All digital camera lenses are designed to send their image through the glass filter pack in front of a digital sensor. This gave me the idea to use reversed digital camera lenses to shoot through water. Most cameras have filter packs that are only about 2mm thick, but Micro Four Thirds cameras have filter packs of 4.2mm thickness. This means I can use MFT lenses in reverse to shoot through the equivalent amount of water, which is about 4.8mm.
For these lenses, there will be a plane of sharpest imaging 4.8mm below the surface of the water, and there will be a tolerance zone above and below that depth. The thickness of the tolerance zone will depend on the aperture. I can estimate the depths of that tolerance zone as a funtion of aperture using the formula I extracted from the Thorlabs graph that we discussed in an earlier post:
https://www.photomacrography.net/forum/ ... sc&start=0
If I want the zone of tolerance to include the surface of the water, then the tolerance zone should be about 4.8mm thick in each direction from the plane of best sharpness. My formula says that the aperture which will give this tolerance is between f/2.8 and f/3.2, depending on how the aberrations vary with wavelength (we aren't sure about that). Thus a reversed MFT lens set to 2.8-3.2 should give a sharp image from a depth of about 9mm to a depth of near zero mm. There are lots of uncertainties about this formula and its validity, but it gives me a starting point to test empirically.
I first tested it on a reversed Four Thirds lens, because I wrongly assumed that Four Thirds cameras used the same filter pack thickness as Micro Four Thirds cameras (since MFT cameras support Four Thirds lenses and their adapters). This is what I got last night, using a tiny new species of orchid sitting on its side in 5mm of water in a glass cell. The lens is the famously sharp Four Thirds Olympus 50mm macro lens reversed, set at f/3.2, on a Mamiya 200mm tube lens. The magnification is 4x, on a FF sensor. Please ignore the stacking artifacts and floating garbage. None of these photos are sharpened or adjusted.
The pointed petal pointing downwards is resting on or near the floor of the cell, at a depth of 4-5mm. If the lens had been corrected for a 4.2mm sensor filter pack, this petal should have been perfectly sharp. Instead it shows the classic smeary cross-shaped aberrations commonly seen when shooting through water, here at 100%:
The petal pointing upward has its point very near the surface. It is fairly sharp at 100%:
So obviously I was wrong about Four Thirds lenses being corrected for a thick sensor filter pack. Too bad, becuase they have great working distances.
MFT lenses have about half the working distance of Four Thirds lenses, but at least they really are corrected for a thick sensor filter pack. I used another flower from the same orchid species today, again in 5mm of water, this time using a reversed cheapo Panasonic 42.5mm f/1.7 MFT lens, used at f/3.2 on the same Mamiya 200mm tube lens:
Now the petal pointing down, which is at the bottom of the cell under about 5mm of water, really is sharp, as predicted by theory (100% crop):
And as predicted by my formula, the tolerance zone at f/3.2 extends to the surface; the petal pointing upwards, whose tip is near the surface, is sharp (100% crop):
In that set-up I have wasted a large part of the tolerance zone; I could do better by raising the water level to 8mm, so that the zone of sharpest imaging cuts through the middle of the flower instead of passing through the lowest edge of the flower. This does turn out to be better, and has the additional advantage of eliminating the effect of the floating garbage seen in the other pictures:
Here is a 200% blowup of the bottom petal, 8mm below the surface, showing cell-level detail:
And the same for the other petal, about 3mm below the surface:
So we now have a way of imaging objects in water that are up to 7 or 8mm tall in the z-direction. This approach has the additional advantage of being telecentric, for all MFT lenses; the ratio of successive scale factors in Zerene is 1.000003. So any size object in the x-y direction can be stitched together perfectly. The Mamiya tube lens with a shift adapter also allows me to make small stitches by moving the FF sensor across the medium-format image circle.
Resolution can be increased by opening the aperture (if the lens is good enough), but this will disproportionately decrease the tolerance depth. Opening the aperture to f/2.8 will increase the resolution by 14% but will shrink the tolerance interval by 33%.
There are surely better MFT lenses for this than the cheap one I chose here. The Pana-Leica Nocticron (same focal length) has much better MTF curves.
