Jason,
Thanks for the tests and analysis -- this is very helpful!
A few comments...
1. I agree that stray light is a huge problem. However, I do not think of the problem in terms of lighting but rather in terms of subject size and its surrounding environment-- more precisely how much bright area lies outside the field that is imaged onto the sensor. If everything outside the imaged field is black, then all the light that gets through the aperture is angled so as to eventually hit the sensor, not the barrel. But light coming from bright areas that are much outside the imaged field is angled so as to eventually strike the barrel. It is those bright areas outside the field, in combination with the reflective barrels, that cause the stray light problem. One implication is that the stray light problem can escape notice when testing this lens with either a small isolated subject or with a trans-illumination system that provides a field stop.
2. Best practice is to avoid stacking in tests with flat resolution charts. The main reason is that the stacking process may do sub-pixel shifting of the images in order to align various frames, and the associated interpolation can mess with the contrast of pixel level detail. The canonical example is when a half-pixel shift turns alternating pixel values of 10,90,10,90 into 50,50,50,50 and the detail disappears completely. It's great to shoot a stack in order to guarantee capturing best focus, but then select just the sharpest individual frames for comparison. In case of curved field or slightly angled test target, you may need to select different best frames in various parts of the field.
3. Sharpness is not just a function of NA. The performance mentioned by lothman where NA 0.5 looks less sharp than NA 0.3 could easily be caused by aberrations that are worse at the wider aperture.
https://www.bobatkins.com/photography/t ... ution.html has some good discussion of this issue. Note in particular that a lens with 0.5 lambda wavefront error retains the same high frequency limit as a perfect lens of the same aperture, but has severely degraded contrast at lower frequencies. The effect can be quite dramatic. At
viewtopic.php?p=136282#p136282 and
viewtopic.php?t=20594 , I show a damaged objective that still makes a pretty good picture of a resolution test chart, even though its performance for real subjects is quite degraded. On the other hand, if the loss of sharpness is well behaved so that the MTF curve does not drop too close to zero at any frequency, then digital sharpening can undo most of the loss, albeit with some increase in noise.
--Rik
Edit: corrected typo that said "NA 0.3 looks less sharp than NA 0.5" .
