Offset microlenses on sensors

[BJL]

I have just noticed that the new Kodak 31MP 33x44mm sensor has microlenses that are offset towards the center, instead of straight in front of the most light sensitive part of the photosite. That is, they effectively look towards the exit pupil of a traditional shortish focal length lens. (Up to 8º off-perpendicular in the corners.) The same is done in the 10MP Kodak sensor in the Leica digital back for Leica R SLR's.

This is potentially the salvation of some more traditional near-symmetric wide angle lens designs from the problems they have with light reaching the corners of the sensor so far off-perpendicular that sensitivity is distinctly less than at center frame (I call it "microlens vignetting"). Presumably, Kodak is trying to save existing Leica R SLR and medium format SLR wides. Maybe Leica and Kodak also have an eye to helping Leica M rangefinder wides next, though rangefinder wides can have even more severe off-perpendicular issues than SLR wides, so maybe this method is not enough.

More generally, this could reduce the need for highly retro-focus wide angle lens designs for all DSLR systems, potentially reducing cost and weight and improving optical quality.

[Jonathan Wienke]

More generally, this could reduce the need for highly retro-focus wide angle lens designs for all DSLR systems, potentially reducing cost and weight and improving optical quality.
[a href=\"index.php?act=findpost&pid=51649\"][{POST_SNAPBACK}][/a]

No, because the retrofocus design is needed to accommodate the reflex mirror for the optical viewfinder, and has nothing to do with film or digital or microlenses. There has to be enough room between the lens and film/sensor for the mirror to move without hitting anything.

[BJL]

No, because the retrofocus design is needed to accommodate the reflex mirror for the optical viewfinder, and has nothing to do with film or digital or microlenses. There has to be enough room between the lens and film/sensor for the mirror to move without hitting anything.
[a href=\"index.php?act=findpost&pid=51960\"][{POST_SNAPBACK}][/a]

The need of standard micro-lensed sensors for a high exit pupil to get a high degree of near telecentricity sometimes goes beyond what comes simply from designing a lens with high enough back-focus distance to clear the reflex mirror, at least with very wide angle lenses.

For example, 35mm SLR's have lens mounts about 42mm to 47mm from the focal plane. Lenses with exit pupil about level with the lens mount send light to the corners of the 35mm format frame at an angle of 24-26º off-perpendicular; if the rear elements protrude into the body, the angle could be larger. At 25º, Kodak's microlensed CCD's (e.g. KAF-5100CE in E-1, KAF-8300CE in E-500) have sensitivity about 35-50% of maximum, about one stop lost. In fact Kodak's graphs for those CCD's stop at 25º, while for non microlensed CCD's they go to 40º. At 25º, Kodak CCD's without microlenses still have about 90% of maximum sensitivity: negligible "vignetting". The 8º microlens offset would reduce the effective angle to 17º, at which sensitivity is again 90% of maximum; back to negligible "vignetting".

Olympus in its descriptions of the FourThirds standard has mentioned a target of light reaching the corners no more than 7º off perpendicular, far stricter than what is automatic in SLR wide angle lens designs, and leading to the use of more highly retrofocus designs in their lenses that cover wide angles, increasing weight and complexity (lots of lens elements.) This microlens progress could eliminate that design constraint.

Aside: apart from off-setting the microlenses, other progress has improved the off-perpendicular sensitivity of the 31MP KAF-31600 to 80% at 25º and 60% at 30º, so perhaps this too indicates that the problem of "microlens vignetting" is going away.

[Jonathan Wienke]

I'm not convinced that "microlens vignetting" ever was a significant issue with DSLRs in general. The original 1Ds has offset microlenses, and the limited comparisons I've made haven't shown any significant vignetting increase with the 1Ds over film. There may be some difference, but I've never seen any real-world comparisons indicating it has any practical significance.

[BJL]

I'm not convinced that "microlens vignetting" ever was a significant issue with DSLRs in general. [a href=\"index.php?act=findpost&pid=52085\"][{POST_SNAPBACK}][/a]

Maybe not, and anyway it certainly seems to be on its way out as a concern for DSLRs, if it ever was! Unless CMOS sensors have stricter angular limits due to requiring stronger microlenses than CCDs: I have not found such data for anything except FFT CCD's.

By the way, another improvement, in off-perpendicular sensitivity, undermines another claimed limitation; that lenses of aperture ratios beyond about f/1.4 will be of little use with digital cameras because light at the outer edges of the broad incoming light cone are too off-perpendicular to be detected well. With the microlenses of the Kodak KAF-31600, all light from even a f/1.0 lens is detected with at least 75% of maximum "square-on" sensitivity. So once again, sensor improvements seem to be pushing the ultimate performance limits back onto lens designs: that is probably where both speed and angular resolution limits will be determined.