Lens Aberration in Superzooms?

Started Nov 28, 2013 | Discussions thread
OP Stephen Barrett Contributing Member • Posts: 683
Re: Lens Aberration in Superzooms?

Thanks for the comments, Jerry.
It was a lot of work, but most of it was avoidable by using one of MTF software programs that I cited. All my tests were in the centre of the lens with the slit 5.7 degrees from vertical.

Loss of Contrast in Photos:
The biggest deviation from theory is at f/2.8, where the contrast drops quickly with frequency. I would think that such a large loss of contrast would show up in photos. I often do have problems with low contrast in photos, in spite of setting the exposure compensation to -2/3. So I often adjust highlights and contrast in post-processing as well. But all of that could be a result of my inexperience as a photographer so I don't dare to blame my SX30.

I think that the rapid drop of contrast with frequency at f/2.8 is real, but feel less confident about deviations from theory at higher frequencies. First of all, the theory has some uncertainty at higher frequencies , such as OLPF and Bayer-demosaicing parameters. Also, there are probably some imperfections in my testing technique and some statistical error because of my small sample sizes. The latter doesn't seem to be a big factor because I have repeated some of the tests and got very similar results as the first time.

I am not sure that doing slanted-edge tests (as opposed to my slanted-line tests) with commercial software would get rid of the uncertainties in resolution. The recommendation for slanted-edge tests is to use low-contrast targets, but that does not seem like a good solution to me, as it only makes the halos less visible. Perhaps some sort of apodization to reduce the halos in the LSF would be better. I tried completely removing halos artificially. While it was instructive, it went too far because there should be some ripples from diffraction.

Using the USAF 1951 target at 60 m and full optical zoom at f/5.8 on my SX30, I measured 30 microradian resolution, which converts to 0.316 c/px (approx 9% CTF or 7% MTF).
Using a slope of -200% MTF per c/px, 9% MTF would correspond to 0.306 c/px.
I would estimate that this is accurate within 10%.

From my slanted line test at 10 m full optical zoom (150.5 mm) and f/5.8 0.292 c/px. So the agreement is within 5%, which is not bad, but I trust the bar chart more. I had hoped that slanted-edge or slanted-line tests would get me more accuracy and precision of resolution but there is the problem of "spurious resolution", so I don't feel I've gained anything there.

Halos and Contrast:
I now feel that the halo and contrast problems are more important than precise resolution numbers. For my photography, I am satisfied with the kind of resolution I get with closeup lenses. In fact, I often back away from full zoom where resolution is highest, in order to avoid the dreaded halos, which take hours per shot to remove by cloning.

Testing More Cameras:
Both you and Hatstand have suggested doing similar testing on other cameras. I agree. It doesn't have to be anything fancy, just some cropped shots of a black-white edge for various combinations of f-number and zoom.

 Stephen Barrett's gear list:Stephen Barrett's gear list
Canon PowerShot SX30 IS +3 more
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow