Originally posted by TDvN57 I did a bit of reading on power spectral density and it appears that each lens, or perhaps lens batches or even perhaps types will have their own PSD characteristics.
This brings a totally new variable into the selection of lenses. I doubt that any lens manufacturer would publish this data and a calibrated measurement of each lens does not seem practical.
I wonder if the comparison between the f stop and t-stop values might be a layman's check to come to the same conclusion as a full psd analysis. I know these are two totally different issues, but sometimes some unrelated characteristics very in unison with others under similar conditions. Or perhaps best yet is just take a picture and compare to a known reference. :-)
Kaseki, any thoughts on this?
I let some EE drift into OE. It would be better to think in terms of modulation transfer function MTF, which is a non-statistical (deterministic) measure of the same thing. The EE PSD has an abscissa of cycles per second, or radians per second; the MTF (at the focal plane) abscissa is cycles per mm. One can download from Kodak, for example, the MTF (by color) of films such a Portra. One can read up on MTF of spatial filtering, such as the array we are sensing with -- a sinc-squared function, as I recall. There is an MTF equation for diffraction. There is an MTF for the lens assembly that certainly varies with zoom and f/no. The lens assembly MTF can be measured with a decent optical bench, a Nikon autocollimator, and a focal plane microscope, for example. What I was trying to hint at was there are a lot of differently shaped MTF functions that pass the value 0.5 at the same spatial frequency, so they appear to have the same nominal bandwidth, but may yield images of varying character. This is also true of the 90% and the 10% points.
The product of the MTF functions is the overall MTF of the entire image chain. Camera processing and post processing can further modify the overall MTF shape.
Sometimes magazines make such measurements and report them. I recall such a source in Germany that reported on film scanners, and maybe other things.
For a technical description, the lowest price unused source I can think of, besides the Internet, is Edward O'Neill's
Introduction to Statistical Optics, reprinted by -- who else -- Dover Books.
A less hairy, or maybe equally hairy but less grad student level would be the likely out-of-print
Digital Image Processing, by Wm. Pratt.
As I recall, T-stop is just F-stop with transmission included. Unless the effect in the images is due to a lot of internal scattering, I wouldn't expect the T-stop differences between lens assemblies to inform us much about high spatial frequency resolution.
On top of all this, it occurs to me that auto focus systems (including one's eyes and brain) might maximize contrast in different ways that could move focus from where the finest detail contrast is maximized to where some other spatial frequency contrast is maximized. And this effect could be MTF shape dependent.