In a recent round of testing, I spent an embarrassing lot of time shooting a moth wing, only to realize after I had done several lenses that the wing really wasn't working very well. I know, you'd think I would have noticed sooner, but hey, moth wings had worked well before.
There were basically three problems. First was variability across the field. I was shooting mostly at 4X, field size about 3.7 x 5.5 mm, and over that large of a field the scales were changing color, orientation, shape, and texture. If a corner looked bad, I had no way to know whether it was the lens or the subject. Second was that the moth wing happened to be missing detail in a critical size range that I cared about. All of the lenses were resolving the shapes of the scales and the large rippling within scales, but none of the lenses were resolving the fine striations. So it was like having a resolution test chart with a bunch of line groups missing at exactly the most unfortunate places. And finally, moth wings really aren't very flat, which meant my stacks were getting big. Hmmm...
After much thinking and several experiments that mostly revealed more problems, here's what I came up with (shown with one of the lenses tested -- an MP-E 65).
The target is a piece of 600-grit silicon carbide sandpaper, taped to a microscope slide, and scored on a fine and not terribly regular grid using a spring steel insect pin which had been worn to a chisel point in earlier experiments along the same lines.
Let's work our way in and see what this thing looks like.
This is the whole target, with a marquee drawn to indicate the area covered at 4X using my Canon T1i camera (sensor size of 22.3 x 14.9 mm):
Here's the full 4X image, with a marquee showing the 10X area:
The full 10X image, with a marquee showing the 40X area:
The full 40X image, with a marquee showing a crop to be enlarged. The lower left portion of this frame is an unscored area, consisting of 600-grit uncrushed particles sitting in what must be a layer of glue.
Crop from a 40X image (NA 0.50), shown at 50% of actual pixels:
I'm pretty pleased with how well this target worked out. The scoring process does two things: 1) it creates a recognizable pattern that is easy to line up after changing lens setups, and 2) it crushes a lot of the silicon carbide particles, producing a huge amount of detail at all size scales down to smaller than my lenses will resolve. In addition, the target is generally flat but still has enough depth to reveal color aberrations, and it should hold up well in storage for subsequent tests, at least as long as I can avoid touching the surface.
I was interested to see what appeared as I zoomed in closer and closer. At low magnifications, the sandpaper is mottled gray. But at higher magnifications, its true nature is revealed as a bunch of shiny dark crystals. Light versus dark gray is a matter of how many crystal surfaces are oriented to reflect the light to wherever your eye happens to be. The score lines are brighter than undisturbed surface under most conditions, but when illuminated and viewed from certain angles, they can be almost invisible or even darker than the undisturbed surface.
So, the target really consists of a random collection of bright specular reflections floating in a sea of darkness. This combination is merciless about revealing chromatic aberration.
Here for your amusement are several views of the same chunk of target, as seen by various lenses at 4-5X and 100% pixels. These are individual frames, not stacks. Typically stacked results have much less false color than individual frames since most of the OOF information has been thrown away.
One last comment about color aberrations. If you look carefully at the 4X frame, you may notice that has a slight hue in certain areas. That's caused by longitudinal color aberration combined with a nonlinear transfer function between amount of light and pixel values. Essentially, light that gets spread into dark areas "counts more" than light that stays nicely focused into bright areas. As a result, the color balance shifts from neutral for very OOF regions, to reddish just on one side of focus, neutral again at focus, to greenish on the other side of focus, and back to neutral again for very OOF.
This pattern is particularly easy to see in a thumbnail rendition of a stack shot with a slightly tipped target so that focus and the aberrations slide across the frame in a predictable manner. Here's one I shot with the Olympus 38 mm. I've cranked up the saturation on the right so you can't miss it. The Olympus 38 crop in the previous panel comes from IMG_0083, a small area at image center where focus is perfect and the color shift is passing from reddish to greenish.
I hope this is helpful, or at least interesting.
--Rik
Edit: to clarify CA in single frames vs stacks
