The different lighting thing makes perfect sense - although in thinking about it I\'m surprised we don\'t see it more often.
The phase detection sensors have lenses that split the light from each side of the image, then bend it back to focus on a pair of sensors. I think it\'s safe to assume those tiny lenses are not apochromatic, that is they aren\'t correcting so that different wavelengths of light are all bending in the exact same manner.
Halogen lights, like most artificial lights, are going to put out a very narrow spectrum. This is an graph from wikepedia for streetlights, but it gives the idea.
Sunlight, on the other hand, gives a broad peak of multiple wavelengths.
I assume that the AF sensor in a Canon camera is tuned to one wavelength of light, probably something in the green spectrum, but I don\'t know for sure. If the coatings on a Sigma lens are a bit different than Canon (they probably are) then they\'ll pass a slightly different spectrum through the image.
It probably won\'t matter with sunlight, where there spectrum is very broad without narrow wavelength peaks, but could matter with artificial light - halogen, LED, some fluorescents. The different spectrum would bend a slightly different amount than the focus algorithm in the camera is planning.
If this is the case, I\'m curious if we will see something similar with other third-party lenses (although it probably will require a sharp third-party autofocus f/1.4 lens to see it, so that limits our options to observe it). It would also seem that a Nikon camera with a Sigma lens might not exhibit this behavior but a Canon camera might or vice versa.
Then again, I\'m probably way off base. Is there a doctor (of optical physics) in the house?
The different lighting thing makes perfect sense - although in thinking about it I\'m surprised we don\'t see it more often.
The phase detection sensors have lenses that split the light from each side of the image, then bend it back to focus on a pair of sensors. I think it\'s safe to assume those tiny lenses are not apochromatic, that is they aren\'t correcting so that different wavelengths of light are all bending in the exact same manner.
Halogen lights, like most artificial lights, are going to put out a very narrow spectrum. This is an graph from wikepedia for streetlights, but it gives the idea.
Sunlight, on the other hand, gives a broad peak of multiple wavelengths.
I assume that the AF sensor in a Canon camera is tuned to one wavelength of light, probably something in the green spectrum, but I don\'t know for sure. If the coatings on a Sigma lens are a bit different than Canon (they probably are) then they\'ll pass a slightly different spectrum through the image.
It probably won\'t matter with sunlight, where there spectrum is very broad without narrow wavelength peaks, but could matter with artificial light - halogen, LED, some fluorescents. The different spectrum would bend a slightly different amount than the focus algorithm in the camera is planning.
If this is the case, I\'m curious if we will see something similar with other third-party lenses (although it probably will require a sharp third-party autofocus f/1.4 lens to see it, so that limits our options to observe it). It would also seem that a Nikon camera with a Sigma lens might not exhibit this behavior but a Canon camera might.
Then again, I\'m probably way off base. Is there a doctor (of optical physics) in the house?
The different lighting thing makes perfect sense - although in thinking about it I\'m surprised we don\'t see it more often.
The phase detection sensors have lenses that split the light from each side of the image, then bend it back to focus on a pair of sensors. I think it\'s safe to assume those tiny lenses are not apochromatic, that is they aren\'t correcting so that different wavelengths of light are all bending in the exact same manner.
Halogen lights, like most artificial lights, are going to put out a very narrow spectrum. This is an graph from wikepedia for streetlights, but it gives the idea.
Sunlight, on the other hand, gives a broad peak of multiple wavelengths.
I assume that the AF sensor in a Canon camera is tuned to one wavelength of light, probably something in the green spectrum, but I don\'t know for sure. If the coatings on a Sigma lens are a bit different than Canon (they probably are) then they\'ll pass a slightly different spectrum through the image.
It probably won\'t matter with sunlight, where there spectrum is very broad without narrow wavelength peaks, but could matter with artificial light - halogen, LED, some fluorescents. The different spectrum would bend a slightly different amount than the focus algorithm in the camera is planning.
If this is the case, I\'m curious if we will see something similar with other third-party lenses (although it probably will require a sharp third-party autofocus f/1.4 lens to see it, so that limits our options to observe it). It would also seem that a Nikon camera with a Sigma lens might not exhibit this behavior but a Canon camera might.
Jan 13, 2013 at 12:33 PM
Previous versions of RCicala's message #11265597 « Sigma 35mm f/1.4 DG HSM - AF »
