I didn't realize rectilinear lenses created MORE problems in stitching. For some reason I thought it was fisheyes that were tougher to un-distort. I'm glad I asked! (Though, Ken Rockwells review of the Tokina 11-16mm f/2.8 AT-X 116 Pro DX stated the distortion was "simple" and easier to remove than the Nikon equivalent)

The main reason I was gravitating toward a rectilinear lens was the issue of the fixed NPP. From what I've read, fisheyes have an NPP that shifts with the view angle.  I was also interested in the extra speed of the Tokina 11-16mm f/2.8  compared to any other lens in this class.

However, As long as the NPP shift isn't to severe, I get the coverage/density I need, and I can stitch the shots successfully, I'll be happy.

Does anyone  have  experience with the Tokina 10-17mm zoom fisheye? How about the Tokina 11-16mm f/2.8 AT-X 116 Pro DX?  (Remember, this is for 360x180 panos)

The above post shows results from future Papyspheric program when submiting data corresponding to:
- Canon 7D
- Sigma 10-20 rectilinear lens at 10 mm
- Tokina 10-17 fisheye at 10 mm

Note: I say this is a future program because I sometimes find bugs in it ! as a consequence those results are to be handled with care (they should be verified by actual tests of both lenses, that is.)

Your question about Sigma 10-20 mm (but, BTW, you could include the new Sigma 8 mm non-fisheye as a possible solution) corresponds to the tests I did with my Nikon 16-85 zoom and Tokina 10-17 mm fisheye zoom both at 16 mm using an APS-C camera.  I used this 16 mm focal length because this is the widest focal length they share. A comparison with Sigma 10-20 would have been much more interesting. I wrote the posts in French language because I doubt rectilinear ultra-wide lenses being the best solution for spherical panos but am afraid of hurting peoples who dislike or despise fisheyes while not bringing solid enough arguments to base these doubts.

http://www.autopano.net/forum/t8381-tok … 6-mm-aps-c

Why am I in doubt about ultra-wide-angle lenses and spherical panos ?

When the goal is approximating a sphere with flat rectangular images what it's important is:
- how many degrees of the sphere surface each source image can cover
- how bad is the spherical image at some location on the sphere surface which corresponds to the worst results being recorded by the camera sensor

There is no doubt a fisheye can record larger vertical an horizontal FOV than a normal lens when focal length is identical. Something I found by observing my test results and which is confirmed by the theory is that next to the image center both lens "work in the same way". The problem is located near the far sides of the sensor and especially in the corners:

 Full frame sensor size image, rectilinear lenses: 

 S1:       Corners stretching along the diagonal
 S2:       Corner stretching perpendicular to the diagonal
 S1 x S2 : Corner versus center ratio of pixels

   Focal     Diagonal    
   Length    FOV          S1     S2     S1 x S2

   200 mm     12 deg     1.01    1.01    1.02   
    50 mm     46 deg     1.19    1.09    1.29   
    35 mm     63 deg     1.38    1.18    1.62   
    28 mm     75 deg     1.6     1.26    2.02   
    24 mm     84 deg     1.81    1.35    2.44   
    20 mm     94 deg     2.17    1.47    3.2    
    16 mm    107 deg     2.83    1.68    4.76   
    14 mm    114 deg     3.39    1.84    6.24   
    12 mm    121 deg     4.25    2.06    8.76   

     Example: the 2.0 ratio value for a 28 mm lens means that
     if image of an object corresponds to 1 pixels when this 
     image is located next to the sensor center, an image of this 
     same object will correspond to 2 pixels when image
     is being recorded in a corner of the sensor.

In a few words: I doubt ultra-wide-angle lenses being a good solution for spherical panos, I have some clues about that but no consistent proof of that.

Last edited by GURL (2010-03-01 12:50:51)

I'm surprised you've considered, and rejected, the Canon 10-22mm EF-S lens.

On the 7D it makes a stunning wide-angle lens and the f3.5 aperture certainly isn't a real issue.

Last edited by Photosbykev (2010-03-01 19:19:16)

I initially stated that I'd dismissed the Canon 10-22mm EF-S based on cost compared to the Tokina 11-16mm f/2.8. But now that I've read more reviews I'm not as put off by the extra cost. It sounds like a really great lens.

However, now that GURL has put these extra ideas about fish eyes in my head, I feel like I need to do more research.

According to reviews, the Canon 10-22mm EF-S is supposed to have minimal distortion, which should help avoid the problems with wide rectilinear lenses both mediavets and GURL have suggested.

As I've mentioned, I'm attracted to the rectilinear lenses based on the fixed NPP. And they also seem like they'd be more versatile for NON panorama use. 

I want to avoid too many dead ends and expense if I can. But maybe I aught to just settle on a lens and see how it works out.

Hello all,

I can't give the figures for the Canon 7D but they can be roughly extrapolated from the next "practical" results for Canon 40D and Canon 5D Mk2 ...
All figures are given for sphericals pano's !

+/- 100 Mb for Canon WA 10-22 mm @10 mm & 40D (10 Mpix) Preset 1@nadir,4@-60, 8@0, 4@+60 = 19 pics
+/- 85  Mb for Canon WA 10-22 mm @ 10 mm & Canon 40D (10 Mpix) Preset 6@-30, 6@+30 1@zenit = 13 pics
+/- 62  Mb for Tokina FE 10-17 mm @ 10 mm & Canon 40d (10 Mpix) "Snake" preset 4@-30, 4@+30, No Nadir, No Zenit       
+/- 60  Mb for "shaved" Tokina FE 10-17 mm @ 12 mm & 5D MkII (21 Mpix)  Preset only 4@0 !

claudevh wrote:

Hello all,

I can't give the figures for the Canon 7D but they can be roughly extrapolated from the next "practical" results for Canon 40D and Canon 5D Mk2 ...
All figures are given for sphericals pano's !

+/- 100 Mb for Canon WA 10-22 mm @10 mm & 40D (10 Mpix) Preset 1@nadir,4@-60, 8@0, 4@+60 = 19 pics
+/- 85  Mb for Canon WA 10-22 mm @ 10 mm & Canon 40D (10 Mpix) Preset 6@-30, 6@+30 1@zenit = 13 pics
+/- 62  Mb for Tokina FE 10-17 mm @ 10 mm & Canon 40d (10 Mpix) "Snake" preset 4@-30, 4@+30, No Nadir, No Zenit       
+/- 60  Mb for "shaved" Tokina FE 10-17 mm @ 12 mm & 5D MkII (21 Mpix)  Preset only 4@0 !

At some point, the sensor may outresolve the lens, so you may have to downgrade these numbers.

hankkarl wrote:

Speed is not really an issue in panos--you will probably use f/11 or f/16 to get lots of DOF -- and sometimes f/2.8 lenses get worse above f/5.6 or so.

Good point. I usually try to shoot at the smallest non-diffracting aperture I can for panos anyway, since the camera is on a tripod... and I want everything in focus.