02-11-2012, 06:58 PM
[quote name='Klaus' timestamp='1328938897' post='15792']
Why ? The two values are averaged during the testing procedure. This is a perfectly valid approach.
[/quote]
Hi Klaus,
I do not doubt, that your tests covey useful info regarding the quality of the lenses tested. However there are many lenses for which radial and tangential structures are reproduced vastly different. All proper testing in the "film-days", where the sensors had no structure of its own, took care of it (e.g. typical USAF test target was always arranged such that the ones in the corners were angled to coincide with that).
With digital things are different, the sensors have regular pixel patterns and at least in the early days of digital the sensor resolution was more limiting in many cases than the lens. I understand imatest has their targets angled with respect to the sensor grid, to avoid artefacts from the sensor. This makes sense in the early days, when sensor was limiting thing in your imaging train. However nowadays digital cameras having pixel pitches of about 5µm, corresponding to a resolution of 100 lpmm (2 pixels to resolve a line pair). These values are way higher than the 30 lpmm or 40 lpmm used as max frequency in a traditional lens MTF chart. So in particular in the outer zones of the frame, where the difference between radial and tangential are typically significant, the lens is these days the limiting factor again. It would make more sense, in particular for a site testing lenses, to follow the symmetry of the lens and not the structure of the sensor.
In that context it might be interesting to study oatao.univ-toulouse.fr/303/1/estribeau_303.pdf , which amoung other things discusses the influence the slanting of the target relative to the sensor has on the measurement result of the MTF.
Why ? The two values are averaged during the testing procedure. This is a perfectly valid approach.
[/quote]
Hi Klaus,
I do not doubt, that your tests covey useful info regarding the quality of the lenses tested. However there are many lenses for which radial and tangential structures are reproduced vastly different. All proper testing in the "film-days", where the sensors had no structure of its own, took care of it (e.g. typical USAF test target was always arranged such that the ones in the corners were angled to coincide with that).
With digital things are different, the sensors have regular pixel patterns and at least in the early days of digital the sensor resolution was more limiting in many cases than the lens. I understand imatest has their targets angled with respect to the sensor grid, to avoid artefacts from the sensor. This makes sense in the early days, when sensor was limiting thing in your imaging train. However nowadays digital cameras having pixel pitches of about 5µm, corresponding to a resolution of 100 lpmm (2 pixels to resolve a line pair). These values are way higher than the 30 lpmm or 40 lpmm used as max frequency in a traditional lens MTF chart. So in particular in the outer zones of the frame, where the difference between radial and tangential are typically significant, the lens is these days the limiting factor again. It would make more sense, in particular for a site testing lenses, to follow the symmetry of the lens and not the structure of the sensor.
In that context it might be interesting to study oatao.univ-toulouse.fr/303/1/estribeau_303.pdf , which amoung other things discusses the influence the slanting of the target relative to the sensor has on the measurement result of the MTF.
enjoy