It was plain that any future modelling efforts would require extensive use of 3D printing, and to my own designs....
The first option I considered was to do these through Shapeways. However the big drawbacks of cost, turnaround time and shipping everything back to Australia made that look impractical, especially given how many different 3D printed objects I was looking at. Even the very real benefits of having a wide variety of materials to choose from and being able to sell some of the less-specialised designs were not enough to make this viable for me.
The second option was to make use of the network of local 3D printing hobbyists and services, via 3dhubs.com or similar. This is cheaper and avoids much of the delay, but is still awkward, particularly when experimenting. I did several trial prints this way, which was enough to convince me that I could design what I needed.
The third and final option was to go out and buy a printer. I settled on the FlashForge CreatorPro, one of the better low-end products as of 2017. I have been using it on-and-off for nearly a year now, and it is serving me well. I chose it for a combination of decent resolution (0.1mm), proper handling of ABS (my material of choice), a reputation for good clean builds, and a build volume adequate to my needs (just under 9" x 6" x 6").
Learning how to print successfully involved downloading all sorts of sample designs from the Web, mainly from Thingiverse.com, and just playing: re-scaling, trying different build orientations and supports, etc. The main lessons learned were:
- orienting or splitting the model to minimise support requirements
- 0.1mm resolution does NOT mean 0.1mm features: 0.5mm is about the minimum practical feature size.
- how to detach the finished models from the build plate, prizing them off with a paint scraper as the plate cooled down through around 70-80 degrees.
The other big decision was choosing a CAD tool. After a few experiments and false starts, OpenSCAD won the day. This lets you build a model by adding a subtracting simple geometric shapes, using a script language based on the C programming language. This is a very simple, straightforward and powerful approach for someone with programming experience, but might be less suitable for those without a software background.
The major pros and cons of OpenSCAD are:
- good for modelling artificial objects: structures, vehicles, brackets, etc.
- poor for modelling naturally curved shapes: people, trees, etc.
- good for parameterised models: varying the size, quantity and arrangement of parts of objects like viaducts, footbridges, buildings, etc. is very simple
- poor speed when rendering complex objects: the math library used cannot take full advantage of modern computer hardware
- free open source software, available for all major computers.