# Field of view question

Hi there, i’m a Production Designer, i use Sketchup to make mostly every single design on my work, From basic props sketches to full constructive plans.
I often use the Advanced Camera Tools for previs purposes as well, given that kind of work has to be extremely precise, i have a question that might be stupid but i need to know the answer:
All camera lenses have three different Field of View angles: horizontal, vertical and diagonal. However, Sketchup uses only one of these parameters, which one is it? i haven’t found that info on the documentation.

Thanks!

I recall camera angles in SU are measured vertically and the camera eye is set at the lens rather than the film plane…

john

I haven’t found it written anywhere (but I didn’t look too hard!), but there’s reason to think that it’s vertical. As you make the window wider or taller, the vertical contents are always visible, but the horizontal contents go off the left and right edges.

I did a test though, where I had two horizontal lines above and below the origin, then drew lines at 17.5 degrees, dropped a vertical line down from where they meet, and then used the Position Camera to see what the view was like from there. The two lines were very close to the top and bottom of the window view.

I was just exploring this myself this past week. I drew myself some frustum shapes to confirm the field of view numbers I was finding on the internet.

The SU field of View does appear to be the vertical angle, though looking at my known Match Photos recently, the numbers (lower right corner) are at tad narrower. In the examples below, 12mm on a Nikon DX body is equivalent to 18mm at full frame (still) = 67.38° for the narrow edge and 90° for the long edge. Close, but not exactly.

Horizontally:

…and turned vertically:

I would think camera location is the optical center of the lens, but who knows.

unless they actually changed the ACT code from the original when it was overhauled…

``````    # Create the lens
# REVIEW:  I am creating the camera geometry so that the front of the lense is
# at the eye position.  This is probably not really correct.  I should probably
# have the film plane at the eye position.  I am doing it this way for now so
# that the camera geometry will never be visible.  I could control this using
# the near clipping plane, but we don't currently have any way to set that.
``````

it may also have some clues as to what else happens…

john

The critical point of interest isn’t the film plane or the front of the lens, but the “optical center” of the lens I think it’s called. For a pinhole camera, it easy, that’s the pinhole. I think, for a glass lens, that’s where the field of view lines come to a point and that doesn’t necessarily correspond to the theoretical focal point. (That’s also the point to rotate a camera when doing careful panorama stitching.) What actually happens inside a lens gets complicated, making the two focal points different, but I don’t think the use of SketchUp needs to know any of that stuff.

Edit again: “Entrance Pupil” or maybe “Nodal Point” seem to be the correct terms I’m looking for rather than “optical center” and quickly getting sucked into too much info researching it. Maybe later. Here’s one link.

Match Photo views are special just like 2-point perspectives. Both have their own way of calculating the image with some differences to the standard SketchUp perspective camera. When you orbit even slightly in these views, the camera switches to standard, and you see a “jump”.

Nodal Point is indeed the term I’ve encountered in the past to describe the location from which the camera appears to “see.” As you mentioned it is the point about which pans and tilts should be done to avoid perspective changes.

Of course, lenses have all manners of distortions that affect the 2D projected/captured image which may need to be accounted for when doing accurate photogrammetry and such.

Vertical?

There was a discussion about 2 years ago where the consensus was that FOV is the vertical extent.

I think that Advanced Camera Tools are treated differently, however.

I made a sample scene with a spatial reticle doohickey in it and placed a camera at the apex. When looking through that camera, there is an unambiguous angular scale in view which also serves as position check for the camera.

I chose the camera type to be full-frame still camera, 36 mm width, aspect ratio 3/2 = 1.5. In the camera’s scene, I modified the focus (Opt + Up or Down on Mac) to get a FOV of 45° vertically, using one of the rings in my reticle doohickey. At that point, ƒ is reported as 29 mm. Basic trigonometry confirms that this is consistent.

Choosing Camera → Field of View at this time will populate the Measure field as “63.65 deg.”, which happens to agree with the FOV as read from my reticle along the horizontal axis:

I repeated the exercise by choosing the vertical FOV to be roughly 30° by setting ƒ = 45 mm by Opt+Up/Down in the camera view, after which Camera → Field of View reports “43.60 deg.” Therefore, this appears to be the intended behavior.

The horizontal sweep angles of the frustum volume also agree with the reported FOV.

There appears to be a mild inconsistency: The reported FOV does not change when changing the Aspect Ratio in Edit Camera. It appears a fixed aspect ratio is used to report FOV, perhaps the ratio implied by the camera type chosen from Select Camera Type.

Regular camera: Vertical

Alas, now choosing Tools → Advanced Camera Tools → Reset Camera will zoom out and create a view with indeed 63.65° vertical FOV, confirmed by the reticle:

Dolly Zoom

There is another indication that the regular SketchUp camera is treated differently from Advanced Camera Tools, namely, the Camera → Zoom tool performs a dolly zoom. It breaks out of an Advanced Camera view (if active), and clearly does not alter the focus length of an Advanced Camera as one might expect. SketchUp’s Zoom action being a dolly zoom always alters the camera position.

I have a feeling there are historical reasons for the behavior of the two camera concepts.

3 Likes

The advanced camera tools in itself does not give an angle of view but a lens focal length in mm. Invoking the “zoom” tool gets you the FOV info which according to my own use delivers the horizontal FOV. Most often lenses, while covering a circular field of view, will ordinarily have to be configured to cover all of the film’s “rectangular” plane or sensor and (unless its a square sensor?) will be configured to the (larger) horizontal coverage to avoid vignetting that would happen at the sides if it were only configured to the smaller vertical size. So in cinematography, the horizontal angle of the lens is dominant as the vertical angle is more often a factor of the aspect ratio of the composition. i.e. 1:85 will have the top and bottom cropped of the 4:3 full frame, in both instances the horizontal FOV would be the same for a particular lens and the vertical angle of view determined by the aspect ratio. The focal length of a lens remains the same as it is the measurement of its optics, however, it will “appear” to act like a “longer” or “shorter” lens depending on the film sensor size it is being used with. e.g a 50mm lens will appear to look more telephoto when used on a smaller sensor.

The nodal point, is the point to rotate the camera around in all axes to avoid any “parallax” issues which was critical in foreground glass matte paintings when a camera pan or tilt was required and currently with multiple shots that might require stitching. Without the nodal move, the matte painting would shift out of alignment from the distant objects. This point is usually marked on a camera with a symbol at the file/sensor plane (a circle with a line through it). The rotation point will be directly below this plane on the lens center for panning if the camera is level and will no tilt. To tilt about this point a “nodal head” mount is needed, which looks like a large up facing “arc” on a geared mechanism. This allows for the nodal point of the camera to be aligned with the center of this “arc” so that the tilt will rotate the camera about the nodal plane and avoid any parallax shift.

In SU, holding the shift key while zooming will change the angle of view while keeping the camera in the same spot…otherwise the “zoom” tool acts like a dolly where the FOV remains the same but the camera position changes.

As mentioned above, the SU nodal point is in the center of the front of the lens. If it were mimicking a real camera that point would be behind the lens and a little way into the camera body.

1 Like

I’ve always understood that symbol to mark the film plane, which would be different from the nodal point. I believe the nodal point varies with each lens, so I would think it’s possible to mark the nodal point somewhere on the lens.

Oh, I like that. More useful features than what I made.

BTW, Canon has a nice web based resource for calculating FOV here.

Hi Illking!

I have a question for you related to Advanced Camera Tools. I am also a Production Designer. I’m using the ACT for the first time. Before the free trial runs out (2 days!) and I drop the \$700, I was curious:

Do you experience image clipping in your models when looking thru the ACT cameras? My image clipping is severe and seems to originate from the frustrum volumes even when they are “hidden.”

I have read other forum complaints about this, but no solutions. I’m curious if you’ve experienced this and if you’ve found a work-around.

PS: here is a good calculator for FOV (with entries for custom image sensor sizes, etc):
FOV calculator

1 Like

(For the custom image sensor size scroll to the bottom of the “sensor size” tab. You’ll want this for cameras not listed like Arri Alexa etc)

Hi there!
Yes I have experienced the same glitch on some projects under certain circumstances. I’ve noticed a relationship between models with huge amounts of complexity (Street blocks with fully detailed buildings, mostly) and this glitch. I work on Mac, don’t know which OS you have but it seems to be a rendering issue, perhaps some kind of VRAM bottleneck…

How many cameras do you have in the scene? I’ve noticed the issue seems to worsen the more you have.
Assume you’ve tried all the z clipping remedies?

Yeah, I’m on a Lenovo Y50-70. Running Windows 10. Discreet NVIDIA GeForce GTX 860M graphics card. Built for gaming. I haven’t noticed any other rendering lag, but maybe it’s still not quite up to snuff. Darn!

I get this problem in all the models I’ve built so far. All simple 4 wall sets with various set dec.

Hi! Yes, it does seem to get worse with more cameras, but I still get it with just one. It’s not as severe with just one, but definitely still a problem.

I have run through the fixes in that exact video! Haha!

Maybe it’s my system. I dunno. I don’t suffer any clipping from the normal camera, just the ACT ones.

I think it’s a bug. The only workaround I’ve found when it gets bad is to just save a particular camera’s view as a regular scene ( kinda defeats the idea I know) and delete the camera, you end up with the same view without the physical camera and you can add the aspect ratio frame in as a watermark, or leave the camera and delete the FOV lines and volume “layers” and see if that helps ( that at least leaves the aspect ratio border and info). Using a renderer is another way of maintaining an aspect ratio frame.

Thanks for the tips, White Rabbit. Those are definitely doable options.

I think the best solution I have now is using the FOV calculator to create a cheat sheet of the various vertical FOV angles for my given camera and focal length, and manually entering those. I also created a keyboard shortcut for “Look Around” so most of the basic camera controls are there. It IS helpful that the Aspect Ratio Frame stays on screen after you delete a camera! Haha.

I’ll just have to run a separate model if I ever want to look at the camera and frustrums in the set. But the main use is looking THROUGH the camera.

If you do delete the FOV lines and volume from the layers palette, you still get left with the camera object (that you can still click to look thru) and the aspect ratio. Might be worth testing on the model you were having trouble with to see if it makes any difference. i.e. if its the FOV info that’s causing the problem?