tilt-shift lenses and digital camera sensor

[robijanto_soetedjo]

The micro-lenses on dSLR sensor require the light to hit the sensor

close to 90 degree angle. Would this restriction cause problems with

tilt-shift lenses? I imagine that the rays from T/S lens come at a

sharp angle.

[les]

Even with a normal lens (not tilt-shift), at any point on the sensor the light comes from many different angles. Each pixel gets the light from the whole circle defined by the current aperture. The angles are larger with wide angle lenses, smaller with telephoto. The range of tilt is small compared to these angles, particularly on wide angle lenses, so there should be no problem (or if there is, it is not because of tilt-sahift).

[mark u]

It seems that the 24mm TSE has acquired a reputation as a less successful lens with digital (compared with the 90 TSE). I think this is partly due to the interactions of the imaging rays and the sensor optics, including the filters in front of the sensor.

 

If you consider pure shift, then there is a portion of the image that lies directly in line with the lens axis, where quality will be the same as if the lens was centred normally. However, the other edge of the frame will be closer to the edge of the imaging circle. The lens was designed to operate with film, with rays focussed in the thin light sensitive layers. With digital, these rays first hit the IR, UV and anti-alias filters and different wavelengths can be diffracted by different angles, leading to sensor optics induced CA. Depth of focus (effectively the accuracy with which the sensor has to be aligned) is of the order of 0.1mm at wide apertures - a distance that is not insignificant in the context of the sensor filters. Stopping down can help to mitigate these effects, because it both narrows the light cone focussed onto each point of the sensor and it increases the depth of focus. The approximately 30 degree acceptance cone of the microlenses can also lead to vignetting over and above the vignetting that might be apparent with film use. The extent of these problems is dependent on the crop factor and the degree of shift (i.e. what portion of the image circle you are using). A similar analysis applies to tilt, although the geometry is more complex.

 

Of course, there is no other way of achieving tilt effects, although "converging verticals" corrections equivalent to shift can be applied in software at a cost of some image degradation due to the deformation.

[chrisjb]

Hi, just a side thought after reading Marks post, I wondered if the Zork shift adapter would be more succesful with digital, considering it requires a MF lens with a larger inner circle. Would this reduce the angles of light onto the sensor surface and inprove image quality.