"Low Budget" focus stacking setup.
 

Thanks for posting that. For those interested in description of pretty much the same set-up for botany, see
Geiger, D. L. Imaging small orchid flowers using visible light. Orchid Digest 77: 112–123.

pdf available from
http://www.vetigastropoda.com/abstra.../poppapers.php

A couple of points, may have more later. The authors make a critical mistake by using an APS size sensor on the SLR camera. A full size sensor gives superior results for larger reproduction sizes (for web, it won't matter). Second, magnification of MP-E 65 is limited to 5x (with teleconverter to 7/10 x), while the Leica microscopes go quite a bit higher. I haven't used the Leica, but my Zeiss system with high quality planapo lenses is way better than the MPE 65 at higher magnifications. Cropping won't work.

With straight MPE 65 one is limited to object area 5x7 mm in size. Most orchid flowers are larger, but some are smaller, so it does matter.

Agree fully on preference for Zerene. Have used overpriced automontage, as well as HF (OK), but not combine z. Zerene supports multicore processing, which is a big help for large stacks.

As promised here some more details. The authors conflate file size (pixels x pixels) with resolution, the latter is the distance between two points that can be seen as separate, or measured in line pairs per mm (not pixels x pixels). This is pervasive throughout the paper.

Table 4 has some strange entries in that pixels per mm is not # of pixels divided by sensor size; it is by a factor of around 10 too high. #µm/pixel (aka pixel size) is too small by a factor of 10; should be in the single digit micro meters, not in the hundreds of nanometers.

They completely ignore the problem of pixel size at the sensor, which is critical. If they would understand it, the 10x error in Table 4 would not have crept in. 200 nm would be theoretically unresolvable with light (wavelength/2, where wavelength is 400-700 nm), even disregarding any additional limitation due to lenses (see below).

In terms of overall cost assessment the paper is fine, but the underlying mechanisms are not properly addressed. The real fun with resolution limitation does not start until you pass the 5x threshold. At 20x and above, it becomes critical to properly understand, such as resolution limit = 1.22 x wavelength divided by numerical aperture. Otherwise you waste lots of pixels (storage space, computer time) on diffraction blur circles.

It also seems that they use for stacking steps the full depth of field, whereas most z-stackers would use 70% of that value to have a bit of safety margin. Using thinner slices does not help; I've done some experiments on this down to 0.1 of DOF on a 200+ slice stack.

I do use the Cognysis System with a Canon DSLR and the MP-E 65 mm lens, so know precisely how it works. I'm tempted by the new X3 controller. There is also an iPad application to control camera more easily, don't remember the details off the top of my head.

There are some other, minor issues. If anybody is interested in particular aspects, I may have more to say. I wonder who reviewed that ms.