[interested_observer]

Up till now I have been somewhat lazy over the last few years and have asked Pentax's technical support several times (phone and emails) what the actual horizontal and vertical Angle of View (AoV) for the DA 10-17 fisheye lens. Each time Pentax USA tells me that that its 180 degrees (10 mm) and 100 degrees (17 mm), and I say no, that is the diagonal Angle of View. Then they tell me that they need to bundle the request off to Pentax Japan for an answer - and then it goes into the deep dark black hole never to be heard from again.

I finally decided to figure it out for myself. I tried to keep it simple, so for anyone else interested - here it is....

Focal Length -- Width ------------ Height
10 mm FE --- 150 degrees --- 100 degrees
17 mm FE --- 83 degrees ----- 55 degrees

Attached Images

 

---

Last edited by interested_observer; 11-24-2011 at 01:26 PM.

[Adam]

Assuming the proportions are linear that sounds kosher to me, but considering that the DA 14mm has a horizontal FOV of 81 degrees, I would be inclined to think that the fisheye would have more. I will do an empirical test to verify and post my findings!

[interested_observer]

Assuming the proportions are linear that sounds kosher to me, but considering that the DA 14mm has a horizontal FOV of 81 degrees, I would be inclined to think that the fisheye would have more. I will do an empirical test to verify and post my findings!

That is exactly why I asked Pentax. In order to do this to some degree of accuracy, you have to make the assumption that its linear, else-wise you have absolutely no idea. The only folks that really know for sure, would be the actual lens designers (and their engineering notebooks) - apparently they are not divulging anything....

With respect to the 14mm which is a rectilinear lens, we are comparing apples to oranges to a degree. However, the 14mm should be narrower, then again you go back to the fisheye optical formulation and the zoom design implemented.

So, for what its worth, this is essentially an approximation. But - for what I was looking for, its better than what I had before, which was nothing.

An empirical test to verify and posting your findings, would be great!

[RioRico]

IIRC lensmakers have seemingly-valid reasons for specifying diagonal coverage, rather than horizontal+vertical. I'm too lazy to look that up now. And it don't worry me.

But playing with projections is fun. My favorite toy is a Kenko 180 Degree Fisheye adapter, mounted on a short zoom. I first mounted it on a Tak-A 28-80, then an M40-80, but both were too loose for reproducible work. So now I usually put it on the F35-70 or FA28-80. Some AF or A-type multi-ring 35-135 would be best. Too bad my Tokina RMC 35-135 is another loose one-touch...

Anyway, depending on the exact host lens, at around 40mm it produces a full-circle image, and from about 60mm it's frame-filling. Going from 60mm onwards produces various frame-filling projections. I am entertained and enlightened by the changes. And I don't worry about the exact AOV at any setting because it's so easy to tweak.

[panoguy]

Just to add some full-frame confusion to the mix, Dr. Thoby (an actual rocket scientist and fellow pano-enthusiast!) has removed the sun-shield from his Tokina version and mounted it on a full-frame camera, then calculated the limits of the FOV. Doesn't answer your question directly, but has a lot more information to think about!

The Tokina 10-17 design is apparently by Pentax (who else has made zoom fisheyes before), and it is not a straight fisheye projection. Apparently it goes almost to 190 degrees on larger sensors! However, most panorama-stitching software calculates the "real" FOV starting from the theoretical (180 deg. diagonal) and then matching features in two or more images, which were taken while rotating the lens around the optical center of projection ("no parallax point"). My software successfully stitches 10mm shots all the time, and indicates an undistorted FOV of 169 degrees at 10mm, but this is calculated backwards from what the lens designer has done, and is influenced by the vagaries of computer vision algorithms.

So, yeah, fun to contemplate, but beyond me to definitively calculate.

BTW, I have shot (at last count) over 48,000 images with the DA 10-17mm fisheye. Ah panoramas...

[SpecialK]

Should be easy enough to to an actual test, such as putting the camera 10 feet from a fence, marking the edge of the frame on the fence, measuring the width, then doing the math.

Or put a protractor on top of the camera and do some eyeballing.

[RioRico]

...and it is not a straight fisheye projection.

I've read that many fisheye projections are possible. It's like the cartographers' problem of projecting a 3D world onto a 2D plane (map). Every new projection introduces a different distortion.

And yes, the real answer for the OP is to test. I'd skip the protractor; that's a bit rough. But with known distances (subject-to-lens and FOV) the calculation is trivial. Whip out the slide rule!

[aurele]

so far @17mm it is almost the same as what the page of the 10-17 tells here for horizontal field of view

Pentax-DA 10-17mm F3.5-4.5 Fish-Eye ED [IF] Reviews - DA Zoom Lenses - Pentax Lens Reviews & Lens Database

[interested_observer]

Just to add some full-frame confusion to the mix, Dr. Thoby (an actual rocket scientist and fellow pano-enthusiast!) has removed the sun-shield from his Tokina version and mounted it on a full-frame camera, then calculated the limits of the FOV. Doesn't answer your question directly, but has a lot more information to think about!

The Tokina 10-17 design is apparently by Pentax (who else has made zoom fisheyes before), and it is not a straight fisheye projection. Apparently it goes almost to 190 degrees on larger sensors! However, most panorama-stitching software calculates the "real" FOV starting from the theoretical (180 deg. diagonal) and then matching features in two or more images, which were taken while rotating the lens around the optical center of projection ("no parallax point"). My software successfully stitches 10mm shots all the time, and indicates an undistorted FOV of 169 degrees at 10mm, but this is calculated backwards from what the lens designer has done, and is influenced by the vagaries of computer vision algorithms.

So, yeah, fun to contemplate, but beyond me to definitively calculate.

BTW, I have shot (at last count) over 48,000 images with the DA 10-17mm fisheye. Ah panoramas...

I had gone web surfing on the topic before looking for an answer. Why I did not run across these calculators before - I don't know - there are several, apparently all based on the work of one person.

• Pano Calculator

The estimates that I came up with are close enough. Apparently the calculators are using a standard spherical fisheye, and I do think that the Pentax design is a bit different - what it actually is, I don't have the slightest idea.

The lens is wide enough that a ballpark estimate is close enough. Its like going fishing with a thermo nuclear weapon - you don't need pin point precision.

I did come across this, which has some nice descriptions and observations...

• Computer Generated Angular Fisheye Projections

48,000 panos! That is at least a shutter's lifetime.....

so far @17mm it is almost the same as what the page of the 10-17 tells here for horizontal field of view

Pentax-DA 10-17mm F3.5-4.5 Fish-Eye ED [IF] Reviews - DA Zoom Lenses - Pentax Lens Reviews & Lens Database

That's true to an extent. The diagonal FoV is listed, however the Horizontal FoV is just for 17mm focal length. You need both the Horizontal and Vertical FoV for at least both ends of the focal length range - the 10mm and 17mm. Originally I was just looking for the vertical FoV and I would have gone away happy as a clam...

I've read that many fisheye projections are possible. It's like the cartographers' problem of projecting a 3D world onto a 2D plane (map). Every new projection introduces a different distortion.

And yes, the real answer for the OP is to test. I'd skip the protractor; that's a bit rough. But with known distances (subject-to-lens and FOV) the calculation is trivial. Whip out the slide rule!

There are many possible projections. I have used a lot of mapping software - in particular ESRI's Arc system. They have a wonderful book of projections, and several websites on the topic. The second link is ESRI's on line "Book" on projections and about the 60 or so most useful....

• Mapping Center : Map Projection Animations
• http://www.skidmore.edu/gis/Understanding_Map_Projections.pdf

Also, more that what I really wanted to know, but also an interesting read on the topic...

• Fisheyes, by Pierre Toscani

Should be easy enough to to an actual test, such as putting the camera 10 feet from a fence, marking the edge of the frame on the fence, measuring the width, then doing the math.

Or put a protractor on top of the camera and do some eyeballing.

There is a range down in Sierra Vista that I use to use a bit, that was all setup for this kind of stuff. 10 seconds and 2 shots would have answered the question. Well all of that was before I got the lens, and would have been overkill - way over the top.

[panoguy]

48,000 panos! That is at least a shutter's lifetime...

Not that many panos - that's just the number of source images shot with the DA fisheye, and I currently use 45-50 source images *per* HDR pano. Plus, that is spread out over 6 years and 3 Pentax camera bodies. Prior to that I was using Canon with Sigma 8mm and it was never as good. The K-5 is currently at 22,000 clicks after 10 months, so I'm not slowing down...

Any particular reason you'd like to know the FOV of the fisheye, or just curiosity?

[interested_observer]

Not that many panos - that's just the number of source images shot with the DA fisheye, and I currently use 45-50 source images *per* HDR pano. Plus, that is spread out over 6 years and 3 Pentax camera bodies. Prior to that I was using Canon with Sigma 8mm and it was never as good. The K-5 is currently at 22,000 clicks after 10 months, so I'm not slowing down...

Any particular reason you'd like to know the FOV of the fisheye, or just curiosity?

I was reading it as 45,000 finished panos.... Regardless, that is still an awful lot of work!!

As I was writing the last response, it occurred to me that I had a measurement "range' nearby. So, I pulled out some shots, and aligned the landmarks to Google Earth and used the tape measure to form a triangle of 3 known distances and then let WolframAlpha site do the heavy lifting.

• a triangle with sides of 106, 143, 171 calculate interior angles - Wolfram|Alpha

For 10 mm I was measuring 130 degrees wide (landscape) against the 150 degrees calculated. I could believe that each side of the rectangle cut off 15 degrees from the image circle, but lopping off 25 degrees was a bit startling. So I took camera and went back this afternoon and took 4 sets of images - this time much more carefully, making sure that I had distinguishable landmarks that I could identify in Google Earth's overhead views, to enable easy measurements.

So here is what I came up with on the more carefully performed measurements...

• 10 mm landscape - 135 degrees against the 150 degrees calculated
• 17 mm landscape - 85 degrees against the 100 degrees calculated
• 10 mm portrait - 85 degrees against the 83 degrees calculated
• 17 mm portrait - 56 degrees against the 55 degrees calculated

There are probably several feet of measuring error possible on each of the 3 measurements, which would probably translate into +/- 2 degrees.

• 10 mm we have a FoV of 135 degrees wide by 85 degrees high
• 17 mm we have a FoV of 85 degrees wide by 56 degrees high

Overall, I would have expected a lot wider and higher especially at 10mm. A 12mm rectilinear lens provides 88 degrees wide by 66 degrees high, which essentially matches pretty well the fisheye's coverage at 17mm.

Why my interest? Well a bit of morbid curiosity I suppose, along with I have a table for my lenses for the Nodal Ninja 3 rotator settings that I was updating and always had a guess for the fisheye. Also, I was thinking of going out this evening and.or tomorrow morning and doing a 360 panos. I had never done one and was wondering what I should set things to.

[Smeggypants]

Up till now I have been somewhat lazy over the last few years and have asked Pentax's technical support several times (phone and emails) what the actual horizontal and vertical Angle of View (AoV) for the DA 10-17 fisheye lens. Each time Pentax USA tells me that that its 180 degrees (10 mm) and 100 degrees (17 mm), and I say no, that is the diagonal Angle of View.

I have a 10-17. The pentax specs I have always read have stated that it is 180 -> 100 deg diagonal angle of view. In practise by eyeballing I've found this to be correct.

can't see the dilemma

[panoguy]

Why my interest? Well a bit of morbid curiosity I suppose, along with I have a table for my lenses for the Nodal Ninja 3 rotator settings that I was updating and always had a guess for the fisheye.

No need to guess. At 10mm, you should start with the bottom rail set to 44.5 and the top rail set to 95. Then rotate the head so your camera is pointing straight down, and check (w/ Live View - zoom in or turn on the grid) if the center of the NN logo (white triangles) is in the center of your view. If not, adjust the bottom rail distance. Also, looking at it from the side in portrait orientation (the default), the edge of the lens glass where it meets the barrel should be just over that same center of rotation.

Also, I was thinking of going out this evening and.or tomorrow morning and doing a 360 panos. I had never done one and was wondering what I should set things to.

Again, at 10mm, shoot 6 around (60 degree increments) and then one straight up, and if you want to see the ground and your tripod, one down. If you don't like having the vertical arm in your photos (see last image below), shoot a second "down shot" after you've rotated the head 180 degrees - you can then layer the two shots in Photoshop (before stitching) and remove the vertical arm. Stitch these with a decent, fisheye-capable stitching program (Autopano Pro, Stitcher 2009, Hugin (free!), PTGui) and any slop will be taken care of in blending unless you're really way off the no parallax point.

Here is what those 8 images (with a down shot) would look like if you set it up in my friend's loft (NB: that's an NN5 I'm using there):



Quick extra 360 pano tips: set your camera to MF and pre-focus to infinity if you're outdoors, or just below for large indoor spaces. Also, use M mode so your exposures are identical (or autobracket like I do and enfuse/tonemap them together).

Have fun!

[interested_observer]

Hi Pano,

Thanks for the information. Of my choices of 120, 90, 60 or 45 - 60 made the most sense, but I never knew for sure. The quick overview was wonderful and very insightful - especially on getting rid of the arm in the downshot. I was going to get up early during my week off and give this a try - but I never did. I wound up taking our oldest son back up to NAU in Flagstaff yesterday and I did find a wonderful vantage point to shoot from in Sedona. One half was spectacular, however the other half was the elementary school's junk yard and bus parking lot - so I passed on the 360 degree set of shots. Up till now I have never really needed a rail, when just using the ballhead. There was one excellent shot with a tree in the foreground. I guess I will be getting a rail also.

[Smeggypants]

Thanks for the tips Panoguy - I came across this thread again as I just shot today ( for the first time ) some pano elements with the 10-17 at 10