4

Consider a photographer with an SLR camera and a PC lens attached in its default, unshifted position, standing on the ground at the base of a columned building whose height is X times that of the person and is located Y feet away from him.

The photographer composes a frame with the center of the building in the center of the frame with some empty space above it, to do so they have to pitch the camera slightly upwards which creates a keystone effect with the columns appearing to converge vertically towards a point above the frame.

Next the shooter maintains the camera orientation levels the camera and adjusts the shift of the PC lens until the columns appear parallel in the image.

Is there any change in focus at the bottom of the columns and the focus at the top of the columns before and after the shift ?

mattdm
  • 143,140
  • 52
  • 417
  • 741
SirPeepsalot
  • 41
  • 2

4 Answers4

1

The shift will not change the convergence of vertical lines. The verticals converge because the lens isn't horizontal. The purpose of the shift is to allow you to frame the building while maintaining the lens axis horizontal.

If you focus to the middle of of the building, and then make the camera horizontal, then the focus has changed: in the worst case scenario, if you are right next to the wall, you focus at half the building height, but when you take the picture you are just a few inches away. Of course for a real shot it will depend how far you are and how high is the building.

xenoid
  • 21,297
  • 1
  • 28
  • 62
1

The confusion here is caused by the way so many people incorrectly describe what the "shift" function actually does. Shifting the lens does not alter the scene perspective at all, it simply causes the image sensor to be moved to a different part of the image circle being produced by the lens.

For example, the Canon TS-E 24mm f/3.5L Mark II creates an image circle which is so large that it actually represents the same view that would hit the image sensor if you were using a 17mm lens (to the best of my recollection; possibly it's a 16mm). But when using the TS-E lens only a small part of the image circle is hitting the camera sensor (or 35mm film frame if you're old school). Shifting the lens up effectively causes the image sensor to be moved to the part of the image circle which represents the scene further from the ground (the top of the building, for example). Shifting the lens to the left effectively causes the image sensor to be moved to the part of the image circle which represents the scene to the left of centre.

In other words, shifting a TS-E 24mm lens is exactly the same as capturing the same scene with a standard 17mm lens, and then (in post-processing) drawing a selection marker (with 3:2 aspect ratio and a width and height being 70% of the full size) and moving that marker around the scene (up/down/left/right, or diagonal) until it contains just the relevant part of the scene (the top of the building, without the road/pavement, for example) and then selecting "crop to selection". Shifting the lens does not alter the perspective of the scene at all, just which part of the scene you actually capture on the image sensor/film. The benefit of a lens "shift" is that it does this scene cropping without losing any of the resolution of your sensor/film, while the post-processing software crop will leave you with only 49% (0.7 × 0.7) of the megapixels your sensor can capture.

This moving of the image sensor around the image circle is what allows you to take multiple shots by shifting up/down/etc to capture more pieces of the scene, and then stitch the pieces together to create a view which is closer to the angle you'd get from a 17mm, but with higher resolution than you'd capture with a standard 17mm (because you're adding pieces together).

As scottbb mentions in the comments, shifting the lens does slightly alter the view of the scene. This will generally cause no problem, unless you're trying to capture multiple images to stitch together, in which case the change caused by moving the lens may cause parallax errors (especially where your scene mixes close items on top of far-off items, such as stair rails, chain-link fences, glass with writing or graphics on it). To avoid this problem you need to make sure that you move the camera body (the sensor/film) instead of the lens, and there are various tripod utilities which make this possible, either by holding the lens fast, or by allowing you to shift the camera body in the opposite direction to the lens shift so that the lens ends up in the original position.

As for the plane of focus, this depends upon two things: is your camera sensor/film plane perfectly parallel with the columns in the scene, and is your lens of a high enough quality that it doesn't suffer too badly with field curvature. If your camera is parallel with and focused on the columns, then they should (in theory) now be in focus from bottom to top no matter how tall they are, and no matter in which direction you shift the lens. But no lens is perfect, and field curvature means that the large image circle may suffer from an imperfect focal plane, such that instead of being a perfectly flat plane it is slightly concave or convex. But that's somewhat out of your hands, and if you do detect this being a problem then stopping the lens down to a narrower aperture may alleviate it.

Oh, and the last thing to mention is that if you shift a lens to its limit (such as ±12mm on the Canon TS-E 24mm) then the extreme edge is likely to be rather soft. This is somewhat inevitable (because you're then capturing the very edge of an unusually large image circle) and tends to be visible well before you view the end result at 100%. This could be mistaken for the top of the column being "out of focus", even though that is not the cause. So avoid shifting to the extreme unless you simply can't capture the relevant part of the scene without doing so.

Bobulous
  • 647
  • 1
  • 7
  • 19
  • 1
    Good answer. But a technical nitpick: "Shifting the lens does not alter the scene perspective at all, it simply causes the image sensor to be moved to a different part of the image circle being produced by the lens." By and large, that's true enough. However, if the body is held still and the lens is shifted, then that means the entrance pupil has shifted. And the camera's perspective is defined by the location (x, y, z coordinates) of the entrance pupil. Will that make a difference in a landscape shot? No. In a macro shot? Quite possibly, yes. But if the lens were held fixed in place... – scottbb Feb 22 '20 at 18:22
  • ... and the body shifted (i.e., just like rear / film plane shift in a view camera), then there is absolutely no movement of the perspective point. – scottbb Feb 22 '20 at 18:23
  • 1
    Yes, @scottbb, you're right, but I was glossing over that for simplicity. For anyone concerned about the way the view of the scene can change slightly when the lens is shifted instead of the body, the Northlight Images review of the Canon TS-E 17mm f/4L has an interesting section about avoiding parallax error by using a tripod which lets you move the camera body and hold the lens in exactly the same place, vital for when you're capturing multiple images to stitch together and have fences or balusters relatively close to the lens. – Bobulous Feb 23 '20 at 15:47
  • Yeah, I've always wished TS/PC lenses had 1/4-20 UNC threads on the lens body in front of the movements, so that the controls sort of turn the setup to rear tilt and shift. Of course, if I really wanted to dive in, I could just go whole hog into something like a Cambo Actus type system. =) – scottbb Feb 23 '20 at 18:03
  • 1
    I've updated my answer to add a paragraph about parallax error. – Bobulous Feb 24 '20 at 22:26
0

Possibly. That is because when the lens is shifted it is using the periphery of the objective lens and not the center. The actual result would depend on the focus field curvature of the lens. It is typically only specialized macro lenses that are designed to be flat field.

That said, I have not noticed any such effect using Nikkor PC-E lenses, but I also tend to use them tilted more than shifted.

Steven Kersting
  • 17,087
  • 1
  • 11
  • 33
0

Once both sensor and lens are parallel (and the lens is sufficiently "perfect"), the focus of the image without falling lines will be uniform.

If you want a non-uniform focus while retaining non-falling lines, you tilt the lens. The ability to separately shift and tilt the lens (and the sensor) allows you to choose the perspective distortion independently from the unsharpness. Of course the actual process of getting both as you like them is a bit of an iterative one, so "independently" is a euphemism for "differently".