Page 1 of 2

FOV/angle of view PESPECTIVE

Posted: Sat Jan 21, 2012 7:59 pm
by chucky
Hi guys, I have what I consider to be an important topic to discuss.
Field (or angle) of view and focal length in ASPRO.
Can anybody tell me what the zoom what is AS equivalent to?
I come from a film background and do a lot of storyboarding and lenses are very important to visual storytelling.
There is a standard way of measuring an FOV using mm in those fields using focal length ( I'm sure most are familiar with this , I am not being at all patronising just bear with me)
Also forgive me for any inaccuracies and feel free to correct me when I am wrong or mistaken in this post .
Ranging from small numbers ...
...Fish eye 16mm -very wide angle lens
...Human eye (roughly and arguably- yes there's more periphery ,lets not get into that)50mm
...zoom -narrow lens... 85 -120 (ish) good for portraits too.
....orthogonal- I think that might be infinite but 5000 should be enough.

So what is ASPRO's lens measured by? I just don't know what that number represents is it angle of view or what , it certainly isn't focal length.

Here's some fairly standard ways of representing a cameras basic properties.


Can we not have ASPRO standardised in this department?

With vector and image layers in a 3d space can we have the option to keep in them one point perspective?

I don't think of this as asking for a feature as in my mind it should be a default behaviour in most circumstances ( unless the layers are used to construct buildings or boxes or using the extrude/ lathe etc, features).
I am getting very odd results when layers (that should behave like sprites) don't render or display flat to camera under many circumstances.
Especially if wide angles are used.

I know there is a ignore camera movements option but this really doesn't do the trick - it would be desirable if the layers could keep their relative parallax but maintain their 'squareness' to the picture plain at all times.
My description here is not great I know but in essence objects like characters or illustrated layers that create layered back grounds most often need a one point perspective rather than 3 to keep this squareness.

This video uses the rectangular cards to illustrate how the 2d layers respond to camera movement.

The cards in the row on the right are set to 'rotate to camera'

The cards in the row on the left are left at default to camera.


It would be most desirable for those cards to remain completely flat to the image plain, the middle row there is set to 'rotate to camera' but distorts much more than is useful.

Also an orthographic option would be good especially for top/front/left/right views.

Actually while trying to illustrate this on ASPRO I just discovered a couple of bugs when using 'sort by true distance' and one you can see here involving line thickness.
Those are other matters though.

Does this make all sense to everybody/anybody?
FOV .anme

Posted: Sat Jan 21, 2012 8:56 pm
by hayasidist
Hi Chucky. Makes a lot of sense - I've met the same questions Had me floored at first too (I am more used to 35mm camera lens "standard" focal length too!) BUT .... in AS: zoom is field of view in degrees.

The focal length examples you give work for 35mm format; but not, for example, for DSLRs. The difference is because a 35mm frame is much bigger than a typical CCD... In a "typical" DSLR the "human eye" equivalent is about 30mm.

In AS zoom default is 60 degrees. then there's the distance between camera Z and image Z. In the default set-up camera ia at z=1.732 and image is at 0. If you're into maths - that will be instantly recognisable as an equilateral triangle of side = 2 with the lens, top centre of frame and bottom centre of frame as the apexes.

what I then found (glimpse of the b---g obvious really!) was that there is no way of telling AS what that "frame height = 2" maps to in the (depicted) real world. It could be "a face" or it could be "a mountain". In the "face" case (imagine that's 40 cm top to bottom) the camera by default is about 35 cm away. Not impossible or implausible - but typically you might prefer to be (say) 3.5 metres away with a long lens!?

Then there's the way that AS works out what a "2d" shape will look like when you pan / tilt the AS camera. Put simply, it applies perspective to the shape. Stuff that is close to camera distorts more than stuff that is a long way off; a fish-eye lens will distort more than a long lens. To get round these "issues" there are layer options (that you know about) such as "immune to camera" and "rotate to face camera".

I got a bit motion-sick watching your dancing cards - so I'm not quite sure what I'm meant to be looking at. BUT -- and not sure if this will actually help with what you want - maybe try different Z for the cards?

Is this going in the right direction / helping?

Posted: Sat Jan 21, 2012 9:05 pm
by chucky
Hayasidist, thanks for your info.
I'm not after help in using ASPRO's camera, if so I would post in the 'how do I' thread , I am in fact very proficient in camera use across many applications what I want to start a discussion about is.... the neccessity to change how the camera is implemented in ASPRO.

Take a look at the cards, as I said in the post, the cards in the middle row are set to rotate to camera, but they distort like #@%*.

I could have used various zoom lengths and distances but this one illustrated the point clearly without getting too confusing.

The cards in the row on the left are left at default to camera.

Neither of these are good for most 2d elements.

When using ignore camera movements the elements do not share a spacial relationship to each other when the camera moves ( yes, obvious).

So there is no good way of using ASPRO's camera feature.
Because of that the 3d camera is not a feature so much as a gimmick IMHO.
In practice the camera needs to work with both wide angle and long lenses especially to earn the label 'pro'

You also bring up an interesting point...
there is no way of telling AS what that "frame height = 2" maps to in the (depicted) real world. It could be "a face" or it could be "a mountain".
I was tempted to discuss this factor in my initial post.....scale....
AS well as standardising lens, scale too could be of some more useful gauge.

Posted: Sat Jan 21, 2012 11:36 pm
by hayasidist
I've taken your .anme and had a play with it ... camera at z= 30 and zoom = 24 gives the same(ish) screen-filling shot - but makes a big difference to the relative distortion.

Consider if you track wide angled camera across someone's face (say) 30cm from their nose -- I'd expect to see "odd" effects because the relative distances from lens to left side / centre / right side of face change proportionally quite a lot (15% longer to the edge than to the centre on a 60 degree FOV) ... but the same shot on a long lens still fills the frame but has a much smaller ratio (only 1.5% longer on a 20degree FOV). I can see the artisic value of both ...

Also, (need to check this - but I think this is right) I think that the 60 degrees is on the "short" aspect ratio - so if you're widescreen (say going on 2:1) then in fact by default you've got nearly 100degrees FOV on the long dimension (and 50% longer to edge than to centre!): very fish-eye indeed!

Posted: Sat Jan 21, 2012 11:51 pm
by slowtiger
I agree that it's frustrating to get a certain focus length.

To avoid the distortions you mention I usually do this: I place the camera quite far away from the layers, like setting z=10 (tool 4) when the top visible layer is z=0, and all following will be z=-1, -2, -3 and so on (this done with tool 1). Then I select tool 5 (camera zoom) and set it to 10 or 20. When I now use tool 4 (camera tracking) the layers appear undistorted.

I use pan/tilt (tool 7) quite cautious, mostly with directly entering values (in 5° increments).

Tool 5 (zoom) certainly is focus length, but with counter-intuitive values: the higher the numbers, the shorter the lens. I hit reset quite often when working with camera ...

Posted: Sun Jan 22, 2012 2:26 am
by hayasidist
Hi Slowtiger - I guess we're saying the same thing - to reduce distortion we need a bigger distance between camera and subject and "smaller number" zoom.

The key is the separation between camera and image - so you could leave the camera at its default 1.73 and put the "closest" layer at -20 or whatever...

there's a mathematical relationship (1/(tan(FOV/2)) between FOV (which is what AS uses as "zoom") and lens focal length. As I mentioned before this also depends on the size of "film"... however, I've done some number crunching and (again - need to cross check these - and not sure if they depend on relative z ... but ...) by "rescaling" the AS canvas to be "35mm film" the following approximation "seems to work":

AS zoom , "35mm focal length"

1, .... 1100
3, .... 400
10, ... 110
20, ... 56
22, ... 50
30, ... 40
45, ... 24
60, ... 17
80, ... 12
100, .. 8
160, .. 2

Posted: Sun Jan 22, 2012 4:33 am
by Rhoel
I agree with this ... for 3DTV, the focal length is important, the standard wide-angle setting produces keystoning. Backing the camera off fixes the distortion problem.

Having some kind of chart or feedback on the FOV angle will help a lot.


Posted: Sun Jan 22, 2012 5:40 am
by chucky
Thanks hayasidst, I will check you numbers to see if anything starts looking right.

Posted: Sun Jan 22, 2012 6:13 am
by chucky
Ok using a combination of your chart and by taking the layers out of a group ( thanks to Wes' observation), I can get much more predictable and pleasing results.
I can see now that my initial animation was on a a crazy 16mm lens, no wonder I was getting such bizarre results, without a scale on the objects I had no idea, although I think I may have suspected ( sure I have posted in the past about the default camera being a crazy wide angle, but knowing it's default is pretty much 'endoscopic':wink: really horrifies me). I am more determined to convince the powers that be to set it at a more sensible level, I feel sorry for the full noobies when they try moving the camera ).

Not being able to group stuff is going to be a problem though, but this is all very useful. I can see a way out of this briar patch already.
We might need a feature that allows layers to be in groups without affecting the way they RTC.
Also maybe one of the script magicians could make an altered camera tool that shows a lens conversion ( I don't think it should be up to users to do this though)
Thanks for the discussion and observations on this topic everybody, it has been very helpful to me anyway and I hope some others.
From here I can make a more useful default file with a more neutral starting camera lens and position and I'll put those conversions in a note layer for a reminder.

Posted: Sun Jan 22, 2012 10:47 am
by slowtiger
I assume the default setting of "60" for camera zoom was chosen to easily create a convincing multiplane effect, which works nicely if you move either layers or the camera just horizontally without any tilting of planes.

Posted: Sun Jan 22, 2012 12:14 pm
by hayasidist
camera / lens setup guide:

I think the question is how you handle "3d" aspects... if you draw everything as though you were painting on flat cels then you, the artist, deal with perspective and the default camera set up is just what you need. I think it matches, as far as is practical and straightforward to implement in software, the set-up of a rostrum camera.

However, If you use AS to do the 3d, and you're using the camera as though you were filming on a live set, then I think this is where the "scale" discussion comes back in. By default the height of camera is at the midpoint of the shot. So where the subject is 2 metres away the camera is 1 metre off the ground. But if the subject is 2 miles away the camera is 1 mile off the ground. Arguably, a better default would be to keep the camera "1 metre" off the ground irrespective of the distance to subject - which would change how the ground between camera and subject is seen etc etc. Obviously, you can manually set the parameters in AS to do this. But I can for sure see the value of a tool to help do this positioning and lens selection.

Posted: Mon Jan 30, 2012 6:51 am
by chucky
I think the question is how you handle "3d" aspects... if you draw everything as though you were painting on flat cels then you, the artist, deal with perspective and the default camera set up is just what you need. I think it matches, as far as is practical and straightforward to implement in software, the set-up of a rostrum camera.
absolutely disagree with this assumption, this is the basic reason for this entire thread.

2d layers in a 3D environment.

I am not going to explain this tried and true technique, I have to assume you know of it.

Think of a pop up book with a moving camera, using a 3d environment allows for accurate spatial relationship and animation as the camera moves, furthermore if the camera changes then all elements move in concert without having to re-animate every layer.

These 2D elements MUST remain flat and without distortion in most situations so as not to reveal the effect. Also a variety of lenses should be available as with any set.

Posted: Mon Jan 30, 2012 6:00 pm
by hayasidist
sorry - misunderstanding -- what I was trying to say was that if you had _just_ flat cels lying on the table of a rostrum camera then the AS camera set-up is exactly what you need.

When you put your (2d) objects popping up out of the rostrum camera table (like a pop-up book - as per your example) then - to agree with you - the camera set-up is less than ideal.

Posted: Mon Jan 30, 2012 6:50 pm
by chucky
I think you might still be misunderstanding me hayasidist, never mind. I'm probably not explaining this eloquently or elegantly enough.

I'm not talking about a new technique or anything out of the ordinary.

It is pretty standard now in these days - post rostrum camera.

I don't mean to sound flamey , please don't think that I'm getting frustrated. 8)

Here is what I'm talking about in this demo by Adam Phillips (on that 'other' software) notice that there's no layer tilt or distortion even though the lens is wide enough to accommodate some perspective as the camera tracks back.

It is not an extreme tracking shot but I think it might explain better by way of visual illustration.

Posted: Tue Jan 31, 2012 5:05 pm
by hayasidist
:idea: ah -- ok -- watched the vid -- so do you mean multi-plane and how a layer (in AS) will distort when closer to the (AS) camera as it's "fish eye" lens??? So it'd be better if AS defaulted to a more "human eye" lens?? (and no offence / flame taken or intended - I know how hard communication by written word alone is... ) (oh and btw - about the "post-rostrum era" ... even though my stack of blank cels is gathering dust :wink:, I still use real cameras in the stop motion world - and elsewhere!)