For more than a decade, I’ve been fascinated in making machines that make things. As such, I was destined to encounter the concept of RepRap at some point in my life.
Current state of FDM printers are culmination of RepRap. Indeed, they’re so mature now that I only own Bambu printers. Far from original RepRap philosophy, but very useful as a tool.
However, within the things I want to make, I see patterns of hard-to-make things
- HCI devices (data gloves, HMDs): actuators, linkages, optical housing, etc.
- Swarm robots (like ones in Scaffold): ditto
Aside from obviously difficult and niche parts like lenses or fabrics, common in these are small mechanical and/or electromagnetic parts with high accuracy.
Many people would enjoy having such prototyping capability in their homes or workshops.
Thus the search started…
Powder Bed Fusion and CNC Mills
When talking about versatile metal manufacturing, powder bed fusion looks promising. De-facto method for metal additive manufacturing, and simple software.
However, when thinking more deeply, it quickly loses attractiveness:
- Inaccurate: Surface finish of the highest-grade machines seem to be on par with FDM.
- Unsafe: Metal powder, metal vapor, purge gases, high-power lasers.
- Difficult: Performance most likely depends on laser optics and powder physics. They’re not easy.
- Inefficient: Pre-heating can take 30 min+, and powder cannot be reused 100%.
I still think it’s physically viable to miniaturize PBF machine to a home-friendly form factor. But I can’t imagine this kind of machine evolving in RepRap way.
Then, what about CNC mills? Maybe it’s possible to automate the CAM to get “3D printing” feeling. However, my experience of owning both desktop mills (both non-CNC and CNC) suggests otherwise:
- Unfavorable physics: bigger & heavier → faster & more accurate
- Metal chips: Very difficult to contain them 100%. They also tend to prevent 100% automation, by clogging or breaking tools.
- Limited tools: Many specialized tools are needed to be versatile.
Maybe chips and tools could be made OK. But the underlying physics is unavoidable.
These methods, be it drills, mills, lathes, grinders, all depends on slamming hard blades or grains to deform - and eventually cut off - tiny pieces of the work. Both speed and accuracy come from how much energy can be put into small targeted regions. When that’s dependent on contact force, of course machine needs to be rigid.
For miniaturizaton, we need to look into more esoteric means of operation.
Electric Discharge Machining
There’s a method called EDM (Electric Discharge Machining). You can watch a nice video by Applied Science to get an idea.
Within so called non-conventional machining, EDM is much less known compared to more flashy ones like laser cutting. However, I find EDM to be surprisingly nice to work with.
- More than half a century of history
- Extensive research literature
- Constructible with decades old tech
- Fundamental patents are expired
- Can work with almost any shape of work & tool
- Tractable at small scale, as shown in BAXEDM and Rack Robotics
Basically, EDM transforms any conductor into magical plasma tool that removes the work on contact. Like a boolean “remove” operator in modeling software.
If it’s so good, why we don’t see them used everywhere?
Big elephant in the room is that discharge plasma works both ways; it also erodes the tool!1 This means tool shape can change over time, and they also needs replacement.
Thus, currently common form of EDM are limited to applications that can circumvent tool wear.
- Wire EDM: Cuts metal slabs using a wire as the tool.
- Workaround: cotinuously fed wire
- Die-sinker EDM: Copies die shape.
- Workaround: die-specific tuning of parameters. dies are also consumables.
- WEDG & micro-drilling: Create tiny high-aspect-ratio holes.
- Workaround: 2-pass EDM; one to form the drill tool using wire EDM, another to use the ephemeral tool to drill.
Electrode wear if often measured in a metric called EWR (Electrode Wear Ratio).
EWR = (volume of electrode removed) / (volume of work removed)
In a very specific setup (e.g. material combination, waveform), EWR can reach less than 1%. But if we aim for more general and/or faster setup, EWR of a few % ~ even 30% is commonplace.
Are we stuck? I think not. WEDG shows the way. With careful control, we should be able to extract much more from the process.
TBD…
The machine does not work yet, better make it than talking!
- Leveraging Cheap Compute
Footnotes
-
Another big reason is EDM is less efficient in (material removed)/(energy spent). ↩