Large format angle of view

I worked through some math and found some elegant things related to field of view at 45 degrees (and interestingly 1 + sqrt(2)), 90 degrees (f = 1.0, so X' = f*X/Z has no scaling). Then I decided to use the 1DX II to do an experiment with a human in the loop. I put the 24-70 F2.8 II on and with both eyes open looking through the optical viewfinder adjusted zoom until the width of a door handle matched in both places and stereo fusion was easy. I got an interesting number (not 43mm).

I repeated the experiment with the 24-105 F4L on the EOS R and got the same result.

Everyone with a full frame camera with a 24-70 range zoom lens can try this themselves. The result will settle the normal lens debate (don't expect results to match exactly, however the range = normal, vs wide or tele).
When set correctly, you can pan your head and both views will match. If too wide or tele, the views won't match. The range where it matches = normal.

EDIT: tried this with the 80D and optical viewfinder- surprisingly got the same result (in mm), the difference being a narrower image crop (took photos with FF and crop cameras and compared).
 
Last edited:
jcs said:
I worked through some math and found some elegant things related to field of view at 45 degrees (and interestingly 1 + sqrt(2)), 90 degrees (f = 1.0, so X' = f*X/Z has no scaling). Then I decided to use the 1DX II to do an experiment with a human in the loop. I put the 24-70 F2.8 II on and with both eyes open looking through the optical viewfinder adjusted zoom until the width of a door handle matched in both places and stereo fusion was easy. I got an interesting number (not 43mm).

I repeated the experiment with the 24-105 F4L on the EOS R and got the same result.

Everyone with a full frame camera with a 24-70 range zoom lens can try this themselves. The result will settle the normal lens debate (don't expect exact results to match exactly, however the range = normal, vs wide or tele).
When set correctly, you can pan your head and both views will match. If too wide or tele, the views won't match. The range where it matches = normal.

EDIT: tried this with the 80D and optical viewfinder- surprisingly got the same result (in mm), the difference being a narrower image crop (took photos with FF and crop cameras and compared).
Click to expand...

I have done basically this before with my 5D and, interestingly enough, it was ~70mm, if memory serves me correctly.
 
Run&Gun said:
I have done basically this before with my 5D and, interestingly enough, it was ~70mm, if memory serves me correctly.
Click to expand...

I was able to fuse 70mm too, however as soon as I panned it was clearly too much. Try it with your F55.

Putting a loupe on the iPhone X 1.6x zoom matched in the same way (1x = wide lens, 2x = zoom, so 1.6x makes sense as normal: ~26mm (FF eq) * 1.6 = 42mm. https://improvephotography.com/5546...th-of-an-iphone-camera-and-why-should-i-care/.

On all 3 Canon cameras I got ~52-55mm. The 80D matched magnification and perspective at around 52mm, and the image was cropped as expected (optical viewfinder and photo taken compared to FF). Testing again on the EOS R in 4K (crop), I got 35mm, which if the crop is 1.75 = ~61.25mm. Without further analysis into the full optical view chain, we're getting between 40-60mm for 'normal' as noted elsewhere.

When doing this experiment, when zoomed correctly you should get a good 3D image, and when panning it should look correct and natural (aka normal). Important to turn off all IS.

Thus it's clear that focal length for a 'perfect lens' where magnification comes from focal length is what determines wide, normal, and tele in terms of matching the human eye perspective. We can alter the crop factor and it won't affect the perspective at all. For example, shoot full frame and crop the image to S35: perspective projection won't change, only the field of view. Optics change the magnification (vs. e.g. a pinhole camera where it's only focal length).

Perception of the visual field does change based on concentrated focus and 'peripheral vision' types of awareness (it's possible to do both at the same time; some people don't use peripheral vision very often (tunnel vision); there are techniques to improve it e.g. using your hands and paying attention as they go wider until you can't see them anymore (taught in NLP psychology)- possible to get close to 180 degrees visual awareness).

This is kind of the opposite test used to prove the full frame look is a myth: we're not matching final field of view, instead we're matching magnification and perspective of an object matched to the eyes (where e.g. 80D image will be cropped (making final image look tighter), however perspective will match with FF).
 
Last edited:
combatentropy said:
Can you elaborate? The traditional normal lens has a diagonal angle of view of about 45 degrees.
Click to expand...

Was curious if there was anything happening related to right angles since AOV of 90 degrees has a perspective projection scaling factor of f = 1.0, X' = f*X/Z, or X' = X/Z. Yeah, the sensor diagonal defines the image circle and with a focal length equal to that we get an AOV of 45 degrees, which is f = 1 + sqrt(2). Not like the Golden Ratio, just interesting that it appears.

I was curious why folks were writing for FF the diagonal is 43mm and the AOV is 55 degrees (not 45), do I decided to test with real cameras. I found that on Canon FF cameras I need ~52-55mm to match my eye (both optical and digital VFs). Seeing that the 80D also needed 55mm and thinking about it more, realized that focal length (effective lens magnification) was what was important to match what the eye sees. Sensor size then determines the field of view. If we were to optically or digitally add additional zoom to the viewfinder, then match the camera zoom so the subject (somewhat) matched the open eye, the perspective would no longer match. Need to think about how to explain this more: why the crop 80D still worked for perspective and magnification (with cropped edges) for an object in center frame at 55mm compared to FF, however obviously we can't put a ~50mm lens on a 5.6mm wide iPhone X sensor (the tele is stated at "6mm = 52mm FF equivalent" (I get 52/7 (crop factor) = 7.43mm)) and get a normal look: a 50mm lens is not a 50mm lens everywhere.

In summary: ~40mm to ~60mm (FF equivalent) testing with real cameras and the eyes does match well with published results for normal lenses.
 
JCS
Yep looking into the high end DSLR one can really see it. Because your 'monitor' is exactly the right size and ones eye positioned correctly - you place your eye correctly and the gg screen is of the same size as the capture sensor.

Even cheap DSLRs often have a pentaprism that is '80% scale' - the high end DSLRs have 100% scale pentaprisms (which are more costly to make and larger)

Also often (cheap) lenses marked wrong - I would not trust the 24-105 as a reference tool. You might buy a '24' and actually it is a 24.9 for example. This is why one can get results of maybe 45-55 giving a 'normal' result.

And cameras sensor size changes - ie not all 'S35' chips are the same size - to my memory my FS100 had a smaller chip than the F3.

I dont think one is going to get a really good answer without going in the lab.
 
The zoom set for one eye only encompasses that one eye, vision is stereo and both eyes need to be factored into what looks correct or "normal". The field of view with both eyes will likely be very different from one eye and one lens.
 
Greg_E said:
The zoom set for one eye only encompasses that one eye, vision is stereo and both eyes need to be factored into what looks correct or "normal". The field of view with both eyes will likely be very different from one eye and one lens.
Click to expand...

Difference between one and two lenses for perceiving scale drops with subject distance. Close on a 14 we see huge distortion in camera (no second eye) - a passing jet plane and our eyes are too close too use the two views to calculate size - hence our inability to perceive a jet being much bigger than a large bird; photographic single lense view is basically the same as binocular view.
 
@morgan- yeah it’s not perfect though we can see what works with a real test vs. armchair photography/video :).

@Greg- for this test we match magnification and perspective, allowing the brain to fuse the stereo pair, even though FOV is greater for the eye only view. I’ve done a lot of work in stereoscopic systems including shutter glasses, VR with 3D and head trackers, and solving the color anaglyph problem (stated by many to be impossible (abandoned the patent when not enough industry interest for licensing- still secret)). Thus I realize I may be able to fuse image pairs that could be hard for those with less practice. In any case, we’ve got a simple test everyone can try at home to learn first hand what a normal lens is.
 
That viewfinder exercise is just finding what's the viewfinder magnification of the camera you're using. On the 1DXII it's quoted as 0.76x. And guess what that's defined against? Exactly.
 
Samuel H said:
That viewfinder exercise is just finding what's the viewfinder magnification of the camera you're using. On the 1DXII it's quoted as 0.76x. And guess what that's defined against? Exactly.
Click to expand...

Yup- 50mm! The 80D is .95x which performs the extra magnification to make the viewfinder cropped (matching the 1DX II with the same focal length)- makes sense now. Both the FF and crop cameras see what the unaided eye sees (approximately) at 50mm. Thanks for that data point!
 
Yeah, but now we're back at square one: 50mm on FF is what people consider "normal" and it's the point of reference for a lot of things. Which may or may not mean anything.
 
Just tried out the Leica SL2 with 24-90 2.8-4: got ~70mm to match the VF. Then I asked the saleslady to try- she got 52mm. I was wearing contact lenses- while glasses 'make the world smaller' (near sighted), didn't think contacts did that so much (much less than glasses)- though that could explain the difference.

Sure- 50mm lenses on FF can 'chameleon' based on frame and distance to subject.

The essence of ~50mm on FF being labeled normal is that it closely matches the unaided eye with a similar perspective (projection) and lack of perceived distortion and compression. It's more of a geometric technical label than a label of what's perceived as normal, since unless viewing playback in a mirrorless camera in the VF, it won't necessarily match viewing in other sizes or distances: monitor, large screen, print etc.
 
You must log in or register to reply here.
Back
Top