM?

________________________________________________________________________________________________________________________________________________________________________________________

Is the bottom shot from an I-phone or mobile phone?

 

[Jacek Zakowicz]

Regardless of the choice, if it is so close, the point of "large format perspective compression" advantage is quite moot......

 

[Eric Lange]

Regardless of the choice, if it is so close, the point of "large format perspective compression" advantage is quite moot......

Aye...

But that's where the aperture argument 'Kicks in". I.e. Relative aperture versus actual physical diameter of the aperture.

So for a 24 mmm wide sensor versus sensors that are 40 mm ++ wide, generally the lenses/glass for equally fast lenses on each format... the lenses specifically designed for a 60 mm image circle may have a correspondingly larger diameter for being wide open, versus the physical size of aperture wide open for a lens that is specifically designed for a sensor that is only 24 mm wide. It's still possible to have considerable functional overlap between the two formats.

In the Yedlin article he sates that comities decide the diameter of the circle of confusion/ blurr (which in part is true), but the equations we used to use you can actually determine what you want the size of the "circle of confusion" or " blurr circles" to be and work around that.

So for Bokeh I think it still comes back to physical size of aperture rather than relative size of aperture like your f numbers and related T stops.

_____________________________________________________________________________________________________________________________________________________________________________

Long story short, it's the physical size of the largest aperture not the size of the format for Bokeh/Blurr circles. It just so happens (on medium format cameras that have lenses specifically designed for them, the physical size of the aperture is going to be larger than lets say the diameter of an aperture for a lens on a video camera).

 

[Jacek Zakowicz]

I'm not sure I understand- the circle of confusion, to be on the safe side- could be the size of the pixel as the pixel is the smallest light gathering entity in the sensor. With the blurr (OLFP) filters it could probably be even bigger. (but then again- some cameras do not use OLFP filters) There fore, for different formats carrying the same resolution, the COC should ideally vary...

 

[Eric Lange]

I'm not sure I understand- the circle of confusion, to be on the safe side- could be the size of the pixel as the pixel is the smallest light gathering entity in the sensor. With the blurr (OLFP) filters it could probably be even bigger. (but then again- some cameras do not use OLFP filters) There fore, for different formats carrying the same resolution, the COC should ideally vary...

Yup that's right.

So again MTFs for each element of a system can be really important.

When I started out with digital medium format in the early 90's photosites we had were B/W (and non-bayered) had to use filter wheels with Kodak wratten gel filters for each channel. The photo sites were between 9 to 15 micron (peltier cooled sensor). Not super challenging to find suitable lenses for that.

Nowadays with photo sites of the order of 3 micron and bayer pattern that can tax the resolution limits of what a lens can deliver. (In my case we need wider angle lenses of the order of 90 to 110 degrees (HFOV))... Longer lenses + variants of telecentric lenses can serve smaller pixels /photo sites p[retty well (as you know Jacek).

That does raise the question with a larger sized sensor with larger photo sites that lens wise it can be easier to hit certain goals for MTFs (like the Arri large format sensor 65-ish) IF you have glass that genuinely serves that image circle but the overall number of resolvable elements in the object field may be lower than other sensors that have a high total number of photo-sites.

On the other hand I think Michael Cioni is also making the argument that more pixels per blurr circle (He shows simulated lense coma diagrams... interesting), can actually do a really good job. I agree with that but I also think more photosites per "Blurr circle" should be able to achieve better color accuracy, but I think Michael is trying to show smoother and better computed gradients after de-bayering? So I think that's why Michael has a graphic showing 'Effective 2.5 micron" sensor... Haven't jumped down that rabbit hole quite yet.

Personally I still think the stuff that Phil Holland did with the Zeiss OTUS and their special formulations and special dispersions are near miraculous.

____________________________________________________________________________________________________________________________________________________________________________________

http://it-book.org/pdf/modulation-transfer-function

^^^This free download book "Modulation Transfer Function In Optical And Electro Optical Systems"* by Glen D Boreman is a pretty quick read and pretty understandable for the non-mathematically inclined too. In the book Combined system MTF is pushed pretty hard. [This volume can be downloaded elsewhere too if you google about. ].


* no affiliation.

 

[Hrvoje Simic]

That is impossible to say.

No it isn't if one has been paying attention in this thread.





Nice shots Aaron, thanks for sharing.

 

[Tom Gleeson]

Aaron,

Interesting comparison as there are clear geometric differences between the images. Image 28403996899_c72049c0b6_o.jpg looks like a flatter perspective in comparison. I assume you attempted to keep the nodal point of each lens in the same place and compensated with focal length and stop? I can see a small visual displacement difference between the the first and second pawn that indicates the two lenses were just a little off? In my test I was surprised by how different the lenses looked due to the barrel distortion or mapping distortion each lens and focal length brings to the image. Not being pedantic but I think in these sort of tests both lenses should be mapped and a correction applied in post to remove this variable before comparing.