[rdenney]

I'm wondering if everyone posting on this thread noticed that your are testing 645 format lenses and I have to assume that your teleconvertor is also for 645 format.

Given that I posted it in Pentax Medium Format, and used a 645Z for the test, I didn’t think that needed to be reiterated. But I do realize there are ways to see a post without going through the forum structure, so I’ll be more explicit in future installments.

Rick “now noticing the confusion” Denney

[normhead]

Looking at the differences in the images, I'm not sure that's need to know. After all, each combo needs to be evaluated on it's own. I'm just seeing some image degradation I wouldn't have expected.

[bull drinkwater]

I am one who uses tc's all the time. I use both the 1.4 and the 2.0. with the 645 200mm,the 400mm,and the 6x7 300mm I've go some pretty good images, for me it's about matching good components. on a side note, shooting with my Sony a-900, using Minolta maxxum 200mm apo G and 80-200mm apo G plus their 1.4 and 2.0 tc's. I've got some "excellent" images. I hope there are others out there who have had good results with tc's.

[kaseki]

I can't remember whether I read this resurrected topic when it was fresh, but I didn't comment if I did. I do have a few comments FWIW.

Agree that art and precision are different aspects.

Agree that lens testing should evaluate the best the lens can do.

Suggest that the ideal system (excluding any cost considerations) is a diffraction limited lens and a sensor that over-samples the image field.

The horizontal and vertical counts of the 51M pixels of the 645Z should not be taken as establishing the Nyquist frequency for sampling. We do not have each pixel acting as a tri-stimulus (3-color) measurement system. Instead we have green red and blue sensitive pixels in an array feeding a Pentax defined algorithm for estimating what a tri-stimulus sensor of 51M pixel count would generate. This algorithm cannot estimate color values for bidirectional pixel-sized details.

I suggest that it is analytically safer to assume that the 645Z sampling at the limit of resolution is 25.5M pixels in the green, and 12.25M pixels each in the red and blue. If a lens can do better than this at the limit of resolution at some contrast then one doesn't have an oversampling sensor relative to the lens. (2% contrast is used for atmospheric visibility range -- I'll accept any choice based on CTF or MTF because I don't think it affects the generality of my argument.) The lens-camera system, of course, may over-sample the image relative to a print or a human eye observing the print.

I think Rick's lens measurements have consistently adopted this last point. Small variations in image quality just detectable when viewed at the seven-foot-wide print magnification are negligible at one-foot-wide print magnification.

kas