PDA

View Full Version : Understanding entrance and exit pupils, camera design, and image distortion



Theodoros
13-May-2016, 00:45
MODERATOR'S NOTE: Discussion has been moved from this thread:

http://www.largeformatphotography.info/forum/showthread.php?130626-Looking-for-feedback-on-the-Arca-Swiss-4x5-F-Metric

__________________________________________________________________________________________________________________________________


I guess I'm not seeing the "distortion" to which is being referred. The need on a front standard is the ability to place the lens axis at any desired position, and at any desired 3-degree direction. The Arca, and similarly designed cameras, enable this capability.

I directly compared a Sinar Norma standard (base tilt above shift/swing) with a Sinar F (base tilt below shift/swing), and the worst I observe is that the latter rotates the lens a bit. But obviously, this does not cause distortion of the image, since lenses are axially symmetric. (Round.)

As for whether or not tilt is on axis, or at the base, it's always possible to use the focus knob in conjunction with other movements, to achieve the same position/orientation of the lens axis with base tilt that can be achieved with axis tilt.

Hi Neil... There is a certain point on the lens Axis around which all rotations of the Axis should happen (it being constant with respect to tilts/swings). This point is the exact position of the lens entrance pupil, the purpose of a view camera is to shift/side shift/swing/tilt with respect to the entrance pupil, not the axis of the lens. Think of it as a sphere that the image area intersects, if the entrance pupil is not exactly at the centre of the sphere, the shape of the intersection is not circular, but elliptical... This is not me saying... it is physics.

In other words, you will still have the axis of the lens penetrating the sphere by its centre, but while the image area plane will be an intersection by the centre of the sphere, the image projected by the lens will have a different centre (on the same axis) where the entrance pupil will be repositioned... This will cause the intersection not to be of circular shape (as it would be if the entrance pupil was positioned exactly at the centre of the where) but elliptical.

Let me give you a practical example (which will make it - I think - easier to understand). Say that you are using a recessed board on your camera, now by doing this, you retain the lens axis constant, but you've moved the entrance pupil of the lens backwards... If you now tilt the lens, the entrance pupil will drop with respect to the image area's centre... The image area will then record the intersection plane of a (hypothetical) sphere, the centre of which is where the original lens entrance pupil position would be, if no recessed board would be used. But the lens will project towards the image area a cone which is part of a different sphere, which now has a new centre, that being the position of the entrance pupil of the actual lens used on the recessed board. Thus the image area will intercept a plane that is of elliptical shape.

B.S.Kumar
13-May-2016, 06:04
Theodoros, just curious. Are all lenses designed so that their entrance pupils are at the correct position when they are mounted on flat boards from different manufacturers? Some are thicker than others, so would that make a difference? Also, concerning recessed boards, even Sinar makes such boards for wide angle lenses. Are Sinar boards designed to ensure that the advantage of using a P2 is not lost? Some lenses simply cannot be used with a flat board.

Kumar

Theodoros
13-May-2016, 10:49
Theodoros, just curious. Are all lenses designed so that their entrance pupils are at the correct position when they are mounted on flat boards from different manufacturers? Some are thicker than others, so would that make a difference? Also, concerning recessed boards, even Sinar makes such boards for wide angle lenses. Are Sinar boards designed to ensure that the advantage of using a P2 is not lost? Some lenses simply cannot be used with a flat board.

Kumar

Hi B.S., all good points that need to be discussed extensively.

Most (if not all) lenses that have been exclusively designed to be used on a view camera, meaning lenses that are of two cells (front and rear) despite if they are of symmetrical or of retrofocus design, have (to my knowledge) their entrance pupil positioned exactly where their mounting plane is. Most (if not all) other lenses that have been originally designed as to be used on a different camera, (meaning most lenses with focusing mechanism on them) have a variable entrance pupil (some people call it nodal point) the position of which varies with the focusing distance... If they are used on a view camera (like some use Mamiya RZ or Hasselblad V lenses) they are always focused to infinity and thus the entrance pupil doesn't vary with the focusing distance. However, the entrance pupil on these later lenses is never on their mounting plane and can be several centimetres from it... (from 3cm up to 7cm... even more sometimes).

The entrance pupil position of a lens, can be easily found by the user himself with a simple process... (put two bars 10m apart from each other, then the camera another 10m apart, and then swing the camera right and left and adjust the lens until the bar at 20m is invisible by the lens, hidden behind the 10m bar... the swinging point found if projected in the lens barrel, is the entrance pupil of the lens). So, a test can be perfumed once and for all lenses one may have, as to exactly know the issues involved with his (correct) use of equipment.

Now, as far as mechanical design of cameras are concerned, the tilt and swing mechanisms should be all applied by keeping the entrance pupil of the lens absolutely constant (meaning that the lens should rotate by its entrance pupil when swings and then tilt by it), while sifts and side shifts should move the entrance pupil (and hence the axis of the lens) horizontally or vertically. If all the before mentioned are taken care off, then one can be certain that his system performs to its maximum.
Therefore, it is obvious that if a shift mechanism is placed by a maker above the tilt mechanism, when one shifts, he will change the distance of the entrance pupil (and therefore the radius) with respect to the arc that the tilt mechanism has been designed to work... hence the entrance pupil won't be anymore at the centre of the circle of which the arc is part off and there fore the entrance pupil won't remain constant... Obviously, if the shift mechanism was below the swing mechanism, it would shift the tilt mechanism as well, thus keeping the entrance pupil & tilt mechanism geometry constant. :o Same thing would happen if side shift mechanism is above the swing mechanism... Imagine... you side shift your entrance pupil and then swing on the axis that it previously was! It moves on a circle doesn't it? :confused:

Now, (nearly) all modern cameras claim to be yaw free... (Yaw free means having always the entrance pupil constant when tilting or swinging the lens) but they "forget" to mention that if shift mechanisms are above the tilt/swing ones, "yaw free" is lost when one combines the two movements... Further more, L-type cameras and Arcas (Arca's Orbix does tilting exactly like an L-type of camera - by rotating the lens mount on an axis perpendicular to the lens axis) forget to mention that yaw free applies to the tilt mechanism only if the lens used has its entrance pupil positioned at the centre of the lens board... If a recessed board or a lens that has its entrance pupil in a different position than its mount is used, then... bye-bye "yaw free"!!! ...and in addition, the (geometrical) error caused is much more than monorails with arc type tilt mechanisms... simply because the tilting arc is by far more "tight" than them...

OTOH, the Arc type tilt mechanism monorails, they do tilting on a plane parallel (a few centimetres, but it is still not right) to the mounting board plane... This of course means that the entrance pupil of the lens (if a lens that has its entrance pupil on the lens board is used) is never on the centre of the circle of which the tilting arc is part of... and further more, swing mechanism axis is in most cases in the centre of the blog, rather than being a line on the same plane that the lens board is...

Now, add to this that all the above applies for the rear standard too, only difference being that instead of the lens entrance pupil, one has to consider the centre of the image area having the same (exactly) properties...

My conclusion then?

1. There are poor cameras, there are good cameras, there are better cameras, but there is no "perfect" camera....
2. It is better for somebody to chose one of the better cameras and convert it for being perfect...
3. Ideally a "perfect" camera should have its lens entrance pupil and image area centre point adjustable so that the user can position them for "yaw free" operation... Also, a perfect camera should be sturdy and have the shift mechanisms bellow the tilt/swing ones...

So why the Sinar P2 for me? Simply because the above criteria are only fulfilled by the Sinar P-series and the Arca Monolith cameras and out of the two, the Sinar is easier to convert and cheaper to find...
My conversion consists of new frames for the front and rear standard that can both slide back and forth on the standard (for one to position the entrance pupil and centre of the image area exactly for yaw free operation despite the lens used or the image area type) and - off course - new bellows as to avoid using recessed boards at all despite the mounting distance required for the lens. Needless to say that I'm "killing" Sinar's parallel plane off-axis system too.... the new frames are both centred on the blogs.
For one to convert the Sinar, he only has to bolt the new (sliding) bases of the frames on the existing threads and then slot the new frames in them and attach the bellows... It is as easy gets... especially if one will buy longer Arca type bases with the respective (longer) camera plates that slot into them and then bolt the "female" part on the blog and use the "male" part as frame base (so that the frame can move and adjusted back and forth). Care has to taken for the new frames centre to be retained at the centre of the tilt mechanism's arc (12cm radius for the Sinar)...

Emmanuel BIGLER
14-May-2016, 06:01
From Theodoros:
the lens should rotate by its entrance pupil when swings and then tilt by it

Well, I'm not sure that I clearly understand Theodoros' point (certainly other here do understand, not me for the moment).

In short my understanding is:
- I do not see why rotating the lens far from the entrance pupil could induce any image distorsion;
- rotating the lens itself or the whole camera around the entrance pupil keeps the center of perspective fixed but is not related to image projection of image distorsion;
- for lens tilts, rotating around the rear nodal point has the advantage to keep the image stationnary with respect to the subject, this is not related to the location of the entrance pupil;
- yaw effects used to be irritating to the LF studio photographers in the past, not sure that this is still a major issue today!
Doing mostly landscape shots with small tilts, yaw is basically a non-issue to me ...

Now in detail:

To me the only advantage or rotating any lens around its entrance pupil is when you want to make panoramic stitching by rotating the whole camera, so this is not related to the camera's tilts and shifts; or has the advantage that the view point and perspective center does not change when you tilt the lens, this is (IMHO) a very marginal advantage, except may be for close-up work??

Yaw effects, as far as I have read here,
http://www.largeformatphotography.info/yaw.html
refer to the fact that after combining movements, a vertical object is no longer projected parallel to the vertical image edges.
This can be very irritating and time-consuming when doing studio work, taking pictures of objects for which you have to combine all movements in order to get both the image sharp and the projected geometry right.
This yaw effect happens in LF cameras with axis tilts. For this reason several cameras have base tilts.

But may be Tuan Luong has to correct the above mentioned document, where no mention is made of the entrance pupil?

Regarding the entrance pupil: in the past, I am convinced that most professionals using large format cameras with film, never made any panoramic stiching, hence most view camera users never cared for the entrance pupil location!
But this is another issue, panoramic stitching by rotating the camera as a whole is not related to camera settings (tilts & shifts).

However there is some true and proven advantage of rotating the lens around the rear nodal point N' or image principal point H'; doing this you ensure that the image of a far-distant object does not move very much, i.e. remains stationnary on film with respect to the object after focusing, when you tilt the lens.
If you are doing a close-up at 1:1 ratio, in the 2f-2f focusing set-up, the best lens rotation axis to avoid image shift, for a symmetrical lens, is to rotate around the center of symmetry of the lens. Another rotation point!
So, regarding this important issue of keeping the image in focus or stationnary when rotating the lens with a fixed read standard, the entrance pupil plays, IMHO, no role.

Of course in non-symmetrical lenses, the entrance pupil is not located at N, N', H or H' , so you have to choose, either you rotate the lens around the entrance pupil, to keep your center of perspective fixed, or you rotate around any other rotation axis optimized for image stationnarity, you can't get both.
In quasi-symmetrical lenses, with the iris is located close to the center of symmetry, the entrance pupil is always located very close to N or H; moreover the distance NN'=HH' is small [i](less the 3% of the focal length in apo ronars, trying to show that the entrance pupil is not located close to the front or rear nodal point is a real challenge for apo ronar users ;-) ), hence in most cases of quasi-symmetrical lenses you can consider that the entrance pupil is located close to the shutter blades, close to the lens board ... and even close to the rear nodal point N'. No mystery at all.

Base tilts allow a yaw-free operation for combined movements, but when you use a base tilt, you loose your focusing set-up.
And by rotating around any axis far from the exit nodal point N', your image will moves faster and faster when your axis is further away from N'.

This is my understanding, those issues are very subtle and not always visible to the LF practitioner doing landscape shots with small movements.

Yaw effects were important for studio photographers in an era when no digital post-processing of images was possible. Hence everything had to be perfectly right when depressing the shutter release.
Correcting a small amount of image rotation due to yaw does not seem to me as a major issue when you can perform digital image post-processing.
Sure you loose some image quality due to re-sampling ...

Another remark regarding projected geometry: in all our beloved quasi-symmetrical lenses with their pupils located in the principal (or nodal) planes, except for a small amount of intrinsic barrel or pincushion distorsion (this is very small for our modern lenses, and zero for symmetrica lenses used at 1:1 ) there is absolutely no difference in the projected geometry between a symmetrical lens and a pinhole camera, even for out of focus images.
Hence I do not see how camera movements, per se could induce any distorsion which is not described by classical laws of conical perspective, something well-known since Renaissance times!
And yes, there are some non-standard distorsion effects (i.e. not explained by basic laws of conical perspective) for out of focus images projected by non-symmetrical lenses (for which the pupils are located far from the principal planes), but, again, this is not at all related to camera movements.

Theodoros
14-May-2016, 09:23
Hi Emmanuel,

I'll try to explain better...
1. There is no way in photography where one can see (linear or angular or proportion) distortion through his lens and this distortion will not be transferred in the image, in photography what you see through the lens is what you get.
2. There are types of photography that the presence of distortion is more crucial than other kinds, or kind of photographers that don't care, or that only care for the more obvious of distortion... It is of course personal, but it doesn't change the fact that distortion is there, or that the same capture could have been done by eliminating it. In addition, when one stitches distortion presence doesn't only effects the geometry, but (significantly) the image quality too...
3. The description of yaw free you mention, is correct and is the same with mine... One describes the result, the other the cause behind it.
4. Digital distortion correction of images, can't correct for DOF or for perspective but only for some of the geometry. The geometry correction they can perform is with respect to the (wrong) position of entrance pupil that was applied during the capture, they can't correct for the distortion caused by the entrance pupil repositioning... Hence, there is linear distortion left in the final image (fatal for architectural work). In addition, digital "correction" of distortion causes further degradation of image quality.

Never the less, I can't see why one would use a view camera at all and use it (despite how demanding his work is for accuracy), if A.) He could have his camera set up right in the first place and yet he doesn't do so, B.) He could have used an SLR instead (even with a T-S lens) and "correct" things digitally later anyway...

I can see your position for considering things as being less critical than they are for the type of photography you do... But mind you... there is much more demanding photography for accuracy and then the use of a correctly set-up view camera is absolutely critical... Take my work for instance (which is what led me to analyse in depth the camera issues involved with it), I do (mostly) art reproduction for a living, which includes reproducing wall paintings in ancient byzantine monasteries and then internal and external architectural work for ancient monasteries and churches and even statues and architectural work with ancient monuments... Just try to capture and accurately reproduce then a four square meter wall painting masterpiece that is positioned higher up on a curved merging wall that extends towards the dome of the building... Or try to capture the front half part of the interior of an ancient Byzantine monastery (inc. half of the dome in height) and then (looking back) the rear half...

Sometimes it's best for one to admit "hey I didn't notice all this was happening all that time", than trying to pretend that there is no issue or that the issue is of limited extend instead... There is only one way for one to judge how crucial a camera design issue might or might not be and this is only if one compares equipment that has the issue, with other that doesn't.

Bob Salomon
14-May-2016, 10:35
Hi Emmanuel,

I'll try to explain better...
1. There is no way in photography where one can see (linear or angular or proportion) distortion through his lens and this distortion will not be transferred in the image, in photography what you see through the lens is what you get.
2. There are types of photography that the presence of distortion is more crucial than other kinds, or kind of photographers that don't care, or that only care for the more obvious of distortion... It is of course personal, but it doesn't change the fact that distortion is there, or that the same capture could have been done by eliminating it. In addition, when one stitches distortion presence doesn't only effects the geometry, but (significantly) the image quality too...
3. The description of yaw free you mention, is correct and is the same with mine... One describes the result, the other the cause behind it.
4. Digital distortion correction of images, can't correct for DOF or for perspective but only for some of the geometry. The geometry correction they can perform is with respect to the (wrong) position of entrance pupil that was applied during the capture, they can't correct for the distortion caused by the entrance pupil repositioning... Hence, there is linear distortion left in the final image (fatal for architectural work). In addition, digital "correction" of distortion causes further degradation of image quality.

Never the less, I can't see why one would use a view camera at all and use it (despite how demanding his work is for accuracy), if A.) He could have his camera set up right in the first place and yet he doesn't do so, B.) He could have used an SLR instead (even with a T-S lens) and "correct" things digitally later anyway...

I can see your position for considering things as being less critical than they are for the type of photography you do... But mind you... there is much more demanding photography for accuracy and then the use of a correctly set-up view camera is absolutely critical... Take my work for instance (which is what led me to analyse in depth the camera issues involved with it), I do (mostly) art reproduction for a living, which includes reproducing wall paintings in ancient byzantine monasteries and then internal and external architectural work for ancient monasteries and churches and even statues and architectural work with ancient monuments... Just try to capture and accurately reproduce then a four square meter wall painting masterpiece that is positioned higher up on a curved merging wall that extends towards the dome of the building... Or try to capture the front half part of the interior of an ancient Byzantine monastery (inc. half of the dome in height) and then (looking back) the rear half...

Sometimes it's best for one to admit "hey I didn't notice all this was happening all that time", than trying to pretend that there is no issue or that the issue is of limited extend instead... There is only one way for one to judge how crucial a camera design issue might or might not be and this is only if one compares equipment that has the issue, with other that doesn't.

Yaw occurs with any camera when the swing point is below the tilt point. To make a yaw free camera with either base or axis movements the modern solution is to have two tilt points on each standard, one below the swing point and the other above the swing point. Examples of axis tilt cameras with this feature are the Linhof GTL and for base tilt (in this case asymmetrical) the Sinar P system.

As an aside, any camera that is yaw prone, with any type of movements, becomes yaw free when it is rotated 90° so it's swing movement becomes the tilt movement. A Linhof TK or TKS even is equipped with an extra bubble level just for this purpose.

Lastly, regardless if a camera is yaw free or yaw prone, when equipped with the same lens the final result will be identical.

Peter De Smidt
14-May-2016, 11:18
Theodoros, by "distortion" do you mean the changing of the positions of near and far objects relative to each other in the image as movements are used?

Theodoros
14-May-2016, 11:53
Lastly, regardless if a camera is yaw free or yaw prone, when equipped with the same lens the final result will be identical.

True, but it can be totally misleading as to create the impression that one can use a yaw prone camera and end up having the same result as a (totally) yaw free system that one can even correct for the entrance pupil position of the lens and the centre of the image area...

Off course if an image area ends up to be at exactly the same position behind the same lens, the result would be identical... The question then should rather be if one can achieve that by using a yaw prone camera at all... "Hardly the case" is the only applicable answer... Unless if one uses a "perfect" camera, sets it up within seconds, then sets up another camera next to it, uses the same lens on it and then ....spends a couple of hours to recreate the exact same frame (for which the values used on that second camera will be totally different on both standards)...

If one uses a yaw prone camera in the first place, he will end up with a totally different result since he is unaware of the errors he creates and the magnitude of them... especially since he lucks the correct result to use as a reference...

Theodoros
14-May-2016, 11:59
Theodoros, by "distortion" do you mean the changing of the positions of near and far objects relative to each other in the image as movements are used?

Hi Peter,

I mean all direction distortions (of both linear or curved lines) distant or on a parallel plane.... including these you mention...

Emmanuel BIGLER
14-May-2016, 12:10
Thanks, Theodoros, for explaining your point. I have understood your points 1/ 2/ and 3/ and agree with them.
I'm happy if your definition of yaw-free is the same as explained here on this web site.

Now I have some comments on your point #4, quoted below :
4. Digital distortion correction of images, can't correct for DOF or for perspective but only for some of the geometry. The geometry correction they can perform is with respect to the (wrong) position of entrance pupil that was applied during the capture, they can't correct for the distortion caused by the entrance pupil repositioning... Hence, there is linear distortion left in the final image (fatal for architectural work). In addition, digital "correction" of distortion causes further degradation of image quality.

I agree that digital correction has hard times to re-create sharpness for blurred images, but I'll concentrate on geometry since I'm not sure that we speak about the same things.

I'm not sure that we use the term "perspective" in the same way, so we have to agree on the following.

If we neglect, or correct digitally, any distorsion (barrel or pincushion) generated by the lens, then the model for image projection is simply conical projection if we use a quasi-symemtrical lens, i.e., the same as for a pinhole camera.
And even if the images are blurred because we did not apply Scheimplug's rules properely, out of focus images obey the same rules as in classical conical projection.
If we do not agree on this, it will be hard to continue the exchange.
I assume that you do not use retrofocus lenses, they are affected (for out of focus images only) by a very special kind of distorsion unknown to symmetrical lenses. I can elaborate on this issue if required.

Hence with a symmetrical lens, provided that the entrance pupil is located where you want, all images projected on film through the lens exhibit a final geometry defined by basic laws of conical projection, exactly like through a simplified camera composed of an entrance pupil and an exit pupil of very small diameter separated by a small distance, ray tracing being the same as a pinhole camera, except for the separation between both pupils.

Of course, digital post-processing has hard times, or is totally unable to, re-create the desired image if the entrance pupil is located not where you want, because alignments as seen in object space differ if you move the entrance pupil sideways.
But if entrance pupil is properly located, and if the detector, behind the lens, is simply mis-oriented in any direction, and if the lens is quasi-symmetrical, since in-focus and out-of-focus images are projected like in a pinhole camera, I do not see any weird or peculiar distorsion there, that could not be modelled by classical conical projection.
And if easily modelled, that could not be digitally corrected by post-processing. At least for a small degree of blur, but for a pinhole like for a symmetrical lens, out of focus images are projected identically.

The major difference between a pinhole camera and a lens camera is of course that the price to pay for the extreme sharpness delivered by a modern lens is shallow DOF ... and in some circumstances, there is no other solution than digitally stacking images, a technique now widely used for macro work with a DSLR.

Theodoros
14-May-2016, 13:33
I agree that digital correction has hard times to re-create sharpness for blurred images, but I'll concentrate on geometry since I'm not sure that we speak about the same things.

I'm not sure that we use the term "perspective" in the same way, so we have to agree on the following.

If we neglect, or correct digitally, any distorsion (barrel or pincushion) generated by the lens, then the model for image projection is simply conical projection if we use a quasi-symemtrical lens, i.e., the same as for a pinhole camera.
And even if the images are blurred because we did not apply Scheimplug's rules properely, out of focus images obey the same rules as in classical conical projection.
If we do not agree on this, it will be hard to continue the exchange.
I assume that you do not use retrofocus lenses, they are affected (for out of focus images only) by a very special kind of distorsion unknown to symmetrical lenses. I can elaborate on this issue if required.

Hence with a symmetrical lens, provided that the entrance pupil is located where you want, all images projected on film through the lens exhibit a final geometry defined by basic laws of conical projection, exactly like through a simplified camera composed of an entrance pupil and an exit pupil of very small diameter separated by a small distance, ray tracing being the same as a pinhole camera, except for the separation between both pupils.

Of course, digital post-processing has hard times, or is totally unable to, re-create the desired image if the entrance pupil is located not where you want, because alignments as seen in object space differ if you move the entrance pupil sideways.
But if entrance pupil is properly located, and if the detector, behind the lens, is simply mis-oriented in any direction, and if the lens is quasi-symmetrical, since in-focus and out-of-focus images are projected like in a pinhole camera, I do not see any weird or peculiar distorsion there, that could not be modelled by classical conical projection.
And if easily modelled, that could not be digitally corrected by post-processing. At least for a small degree of blur, but for a pinhole like for a symmetrical lens, out of focus images are projected identically.

The major difference between a pinhole camera and a lens camera is of course that the price to pay for the extreme sharpness delivered by a modern lens is shallow DOF ... and in some circumstances, there is no other solution than digitally stacking images, a technique now widely used for macro work with a DSLR.

Hi Emmanuel,

1. What you state about digital distortion correction, would be true if the "cone" you state was part of a sphere of which the (correct) entrance pupil of the lens would be the centre, ...if you use a yaw prone camera instead you create a new cone (out of the same scene) that has been a creation out of a cone that is part of a different sphere with different centre (that being the wrong position of the entrance pupil)... Hence the distortion you correct is true for the (false) cone that was captured... The distortion captured by changing the position of the entrance pupil will still remain because the software assumes that the entrance pupil is in its proper position... Besides, the projection of the lens on the image area, won't be of circular shape anymore... It will have an "eggy" shape....

Do the following test... Use a yaw prone camera that has its tilt mechanism bellow the shift one and its swing mechanism bellow the side shift one... Set all mechanisms to "0", then mount a ...35mm (yes) lens... then shift for 20mm and side shift for another 20mm, then swing by 10 degree and then tilt for another 10 degree... Shoot the sky on 4x5 film....

Q1: Will you see a circle? ...I'm afraid far from it!

Q2: Can you correct the "egg" that will be captured digitally as to recreate a circle? ...Can you do that by retaining the original proportions? ...well, here you are!

Theodoros
14-May-2016, 15:01
Aaaah... Emmanuel,

I forgot to mention one more very important thing that may help you to improve your landscape photography significantly... Before that, I should mention that if you do the above test with a (truly) yaw free camera, you'll see a circle on the 4x5 sheet of film (but moved for the shift and side sift). This is important to know because it directly relates with the suggestion that follows...

A "perfect" view camera is the ultimate tool (by far) for one to do stitched landscape panoramas... "How is this?" you may ask... Well, if one will mount the camera on a tripod so that the mounting point is aligned in the same axis as the lens entrance pupil, then by simply rotating the tripod's head the panorama will be "perfect" too won't it? This is the obvious part though... What is not obvious, is that one can then swing the lens between shots with respect to the plane (or planes) that are captured (without refocusing) for each different shot and then stitch them.... Try this (but only if you ensure that your camera is fully yaw free corrected) and you'll remember me! You'll never go back and shoot panoramas in another way...

Theodoros
15-May-2016, 10:59
Yaw occurs with any camera when the swing point is below the tilt point.



This is (obviously) a misleading invention! If the swing point, the tilt arc's centre and the Entrance pupil of the lens are all aligned in an axis and remain aligned on that axis despite the shift or side shift movements, the camera is (completely) yaw free!



To make a yaw free camera with either base or axis movements the modern solution is to have two tilt points on each standard, one below the swing point and the other above the swing point. Examples of axis tilt cameras with this feature are the Linhof GTL and for base tilt (in this case asymmetrical) the Sinar P system.



Hardly the case... Two or three or four or 100 tilt centres (not "tilt points" as you call them - the entrance pupil can be called a "tilt point") will only make things worst... Neither the Linhof or the Sinar have two tilt centres (nor any other camera), The Linhof M679 has a tilt mechanism in its base, that tilts the whole camera... (nothing whatsoever to affect the lens tilt or swing) and the Sinar P series of cameras have a mechanism to tilt the whole standard so that one can have his standards vertical when he uses the rail in an angle...




As an aside, any camera that is yaw prone, with any type of movements, becomes yaw free when it is rotated 90° so it's swing movement becomes the tilt movement.

...and I'm an astronaut that has travelled all over the galaxy in secret missions....

neil poulsen
19-May-2016, 05:09
. . . To me the only advantage or rotating any lens around its entrance pupil is when you want to make panoramic stitching by rotating the whole camera, so this is not related to the camera's tilts and shifts; or has the advantage that the view point and perspective center does not change when you tilt the lens, this is (IMHO) a very marginal advantage, except may be for close-up work?? . . .

I hadn't thought of this, and it's an interesting aspect of panoramic stitching.

In this discussion, aren't we speaking primarily about front standard movements? It's been my understanding that panoramic stitching is best done by shifting the rear standard. For panoramic photos, seems like having shift on top of tilts and swings would be most desirable for the rear standard.

One thing is for sure. Thinking about, and discussing yaw, tilts on shift versus shifts on tilts, etc., and their ramifications sure challenges ones ability to think spatially. I've always told my wife that I have spatial abilities. But, maybe not.

Theodoros
19-May-2016, 14:07
I hadn't thought of this, and it's an interesting aspect of panoramic stitching.

In this discussion, aren't we speaking primarily about front standard movements? It's been my understanding that panoramic stitching is best done by shifting the rear standard. For panoramic photos, seems like having shift on top of tilts and swings would be most desirable for the rear standard.

One thing is for sure. Thinking about, and discussing yaw, tilts on shift versus shifts on tilts, etc., and their ramifications sure challenges ones ability to think spatially. I've always told my wife that I have spatial abilities. But, maybe not.

Hi Neil,

Panoramas are best done by rotating the camera instead of stitching inside the image circle that the lens projects because: 1. You only use the best part (the central) of the lens, 2. You create as large an angle as you wish...
Having the lens positioned so that its entrance pupil is in axis with the swing and tilt movements, has the same distortionless result as with panorama picture taking... (it's the same principal)... You can even tilt or swing your lens between the shots of the panorama (if you know what you are doing) thus creating a unique effect. One of my favorites is to create a 5 or 7 pictures stitch panorama and swing the lens in between shots for half the angle of the planes with respect to the subject having the lens parallel only for the middle shot... (very effective for UWA architectural interiors too).

sanking
19-May-2016, 18:49
What about panoramas made by rotating the lens around the film plane? The best panoramas I have seen were made with these type of cameras, Noblex for example. Sharpness on the far corners of the field is just as good as at the center of the frame.

Sandy


Hi Neil,

Panoramas are best done by rotating the camera instead of stitching inside the image circle that the lens projects because: 1. You only use the best part (the central) of the lens, 2. You create as large an angle as you wish...
Having the lens positioned so that its entrance pupil is in axis with the swing and tilt movements, has the same distortionless result as with panorama picture taking... (it's the same principal)... You can even tilt or swing your lens between the shots of the panorama (if you know what you are doing) thus creating a unique effect. One of my favorites is to create a 5 or 7 pictures stitch panorama and swing the lens in between shots for half the angle of the planes with respect to the subject having the lens parallel only for the middle shot... (very effective for UWA architectural interiors too).

Theodoros
20-May-2016, 03:07
I'm not familiar with the Noblex, but I'm very familiar with the Roundshot 220VR... I must say, it is the sharpest and most distortion free thing I've ever come across with for panoramas. It doesn't rotate the lens around the film plane, it rotates around the entrance pupil of the lens and then "scans" the image on the film plane through a slot that is always in axis with the lens entrance pupil and then the axis is always perpendicular to the captured plane...

I guess if one rotates the lens around the (center of the) image area, the result should be the same as if one rotates the lens around its entrance pupil, but then the rotation axis should again be aligned (exactly) with the center of the image area and if one uses movements for the capture, then he should apply movements on the rear standard and make sure that the tilt and the swing axis are also aligned with the rotation point and the center of the image area...

In other words, for a view camera (any view camera, used for any task) to be perfectly set up, the entrance pupil of the lens AND the center of the image area must be aligned with the swing point and the tilt arc's center on both standards (that's why I convert my Sinar as to have sliding frames with back and forth movements on both standards - so that one can adjust the entrance pupil & the center of the image area despite the lens or type of image area used) and on top of that, all shift mechanisms should be below all tilts and swings so that the entrance pupil and center of the image area remain in the yaw free position (and on the same plane as shift movements are) if movements are combined.

Any deviation out the above, causes inaccuracies (and thus distortions) despite if the user notices them or not.

Think about it... if one focuses and then tilting or swinging causes him to refocus, it means that the proportion between the distances have changed... doesn't it? It may be that the error is less evident with distant subjects in analogy, but then the error can be huge the nearer the subject and the wider the captured angle (stitched or not)... So whether it is panoramic captures or using movements on a view camera, the principal behind the theory is the same and must be applied to both.

Emmanuel BIGLER
20-May-2016, 11:03
Well, the discussion is taking us quite far from the original subject, I wish to apologize to the original poster who can contact me by the forum's messenging system if he has any specific question regarding Arca Swiss cameras.
Or he can click here (http://farm2.static.flickr.com/1382/1275439235_e2f26a3c5c_m_d.jpg)

--------------

Well, I'm afraid that I have to disagree with Theodoros on his explanations regarding panoramic stitching. And I apologize for a very long post.

There are different cases, and those different cases should not be confused!

1/ Panoramic cameras for which the film is fixed with respect to the landscape, and have their lens rotating in synchronism with a slit defining a small strip of light just in front of a curved film, need to rotate around the exit nodal plane N'=H' of the lens, located exactly one focal length ahead of film. And nowhere else, otherwise the image would be blurred by the lens movement.
So far, no curved Silicon available, so this kind of panoramic camera has not become "digital" yet; but one never knows, there are now flexible solar cells under development!

And yes, those cameras do NOT rotate anything around the entrance pupil of the lens!

This condition, rotating around N'=H', is required, because for far-distant objects, the only point around which you can rotate the lens with a film fixed with respect to the landscape, is N' and not the entrance pupil. Fortunately, for a quasi-symmetric lens like a tessar design used in Noblex cameras, the entrance pupil is located not very far from the exit nodal point N', hence parallax errors are very small when the lens rotates.
We should add that those cameras do not need a wide-angle lens, the vertical angle of view is not extreme, hence a tessar design is perfect for the job.

2/ Creating a panoramic stiching by moving the sensor in the image plane has nothing special, this has become a favourite way of using the whole image circle of a large format lens attached in front of a small sensor, e.g. one of those extra-flat mirrorless digital cameras now available at a small fraction of the cost of a medium format sensor. The only problems you will encounter are when sensor shifts become so extreme that light rays fall at a high angle on the sensor. If the sensor is a consumer grade sensor, it is fitted with micro-lenses and the rays will not be detected if the incidence angle exceeds a certain value. And you should expect some really weird color effects at the edges of the image with slanted rays.
Another brutal problem, but more easily understandable than color fringes generated by a Bayer pattern, arises when using a mirror SLR body, since the camera chamber itself will cut slanted rays "mechanically" and very abruptly. Mirrorless camera bodies are more tolerant to this aspect.
Well, in this situation, scanning the image plane with a sensor smaller than the image circle, you get what you would expect by placing a whole piece of film in the image plane, and distorsions are what you get as usual, frames will overlap without any problem even in the worst case of image distorsion generated by the lens or by the laws of conical image projection. You'll get, extra, weird color effects in the edges due to the sensor itself, that you would not see with color film, that's all.
Camera settings are the usual routine of any LF photographer, except that the rules of the game consist in detecting images on small stamp-sized piece of Silicon instead of using a silver-gelatin detector available in much larger sizes that 24x36 mm ;)

3/ Creating panoramic shots by stitching different images recorded by rotating the (lens + camera + sensor) as a whole is a completely different problem!
This technique had the following drawback, prior to digital post-processing: long straight lines were rendered after crude stitching as polygonal lines. This was unacceptable for architecture shots, in the past.
And also: if you do not rotate the whole camera, or lens, around the entrance pupil of the lens, different projected images do no share the same view point, hence alignments in object space are rendered differently from one frame to the next one, and stitching can be made impossible.
But this is only relative to the displacement of the entrance pupil compared to the actual distance to the object. If the displacement of the entrance pupil, expressed in angular terms, as seen from one point of the object, is smaller that the angular resolution of the image itself, you'll never notice anything and you'll be able to stich your images easily.

My point here is that the condition of rotating around the entrance pupil is more or less stringent according to the operating conditions. For a mountain landscape at infinity, you do not care for the entrance pupil. For a 3-D scene in close-up, if you do not care, you won't be able to stitch your images.

If your object is located at infinity, you can rotate the camera anywhere you want.
The closer you are to your object, supposed to be 3-D object with various planes stacked in depth, the more you have to be strict about your rotation point, which should be as close as possible to the entrance pupil of the lens.

So there are at least 3 different ways to create panoramic shots by stitching, and the location of the entrance pupil of the lens with respect to the rotation point of the whole camera or the lens alone is important only in this particular kind of panoramic stitching, #3.

Another point where I have to disagree with Theodoros, is the use of the word 'distorsion'.
If you attempt to make a panoramic stitching by rotating the whole camera or simply by rotating the lens of the view camera around an axis located far from the entrance pupil, you do not generate any peculiar distorsion in your frames. Only usual "distorsions" common to any conical projection to a plane detector.
Each frame is a classical conical projection, the geometry of the frame can be automatically transformed by the stitching software. From projections to different planes, you can re-construct an equivalent projection to a cylinder, a sphere or whichever geometry suits your project.
Doing so you eliminate polygonal lines, this is really great for architecture shots.
Except that straight lines become curved if you choose to re-generate the equivalent of a cylindrical or spherical projection instead of a wide-angle, rectilinear projection.

Millions of digital "panoramic stitchers" do it every day! Except that probably very few of them do it from images recorded on film with a large format camera ;)

The only problem is that your frames might not match on their edges if you have differences in the point of view, when alignments of objects are different from one frame to the other, when you rotate the camera 'carelessly'.
For example on one frame, two distant trees could be perfectly aligned, on another frame, you can see the two trees separately, simply because the viewpoint has changed by rotating either the camera or the lens around something which is not located at the entrance pupil.

No specific distorsions here !!! Only usual distorsions, but applied to different frames that cannot perfectly overlap simply because they do not share the same viewpoint, or center of perspective.

Simply the good old laws of conical projection, the same as you can experience with a pinhole camera.
And, sorry for explaining it again, the pinhole camera explains everything when you use a quasi-symmetrical lens, whether your images are in focus or out of focus, the projected geometry is the same as in a pinhole camera and can be corrected by software.
And camera movements or settings are, obviously relevant to this question of projection, but all cameras fitted with the same lens located in front of the same subject, with a film plane located at the same place, with the same orientation, will record the same projection, same geometry, same distorsions, same lens-induced defects, whether the camera is yaw-prone, yaw-free, cheap or expensive, new or old, monorail or folding, wooden or metal ... whichever kind.
Only out of focus images generated by retrofocus lenses, in certain conditions, generate distorsions that cannot be modelled by the conical ray tracing known to painters since Renaissance times.

Theodoros
20-May-2016, 13:34
Hi Emmanuel,

Thanks for pointing (and backing up) that it is not the entrance pupil (as I thought it was) but the exit pupil that the 220VR's adjustment is for... I use the 220VR out of a friend who is a collector of Roundshots and is kind enough to let me use it whenever I need it (lucky me)... Indeed I do avoid using UWa lenses with the 220vr, the 85mm f2.8 PC Nikkor and the old 35mm f2.8 PC are my favorites with it depending on the height of the image I want and then I use as much length of 120 (or 220) film depending on the angle I want... I even dropped scanning the film and now use my Sinarback 54H in "true colour" 16x multishot mode and a Kaiser lightbox to do all my film scanning... (amazing results - really extracts as much as it can be extracted out of film and if one stitches four 16x multishot captures as to scan a 6x8 negative out a Fuji GX-680 he can then print bilboards with all the detail in the world).

You keep saying "disagree with Theodoros" in your post, but the more I read it, I find we agree on anything you say there... Actually I find it confirming me in detail as to why the use of a badly set up (or designed) camera for close captures can turn to a mess... Of course you do mention that the error with longer subjects is eliminated, but other than that being a different estimation of the error's importance (which is personal) it is only a part that concerns (some) of the landscape photographers and then not in all their shots... There are many cases where a landscaper uses his camera for wider angle near subjects and then the errors related with wrong use is near or as much as with architectural photography. Of course for studio photography or table top, or macro, the demand for a correct set up maximizes.

IMO, if Officer Sheimpflug was still around and he was asked to value the cameras in the market, his marks (for today's cameras) would range from 2 up to 8 (8 for the Monolith and only if used with lenses that have their entrance pupil exactly positioned on the frame's plane as well as the image area on the same plane on the rear frame) with most folding stuff that people buy marked towards the lower end of the scale.... But one has to admit... It is surprising so many "famous" cameras on the market that are in production (or have been in the recent past), have design errors that could have been easily avoided... Take the Universallis for instance... it corrects on the wrong positioning of the shift mechanism of the original F-line and thus makes the mechanism useful if combined with tilts, ....doesn't it?

Emmanuel BIGLER
20-May-2016, 23:17
but the exit pupil that the 220VR's adjustment is for.

No, not the exit pupil, for a panoramic camera with a rotating lens with respect to a fixed subject, it is the exit nodal point N' or or principal point H', otherwise the image would move with respect to film and would be blurred. Pupils are totally irrelevant in this problem.
But there is no difference for a symmetrical lens design with its diaphragm located at the center of symmetry: in such a lens design, pupils are located at the principal or nodal points.
Most of our LF lenses are quasi-symmetrical except telephotos.

BTW there is an interesting variant of the panoramic film camera with rotating lens: the Alpa Rotocamera, able to record more than 360° shots.
The Alpa Rotocamera rotates as a whole around the exit nodal point of the lens and film is wound at the proper speed, so that film stays fixed with respect to the landscape.

------

Configurations #1 (panoramic film camera with curved film and rotating lens) and #3 (panoramic stitching by rotating the whole camera or the lens alone) are completely different.

In configuration #3 the key point is how alignments in object space are projected in image space.
And then enter the pupils on the scene, because we are dealing with out of focus images of a stack of object planes.
Out of focus images are generated by conical projections of the exit pupil, hence pupils play an important role in the geometry of the whole image, generated by stacking all projections of object planes to a single film plane.

In situation #1 the subject is at infinity, its image is projected sharp at the focal plane, no parallax exists for objects located at infinity.
Here we do not care for pupils at all, but we do care for the image movement, the image should not move when the lens rotates. This is a completely different problem.

And regarding your use of the word "distorsion" you probably mean 'wrong projection' or 'projection not acceptable according to certain standards'.
For example, for classical human portraits, we generally avoid to set the entrance pupil of the lens at a distance of one foot from the nose of our subject, except if we are working for a pop/rock group of the last century, where all covers for vinyl records should, on the contrary, show a group portrait shot with a wide-angle lens located at a distance of one foot from the musicians' noses ;)

From a geometrical point of view there is no question of being correct or incorrect; there exist only an infinite choice of projections, geometry of the projected image being defined by the well-known laws of conical (pinhole-like) projection ... at least for quasi-symmetrical lens designs.

For example classical, or should I say: canonical rules for images of buidings with vertical and parallel straight lines lines, are that those lines should appear parallel and vertical in the images.
This is an aesthetic or cultural rule. With a pinhole camera, you simply need to set your film plane vertical and parallel to the verticals of the building, projection of rays through the pinhole takes place between two parallel planes. Doing so, rectangles are projected as rectangles and circles are projected as circles. But spheres are projected as ellipses on the edges of a wide-angle image!
Those rules for keeping rectangles rectangular in the image were strictly applied by XIXst century photographers who would never allow parallel and vertical lines to be projected as converging lines, they simply followed rules applied by painters. Eugène Atget prefers to let vignetting appear on top of his images (due to an excess of front rise) to allowing convergence in verticals.

Same setting applies to a quasi-symmetrical LF lens. If the lens board is not parallel to the building and therefore not parallel to film, the image will be blurred, but with a quasi-symmetrical lens, geometry of the blurred image will be only weakly perturbed by rotating the lens axis. Rectangles in the subject will be blurred in the image, but will stay as rectangles.
With a retrofocus or telephoto lens, starting from the same configuration of film plane parallel to the building, if you rotate the lens axis upward, not only you'll get a blurred image, but you'll get an additional distorsion not explained by the pinhole model. This effect is not described in classical texbooks on LF photography dealing with with symmetrical lenses. The effect has been found only recently when some architecture photographers started to use more extensively tilt and shift lenses with a 35 mm DSLR and retrofocus wide-angle lenses. In those lenses, pupils are not located at nodal points, out of focus images are projected in a different way, classical pinhole-like geometry no longer applies to out of focus images projected by a retrofocus lens.

So going back to our quasi-symmetrical lenses or pinhole cameras, an image is considered 'distorted' when the projection is not the canonical projection used by painters.
For a given combination: subject / lens position and orientation / film position and orientation all cameras fitted with the same lens will deliver exactly the same geometry.
The question is simply: how do I achieve the desired projection combined to proper image sharpness?

There, of course, LF cameras differ, some allow you to get the result quicker and more precisely ;)

Theodoros
21-May-2016, 03:52
Hi Emmanuel,

Lets be more practical then and look on how one can fully overcome all the above by setting up his equipment properly (which is directly related to the O/P question).

-If one would have his front and rear standard frames adjustable, so that he can align the entrance pupil for being in axis with the tilt mechanism's arc positioned on the center of the circle of which the arc is part of and then the swing point aligned on the same axis, there would be no problem using any lens, (despite if it is of retrofocus or quassi-symmetrical design...) and if the same was applied for the center of the image area on the rear standard, then the set-up would be adaptable to eliminate errors on all cases.

-And additionally, if the camera's (vertical) shift mechanism is bellow the tilt one, then (vertical) shifts can be applied in combination with tilts without inducing errors and also, if the (side) shift mechanism is bellow the swing one, then side shifts can be safely combined with swings...

Therefore, if one chooses a camera with all shift mechanisms positioned bellow tilt and swing ones (like the Monolith or Sinar P series are) and additionally converts the camera so that the front and rear standards are adjustable back and forth, he would end up with a camera that would be adaptable to work perfectly well with any lens and any image area size... A camera that Officer Sheimpflug would then rate with maximum marks (10)... isn't it?

And of course, if the same camera can have a bellows which allows the front and rear standard as to (almost) collide mechanically and then makes a rear frame that allows the grip of a mirrorless to recess by the frame so that he looses only the 18mm that the sensor is recessed back from the camera's mount, he would then end up with a camera which can work with a mirrorless and even a 35mm DSLR lens fitted on the front standard. Further more, if the same rear frame has enough width as to accommodate an MFDB he then can use the same set up for all digital image areas (and even film up to 6x4.5)... and finally, if one converts the base of a 4x5 rear standard so that it can be slotted instead of the frame used for the digital image areas, he then only has to interchange the rear frame and the bellows and have anything working perfectly, despite the type of the lens or the size of the image area, without having to use any recessed boards. He then only has to care, that in the rare case where centering the image area on the rear standard and the lens entrance pupil on the front standard, the distance is not enough, to slide his rear frame forward (off position) and avoid to use tilts or swings (only) on the rear standard. :) ...Correct? :)