Intro
I’ve always found something quite appealing about the Nilmslo 3D camera. It’s well inside the toy camera category but there’s something distinctly appealing about the wigglegrams it can be used to make. Despite the low-tech nature of the process (and the so-so image quality!) they’re great for revealing a sense of place and scale without needing specialised glasses or headware for the user.
This page is going to be a quick document covering my first iteration at this and I hope some of the dead-ends I ended up exploring might be useful for others going down the same path.
The Build
After some experimentation I ended up settling on creating a focusable dual-lens 3D stereo camera based on a couple of 40mm focal length achromats. It has a Fuji-X mount and I’d like to think the focusing feels pretty smooth. You can get the parts and print it yourself on thingiverse here.
The baseline is tiny (~17mm!). However with enough foreground to see stereo separation it’s possible to get some fun wigglegrams out of it.
Behind the lenses is a baffle with helps manage light leaking from one half to the other (thanks to geo.moua for the inspiration there!) These project on to the image sensor with some pretty minimal overlap. The images only need a little post-processing to align and get a wigglegram out.
(failed) Quad-lens version
I came into this project absolutely convinced I could add four lenses and project these images on to the sensor to get not only stereo but several plenoptic views to get kind of a circular wiggle for a wigglegram. Initially: the build looked good!
It even made pictures – but there was vanishingly small overlap between the frames, as well as significant leakage between neighbouring lenses.
What was going on? In short, even though I was able to get four pretty decent sized projections, their centres were far towards the corner of the sensor frame and so the actual overlap regions between the lenses was unusably small.
(better) Dual-lens version
By switching to just dual lenses, I was able to push the image centres together so they utilised more of the sensor and had a better usable overlap area. I also added a baffle at the rear of the lens to crop down the image circles so that they interfere less (thanks to a great tip from geo.moua on IG).
The lens chassis has several pins which fit into grooves coming perpendicular out of the lens body; these pins then slot into spiral grooves in the focusing ring. Turning the focusing ring moves the chassis forward without rotating it; any rotations would break the stereo.
The design worked out pretty well, and the focusing mechanism is smooth and doesn’t require me to look at what is happening to be confident of the result. Image quality is toylike, but thats par for the course for a single not-particularly-well-corrected achromat lens.
Whats next
The stereo baseline on this is tiny; you need to get right up close to subjects (i.e. have them in the foreground) to demonstrate any sort of meaningful stereo effect. I’d like to play with mirrors on the front-side of a conventional lens to get a wider stereo baseline; moreover I think it could be extended to have several optical paths to have a plenoptic image paths (i.e. maybe revisit the quad lens idea).
