The Flat-Pack Frontier, Pt. 1
A Case Study in Distributed Manufacturing
Digital fabrication and distributed manufacturing have been garnering a lot of pixels lately, cropping up on design blogs and the front pages of major news sites. The concept — a 3-D printer for every home, a CNC mill for every garage — has broad appeal. Libertarians want to unshackle the means of production from government regulation. DIY woodworkers want to build their own milling machines from open-source plans. Surgeons want to scan bodies to practice operations before putting knife to flesh.
The basic technology behind computer-controlling milling (subtractive manufacturing) has been around since the 1950s, first used for producing aluminum ribs for airplane wings from punch-card instructions. About ten years later, the aerospace industry also pioneered laser cutting for hyper-accurate metal machining. 3-D printing (additive manufacturing) was born in the late 1980s. The last five years have seen a burst of consumer-grade machines come to market, fueled by crowd-sourced capital, expiring patents, and the seductive promise of friction-free micro-factories sprouting in every spare bedroom in America.
Trained as an architect, I’ve spent the last ten years engaged in designing and building furniture using traditional methods. Along the way, I documented my work on Instructables.com, an open-source how-to platform started in 2005. Instructables can be a bit chaotic, but its democratic, non-hierarchical nature has opened up the distribution of design instructions, moving DIY writing from the back pages of Popular Mechanics and into your phone. The advent of similar sites — Make, Thingiverse, Ponoko, Shapeways, Materialise, WikiHouse, Open Source Ecology, Tindie, Milkster, Etsy, a host of sub-Reddit threads— have made it easier than ever for designers to distribute their work and makers to build it.
All of these trends got me thinking . . .
Are we really in the midst of a maker revolution? Or is the promise outpacing the reality? Are these new tools and platforms really leveling the playing field or merely re-distributing power into the hands of an affluent class of consumer? Is it really faster, cheaper, greener, or better than making things with traditional means?
I set out to test these claims using open-source tools, networked documentation, and available consumer parts. Over six months, I researched, designed, built, tested, and distributed a chair in an effort to understand the implications of our flat-pack future.
1. Means of Production
Before I started designing, I needed to determine a manufacturing process, as that would govern many subsequent choices. 3-D printing, while progressing rapidly, is hobbled by small build volumes and insufficient strength for load-bearing applications. Laser-cutting was also out—industrial-grade machines able to work larger areas exist, but they were outside of my price range and consume enormous amounts of energy.
CNC technology has several advantages: being the oldest digital fabrication system, it is the most robust and thoroughly de-bugged; it is able to handle furniture-sized stock; it is generally not much more power-hungry than conventional woodworking equipment; and, crucially, I had access to one through Fab Lab Baltimore. A franchise of MIT’s Center for Bits and Atoms, the Fab Lab is embedded in a local community college, providing a variety of digital fabrication equipment and software to Maryland residents.
Choosing a CNC router as my output immediately put some constraints on the design: material was restricted to sheet goods; size was constrained to 4'x8'; and only one side of the workpiece could be machined. In addition, I set some other goals for the project: avoiding finicky joinery that was hard to assemble; keeping assembly tool-free; keeping the number of pieces to a minimum; restricting the cut time to an hour or less; and keeping the parts within a half-sheet of plywood.
2. Research
Once I settled on a manufacturing method, I plunged into research. I visited the Fab Lab, talked to the staff, and took the introductory class that is a prerequisite for using the machinery. Online, I found several great articles on CNC-specific wood joinery. Interested in the evolution of ready-to-assemble (RTA) flat-pack furniture, I dug up what I could on IKEA. Rooting through a vast daisy-chain of links, I stumbled upon some clever furniture that used zip ties as fasteners. Widely available and tool-free to use, I decided to experiment with zip tie joinery instead of complex friction joints or IKEA’s headache-inducing barrel bolts.
3. Design
I started designing by hand, iterating a dozen quarter-scale models. At first, I attempted to make a self-supporting shell of panels joined at the vertices. Base designs for such a scheme proved problematic, so I moved on to a box-shaped frame supporting a “hammock” of articulated panels. Testing that theory, I made two full-size models out of dumpster-rescued corrugated cardboard. However, the cardboard was too light to bear weight and evaluate how the panels moved with respect to one another. I scrounged up some scrap plywood and spent a Saturday afternoon cutting, fitting, and figuring.
After all that head-scratching (a good four months of nights and weekends), it was satisfying to have a full-blown prototype sitting in my living room. The ergonomics were surprisingly nice, it seemed structurally sound, and it was roguishly handsome. I sat down one night and started putting it into the computer, modeling it in both SketchUp and AutoCAD. Problems cropped up: the pieces wouldn’t fit inside a 4'x4' envelope, a fact that had escaped me as I scrapped together V1.0 from random pieces of plywood. I tweaked parameters until everything fit, crossed my fingers, and hoped it would work.
4. Build
I live in the city and drive a little hatchback, sans roof rack, so in order to get plywood I had to trek up to the suburbs and borrow a minivan from my folks. Keys in hand, I drove to the big-box store and bought a sheet of plywood (for the chair) and a sheet of MDF (for the spoil board), each cut in half to fit in the van. From the hardware store, it was a good 20 miles to the Fab Lab, where I had reserved an hour on the router.
By the time I got there, double-parked, hustled the wood through the hallways, re-parked the van, got back, translated the AutoCAD files into PartWorks (ShopBot’s proprietary software), set the spoil board, screwed down the plywood, changed the router bit, calibrated the depths, set the x/y home point, and fired up the machine, I’d been there nearly two hours. The cut itself clocked in a little under an hour. I unloaded the machine, cleaned up, deleted my files from the computer, and got out just before closing time.
But the work had just begun. Since there wasn’t a vacuum table to hold the plywood down, the workpiece has to be screwed in place and sacrificial tabs maintained between the parts to keeps pieces from vibrating loose. The Saturday after my first cut, I spent a couple of hours cutting the tabs, sanding the faces and edges, and waxing the wood. Routed channels for the zip ties had to be chased out with a chisel because bowing in the workpiece had thrown off the calibration of the CNC. So far, “digital fabrication” seemed to involve an awful lot of handwork — at least a 50-50 ratio. Actual assembly with the zip ties only took about fifteen minutes.
The first sit, usually a celebratory moment in a build, was disappointing. The ergonomics had shifted from V1.0 — all the tweaking to fit pieces into the 4'x4' envelope — and now the seat sloped too aggressively, forcing my shoulders to hunch and compressing my abdomen. It was a chore to get in and out of, the front edge of the seat cut into my thighs, and I was frustrated that it took so long to hand-finish after the machine had done all the “work.”
5. Design (Again)
Back in AutoCAD, I made a new version, 2.1. In order to correct the seat-slope problem, I changed the position of the holes in that attached the “hammock” to the frame in an effort to stretch the seat into a more upright posture. I narrowed the arm pieces, shortened the “wheelbase”, and made the middle seat panel wider. Everything dialed in, I made the trip back out to the Fab Lab and cut again.
The Zip Tie Lounger 2.1 also failed; I changed too many variables in the computer without checking them against real-world models. Though marginally closer to a solid ergonomic shape, the seat needed to come still further horizontal. It was back to the drawing board. I lowered the front support, adjusted the attachment points in the “hammock”, and modeled the changes in the computer to track my changes more accurately.
Despite a cut plagued by two broken bits (all user error, I must admit), V2.2 of the chair finally came out as I had imagined. The side pieces of the box frame were changed to a more asymmetric profile, creating a hierarchy between front and back. The “hammock” finally conformed to a perfectly comfortable slope. With so much practice, I had become much quicker at the sanding, finishing, and assembly.
The final product, from a pure design standpoint, hit most of my self-identified criteria. Just shy of twenty pounds, the parts pack into a slim 2'x2' footprint for shipping. As an object, it does fail to surmount flat-pack design’s fundamental paper-cutout aesthetic, though it manages to ameliorate some of the boxiness associated with CNC design pieces. Ergonomically, it is remarkably comfortable. The seat, while made of flat plates, leverages the flexibility inherent in zip tie joinery for a pleasant, body-conforming articulation.
And then there are the zip ties themselves : an artifact invented for an entirely different purpose, now adapted to new ends. The design successfully exploits three critical qualities — cheapness, ubiquity, flexibility — to improve the final object. As our consumer-driven economy produces more and more stuff, to more and more questionable ends, it is critical to leverage unexpected resources into sustainable solutions.
. . . stay tuned for Part 2, where I will examine the final chair, the means of distribution, and quantify the friction points that plagued the process.